Your SlideShare is downloading. ×
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Community
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×
Saving this for later? Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime – even offline.
Text the download link to your phone
Standard text messaging rates apply

Community

4,195

Published on

colombia guerra

colombia guerra

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
4,195
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
55
Comments
0
Likes
0
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide

Transcript

  • 1. Community Ecology
  • 2. This page intentionally left blank
  • 3. Community EcologyProcesses, Models,and ApplicationsEDITED BYHerman A. VerhoefVU University, Amsterdam, Department of Ecological Science,the NetherlandsPeter J. MorinRutgers University, Department of Ecology, Evolution & NaturalResources, USA1
  • 4. 3Great Clarendon Street, Oxford OX2 6DPOxford University Press is a department of the University of Oxford.It furthers the University’s objective of excellence in research, scholarship,and education by publishing worldwide inOxford New YorkAuckland Cape Town Dar es Salaam Hong Kong KarachiKuala Lumpur Madrid Melbourne Mexico City NairobiNew Delhi Shanghai Taipei TorontoWith offices inArgentina Austria Brazil Chile Czech Republic France GreeceGuatemala Hungary Italy Japan Poland Portugal SingaporeSouth Korea Switzerland Thailand Turkey Ukraine VietnamOxford is a registered trade mark of Oxford University Pressin the UK and in certain other countriesPublished in the United Statesby Oxford University Press Inc., New York# Oxford University Press 2010The moral rights of the author have been assertedDatabase right Oxford University Press (maker)First published 2010All rights reserved. No part of this publication may be reproduced,stored in a retrieval system, or transmitted, in any form or by any means,without the prior permission in writing of Oxford University Press,or as expressly permitted by law, or under terms agreed with the appropriatereprographics rights organization. Enquiries concerning reproductionoutside the scope of the above should be sent to the Rights Department,Oxford University Press, at the address aboveYou must not circulate this book in any other binding or coverand you must impose the same condition on any acquirerBritish Library Cataloguing in Publication DataData availableLibrary of Congress Cataloging in Publication DataLibrary of Congress Control Number: 2009934138 ¨Cover illustration by Janine MarienTypeset by SPI Publisher Services, Pondicherry, IndiaPrinted in Great Britainon acid-free paper byCPI Antony Rowe, Chippenham, WiltshireISBN 978–0–19–922897–3 (Hbk.)ISBN 978–0–19–922898–0 (Pbk.)1 3 5 7 9 10 8 6 4 2
  • 5. ContentsPreface xiList of Contributors xiiiIntroduction 1Part I Shape and Structure 51 The topology of ecological interaction networks: the state of the art 7 Owen L. Petchey, Peter J. Morin and Han Olff 1.1 Introduction 7 1.1.1 What do we mean by the ‘topology’ of ecological networks? 7 1.1.2 Different types of ecological networks 9 1.1.3 Three general questions 11 1.2 Competitive networks 11 1.2.1 Structural regularities 11 1.2.2 Mechanisms 13 1.2.3 Unresolved issues 15 1.3 Mutualistic networks 15 1.3.1 Structural regularities 15 1.3.2 Mechanisms 16 1.3.3 Unresolved issues 16 1.4 Food webs 17 1.4.1 Structural regularities 17 1.4.2 Mechanisms 18 1.4.3 Unresolved issues 21Part II Dynamics 232 Trophic dynamics of communities 25 Herman A. Verhoef and Han Olff 2.1 What types of dynamics can be distinguished? 25 2.1.1 Stable equilibria 25 2.1.2 Alternate equilibria 26 2.1.3 Stable limit cycles 26 2.1.4 Chaotic dynamics 26 2.2 Dynamics of food web modules 26 2.3 Internal dynamics in food web modules or simple webs 29 2.4 Dynamics enforced by external conditions 32
  • 6. vi CONTENTS 2.5 Equilibrium biomass at different productivities 33 2.6 Dynamics of complex interactions 34 2.7 Conclusions 343 Modelling the dynamics of complex food webs 37 Ulrich Brose and Jennifer A. Dunne 3.1 Introduction 37 3.2 Simple trophic interaction modules and population dynamics 37 3.3 Scaling up keystone effects in complex food webs 39 3.4 Diversity/complexity–stability relationships 40 3.5 Stability of complex food webs: community matrices 40 3.6 Stability of complex food webs: bioenergetic dynamics 41 3.7 Stability of complex food webs: allometric bioenergetic dynamics 42 3.8 Future directions 444 Community assembly dynamics in space 45 Tadashi Fukami 4.1 Introduction 45 4.2 Determinism and historical contingency in community assembly 46 4.3 Community assembly and spatial scale 48 4.3.1 Patch size 48 4.3.2 Patch isolation 49 4.3.3 Scale of environmental heterogeneity 51 4.3.4 Synthesis 52 4.4 Community assembly and species traits 52 4.5 Conclusions and prospects 53Part III Space and Time 555 Increasing spatio-temporal scales: metacommunity ecology 57 Jonathan M. Chase and Janne Bengtsson 5.1 Introduction 57 5.2 The varied theoretical perspectives on metacommunities 58 5.2.1 Neutral 60 5.2.2 Patch dynamics 60 5.2.3 Species sorting 60 5.2.4 Mass effects 60 5.3 Metacommunity theory: resolving MacArthur’s paradox 61 5.4 As easy as a, b, g: the importance of scale 61 5.5 Species–area relationships and metacommunity structure 63 5.6 Effects of dispersal rates on local communities 64 5.7 Local–regional richness relationships 65 5.8 A synthesis of metacommunity models 65 5.9 Adding food web interactions into the equation 66 5.10 Cross-ecosystem boundaries 66 5.11 Conclusions 68
  • 7. CONTENTS vii6 Spatio-temporal structure in soil communities and ecosystem processes 69 Matty P. Berg 6.1 Introduction 69 6.2 Soil communities, detrital food webs and soil processes 69 6.3 Soil organic matter 71 6.4 Variability in time in soil communities 72 6.5 Variability across horizontal space in soil communities 75 6.6 Variability across vertical space in soil communities is high 77 6.7 Spatio-temporal scales of community studies 79Part IV Applications 817 Applications of community ecology approaches in terrestrial ecosystems: local problems, remote causes 83 Wim H. van der Putten 7.1 Introduction 83 7.1.1 Issues in applied community ecology 83 7.1.2 Top-down and bottom-up go hand in hand 84 7.2 Community interactions across system boundaries 85 7.2.1 Linkages between adjacent or distant ecosystems 85 7.2.2 Linkages between subsystems: aboveground–belowground interactions 85 7.2.3 Consequences for application: find the remote cause of local effects 86 7.3 Community interactions and land use change 87 7.3.1 Land use change, predictability and major drivers of secondary succession 87 7.3.2 Secondary succession from an aboveground–belowground perspective 88 7.3.3 Consequences for restoration and conservation 89 7.4 Biological invasions 90 7.4.1 Community-related hypotheses that explain biological invasions 90 7.4.2 Mount Everest or tip of the iceberg? 91 7.4.3 Conclusions and consequences for management 92 7.5 Discussion, conclusions and perspectives 928 Sea changes: structure and functioning of emerging marine communities 95 J. Emmett Duffy 8.1 Introduction 95 8.1.1 Fishing as a global experiment in community manipulation 96 8.1.2 Physical forcing and the uniqueness of marine ecosystems 96 8.2 The changing shape of marine food webs 97 8.2.1 Conceptual background 97 8.2.2 Empirical evidence for trophic skew in the ocean 99 8.3 Trophic cascades in the sea 100 8.3.1 Conceptual background 100 8.3.2 Evidence for trophic cascades in open marine systems 101 8.3.2.1 Rocky bottoms 102 8.3.2.2 Continental shelves 103 vii
  • 8. viii CONTENTS 8.3.2.3 Pelagic systems 103 8.4 Biodiversity and stability of marine ecosystems 104 8.4.1 Conceptual background 104 8.4.2 Evidence linking diversity and stability in marine systems 105 8.4.2.1 Comparisons through time 105 8.4.2.2 Comparisons across space 107 8.4.2.3 Mechanisms 108 8.5 Interaction strengths and dynamic stability in marine food webs 109 8.5.1 Conceptual background 109 8.5.2 Empirical evidence 110 8.6 Alternate stable states and regime shifts in marine ecosystems 110 8.6.1 Conceptual background 110 8.6.2 Empirical evidence for regime shifts in marine ecosystems 112 8.6.2.1 Mechanisms 112 8.7 Emerging questions in emerging marine ecosystems 1139 Applied (meta)community ecology: diversity and ecosystem services at the intersection of local and regional processes 115 Janne Bengtsson 9.1 Introduction 115 9.2 A theoretical background 116 9.2.1 A simplified historical narrative 116 9.2.2 Implications of metacommunity theory 118 9.2.3 Metacommunities in human-dominated landscapes: effects of habitat loss and fragmentation 121 9.3 A selection of empirical studies 123 9.3.1 Applied questions allow experimental studies on management scales 123 9.3.2 Biodiversity in human-dominated landscapes: local or landscape management? 124 9.3.3 Local and regional effects on ecosystem services 127 9.3.4 What have we learned in the context of metacommunity ecology? 12910 Community ecology and management of salt marshes 131 Jan P. Bakker, Dries P.J. Kuijper and Julia Stahl 10.1 Introduction 131 10.2 Natural salt marsh: the back-barrier model including a productivity gradient 132 10.3 Effects of plants on herbivores (bottom-up control) 133 10.4 Effects of intermediate-sized herbivores on plants (top-down control) 134 10.4.1 Experimental evidence 134 10.4.2 Effects of herbivores at high marsh 135 10.4.3 Low marsh 136 10.5 Large-scale effects of an intermediate herbivore on salt-marsh vegetation 137 10.6 Interaction of herbivory and competition 137 10.7 Competition and facilitation between herbivores 139 10.7.1 Short-term competition and facilitation between hares and geese 139 10.7.2 Long-term facilitation between herbivores 140 10.8 Exclusion of large herbivores: effects on plants 141 10.8.1 Natural marshes 141
  • 9. CONTENTS ix 10.8.2 Artificial salt marshes 142 10.9 Exclusion of large herbivores: effects on invertebrates 143 10.10 Exclusion of large herbivores: effects on birds 144 10.10.1 Migrating birds 144 10.10.2 Breeding birds 145 10.11 Ageing of salt marshes and implications for management 146Part V Future Directions 14911 Evolutionary processes in community ecology 151 Jacintha Ellers 11.1 Introduction 151 11.1.1 Bridging the gap between evolutionary biology and community ecology 151 11.2 Evolutionary biology: mechanisms for genetic and phenotypic change 152 11.2.1 Benefits and maintenance of genetic diversity at the population level 152 11.2.2 The source and nature of genetic variation 154 11.2.3 The relationship between genetic and phenotypic diversity 154 11.3 Proof of principle: community properties result from genetic identity and selection at the level of individual organisms 155 11.4 Effects of genetic and phenotypic diversity on community composition and species diversity 156 11.4.1 Effects of genetic diversity on community functioning 156 11.4.2 Diversity begets diversity? 157 11.4.3 Phenotypic diversity is also important for community diversity and composition 158 11.4.4 Phenotypic plasticity and invasive success 159 11.5 Effect of community composition on the genetic and phenotypic diversity of single species 160 11.6 Future directions 16112 Emergence of complex food web structure in community evolution models 163 Nicolas Loeuille and Michel Loreau 12.1 A difficult choice between dynamics and complexity? 163 12.2 Community evolution models: mechanisms, predictions and possible tests 165 12.2.1 One or many traits? 165 12.2.1.1 Models in which species are defined by many traits 166 12.2.1.2 Models with a limited number of traits 166 12.2.2 Evolutionary emergence of body-size structured food webs 168 12.2.3 Advantages of simple community evolution models 170 12.2.3.1 Comparison with other community evolution models 170 12.2.3.2 Comparison with binary qualitative models 170 12.2.3.3 Testing predictions 172 12.3 Community evolution models and community ecology 173 12.3.1 Community evolution models and the diversity–stability debate 173 12.3.2 Effects of perturbations on natural communities 174 12.3.3 Models with identified traits: other possible applications 175 12.4 Conclusions, and possible extensions of community evolution models 176 12.4.1 Possible extensions of community evolution models 177 12.4.2 Empirical and experimental implications of community evolution models 177
  • 10. x CONTENTS13 Mutualisms and community organization 179 David Kothamasi, E. Toby Kiers and Marcel G.A. van der Heijden 13.1 Introduction 179 13.2 Conflicts, cooperation and evolution of mutualisms 180 13.2.1 Mutualism can also develop without evolution 183 13.3 Mutualisms in community organization 183 13.3.1 Plant–pollinator interactions 183 13.3.2 Plant–protector mutualism 184 13.3.3 Plant nutrition symbiosis 185 13.3.3.1 Legume–rhizobia symbioses 185 13.3.3.2 Mycorrhizal symbioses 188 13.4 Conclusions 19114 Emerging frontiers of community ecology 193 Peter J. Morin 14.1 Introduction 193 14.1.1 Spatial ecology 193 14.1.2 Complex dynamics 193 14.1.3 Size-dependent interactions 193 14.1.4 Interactions between topology and dynamics 193 14.1.5 Evolutionary community dynamics 194 14.1.6 Applied community ecology 195 14.2 Future directions 196 14.2.1 Biotic invasions 196 14.2.2 Interaction networks beyond food webs 199References 203Index 245
  • 11. PrefaceIn 2001 H.A.V. started a course for second-year turers were accompanied by some of their PhDundergraduate biology students at the Vrije Uni- students, creating an international group of com-versiteit Amsterdam entitled Community Biology. munity ecologists. The course was not intended toThis course has now been running successfully for be encyclopaedic, but rather it focused on the areas8 years. The course was obligatory for all biology of expertise of the invited speakers, many of whichstudents, and it differed from other courses in that share the theme of patterns and processes emergingit was multidisciplinary and provided the students from ecological networks. Participants addressedwith opportunities to perform their own research. the state of the art in theory and applications ofThe multidisciplinarity was emphasized by the dif- community ecology, with special attention to topol-ferent disciplines of the teachers on the course: soil ogy, dynamics, the importance of spatial and tem-ecology, plant ecology, systems ecology, microbial poral scale and the applications of communityphysiology and theoretical biology. The important ecology to emerging problems in human-domi-task of finding a textbook that could link all disci- nated ecosystems, including the restoration andplines and encourage participating lecturers to de- reconstruction of viable communities. The courseliver a unified course was solved by using finished with speculations about future researchCommunity Ecology by P.J.M. That book linked the directions. During the course, it became clear thatdifferent subjects of community ecology, and this international group of students appreciated theintegrated the more theoretical parts on modelling information presented by the various lecturers, de-with the empirical studies, including topics such as spite the fact that research topics exhibited greatbiodiversity and applied studies. Subsequently, diversity. It was during this very stimulating courseH.A.V. and P.J.M. met at an international meeting that the idea for this book took form. H.A.V. andon food webs, and discussed the possibility of par- P.J.M., the editors, convinced most lecturers toticipating in a similarly themed graduate-level transform their lectures into book chapters, andcourse. And, thus, our current collaboration began. asked other colleagues to fill in some gaps. TheIn The Netherlands PhD students from different result captures much of the excitement about com-universities are organized into interdisciplinary munity ecology expressed during the course, andthematic groups, called research schools, that pro- expands the coverage of topics beyond what wevide an intellectual support base for instruction and were able to discuss in an intensive week-longresearch. For example, students working in the field course. We recognize at the outset that certain sub-of socioeconomic and natural sciences of the envi- disciplines of community ecology are not coveredronment belong to the Research School SENSE. In here, and we do not claim otherwise. We know that ´2005, H.A.V., Andre de Roos, Claudius van de Vij- the topics addressed here will be of interest tover and Johan Feenstra organized a PhD course on advanced students and practitioners of communityCommunity Ecology for the SENSE PhD ecology. Ultimately, 19 colleagues participated inprogramme. During this 1 week course held in writing this book. We thank them all for their im-Zeist, leading researchers in the field of Community portant contributions. Writing book chapters,Ecology from Europe and the USA were asked to strangely enough, is less valued than writing arti-deliver lectures on recent and often unpublished cles for scientific journals in some academic circles.developments in their areas of expertise. The lec- Still, like the multidisciplinary course mentioned xi
  • 12. xii PREFACEabove, we find that the interactive writing that hap- well as the students who have taken the communitypens when people from different subdisciplines ecology course that he has taught at Rutgers Uni-work together is a fascinating, synergistic and pro- versity since 1983. Their collective comments andductive process. feedback have helped to refine his perspectives We would like to thank friends and colleagues about the nature of community ecology over thewho were indispensable during the process of years. Thanks also go to participants in a recentwriting: H.A.V. thanks Nico van Straalen, who by seminar on Ecological Networks for critical feed-writing his book Ecological Genomics for Oxford back on some of the writing that appears here,University Press acted as an instigator for this including Mike Sukhdeo, Maria Stanko, Waynebook. H.A.V. also thanks his colleagues who made Rossiter, Tavis Anderson, Faye Benjamin, Denisethe Community Biology course a success for so Hewitt, Kris Schantz and Chris Zambel. ¨many years: Wilfred Roling, Bob Kooi, Matty Berg, H.A.V. and P.J.M. both thank Ian Sherman ofWilfried Ernst, Tanja Scheublin, Diane Heemsber- Oxford University Press, who was immediately en-gen, Stefan Kools, Marcel van der Heijden, Susanne thusiastic about this book project, and Helen Eaton,de Bruin, Lothar Kuijper, Rully Nugroho and Henk who as assistant commissioning editor played avan Verseveld. H.A.V. acknowledges colleagues crucial role in the development of the book.who were directly involved in the organization of H.A.V. thanks Emilie Verhoef, without whom ´the PhD course: Andre de Roos, Claudius van de this book probably would never have been pro-Vijver, Johan Feenstra and Ad van Dommelen. H.A. duced. P.J.M. thanks Marsha Morin for her under-V. is very grateful to his critical friend, John Ash- standing and support during another extendedcroft (Durham), for supportive focusing. writing project. P.J.M. thanks the many students who partici- Herman A. Verhoef, Amsterdampated in the Zeist Community Ecology Course, as Peter J. Morin, New Brunswick
  • 13. List of ContributorsJan P. Bakker, Community and Conservation Tadashi Fukami, Department of Biology, Stanford Ecology Group, University of Groningen, PO Box University, Stanford, CA 94305, USA, 14, 9750 AA Haren, The Netherlands, Email: tfukami@hawaii.edu Email: j.p.bakker@rug.nl E. Toby Kiers, VU University, Amsterdam,Janne Bengtsson, Department of Ecology and Crop Department of Ecological Science, Production Science, PO Box 7043, Swedish De Boelelaan 1085, 1081 HV Amsterdam, University of Agricultural Sciences, SE-750 07 The Netherlands, Uppsala, Sweden, Email: toby.kiers@falw.vu.nl Email: Jan.Bengtsson@ekol.slu.se David Kothamasi, Centre for Environmental Man-Matty P. Berg, VU University, Amsterdam, Depart- agement of Degraded Ecosystems, University of ment of Ecological Science, De Boelelaan 1085, Delhi, Delhi 110007, India, 1081 HV Amsterdam, The Netherlands, Email: dmkothamasi@cemde.du.ac.in Email: matty.berg@falw.vu.nl Dries P.J. Kuijper, Mammal Research Institute,Ulrich Brose, Darmstadt University of Technology, Polish Academy of Sciences, ul. Waszkiewicza Department of Biology, Schnittspahnstr. 10, ˙ 1c, 17–230 Białowieza, Poland, 64287 Darmstadt, Germany; Pacific Ecoinfor- Email: dkuijper@zbs.bialowieza.pl matics and Computational Ecology Lab, 1604 Nicolas Loeuille, Laboratoire d’Ecologie, UMR7625, McGee Avenue, Berkeley, CA 94703, USA, ´ Universite Paris VI, 7 quai St Bernard, F75252 Email: brose@bio.tu-darmstadt.de Paris Cedex 05, France,Jonathan M. Chase, Department of Biology and Tyson Email: nicolas.loeuille@normalesup.org Research Center, Box 1229, Washington University Michel Loreau, McGill University, Department of in Saint Louis, Saint Louis, MO, USA, Biology, 1205 avenue du Docteur Penfield, Mon- Email: jchase@wustl.edu; chase@biology2.wustl.edu ´ ´ treal, Quebec, Canada, H3A 1B1,J. Emmett Duffy, School of Marine Science and Email: michel.loreau@mcgill.ca Virginia Institute of Marine Science, The College Peter J. Morin, Rutgers University, Department of William and Mary, Gloucester Point, VA of Ecology, Evolution, & Natural Resources, 23062-1346, USA, 14 College Farm Road, New Brunswick, Email: jeduffy@vims.edu NJ 08901, USA,Jennifer A. Dunne, Santa Fe Institute, 1399 Hyde Email: pjmorin@rci.rutgers.edu Park Road, Santa Fe, NM 87501; Pacific Ecoin- Han Olff, University of Groningen, Community and formatics and Computational Ecology Lab, 1604 Conservation Ecology Group, PO Box 14, 9750 AA McGee Avenue, Berkeley, CA 94703, USA, Haren, The Netherlands, Email: jdunne@santafe.edu Email: h.olff@rug.nlJacintha Ellers, VU University, Amsterdam, Owen L. Petchey, University of Sheffield, Depart- Department of Ecological Science, De Boelelaan ment of Animal and Plant Sciences, Alfred Denny 1085, 1081 HV Amsterdam, The Netherlands, Building, Western Bank, Sheffield S10 2TN, UK, Email: jacintha.ellers@falw.vu.nl Email: o.petchey@sheffield.ac.uk xiii
  • 14. xiv LIST OF CONTRIBUTORSJulia Stahl, Landscape Ecology Group, University of Wim H. van der Putten, Netherlands Institute Oldenburg, PO Box 2593, D-26111 Oldenburg, of Ecology, Centre for Terrestrial Ecology Germany, (NIOO-KNAW) PO Box 40, 6666 ZG Heteren; Email: julia.stahl@uni-oldenburg.de Laboratory of Nematology, WageningenMarcel G.A. van der Heijden, VU University, University, PO Box 8123, 6700 ES Wageningen, Amsterdam, Department of Ecological Science, The Netherlands, De Boelelaan 1085, 1081 HV Amsterdam, The Email: w.vanderputten@nioo.knaw.nl ¨ Netherlands; Agroscope Reckenholz-Tanikon, Herman A. Verhoef, VU University, Amsterdam, Research Station ART, Reckenholzstrasse 191, Department of Ecological Science, De Boelelaan ¨ 8046 Zurich, Switzerland, 1085, 1081 HV Amsterdam, The Netherlands, Email: marcel.vanderheijden@art.admin.ch Email: herman.verhoef@falw.vu.nl
  • 15. Introduction Herman A. Verhoef and Peter J. MorinFirst of all, we must clarify what we consider to be American botanist H.A. Gleason. His focus was onthe subject matter of community ecology. In his the traits of individual species that allow each toclassic textbook, Krebs (1972) described a commu- live within specific habitats or geographical ranges.nity as ‘an assemblage of populations of living A community can then be seen as an assemblage oforganisms in a prescribed area or habitat’. He de- populations of different species whose traits allowscribed this as ‘the most general definition one can them to persist in a prescribed area (Gleason 1926).give’. The ‘prescribed area’ suggests that we are This is a much more arbitrary unit than that envi-dealing with some sort of spatially bounded sys- sioned by Clements. In Gleason’s view, spatialtems, and it is the somewhat arbitrary nature of boundaries of communities are not sharp and thepossible boundaries that has caused continuing species assemblages may change considerably overdebate about the nature of communities (e.g. Rick- time and space.lefs 2008). It is debatable whether community ecol- We feel that the current broadly accepted viewogy is a proper discipline at all if communities do about the nature of communities is much closer tonot exist as natural definable units (Begon et al. Gleason’s opinion. A given species can occur in1996). At the beginning of the 20th century there rather different collections of species, or commu-was much debate about the nature of communities. nities, under different circumstances. CommunityThe driving question was whether the community ecology can, therefore, be described as the study ofwas a self-organized system of co-occurring species the community level of organization, rather than ofor simply a haphazard collection of populations a spatially and/or temporally definable unit (Begonwith minimal functional integration. At that time, et al. 1996).two extreme views prevailed: one view considered As suggested by Morin (1999), we include withina community as a superorganism whose member community ecology the study of two or more spe-species were tightly bound together by interactions cies at a location, including predator–prey inter-that contributed to repeatable patterns of species actions, interspecific competition, interactions ofabundance in space and time. This concept led to mutualism and parasitism. We recognize thatthe assumption that natural communities exist as some ecologists will argue that such pair-wise in-fundamental units, and this made it possible to teractions are simply aspects of population biology,classify communities in a manner comparable to but we also point out that they provide the basicthe Linnaean taxonomy of species. One of the lead- interactions that contribute to patterns involvinging proponents of this approach was the Nebraskan larger numbers of species. We also recognize that,botanist Frederick E. Clements. His view became even in well-studied communities, much of theknown as the organismic concept of communities. It complexity and daunting diversity involving bacte-assumes a common evolutionary history for the ria, fungi, protists and small invertebrates remainsintegrated species (Clements 1916). poorly known. Nonetheless, fascinating patterns The opposite view has been termed the individu- among well-studied groups of organisms demandalistic continuum concept and was advocated by the our attention and beg for explanation. 1
  • 16. 2 INTRODUCTION In this book, we have grouped the various chap- Community ecology has many real-world appli-ters into several areas: (1) community shape, struc- cations deriving from the fact that the abundance ofture and dynamics, (2) communities over space and species strongly depends on competitive and pred-time, (3) applications of community ecology and (4) atory interactions. These interactions are major dri-future directions of community research. We start vers of ecosystem processes, and they are the key towith shape and structure, with a focus on the topolo- the delivery of ecosystem goods and services.gy of networks of interacting organisms in ecological Chapter 7 focuses on applications of communitysystems. Chapter 1 deals with subsets of the full ecology approaches in terrestrial ecosystems: localnetwork of interactions: competition networks, mu- problems, remote causes. In terrestrial ecosystems,tualistic networks and consumption networks (food there is a clear link between belowground andwebs). It involves recent studies and important ques- aboveground subsystems, which is to be comparedtions about the structure and the processes or me- with benthic–pelagic coupling in aquatic or marinechanisms that may result in their structure. Chapter systems. These above/belowground linkages influ-2 follows with dynamics, dealing with the trophic ence many major processes, such as ecosystem res-dynamics of communities. The focus is on types of toration following land abandonment or the fate ofdynamics that can be distinguished, the dynamics introduced exotic alien species. Some examples ofof simple and complex interactions and the internal above/belowground interactions in relation toand external determinants of food web dynamics. changing land usage and biological invasions areChapter 3 considers modelling of the dynamics of com- discussed and related to climate warming. Chapterplex food webs, with a focus on whether embedding 8 reviews the consequences of industrial-scalesimple modules in complex food webs affects the fishing on marine communities and is entitled seadynamics of those modules. The dynamic analysis changes: structure and functioning of emerging marineof complex food webs is organized around the rela- communities. Growing evidence from models andtionships between complexity and stability and empirical time series suggests that fishing pressurediversity and stability. Body-size-dependent scaling can cause relatively rapid regime shifts into distinctof biological rates of populations emerge as a possi- semi-stable states that resist change even after fish-ble solution to the instability predicted for complex ing has been relaxed. Although theory predicts thatfood webs. Chapter 4 forms a bridge between a preponderance of weak interactions stabilizesdynamics and the next section on space and time. food web dynamics, these weak interactions mayThe subject of this chapter is community assembly be unlikely to buffer marine ecosystems from thedynamics in space, with a focus on three elements of impacts of intense exploitation by humans. Thus,assembly dynamics: the rate of immigration to local the systematic reduction in average food chaincommunities, the degree to which the species pool is length documented in oceanic and coastal ecosys-external to local community dynamics and the tems can initiate regime shifts to alternate semi-amount of variation in immigration history among stable states, and is already affecting the provisionlocal communities. The hypothesis is that these ele- of marine ecosystem services to society. In Chapterments together determine the degree of historical 9, which focuses on applied (meta)community ecology:contingency. Chapter 5, on increasing spatio-temporal diversity and ecosystem services at the intersection ofscales: metacommunity ecology, discusses whether a local and regional processes, several expectationsmetacommunity perspective can be helpful in from metacommunity theory on the effects of landsynthesizing community ecology across spatial use intensification are suggested, based on the factscales, and whether it can provide insights into the that both local and regional processes are importantspatial mechanisms causing variation in species co- for diversity and ecosystem functioning. Examplesexistence, strengths of species interactions and pat- drawn from research on organic farming andterns of local and regional diversity. Chapter 6 deals biological control illustrate that metacommunitywith spatio-temporal structure in soil communities and theory combined with good knowledge of the sys-ecosystem processes, and highlights the complexity tem under management is useful to understandexhibited by soil communities at different scales. how human-dominated ecosystems can be managed
  • 17. INTRODUCTION 3at the larger spatial scales that are relevant to ary community ecology is still in its infancy, butmanagers. holds great promise. Chapter 12 deals with the Community-level management is described in emergence of complex food web structure in communityChapter 10 for salt marshes in North-West Europe. evolution models and aims at synthesizing how evo-It deals with interactions among sedimentation, lutionary dynamics may help the understanding ofnutrient availability, plant growth and herbivores. food web structures and dynamics. CommunityNatural succession in these systems results in the evolution models incorporate both the dynamicaldominance of a single tall grass species to the detri- components of food webs and the complexity nec-ment of the natural herbivores such as geese and essary to understand empirical food web data. It ishares. Introductions of large herbivores can reverse shown how this model matches topological proper-succession to facilitate geese and hares. Without ties of empirical data, while giving information onthis management, these systems would develop the dynamics of the food web. It also discusses theinto low-diversity systems of reduced value. relevance of the allometric theory of ecology to the Finally, we consider future directions of community debate on the relationship between stability andresearch. These chapters include the study of the diversity, as well as the evolution of niche breadthevolution of community processes and patterns, and non-trophic interactions. Another promisingranging from local dynamics to large-scale diversity line of research involves mutualisms. Chapter 13gradients, and focus on the role of mutualisms in deals with the role of mutualisms and communitycommunity ecology. Chapter 11 focuses on evolu- organization. Key evolutionary events, such as thetionary processes in community ecology to try to bridge origin of the eukaryotic cell, the invasion of land bythe gap between community and evolutionary ecol- plants and the radiation of angiosperms, are linkedogy by considering the reciprocal effects of individ- to mutualisms. Mutualism probably brings order toual trait variation and community characteristics. the organization, structure and function of commu-Central principles from evolutionary biology are nities by regulating the acquisition of resources andused to extrapolate the consequences of genetic ameliorating stresses. Finally, Chapter 14 specu-diversity to community level and illustrate the ef- lates about emerging frontiers of community ecology.fects of genetic variation on community properties Among various emerging areas, a focus on the im-and vice versa. It is stated that more studies have portant topic of biotic invasions seems crucial, sincebeen performed on the effects of genotypic and they alter the world’s natural communities as wellphenotypic diversity on community properties as their ecological character. Further emergingthan on the effects of community diversity on ge- areas concern questions surrounding the linkingnetic and phenotypic diversity of single species. of the structure of ecological networks to empirical-This fascinating new field of integrative evolution- ly measured community dynamics.
  • 18. This page intentionally left blank
  • 19. PART ISHAPE AND STRUCTURE
  • 20. This page intentionally left blank
  • 21. CHAPTER 1 The topology of ecological interaction networks: the state of the art Owen L. Petchey, Peter J. Morin and Han Olff1.1 Introduction strength of their interactions (Fig. 1.1). The topology of an ecological network is a description of the1.1.1 What do we mean by the ‘topology’ patterns of interactions (‘who interacts withof ecological networks? whom’). Topology is thus the study of the arrange-Many diverse systems can be described as a network ment of links from an information/organizationof linked nodes (Barbarasi 2002); for example, trans- perspective. The distances and angles betweenportation networks of cities linked by roads, net- nodes have no meaning for the topology; they areworks of interacting populations on a landscape, often chosen so that the network can be convenient-networks of interacting individuals in a social net- ly graphically represented (Fig. 1.1a). If the geome-work, networks of neurones linked by synapses, try of the network is included, the Euclideannetworks of interacting genes in the genome or net- distance between nodes has a meaning (Fig. 1.1b);works of biochemical transformations within the for example, in the genetic or trait similarity ofcell. In the paradigm used in this chapter, species species or as the physical distance between speciesare the ‘nodes’ of ecological networks, and inter- within a landscape (e.g. Thompson et al. 2001) suchactions among species are the links. as in a metacommunity context (Leibold et al. 2004). Ecological communities are immensely complex; If the interaction strength is also included (Fig.variation exists at every level of organization (in- 1.1c), all interactions are not equivalent – some aredividuals, populations, species) and these entities stronger than others.interact with each other in any number of ways In this chapter, we focus on the origins and con-(consumption, competition, mutualism, facilitation, sequences of different network topologies. Chaptermodification). Ecological networks provide a tract- 2 will address the origin and consequences of vari-able simplification of this complexity – they can be ation in interaction strength within interaction net-constructed, modelled, experimentally manipu- works. Fundamental questions about the topologylated and analysed with available tools and re- of ecological networks include how many nodessources (Proulx et al. 2005). This chapter focuses (species) there are, how many links (interactions)on the networks of interactions among the species there are in total and how interactions arein ecological communities to highlight the advances distributed among species pairs (Fig. 1.1a).in ecological understanding that research about The possible interactions that can be representedthese networks can provide. in ecological networks theoretically include all of the The links in ecological networks can be charac- basic pair-wise interactions among species, includ-terized through three main aspects: (1) their topol- ing competition via various mechanisms, predator–ogy, (2) their geometry and (3) the direction and prey and host–parasite interactions, interactions 7
  • 22. 8 SHAPE AND STRUCTURE (a) (b) Birds (Larus, Haematopus, Corvus) + + + + − Small starfish (Leptasterias) + + − − Snails (Nucella) + + − − − − Acorn Goose − 0 − − Mussels barnacles − barnacles (Semibalanus) (Pollicipes) (Mytilus) (c) Recycling Spiders Spartina 305 Insects 81 28 34580 6585 Photo. Resp. 23 224 53 5 Mech. Mech. 27995 Mud Crabs 6280 600000 59 4010 Bact. 462 Crabs 90 31 6 3890 Nemas 25 372 Algae 1620 1800 Photo. Resp. Mech. Mech. 3671 180 Net 28175 3890 644 Export 0.6% 4534 100 103 106 563620 32709 kc/m2/yr 93.9% 5.4%Figure 1.1 Examples of three kinds of ecological networks. (a) A food web for a portion of the grassland community nearSilwood Park, UK. Reprinted with permission from www.foodwebs.org using data from Dawah et al. (1995). (b) Aninteraction network showing the signs of interactions among a subset of species in a community from the rocky intertidalzone of the Pacific Coast of northwestern USA. Reprinted from Wootton (1994), with permission of the EcologicalSociety of America. (c) A network of energy flow from a salt marsh in southeastern USA. Reprinted from Teal (1962),with permission of the Ecological Society of America.
  • 23. THE TOPOLOGY OF ECOLOGICAL INTERACTION NETWORKS 9through behavioural interference and positive inter- were recorded (Dawah et al. 1995). However, manyactions including mutualisms and commensalisms. food webs are more complex and less complete,In practice, few representations of ecological net- with less well-resolved taxa, especially in the smal-works include all of these kinds of interac- ler, relatively inconspicuous and taxonomically dif-tions, partly for historical reasons and partly ficult groups such as microbes and micro-because, for reasons of pragmatism and feasibility, invertebrates.the ecologists who describe these networks often One should note the very important assumptionfocus on only a subset of possible interactions, e.g. in the kind of the topological approach shown inwhen they study the importance of a particular type Fig. 1.1: every species is treated equally in the net-of interaction in structuring the ecological commu- work from an interaction perspective. For example,nity of interest. species that are very abundant in biomass and spe- cies that are very rare are treated exactly the same. Only if these differences in abundance result in a1.1.2 Different types of ecological difference in their number of interactions withnetworks other species will those differences show up in theWhat might constitute the links among species in topology of the web. In other words, the centralan ecological network? Very broadly, one can think assumption of the approach is that the key pointof two types of information represented by a link. for understanding the food web is that species in-The first is when a link represents one specific teract, while information on how abundant differ-biological mechanism or process, the second is ent species are and how strong these interactionswhen a link represents the net effect of a variety of are (e.g. expressed in per capita effects of predatorsmechanisms or processes. Links that represent on prey, and vice versa) can be disregarded. This isbiological mechanisms or processes can be recog- of course a strongly simplifying and yet poorlynized by the transfer of something tangible (such as tested assumption. Efforts to attach more informa-biomass, energy, nutrients, information or combi- tion to nodes have led to significant advances innations of these) between the linked entities, where understanding (Cohen et al. 2003). Nevertheless, athis usually requires close physical proximity be- huge body of research concerns the search for pat-tween the organisms involved in time and space. terns in food web topology (Cohen 1978). Food webs are an example of an ecological Plant–pollinator networks also use links to repre-network with this type of link, in which the sent transfers of something tangible among organ-consumer–resource interactions represent transfer isms that share close proximity in space and time.of energy and material and require a physical inter- Pollen moves from plant to pollinator, and again toaction between individuals. Food webs are perhaps the plant, which requires repeated physical contact.the most commonly encountered type of ecological These interactions involve different currencies.network (Elton 1927; Cohen 1978; Pimm 1982; Polis From the plant perspective, the exchange of geneticand Winemiller 1996; de Ruiter et al. 2005). An information through cross-fertilization is the mostexample of a food web, the Silwood Park, UK, important aspect. From the pollinator perspective,grassland food web (Dawah et al. 1995), is shown the resource rewards of visiting the flower are thein Fig. 1.1a and depicts species as ‘balls’ and feed- driving force. Interest in these so-called two-layering links as ‘sticks’. In this particular food web, networks is a rather recent development, at leastthere are plants, herbivores, primary parasitoids of compared with the long study of food webs, andthe herbivores, and hyperparasitoids. Trophic posi- the networks typically describe which plants aretion increases with height and is also coded by visited by which pollinators or seed disperserscolour (from www.foodwebs.org). This is a rela- (Bascompte et al. 2003; Jordano et al. 2003; Rezendetively simple food web (87 species) in which the et al. 2007).identity of each resource and consumer is resolved Networks of energy and material flow also docu-to the species level and sufficient sampling effort ment this first broad type of link, and they have aensured that virtually all species and feeding links venerable place in ecology, dating back at least to
  • 24. 10 SHAPE AND STRUCTURELindeman (1942). Fig. 1.1c shows the quantification termed competition, and can result from consump-of the patterns of energy flow in a salt marsh in tion (a biological process) by two species of a sharedNorth America among the major functional compo- resource, as well as from a number of other non-nents (Teal 1962). Owing to measurement con- consumptive mechanisms, such as direct beha-straints and the assumed functional equivalence of vioural interference among species (Schoener 1983).different species involved with respect to nutrient When links represent effects of one species onand energy transformation, these networks typical- another, they may result from direct effects, indirectly have low taxonomic resolution, and have mostly effects or net effects. Competition via consumptionbeen developed as a way to understand and of a shared resource can be represented as adescribe energy fluxes and material in ecosystems, negative–negative link arising from an indirectand the associated ecosystem functions and ser- interaction between two competing speciesvices (such as primary and secondary productivity, mediated by their joint consumption of a thirdcarbon sink capacity, etc.). That is, the focus is the resource species; that is, there need not be a transferdynamics of energy and material, rather than of anything tangible from one competitor to thethe dynamics of species and their interaction struc- other, yet changes in the abundance of one willture. An exception is the work by Ulanowicz (1997), alter the abundance of the other. Other indirectwho is one of the rare authors who has extensively effects (or interactions) include apparent competi-considered origins and consequences of different tion (negative–negative effects caused by the pres-topological patterns of energy and nutrient flows ence of a shared consumer; e.g. Holt 1977) andin food webs. In his work, he stresses the impor- apparent mutualism (positive–positive effects, e.g.tance of positive feedback among species, resulting in a trophic cascade; Pace et al. 1999).in autocatalytic loops within subsystems of whole Ecological networks in which links represent thewebs, which may even lead to ecosystem-level net effect of one species on another (the outcome ofcompetition between such alternative loops. Also, all direct and indirect effects) are often, and per-the work on soil foods webs by de Ruiter, Moore haps confusingly, termed interaction webs. As thisand others (e.g. de Ruiter et al. 1998) has tried to also describes the general class of ecological net-map the flows of nutrients and energy from a much works in which species interact (irrespective of themore topological perspective than is done in classic nature of their interactions) they should be moreecosystem science. The outcome of this work will be correctly called interaction-sign networks. They de-addressed further in Chapter 2. pict the sign (positive, negative, zero) and some- The second broad type of information repre- times the magnitude of the net impact of changessented by a link is the effect of one entity on another. in the abundance of one species on another. TheseFor example, if individuals of two species each have net impacts can be the result of all kinds of directa positive effect on the growth and reproduction of and indirect interspecific interactions. Interactionindividuals of the other, they would be linked. This webs are less commonly depicted in the literature,positive–positive interaction is termed mutualism perhaps because of the practical difficulty ofand a range of biological mechanisms might under- determining the sign of net interactions, especiallylie this link. Pollination is an example of a mecha- within trophic levels, for which elaborate field ex-nism that can result in a positive–positive periments are often required. However, sometimesinteraction. Associational resistance, in which a the nature of interactions can be inferred fromspecies gains protection from its consumers by spa- large-scale natural disturbances to ecosystems. Atially associating with its potential competitors, is classic example is the effect of rinderpest on theanother example that can lead to a net positive food web structure and ecosystem functioning ofinteraction between species (Olff et al. 1999). Facili- the Serengeti (Sinclair 1979). In this work, abiotictation, such as often found among grazing herbi- components and processes, such as fire, are alsovores (Huisman and Olff 1998; Arsenault and taken into account as ‘nodes’ in the food web, be-Owen-Smith 2002), also belongs to this class of in- cause of the strong effect biotic feedbacks that theyteractions. A negative–negative interaction is receive. One good example of an interaction-sign
  • 25. THE TOPOLOGY OF ECOLOGICAL INTERACTION NETWORKS 11web comes from Wootton’s (1994) work on inter- Hairston (1981) studied a group of six species ofactions among some of the species living in the rocky plethodontid salamanders found in moist forests ofintertidal zone of Washington State in the USA the mountains of North Carolina, USA. Based on(see Fig. 1.1c). The search for patterns in interaction their morphological similarity and similar life-networks has only begun rather recently. For a styles, all being carnivores feeding on small inver-recent review, see Ohgushi (2005), who focuses on tebrates, it seemed plausible that all six speciesindirect interactions in plant-based terrestrial food might compete for food (Fig. 1.2). Over a 5 yearwebs. period Hairston removed each of the two most common species at this site. Only one of the remain- ing five species responded positively to each re-1.1.3 Three general questions moval, and this was the most common species at the site. These results imply that only the two mostFor each type of ecological network, we suggest that common species compete. Hairston suggested thatthere are three general types of questions: those competition might be for nest sites, rather than forabout whether and what structural regularities shared prey, but subsequent research showed thatexist within and among observed networks; those direct territorial (interference) interactions betweenabout which mechanisms are responsible for the salamanders constituted the primary mechanism ofstructure of the networks and of any structural reg- competition (Nishikawa 1985). In this case, the net-ularities that exist; and lastly, those about the gaps in work of competitors is considerably simpler thanour knowledge. The aim here is to define a set of the completely connected system that would resultgeneral contemporary questions that we will use to if all species shared and competed for commonorganize each of the following sections. resources (Fig. 1.2). Another general and important question about the structure of competitive networks is whether1.2 Competitive networks the interactions form a hierarchy, or an intransitive loop. In a hierarchy, for example, species A out-1.2.1 Structural regularities competes species B and C, B outcompetes C, andReal-world competitive networks remain poorly C cannot outcompete either A or B. This will mostlystudied and documented. A fundamental question lead to predominance of a single competitively su-about their structure remains quite unresolved: perior species (A) and therefore will foster relative-given a set of co-occurring species that appear ly low biodiversity. In an intransitive loop, in whichto consume a similar class of resources (such as A outcompetes B, B outcompetes C, and C outcom-different-sized seeds, forming a potential guild, petes A, all three species can potentially coexist.sensu Root 1967), how many pairs of these species Consequently, mechanisms of interaction thatreally compete? That is, how many of the plausible allow intransitive competitive networks might ex-links are actually realized in competitive networks? plain high biodiversity in some ecosystems. Groups of taxonomically similar species are often At first hand, intransitive competitive networks,assumed to be potential competitors, but experi- especially within one trophic level, would be un-ments show that only a small fraction of the poten- likely to emerge, based on the theory of life historytial competitors actually compete. Sometimes this and competitive trade-offs. An organism’s life his-can be generalized to simple integrative traits, tory generally reflects a compromise in the use ofbuilding on the early work on limiting similarity available resources. Species generally specialize(for a review, see Brown 1981). For savanna large along environmental axes of resources axes andherbivores, Prins and Olff (1997), for example, stress factors in such a way that investing in somefound that large herbivores overlap more in life history/functional trait (e.g. the ability to crackresource use when they are more similar in body big seeds in birds) occurs at the expense of somesize. Thus, more similar-sized species are more other function (e.g. the efficiency with which smalllikely to interact. seeds can be collected). In plants, for example, the
  • 26. 12 SHAPE AND STRUCTURE (what is the opponent sensitive to) than energy- Plethodon serratus or nutrient-based (which requires morphological adaptations, such as allocation shifts). Plethodon Plethodon One example of such an intransitive competitive glutinosus jordani network involves different strains of bacteria competing in spatially structured environments. Desmognathus Desmognathus ochrophaeus wrighti Kerr et al. (2002) explored how a mixture of me- chanisms can allow three strains of Escherichia coli Eurycea cirrigea to coexist via a set of intransitive interactions that are akin to a game of rock–paper–scissors. There are three key features of this system – one strain Plethodon produces a toxin, called colicin, whereas the two serratus other strains are either sensitive or resistant to the toxin (Kirkup and Riley 2004). The three strains Plethodon Plethodon interact as an intransitive network in the following glutinosus jordani way. All of the strains compete consumptively for a carbon source, such as glucose, but they differ in Desmognathus Desmognathus competitive ability such that colicin-sensitive ochrophaeus wrighti strains are the best consumptive competitors, followed by colicin-resistant strains followed Eurycea by colicin-producing strains. However colicin- cirrigea producing strains can displace colicin-sensitive strains by poisoning them, but colicin-producingFigure 1.2 Potential (top) and realized (bottom) networks strains can in turn be displaced by competitivelyof competing salamanders in the study of Hairston superior colicin-resistant strains. These patterns of(1981). Dashed lines in the lower graph indicate sequential displacements occur only in an un-unrealized competitive interactions; the heavy solid lineindicates that only two species, Plethodon glutinosus and mixed spatially complex environment, such asPlethodon jordani, actually compete. the surface of a culture plate. In a well-mixed spatially homogeneous environment, such as a well-stirred liquid culture, only one strain per-ability to compete well for nutrients has been sug- sists. It appears that limited dispersal can promotegested to happen at the expense of the ability to diversity by allowing an intransitive competitivecompete for light (Tilman 1990). Also, competing network to exist (Fig. 1.3). A similar role is playedneeds for growth and reproduction result in differ- by dispersal limitation in maintaining diversity inent optimal combinations in different species, de- neutral communities (i.e. where all species are as-pending on the conditions to which they are sumed to be functionally equivalent; e.g. Hubbelladapted. Such evolution-driven trade-offs make 2001).competitive hierarchies much more likely than Qualitatively similar but much more complextransitive networks in most cases. Said simply, it networks may explain the high diversity of bacteriais unlikely that the species adapted to the ‘left-tail’ that manage to coexist in soils, where dispersal isof some underlying niche axis will outcompete a limited. For example, Torsvik et al. (1990) madespecies from the ‘right-tail’ or vice versa. initial estimates using DNA hybridization techni- An important class of exceptions exists in organ- ques that suggested that thousands of bacterial taxaisms in which the outcome of interactions is not could occur per 30 g of soil. These conservativedriven by resource competition, but by chemical estimates were re-evaluated by Dykhuizen (1998),warfare, such as allelopathy. In this case, few who estimated that 30 g of forest soil could containtrade-offs are expected, as the defence and compe- over half a million species, depending on the as-tition is much more strongly information-based sumptions made in the analysis.
  • 27. THE TOPOLOGY OF ECOLOGICAL INTERACTION NETWORKS 13a Static plate resources, might provide a mechanism to explain how so many bacterial taxa can coexist in soils 12 ´ ´ (Czaran et al. 2002). These authors present a theo- 10 retical analysis showing that, by including multiple Log (abundance) 8 toxins and multiple toxin resistance genes, a large number of taxa will persist in a two-dimensional 6 model grid of interacting bacteria. They find that a 4 system with a few toxin genes and many resistance 2 genes can support up to 1000 different strains in a 180 Â 180 spatial grid. This is obviously somewhat 0 0 10 20 30 40 50 60 70 less than the diversity estimates obtained for soils, but it is of the right order of magnitude.b Flask 12 1.2.2 Mechanisms 10 Log (abundance) 8 Intransitive networks of competitive interactions may explain the high diversity of some systems of 6 competitors. It appears that high diversity can be 4 maintained by intransitive competitive networks 2 only in spatially structured environments, such as soils or two-dimensional surfaces (Reichenbach 0 0 10 20 30 40 50 60 70 et al. 2007). In well-mixed environments without opportunities for much spatial structuring, suchc Mixed plate as aquatic systems, other mechanisms must be in- 12 voked to explain the coexistence of large networks 10 of competitors.Log (abundance) 8 What maintains competitor diversity in well- C mixed habitats with only few limiting resources? 6 S One possibility has been suggested by Huisman 4 R and Weissing (1999). Their models show that 2 large networks of many resource competitors, when competing for several resources (four to 0 0 10 20 30 40 50 60 70 six), can persist due to chaotic fluctuations in spe- Generations cies abundances (Fig. 1.4). They suggest that switches along alternative transitive networks alsoFigure 1.3 Dynamics of an intransitive competitor seem to play a role. The presence of multiple limit-network. C is a colicin (toxin producing) bacterial strain, ing nutrients in their model prevents the formationwhich excludes S, a colicin-sensitive strain, which excludesR, a colicin-resistant strain, which in turn competitively of clear competitive hierarchies. This might pro-displaces C. All three strains persist in a static two- vide an explanation for the coexistence of largedimensional habitat (a), but R eventually predominates in numbers of phytoplankton species in well-mixedmixed environments (b and c). Reprinted with permission environments such as lakes and oceans, a problemfrom Macmillan Publishers Ltd: Nature , 171–4, B. Kerr et al. that has interested ecologists for many years# 2002. (Hutchinson 1961). Recently, Huisman and collea- gues found experimental evidence for similar Intransitive networks involving multiple toxins chaotic dynamics in a closed experimental multi-and multiple resistance factors to those toxins, trophic plankton system that was observed for aalong with trade-offs between the ability to produce large number of generations (Beninca et al. 2008).or resist toxins and to compete consumptively for This seems to be the first well-documented example
  • 28. 14 SHAPE AND STRUCTURE a 60 50 5 Species abundances 40 3 30 2 20 4 10 1 0 0 100 200 300 Time (days) b 55 c 150 50 120 Total biomass 45 Species 5 90 40 60 35 30 30 6 25 9 0 20 25 12 0 100 200 300 15 1 Spec 30 35 ci es Time (days) ies 3 40 18 SpeFigure 1.4 (a) Example of coexistence in a system of chaotically fluctuating consumptive competitors. Reprinted withpermission from Macmillan Publishers Ltd: Nature , 407–10, J. Huisman & F.J. Weissing, # 1999. (b) Populations of threespecies orbit around a strange attractor, typical of chaotic dynamics. (c) Although the populations fluctuate chaotically,total biomass summed over all species is invariant over time.of chaotic dynamics in a food web. Whether such limiting resources, leading to an expectation of onlychaotic dynamics also frequently occur in natural few species to coexist, each limited by a differentplankton communities still needs to be seen though, resource. However, different algae species seem toas the experimental set-up prevented many types of have specialized on different wavelengths of theinteractions between biota and abiotic factors (such light spectrum, which therefore is in fact a wholeas sediments). One can imagine that fluctuations in class of resources (Stomp et al. 2007a,b). These ‘col-external forcing factors, such as light and tempera- ourful niches’ seem to play a yet underestimatedture, may overrule the potential for chaotic dynam- role in the coexistence of plankton species.ics in such systems. Of course, other mechanisms that minimize the Recently, another way out of the ‘paradox of the importance of competition among species can alsoplankton’, or the ‘principle of competitive exclu- operate to maintain diversity in spatially structuredsion’ (no more species are expected to coexist than or well-mixed environments. Perhaps the best knownthe number of limiting resources), has been offered of these mechanisms is keystone predation (Paineby the same research group. Classic explanations 1966), in which consumers feed preferentially on com-for coexistence of plankton species have explored petitively superior species and prevent competitivelight, nitrogen, phosphorus and silicon as the main exclusions of weaker competitors (Leibold 1996).
  • 29. THE TOPOLOGY OF ECOLOGICAL INTERACTION NETWORKS 151.2.3 Unresolved issues plant–pollinator systems (Feinsinger 1976; Petan- diou and Ellis 1993; Fonseca and John 1996; WaserExcept for a very few well-studied and relatively et al. 1996), and recently there has been a growth insimple examples, we know little about the structure the use of the network perspective to analyse theof real competitive networks. This is partly due to greater amounts of data collected on mutualisticthe difficulty of clearly establishing whether species networks (see following references in this section).actually compete, if so to what degree, and of identi- For the main part, the two types of mutualistic net-fying which mechanisms are in play. The necessary works examined are plant–pollinator networks andexperiments are often difficult to perform, except plant–seed disperser networks. The conventionunder simplified conditions in the laboratory. used to depict these networks is to show interac- How would inclusion of competitive networks tions in a two-layer network between one group ofinto food webs (networks in which links are con- species (plants) and their mutualists (pollinators orsumption) alter our predictions about web stabili- dispersers), as shown in Fig. 1.5. The species are notty? We can guess, in part, that inclusion of depicted on different trophic levels, as in foodcompetitive interactions into a community matrix webs, and only interactions between the twobased only on predator–prey interactions would groups of mutualists are considered (no direct com-tend to destabilize the network, to the extent that petitive interactions are included).May’s (1972, 1973) analysis holds true. Mutualistic networks appear to have several How common are subnetworks of intransitive types of structural regularity. First, there are nestedcompetitive interactions, and do they explain the sets of interactions (Bascompte et al. 2003; Lewin-persistence of high species richness? Although they sohn et al. 2006). That is, more specialized mutual-have long been recognized as a potential scenario for ists tend to interact with a proper subset of themaintaining diversity in systems (see Connell 1978), species that more generalist mutualists interactwe still have very few well-documented examples of with. One consequence is that there is a set of spe-non-transitive competitive networks of any sort. cies that form a highly connected ‘core’ in mutual-Minimally, such networks would require that com- istic networks. This makes the number of linkspetitors interact through more than one kind of com- across the network required to connect any twopetitive mechanism (Schoener 1983). The recurring species rather short (Olesen et al. 2006). Indeed,problem is that, based on traditions, expertise and mutualistic networks tend to be more nested thanpractical limitations, almost all studies still focus on food webs and to have shorter paths between anysubsets of entire food webs, such as beetle commu- two species than food webs (Bascompte et al. 2003).nities, plant communities, pollinator communities, This ‘core’ of highly connected interactors appearssoil food webs, soil microbial communities, but to have consequences for how mutualistic networkshardly ever study the entire system with the same respond to disturbance, and seems to make themlevel of aggregation. This will require large interdis- more robust to potential perturbations (Bascompteciplinary efforts and most likely new theoretical ap- et al. 2003).proaches to deal with the resulting data. Another structural feature of mutualistic net- works is of asymmetric patterns of interactions (Bascompte et al. 2006; Vazquez et al. 2007). In gen-1.3 Mutualistic networks eral, species with high numbers of connections tend to interact with those that are connected to relative-1.3.1 Structural regularities ly few species. This asymmetry in connectionsFor the most part, mutualisms do not figure promi- within mutualism webs is consistent with the fea-nently in traditional depictions of ecological net- tures of models that confer greater stability on theseworks. This reflects, in part, short shrift given to networks (Bascompte et al. 2006).reciprocal positive interactions in community ecol- Similar to competitive networks, mutualistic net-ogy (Boucher 1985), at least until recently (Brooker works can also show an intransitive structure, alsoet al. 2008). Some of the earliest works consider called hypercycles. Three species may be arranged
  • 30. 16 SHAPE AND STRUCTURE INO1 ABIS MONT CORRPlants AnimalsFigure 1.5 Examples of networks of positive interactions (mutualisms) between plants and their pollinators or dispersersin four different kinds of communities. In each graph, nodes on the left correspond to different plant species, andnodes on the right correspond to animals that pollinate the plants or disperse their seeds. Lines connect positivelyinteracting species. Reprinted from Jordano et al. (2003), with permission from Wiley-Blackwell.in an autocatalytic loop, in which species A pro- plant–pollinator and 23 plant–frugivore networks,motes B, B promotes C, and C promotes A, result- there was a significant effect of phylogeny on theing in indirect mutualism (A helps C through B). number of links of a species in 25–39% of the net-This structure can also be seen as positive feedback works. That is, more closely related species tendedof species A on its own growth, through species B to both be specialists (or generalists) than moreand C. For example, Ulanowicz (1995) describes distant ones. The amount of phylogenetic similarityand models how the carnivorous waterplant Utri- of two species was also found to predict the ecolog-cularia excretes sugars towards its leaf surface, ical similarity (measured as the standardized num-which promotes the growth of microepiphyto- ber of interactions in common) in nearly 50% of thebenthos (algae, diatoms and bacteria) on the leaves, networks. These analyses provide good evidencewhich then attract zooplankton grazers (as cope- that evolutionary history at least partly explainspods), which are in turn caught by the Utricularia some of the structural regularities of mutualisticplant for food. Such hypercycles may have played a networks.role in early prebiotic evolution, in which space andlimited dispersal again seem to stabilize the inter- 1.3.3 Unresolved issuesaction, e.g. against the invasion of parasites in themutualistic network (Boerlijst and Hogeweg 1991). The patterns described above raise additional ques- tions. Why does phylogenetic history apparently influence the structure of mutualism networks?1.3.2 Mechanisms Perhaps evolution, and specifically coevolutionaryInformation about the drivers of the structure of processes, have a stronger role in structuring net-mutualistic networks is beginning to emerge. The works of mutualists. There is considerable variationclearest example of this is a recent study about unexplained by phylogeny in the webs consideredrelationships between the shared evolutionary his- above, so obviously phylogenetic history is not thetory of the species in mutualistic networks (Re- only factor contributing to structure in these net-zende et al. 2007). Across a compilation of 36 works (Rezende et al. 2007). It is also interesting to
  • 31. THE TOPOLOGY OF ECOLOGICAL INTERACTION NETWORKS 17ask to what extent the nestedness in mutualism Lawton 1978; Pimm 1980). These and other appar-networks is a consequence of the structure of phy- ent regularities have been challenged many timeslogenies, or of more ecological constraints? (Yodzis 1984; Paine 1988; Polis 1991); some are sup- So far, the mutualisms depicted in these net- ported by more recent and higher quality data,works are restricted to interactions involving trans- others are not (Pimm 1991; Martinez 1994; Warrenport of pollen or fruits/seeds by animals and the 1994). Rather than review the literature concerningreciprocal energetic or nutritional reward obtained all of the patterns, we will focus on one of the mostby the animals that consume nectar, pollen or fruits. important properties of food webs with ‘qualita-These are only a fraction of the kinds of positive tive’ (presence/absence) links: connectance. (Noteinteractions that characterize mutualisms. Other that there are quantitative versions of some proper-kinds of positive interactions, such as defensive/ ties, including connectance (e.g. Bersier et al. 2002).)domicile mutualisms involving the defence of host In a food web of S species, connectance is theplants by insects (Janzen 1966), or trading in energy number of realized links (L) divided by the numberand nutrients between plants and mycorrhizal of possible links (S2 or S(S À 1) if excluding cannibal-fungi (Schwartz and Hoeksema 1998), have not yet istic links), i.e. when all species interact. It has strongreceived this kind of analysis. effects on many other structural features, such as the Another problem is that effects of mutualisms on frequency distribution of the number of links perthe dynamics of the larger ecological networks in species (Dunne et al. 2002; Stouffer et al. 2005), andwhich they embedded are essentially unknown. has long been known to affect the stability of modelMost models of mutualisms focus on the dynamics food webs (Gardner and Ashby 1970; May 1972;of a pair of species involved in a mutualistic inter- DeAngelis 1975; Pimm 1984). One might say thataction (Dean 1983; Wolin 1985). Consequences of recording and explaining patterns of connectance ismutualisms for the stability of the larger communi- a first step towards understanding many other as-ty network in which they occur remain largely un- pects of food web structure. Earlier researchers out-explored (but see Ringel et al. 1996). lined a number of ideas about how structure depends on the number of species in a food web. In particular, there was much debate about whether connectance1.4 Food webs remains constant or increases or decreases as the number of species in food webs increases.1.4.1 Structural regularities There has been a great deal of research about theFood webs are perhaps the oldest and most fre- magnitude of connectance in food webs and how itquently studied type of ecological network. The changes with the overall species richness of the weblinks in food webs are determined by observing (Sugihara et al. 1989; Winemiller 1989; Havens 1992,evidence of a resource–consumer interaction be- 1993; Martinez 1992, 1993; Bengtsson 1994). Some oftween two species. This can be done in a qualitative this early discussion was hampered by problemsway (presence/absence of a trophic interaction) or with the source and quality of the data used toquantitative way (e.g. through feeding rates, preda- construct food webs. In contrast, over the last twotion rates, conversion efficiencies). Much effort has decades or so, a large amount of data was collectedbeen expended collecting this type of information, with the explicit goal of constructing highly de-especially the qualitative type, for constructing tailed food webs, and we will focus on 15 excep-food webs, and examining whether they exhibit tionally well-characterized food webs.particular structure. One of the earliest ideas was In these webs, connectance varies from about 0.01that food chain lengths (the number of links from a to 0.35, that is, from 1% to 35% of all possible linksspecies at the bottom of the food web to one at the are realized (Fig. 1.6). Furthermore, connectancetop) were shorter than expected by chance (Pimm appears to decrease with increasing species rich-and Lawton 1977; Pimm 1982). Another was a rela- ness, and this pattern seems to be well supportedtive paucity of omnivores, species that feed on re- by recent analyses (Murtaugh and Kollath 1997;sources at different trophic levels (Pimm and Schmid-Araya et al. 2002; Montoya and Sole 2003).
  • 32. 18 SHAPE AND STRUCTURE 0.6 Benguela pelagic Broadstone stream 0.5 Broom Capinteria Caricaie lakes 0.4 Coachella Connectance EcoWEB41 EcoWEB60 0.3 Grasslands Mill stream Sierra lakes 0.2 Skipwith pond Small reef Tuesday Lake 0.1 Ythan 0.0 20 40 60 80 100 120 140 160 Species richnessFigure 1.6 Connectance and species richness of 15 food webs from Petchey et al. (2008). Note how connectancedeclines with increasing species richness.Among the 15 food webs analysed, however, only mechanisms focus on why consumers eat the num-four are from terrestrial ecosystems (Broom, Coa- ber of resources that they do, and how this dependschella, EcoWEB60 and Grasslands); three have rath- on traits of the consumer (such as body size; e.g.er low connectance; and one, Coachella, has very Woodward et al. 2005) and the diversity of availablehigh connectance. This variation among terrestrial resources. Stability-based mechanisms focus on thewebs is unlikely to be due to the nature of data effects of connectance and species richness on dy-collection, and has more to do with the types of namic stability, with the idea that unstable combi-species in the food webs (e.g. seed eaters versus nations of connectance and species richness will beparasitoids). less common in natural systems than stable combi- As in competitive and mutualistic networks, nations. Clearly these are not mutually exclusivetransitive structures play an interesting role also in mechanisms, though from here on we will focusfood webs, based on consumer–resource inter- on foraging-based mechanisms.actions. Soil food webs have been shown to contain Qualitative foraging-based explanations of con-trophic loops that are suggested to contribute to nectance rely on the fact that connectance is derivedtheir stability (Neutel et al. 2002). Also known as from the summed number of resources of each ofhypercycles, such interaction structures may even the consumers in a food web. Consequently, if therehave played an important role in prebiotic evolu- are many generalist consumers connectance will betion (Boerlijst and Hogeweg 1991). higher than if there are many specialists. An exam- ple of this type of explanation of connectance is provided by a study of the natural enemies of1.4.2 Mechanisms aphid species in a meadow (Van Veen et al. 2008).What mechanisms determine connectance, how Aphids are attacked by parasitoids, pathogens anddoes it differ between food webs and how does it predators, and the hypothesis was made that para-vary with species richness? Broadly, one can sepa- sitoid webs would have lower connectance thanrate the candidate mechanisms into two classes: pathogen ones, which in turn would have lowerthose related to foraging ecology and those affect- connectance than predator ones. The idea was thating overall network stability. Foraging-based the intimate and prolonged interactions between
  • 33. THE TOPOLOGY OF ECOLOGICAL INTERACTION NETWORKS 19parasitoid and host would lead to specialization. foraging and resource selection, on the outcome ofWhereas parasitoids can often suspend their devel- evolution and on physical constraints of consump-opment in the absence of appropriate prey, the tion, and these could differ between broad classesfungal pathogens studied do not, and so might be of species and consumption (e.g. Warren 1990; Vanexpected to have a broader host range. Finally, the Veen et al. 2008). Analyses of food web data suggestpredators have a less intimate and prolonged inter- that a range of relationships may exist, but that theaction with their resources and so may be even number of links scales with Sx, where x is not 1 ormore generalized. Observations supported these 2 but somewhere in between (Murtaugh and Kol-explanations of connectance. The parasitoid web lath 1997; Schmid-Araya et al. 2002; Montoya andshowed the lowest connectance, the pathogen web Sole 2003). Most likely there is not a single relationwas intermediate, and the predator web had high- between connectance and species richness, butest connectance (when based on quantitative mea- rather a range of possible relations, depending onsures of connectance). Examination of parasite–host the taxonomic identity and range of organisms con-feeding interactions suggests that food webs con- sidered.structed of these also are quite different from those Stability-based explanations of connectance orig-focused on predation (Leaper and Huxham 2002). inate from observations that model systems with Quantitative foraging-based explanations of food large numbers of species and large numbers ofweb structure require a prediction of the number of connections tend to be less stable than systemsresource species each consumer species will feed with fewer species or connections (Gardner andupon. One classic paradigm for predicting species’ Ashby 1970; May 1972). However, many good ex-diets is optimal foraging theory (MacArthur and amples of species-rich and well-connected systemsPianka 1966; Schoener 1971). Here, species are as- have been discovered (e.g. DeAngelis 1975; Neutelsumed to forage on the suite of resources species et al. 2002; Brose et al. 2006; Rooney et al. 2006).(items) that maximize their rate of energy intake. Furthermore, studies suggest that there are manyWhile optimal foraging theory has its critics (Pierce ways for highly connected speciose communities toand Ollason 1987), for example that it poorly deals be stable (McCann et al. 1998; Neutel et al. 2002;with species interactions, it does seem to make rea- Rooney et al. 2006; Otto et al. 2007).sonable predictions about the connectance of a suite Another enduring problem in food web ecologyof real food webs (Beckerman et al. 2006), and it concerns the factors that limit the length of foodseems likely that models of diet choice or constraint chains embedded in larger trophic webs. Althoughhave value in explaining and predicting food web simple linear food chains are a convenient abstrac-connectance and structure. tion and are absent from most natural systems, one The determinants of diet breadth are central to can still trace chains of energy flow within complexunderstanding how connectance scales with spe- food webs. The fundamental question is what limitscies richness. If consumers feed on a constant pro- the length of food chains.portion of all available species, then average diet A well-accepted and intuitive explanation for thebreadth is equal to kS, where k is the average pro- short length of food chains based on the inefficiencyportion of resources fed upon by consumers. Here, of energy transfer between trophic levels began tothe total number of links L will be kS2, and connec- be questioned in the late 20th century. Simple mod-tance, L/S2, will be constant at k. In contrast, if els of food chains suggested that the slow recoveryconsumers can feed only on a particular number from perturbations seemed to characterize longerof resources, say n (the average across consumers), model chains (Pimm and Lawton 1977, 1978), andregardless of how many different ones are avail- that this aspect of reduced stability might accountable, then L ¼ nS and connectance will be n/S. So for the rarity of longer chains. Subsequent workwhether consumers eat a constant proportion of all suggested that this result was an artefact causedavailable species or a fixed number of species de- by confounding the frequency of stabilizing densi-fines how connectance scales with species richness. ty-dependent population regulation with the lengthPresumably, this will depend on the mechanism of of model food chains (Sterner et al. 1997). Another
  • 34. 20 SHAPE AND STRUCTUREproblem was that the inefficiency of energy transfer the top of the food chain. Decreasing productivityshould predict that longer food chains should occur reduced the number of coexisting species, thein more productive environments, but there was number of trophic links and maximum foodlittle empirical support for this pattern (Pimm and chain length. There seemed to be clear evidenceLawton 1977; Pimm 1982; Post 2002; however, see that energy played a role in limiting food chainOksanen et al. 1981). If anything, longer food chains length in this system.and more complex food webs seemed to occur in Kaunzinger and Morin (1998) used a simple mi-less productive situations, rather than in the most crobial system to test for effects of productivity onproductive environments. This apparent inconsis- food chain length. The longest three-level foodtency must be tempered by the realization that chains consisted of a basal level (the bacteriumthere were very few detailed descriptions of any Serratia marcescens) consumed by a ciliated protistfood chains or webs at the time, and most of them (Colpidium striatum), which was in turn consumedwere from aquatic systems (Cohen 1978). The rela- by a top predator (the ciliate Didinium nasutum).tively small number of highly resolved food webs The system lacks omnivores, so the trophic positionthat have since accumulated (Winemiller and of each species was known without error. Produc-Pianka 1990; Martinez 1991; Dunne et al. 2002) sug- tivity was manipulated by varying the nutrientgest that the empirical data used to motivate these concentration of growth medium consumed by thetheoretical studies were far from ideal. bacteria. Three-level food chains, those containing Experiments designed to evaluate the roles of en- the top predator Didinium, persisted only at higherergy limitation or population dynamics in limiting levels of productivity (Fig. 1.7). At lower productiv-food chain length take two forms: (1) manipulations ity levels the third trophic level failed to persist,of energy inputs with concordant measurements of clearly supporting the role of energy transfer inresulting food chain lengths and (2) manipulations limiting the length of simple linear food chains.of food chain lengths with concordant measure- Patterns of change in the abundance of species onments of population dynamics. In the first approach, each trophic level are also consistent with simpleif more energy allowed longer food chains to devel- prey-dependent models of predator–prey inter-op and persist, then energetics presumably played actions (e.g. Leibold 1996), but are not consistentsome role in setting food chain length. The basic with ratio-dependent models (e.g. Abrams andtheory underlying this idea has been developed by Ginzburg 2000).Fretwell (1977) and Oksanen et al. (1981). Assuming There is scant evidence for comparable patternsthat efficiency of energy transfer between trophic in natural systems, perhaps because of the technicallevels is about 10%, energy availability would have difficulties in unambiguously assigning species toto be manipulated by over an order of magnitude trophic levels or measuring productivity. Post et al.(e.g. at least 10-fold) to see the addition or loss of a (2000) failed to find a relationship between foodtop trophic level. It is essential to be able to clearly chain length and productivity in a survey of naturalplace the organisms involved on particular trophic lakes, but did find that larger lakes tended to sup-levels – this means that relatively linear food chains port longer food chains. Their finding is superficial-without substantial omnivory are required. Very ly similar to the idea that larger predators locatedfew experiments satisfy these requirements, but higher in the food chain should require larger homethose that do suggest an important role for energy ranges (Slobodkin 1960), even disproportionallyin determining food chain length. larger than one allometrically would expect from Jenkins et al. (1992) tested the effects of produc- their size (Haskell et al. 2002). Post et al. relied on thetivity on the relatively simple food webs that indirect measurement of the trophic position of topdevelop in water-filled tree-holes in tropical Aus- predators using stable isotope ratios, rather thantralia. They examined food chain development using knowledge of the structure of the entireafter experimentally reducing productivity over food chain. Post et al. (2000) and Post (2002) have2 orders of magnitude, a reduction that should suggested that the dependence of food chain lengthresult in the loss of at least one trophic level from on energy inputs shown by Jenkins et al. (1992) and
  • 35. THE TOPOLOGY OF ECOLOGICAL INTERACTION NETWORKS 21 variation should lead to higher values of temporal a 9 Trophic level 1 variability for the same species embedded in longer (Log (N+1))/ml 8 food chains. Lawler and Morin (1993) found that the population dynamics of protists in relatively 7 simple laboratory food chains become more vari- 6 able with modest increases in food chain length. Comparisons of the temporal variability of popula- 5 tions of the same bacterivorous protists in short b food chains in which the bacterivores were the 3.5 Trophic level 2 top predators, and in food chains that are just one (Log (N+1))/ml 3.0 trophic level longer in which the bacterivores are 2.5 2.0 intermediate species preyed on by another preda- 1.5 tory protist, point to increased temporal variation 1.0 in abundance in the majority of longer food chains. 0.5 Increased temporal variation in abundance would 0.0 be consistent with longer return times in longer c food chains, as suggested by Pimm and Lawton 1.0 0.6 Trophic level 3 (1977). (Log(N+1))/ml There is also reason to suspect that energy and 0.4 population dynamics can interact in ways not di- rectly considered by Pimm and Lawton (1977), as 0.2 described in earlier models by Rosenzweig (1971) 0.0 in the context of the so-called ‘paradox of enrich- ment’. Rosenzweig found that a number of differ- 0.001 0.01 0.1 1.0 ent predator–prey models became increasingly Nutrient level (g food/l) unstable as systems were made more productive – a consequence of increasing rates of increase orFigure 1.7 Effects of productivity on the abundance ofspecies occupying basal (a; bacteria, trophic level 1), carrying capacities in the models. In this scenario,intermediate (b; Colpidium, trophic level 2) and top adding energy to a simple food chain might desta-(c; Didinium, trophic level 3) levels in simple linear food bilize the system, and shorten the chain. Of course,chains. Three-level food chains persist only at higher levels it is possible that the addition of another trophicof productivity, as shown by the failure of Didinium to level to an energetically enhanced chain couldpersist at lower levels of productivity determined bynutrient levels in experimental microcosms. Solid and offset the destabilizing effects of enrichment,empty symbols in (b) show the abundance of the though the findings of Pimm and Lawton (1977)intermediate trophic level in food chains with and without might argue against this. However, addition ofthe third trophic level, respectively. Reprinted with weakly interacting species to the food web haspermission from Macmillan Publishers Ltd: Nature 395, been suggested to confer increased stability on495–7, C.M.K. Kaunzinger and P. Morin, # 1998. unstable systems (McCann et al. 1998), so some kinds of increased trophic complex may help to offset the predicted destabilizing effects of enrich-Kaunzinger and Morin (1998) may be a feature of ment.relatively unproductive systems. The other approach used to evaluate links be- 1.4.3 Unresolved issuestween population dynamics and food chain lengthinvolves comparisons of the dynamics of species The relative contribution of foraging-based andthat occur in chains of differing length. If the mod- stability-based mechanisms to the connectance ofels are correct, population dynamics should be real food webs remains unresolved. Given the cen-more variable in longer chains, and that increased tral importance of connectance for determining
  • 36. 22 SHAPE AND STRUCTUREother structural features of food webs, this should and whether those effects offset destabilizingbe a priority research question about community effects of increased energy inputs, as some theoriestopology. predict (McCann et al. 1998). Observations and It would be fascinating to see whether the addi- models on soil food webs indeed suggest that thistion of weak interactors can stabilize food chains, is the case (Neutel et al. 2002).
  • 37. PART IIDYNAMICS
  • 38. This page intentionally left blank
  • 39. CHAPTER 2 Trophic dynamics of communities Herman A. Verhoef and Han Olff2.1 What types of dynamics can be corresponding to stable equilibria in mathematicaldistinguished? models. Resource-based competition theory that makes such predictions for multiple speciesHaving examined the geometry and structure of (Tilman 1982) seems not to hold for more thanthe ecological networks in Chapter 1 we will move two species competing for two resources (Huis-on to consider the dynamics of communities, and man and Weissing 1999, 2001). Also, the life spanconcentrate on the analysis of changes in the abun- of the organisms is often too long compared withdance of species in a multitrophic context. In 1927 the length of the study, making the judgement ofCharles Elton stated that the structure of a com- true coexistence across multiple generations prob-munity is determined by the net of feeding rela- lematic (Morin 1999). A good example is presentedtions between trophic units, the food web. The by Lawton and Gaston (1989). Despite natural per-topology of these feeding links (Chapter 1) natu- turbations affecting the community of aboutrally emerges from the dynamics of populations 20 herbivorous insects living on bracken fern, thewithin ecological communities (May 1973; Neutel relative abundances of the species and the taxo-et al. 2002, 2007; Rooney et al. 2006), while the nomic composition remained the same over atopology of the network in turn will affect the period of 7 years. The generation time of thedynamics of the populations it contains (DeAnge- respective populations is about 1 year. Similarly,lis 1992). in a 25-year-long study of large herbivore coexis- One of the central goals in ecology is to discover tence in a tropical savanna, Prins and Douglas-why populations change over time. Much of the Hamilton (1990) found high community-levelattention to this important subject is theoretical stability in species composition, despite fluctua-and the findings and consequent discussions are tions in abundance of individual species. A morebased on models outcome. However, empirical recent study deals with a small microbial fooddata on the different types of dynamics and the web. Changes in the dynamics of a defined preda-underlying mechanisms are increasingly reported. tor–prey system, consisting of a bacterivorous cili-Four main patterns of population dynamics leading ate (Tetrahymena pyriformis) and two bacterial preyto coexistence (or long-term co-occurrence) of dif- species, were triggered by changes in the dilutionferent species can be identified: coexistence at equi- rates of a one-stage chemostat. The bacterial spe-librium, coexistence at alternate equilibria (with cies preferred by the ciliate (Pedobacter) outcom-critical ‘tipping points’), coexistence at stable limit peted Brevundimonas, the second bacterial species.cycles and coexistence at chaos. At relatively high dilution rates Brevundimonas died off by the sixth day, whereas the remaining species existed in stable coexistence at equilibrium2.1.1 Stable equilibria (Becks et al. 2005). Also in this experiment theOnly a few studies address whether collections generation time of the organisms involved wasof multiple species show stable compositions much shorter than the duration of the experiment. 25
  • 40. 26 DYNAMICS2.1.2 Alternate equilibria 2.1.4 Chaotic dynamicsEvidence is accumulating that certain large-scale The long-term persistence of complex food webs iscomplex systems may have alternate equilibria not automatically linked to stability, and manyand critical tipping points. Examples are the well- mathematical models predict that species inter-known trophic cascades in freshwater lakes (Car- actions can create chaos and species extinctions.penter and Kitchell 1993; Scheffer et al. 1993). Despite receiving an overwhelming amount of the-Again, the criticism of studies that suggest the oretical attention, experimental demonstrations ofexistence of alternate equilibria deals with the chaos are rare. Only a few single species systemslength of the study relative to that of the genera- (Costantino et al. 1997; Ellner and Turchin 2005), thetion time of the organisms involved, and the phys- already mentioned microbial food web study, inical characteristics of the different sites at which which at intermediate dilution rates of the chemo-the species were studied (Connell and Sousa 1983). stat systems chaotic dynamics were observedAlso more recently the possible multiple stable (Becks et al. 2005) as well as nitrifying bacteria in astates of a system have been stated to be difficult wastewater bioreactor (Graham et al. 2007), showto be proven experimentally (Scheffer and Carpen- compelling evidence for chaos. Recently, it has been ¨ter 2003; Schroder et al. 2005), despite several shown that in a long-term experiment with a plank-recent new suggestions (Kefi et al. 2007; van der ton community, consisting of bacteria, several phy-Heide et al. 2007; Carpenter et al. 2008). The transi- toplankton species, herbivorous and predatorytion from one state to an alternate state is nowa- zooplankton species, and detritivores, chaotic dy-days called a ‘catastrophic regime shift’, indicating namics also appear (Beninca et al. 2008). The foodthe serious ecosystem-level implications of this web showed strong fluctuations in species abun-phenomenon. The major problem is that ecological dances, attributed to different species interactions.resilience cannot be measured in practice. Models We refer to this study later. As both the structurecan be used as indicators of ecological resilience and the dynamics of a closed, local community are(Carpenter et al. 2001), but the mechanisms of these the result of the interactions among the constitutingtransitions are often poorly known (Scheffer and species, we need population dynamic models to repre-Carpenter 2003). Recently, van Nes and Scheffer sent them. Such interactions may be of different(2007) and Carpenter et al. (2008) have successfully kinds, such as between predators and prey,explored various indicators of an upcoming cata- between competitors for the same resource, andstrophic shift, such as the ‘critical slowing down’ non-trophic interactions, e.g. through environmen-known from physics. tal modification (Olff et al. 2009). In the present chapter we will discuss the dynamics of small food web modules and those of complex interactions.2.1.3 Stable limit cyclesApart from the predator–prey oscillations based on 2.2 Dynamics of food web modulesthe Lotka–Volterra equations which are only neu-trally stable, the Holling–Tanner model (Holling Insights from specific ‘few-species-interaction-1965; Tanner 1975) produces a range of dynamics. configurations’ or modules (Menge 1995; HoltThis model shows no tendency to return to the 1997; Bascompte and Melian 2005) of consumer–equilibrium point, but displays another form of resource interactions have much increased overpredator–prey oscillations: stable limit cycles. In the last few decades. For example, we know muchthe above-mentioned microbial food web study, more now about resource competition (Schoenerobvious stable limit cycles were established at low 1974; Tilman 1982), mutualism (Oksanen 1988), ap-dilution rates of the chemostat systems (Becks et al. parent competition (Holt 1977), indirect mutualism2005). Maxima and minima for the predator and the (Vandermeer 1980; Ulanowicz 1997), intra-guildtwo preys recurred during the whole period of the predation (Polis et al. 1989), positive interactionsstudy. such as facilitation (Callaway 2007), positive
  • 41. TROPHIC DYNAMICS OF COMMUNITIES 27feedbacks (DeAngelis et al. 1986), regulatory feed- levels) that possess explicit dynamics (Fig. 2.1a–e).backs (Bagdassarian et al. 2007), trophic cascades With these simple ‘building blocks’, more realistic(Carpenter et al. 2008) and multiple stable state food webs can be ‘built’ (Fig. 2.1f), which in turn aredynamics (Scheffer and Carpenter 2003). These subsets of the true complexity in trophic interac-may all be considered organizational forces that tions found in real ecosystems. For example,structure food webs, but they may not all be of Fig. 2.2 shows the network of trophic interactionsequal importance. For example, Ulanowicz (1997) as found on intertidal sand flats in the Waddenmakes a strong case for the special importance of Sea, a soft-bottom intertidal ecosystem with com-indirect mutualism as an organizational force in plex trophic structure. This example shows howfood webs, as the resulting feedback loops ‘attract’ exploitative competition, food chains, apparentresources towards them. Other authors, such as Til- competition and intra-guild predation can operateman (1982), have emphasized the importance of simultaneously within the same ecosystem. In this,competition as a key organizational force in ecolog- it should be realized that food web descriptions inical communities. Again others, such as Krebs et al. terms of interaction topology and flows (as in(1999) emphasize the importance of predator–prey Fig. 2.2) generally capture the long-term averagesinteractions in structuring communities. Despite of organism densities and fluxes. The actual abun-the insights gained into such specific processes, dances may vary due to external drivers (such asthe question remains how such modules together varying weather conditions) and internal dynamicsorganize into complex interaction webs, and how to (e.g. limit cycles). To illustrate this point, Fig. 2.3address their relative importance. shows observations of the long-term population Food web modules are characterized by the fact dynamics of some of the bivalve species shownthat they are small systems (two or three trophic in the food web of Fig. 2.2. In this case, winterPredatorPrey–consumerResource (a) Food chain (b) Omnivory (c) Apparent (d) Exploitative (d) Predation competition competition on competing prey (e) Intraguild (f) A fraction of a food web with predation several interrelated modulesFigure 2.1 (a–f) Examples of trophic modules that are found in ecological communities. After Holt (1997) andBascompte and Melian (2005).
  • 42. 28 DYNAMICS 25 20 Birds Worms Carbon flow mgC m–2 d–1 Phalacrocorax Crangon Snails 0.1–1 carbo crangon Fish 1–10 Bivalves Crustaceans 10–100 Unicells >100 25 Calidris 20 20 20 20 20 20 20 20 alpina Pluvialis Haematopus Larus Numenius Limosa Somateria Calidris Larus 20 Pomatoschistus apricaria ostralegus ridibundus arquata lapponica mollissima canutus canus minutus 20 Nephtys hombergii 8 12 small 12 12 12 12 12 12 12 12 capitellidae polychaete Scolopus Meiobenthos Hydrobia Macoma Lanice Arenicola Cerastoderma Mya worms armiger ulva balthica pygospio marina edule arenaria 2 3+5+7 3+5 9 9 sediment sediment suspended micro- phytoplankton bacteria POC POC phytobenthosFigure 2.2 Example of a real food web, as observed on intertidal sand flats in Wadden Sea near the German island ofSylt, showing how the different types of modules from Fig. 2.1 can together form a complex network of interactions.POC, particulate organic carbon. Data from Baird et al. (2007).temperatures are thought to control abundances The relative importance of external forcing ver-through their effect on recruitment (Beukema et al. sus internal dynamics as causes of dynamics in2001). food webs under natural conditions is still hard to Not all modules will be equally important in assess, despite a long history of research on theevery ecosystem. For example, ecosystems that are subject (Pimm 1982, 1991; Loreau and de Mazan-dominated by many species within the same tro- court 2008). Various research lines can be distin-phic level, such as diverse grasslands, may be guished here, depending on their theoreticalstrongly structured by exploitative competition. versus experimental nature, the complexity of theOn the other hand, ecosystems such as the marine system under study and the level of control ofpelagic zone seem dominated by trophic chains. variation in external conditions (exclusion or inclu-Bascompte and Melian (2005) recently compared sion of forcing factors). Some theoretical studiesthe frequency of different types of modules across have investigated the dynamics of speciesnatural food webs. They found that apparent com- organized in simple modules (DeAngelis 1992).petition and intra-guild predation (Fig. 2.1) were Several studies have been performed under experi-generally overrepresented with respect to a suite mentally controlled conditions, in which the influ-of null models, while the level of omnivory varied ence of external variation on populations has beenhighly across ecosystems. mostly eliminated, e.g. as in the study by Beninca
  • 43. TROPHIC DYNAMICS OF COMMUNITIES 29 (a) Cerastoderma edule (b) Mya arenaria 600 1000 500 800 400 600 Number of individuals . m–2 300 400 200 200 100 0 0 1970 1975 1980 1985 1990 1995 1970 1975 1980 1985 1990 1995 (c) Mytilis edulis (d) Macoma baltica 600 1400 1200 500 1000 400 800 300 600 200 400 100 200 0 0 1970 1975 1980 1985 1990 1995 1970 1975 1980 1985 1990 1995 Year YearFigure 2.3 (a–d) Long-term population dynamics of some of the bivalve species shown in Fig. 2.2, as observed on thetidal flats of the western Wadden Sea in the Netherlands. Reproduced with permission from Beukema et al. (2001).et al. (2008). Such studies are expected to mostly the precise type of dynamics depending on modelshow dynamics that arise from the internal struc- formulation (Fig. 2.4). A good example of a studyture of food webs, and have been performed with of a simple consumer–resource interaction is thatboth small modules and more complex webs. Other of the limnetic crustacean zooplankton speciesstudies have explored the dynamics of interacting Daphnia and its edible algal prey. McCauley et al.species populations under field conditions, which (1999) found large- and small-amplitude cyclesallows the assessment of the importance of external in the same global environment, i.e. consumer–forcing factors. However, there is a necessary trade- resource (predator–prey) and cohort (stage-off here. structured) cycles (Fig. 2.5). In cohort cycles, de- mographic stages (usually thought to be the juve- nile stage in Daphnia) are capable of strongly2.3 Internal dynamics in food web suppressing the other stages (adults) by compet-modules or simple webs ing for food. As the suppressing stage matures orIn theory, consumer–resource interactions consist dies, a pulse of reproduction or growth follows inof two trophic levels that can fluctuate for a long the other stage, causing a cycling strongly out oftime, but can also lead to unstable dynamics, with phase (McCauley et al. 1999).
  • 44. 30 DYNAMICS (a) (b) 80 6 5 Population abundance 60 4 40 3 20 2 1 0 0 50 100 150 0 10 20 30 40 50 Time TimeFigure 2.4 The dynamics of two-species predator–prey interactions predicted by (a) the Lotka–Volterra model (Lotka,1926; Volterra, 1926) and (b) the Nicholson–Bailey model (Nicholson and Bailey, 1935). In both models, the dynamicsare unstable. In the Lotka–Volterra model, the dynamics are neutral cycles with the period set determined by the modelparameters and amplitude set by the initial conditions. Predators lag prey by one-quarter of a cycle. In the Nicholson–Bailey model, the dynamics show divergent oscillations with overexploitation by the predator, leading to the extinctionof both prey and predators. Dashed lines, prey; solid lines, predators. Reproduced with permission from May andMcLean (2007). a b 3 3 10 8 2 2 6 Daphnia Biomass (mg/L) 4 Chlorophyll a (mg/L) 1 1 2 0 0 0 c 3 d 3 10 8 2 2 6 4 1 1 2 0 0 0 150 170 190 210 230 150 170 190 210 230 Julian date Julian dateFigure 2.5 Large- and small-amplitude cycles in the same global environment. Population dynamics of Daphnia(triangles) and their edible algal prey (squares) in four nutrient-rich systems from one treatment. (a and b) Examples oflarge-amplitude predator–prey cycles. (c and d) Examples of small-amplitude stage-structured cycles. The initial biomassof the replicates is similar. Daphnia biomass is calculated from estimates of density and size structure, using length–weight relationships measured for the clone used in the experiment. Algal biomass is measured as chlorophyll aconcentration. Reproduced from McCauley et al. (1999), with permission from Nature.
  • 45. TROPHIC DYNAMICS OF COMMUNITIES 31 a Cyclopoids Protozoa Rotifers Calanoids Picophyto– Nanophyto– Filamentous plankton plankton diatoms Nutrients Bacteria Ostracods Harpacticoids Detritus 3.0 120 b c Herbivores (mg fwt l–1) Cyclopoids (mg fwt l–1) 2.0 60 1.0 0.6 12 0.4 8 0.2 4 0.0 0 160 200 Phytoplankton (mg fwt l–1) d e X 10 Nutrients (µmol l–1) 80 100 12 45 8 30 4 15 0 0 8 5.0 Detritivores (mg fwt l–1) f g Bacteria (mg fwt l–1) 2.5 6 0.6 4 0.4 2 0.2 0 0.0 0 500 1000 1500 2000 2500 0 500 1000 1500 2000 2500 Time [days] Time [days]Figure 2.6 Description of a plankton community in a mesocosm experiment. (a) Food web structure of the mesocosmexperiment. The thickness of the arrows gives a first indication of the food preferences of the species, as derived fromgeneral knowledge of their biology. (b–g) Time series of the functional groups in the food web (measured asfreshweight biomass). (b) Cyclopoid copepods; (c) calanoid copepods (red), rotifers (blue) and protozoa (dark green);(d) picophytoplankton (black), nanophytoplankton (red) and filamentous diatoms (green); note that the diatom biomassshould be magnified by 10; (e) dissolved inorganic nitrogen (red) and soluble reactive phosphorus (black);(f) heterotrophic bacteria; (g) harpacticoid copepods (violet) and ostracods (light blue). Reproduced with permissionfrom Beninca et al. (2008). See plate 1.
  • 46. 32 DYNAMICS This interaction between this grazer species and subject to strong external forcing, e.g. where region-its prey can be thought of as an ‘elemental oscilla- al climatic conditions affect local air, water or soiltor’, the basic building block for ecological commu- temperature. Ecophysiological differences amongnities (Vandermeer 1994). Leibold et al. (2005) species in ability to cope with these changes mayworking with a more complex food web, consisting result in species-specific responses (Karasov andof three grazer species (Daphnia, Ceriodaphnia and Martinez del Rio 2007), thus leading to communityChydorus) and edible algae, hypothesized that the dynamics under varying external conditions. Thisdynamics in such complex food webs can be under- external forcing is the key ‘point of entry’ in study-stood in terms of the simple subsets used by ing effects of climate change on food webs, but alsoMcCauley et al. (1999). The behaviour of these how toxic pollutants will affect trophic structuremore complex food webs might be understood by and ecosystem functioning. For example, ecto-thinking of coupled oscillators consisting of many therms (at lower trophic levels) and endothermssuch oscillators with interacting damping and am- (at higher trophic levels) are expected to respondplifying harmonics (Hastings and Powell 1991). very differently to short- or long-term temperatureThey also found both consumer–resource and co- changes. Surprisingly, although there are good rea-hort cycles. This indicates that interactions of zoo- sons to suspect its importance in natural popula-plankton and algae in complex systems still consist tions, e.g. in the level of synchrony between speciesof the same basic elements – in this case consumer– in long-term ecological monitoring (Bakker et al.resource cycles and cohort cycles – and their dy- 1996), environmental forcing has hardly receivednamics can be understood from the dynamics of any attention in the study of consumer–resourcetheir component parts. In the study with a more interactions, food webs or other interaction webs.complex marine web consisting of phytoplankton There is, however, some relevant theoretical work:and zooplankton, Beninca et al. (2008) found strong a mechanistic-neutral model describing the dynam-chaotic fluctuations (Fig. 2.6). Species interactions ics of a community of equivalent species influencedin this food web are indicated as the driving forces. by density dependence, environmental forcing andThe persistence of this food web despite the great demographic stochasticity. The model shows thatdensity fluctuations and unpredictability of the demographic stochasticity alone cannot oppose theabundances is a rarely demonstrated phenomenon. synchronizing effect of density dependence andIt may also be more common than we think. The environmental forces (Loreau and de Mazancourtconstant external conditions used in this study may 2008). Vasseur and Fox (2007) have shown in theirbe an artefact in itself, causing high productivity, model food web study that the synchronization ofleading to a situation that has been called the para- dynamics is the result of environmental fluctua-dox of enrichment (Rosenzweig 1971). Many other tions. This synchrony promotes stability, becauseimportant aspects of this study that distinguish it the maximum abundance of top predators is re-from real food webs under natural conditions are duced by the synchronous decline in the densitythe absence of higher trophic levels and the exclu- of consumers and synchronous increase in consum-sion of interactions between organisms and their er density is changed by resource competition intoabiotic environment, e.g. through local nutrient de- synchronous decline. These authors conclude thatpletion or organism–sediment feedback. Even future studies on food web dynamics should takethough isolated modules of species may exhibit cha- into account the joint action of internal feedbacksotic dynamics, this may be highly dampened or even and external forcing.excluded by these effects in natural systems. Another example concerns recovery after pertur- bation. The reaction of a system to a perturbation depends on the size of the ‘basin of attraction’. The2.4 Dynamics enforced by external basin of attraction is a theoretical measure of theconditions maximal perturbation that the system can absorbIn addition to dynamics that arise internally within without shifting to another state and is often re-communities, species populations are also often ferred to as ‘ecological resilience’ (Peterson et al.
  • 47. TROPHIC DYNAMICS OF COMMUNITIES 331998; Folke et al. 2004). Systems with a high ecolog- where all species at one trophic level are lumpedical resilience can be seen as particularly stable into ‘trophic species’, under the assumption that thesystems, whereas systems with a low ecological ‘food chain module’ in the system (see Fig. 2.1)resilience can be seen as unstable systems close to strongly overrules the dynamics of the systemthe tipping point into an alternate state. Recent with respect to other modules (whether this is truestudies have explored whether there are early is, however, an open question). This approach towarning signals for this shift into another state. simplify systems leads to the concept of trophicThe slow recovery from perturbations may be a cascades. Hairston et al. (1960) and Fretwell (1977)possible indicator of an impending state shift (van mention this trophic cascade phenomenon in theirNes and Scheffer 2007). writings about population regulation, although they did not call the process by this particular name (Morin 1999). Paine (1980) used the term2.5 Equilibrium biomass at different trophic cascade to describe how the top-down effectsproductivities of predators could influence the abundances ofImportant lessons can be learned from comparing species in lower trophic levels, a concept that wasthe configuration of food webs (abundances of dif- developed further by Oksanen et al. (1981)ferent species on different trophic levels) under (Fig. 2.7a). If we increase the length of this trophicdifferent external conditions. A simple approach food chain by adding one additional trophic levelthat has been used is to compare systems subject (top carnivores) (Fig. 2.7b), indirect mutualismto different boundary conditions that impose con- between non-adjacent trophic levels and a decreasestraints on their level of primary production. Such or constancy in the abundance of ‘odd’ levels withdifferences can be caused by variation in tempera- increasing productivity appear.ture or nutrient input into the system. Evidence for top-down trophic cascades is sur- This approach can be used to study simple food prisingly scarce, and comes primarily from aquaticchains consisting of a predator, a consumer and a systems: stream communities (Power et al. 1985)resource species. This approach can also be used for and lakes (Carpenter and Kitchell 1993). There aresimple systems with three levels, or for systems few terrestrial examples: Emmons (1987), Terborgh (a) Three trophic levels (b) Four trophic levels Tertiary consumers C3 max Secondary consumers max Secondary consumers C2 0 0 max Primary consumers Primary consumers Abundance max C1 0 0 max Primary producers Primary producers max P 0 0 Primary production Primary productionFigure 2.7 Equilibrium of three (a) and four (b) trophic levels along a gradient of primary productivity according to theecosystem exploitation theory of Oksanen et al. (1981).
  • 48. 34 DYNAMICSet al. (2006), Marquis and Whelan (1994) and 2.6 Dynamics of complex interactionsJefferies (1999). For example, Emmons and Terborghet al. describe the disappearance of several common Food webs are conceptualized by their basic unit ofspecies of birds from Barro Colorado Island. The interaction, consumption, and this basic process isbuilding of the Panama Canal created an island oscillatory. When these basic units are connected,that was too small to sustain large predators, such the conceptual framework becomes a system ofas jaguars and pumas. Their extinction led to popu- coupled oscillators. This concept generated notablelation increases of their prey species. These meso- patterns in the theoretical literature as describedpredators fed on the eggs and young of ground- by, for example, Vandermeer (2004). The conclu-nesting birds and their increase in numbers was sion that weak interactions can have strong effectssufficient to wipe out many bird populations. on stabilizing ecosystems (McCann et al. 1998; Neu- Another example concerns a terrestrial trophic tel et al. 2002) derives from this concept of coupledcascade from a study by Marquis and Whelan oscillators. According to May (1973), measures of(1994). They found strong effects of insectivorous interaction strength are the elements in a communi-birds foraging herbivorous insects on white oak ty matrix at equilibrium, which represent the directtrees. Birds significantly reduced the abundance of effect of an individual of one species on the totalherbivorous insects on the oaks. Netting around population of another species at equilibrium. Thesesome trees caused exclusion of birds from the in- results suggest that average interaction strengthsects, while other uncaged trees remained available should be weak in species-rich, highly connectedto the birds. Oaks with birds and reduced herbivo- systems. The fundamental question is whetherrous insects had less leaf damage from insects and the configuration or distribution of interactionsubsequently had a higher biomass. strengths within food webs is important for ques- A slight modification of simple trophic cascades tions of community stability. de Ruiter et al. (1995)leads to intra-guild predation (IGP; Polis et al. 1989). linked the differing approaches by deriving valuesThis type of interaction can be seen as an extension of the matrices from empirical observationsof a simple predator–prey interaction: the predator (Fig. 2.8). The data come from a terrestrial foodalso eats some of the consumers, but potentially web study, the Lovinkhoeve Experimental Farmcompetes with uneaten consumers as well for a (Integrated Management) in The Netherlands.common resource. Spiller and Schoener (1989) This study indicates that the patterning of interac-studied interactions between predatory Anolis tion strengths is essential for system stability. How-lizards, predatory web-building spiders and their ever, there is no direct correlation betweenarthropod prey on small islands in the Bahamas. interaction strength and stability. Weak interactionsThe interaction between lizards and spiders can be may be strong in terms of their stabilizing effects todescribed as an IGP interaction, because lizards eat the community. Further, it has been shown thatsome spiders, but lizards also potentially compete long trophic loops contain relatively many weakwith uneaten spiders for small arthropod prey. links increasing food web stability because theyThere is even an effect on the level of anti-herbivore reduce maximum loop weight, thus reducing thedefence of the dominant vegetation on the islands, amount of intraspecific interaction needed forConocarpus erectus or buttonwood (Schoener 1988). system stability (Neutel et al. 2002) (Fig. 2.9).On islands without lizards the leaves have tri-chomes to discourage insect attack; on islands 2.7 Conclusionswith lizards, leaves are without them. Summarizing we can state that in small food The types of dynamics leading to long-termwebs, oscillating consumer–resource interactions co-occurrence of species in communities can beare not only predicted by models but also occur in indicated as stable equilibria, alternate equilibria,natural systems. In chains of three or more levels stable limit cycles and chaotic dynamics. Althoughtrophic cascades are important, but experimental much of the attention to this subject is theoretical,support is limited. empirical data are increasingly reported, although
  • 49. TROPHIC DYNAMICS OF COMMUNITIES 35 Trophic position Resource Top predators Consumer Predatory collembola Predatory mites Nematophagous mites Predatory mites Predatory nematodes Predatory mites Predatory nematodes Predatory collembola Collembola Predatory mites Cryptostigmatic mites Predatory mites Non-cryptostigmatic mites Predatory mites Fungivorous nematodes Predatory mitesBacteriophagous nematodes Predatory mites Bacteriophagous mites Predatory mites Predatory nematodes Nematophagous mites Fungivorous nematodes Predatory collembolaBacteriophagous nematodes Predatory collembola Fungivorous nematodes Nematophagous mitesBacteriophagous nematodes Nematophagous mites Phytophagous nematodes Predatory mites Amoebae Predatory nematodes Fungivorous nematodes Predatory nematodes Flagellates Predatory nematodesBacteriophagous nematodes Predatory nematodes Phytophagous nematodes Predatory collembola Phytophagous nematodes Nematophagous mites Flagellates Amoebae Phytophagous nematodes Predatory nematodes Bacteria Predatory nematodes Bacteria Amoebae Fungi Collembola Fungi Cryptostigmatic mites Fungi Non-cryptostigmatic mites Fungi Fungivorous nematodes Bacteria Flagellates Bacteria Bacteriophagous nematodes Bacteria Bacteriophagous mites Fungi Enchytraeids Bacteria Enchytraeids Detritus Enchytraeids Roots Phytophagous nematodes Detritus Fungi Detritus Bacteria Basal resources 1 100 –40 –30 –20 –10 0 0 0.1 0.2 0.3 0.4 0.5 0 10 2030 40 50 60 0.1 10 1000 per capita effect on prey per capita effect on predator (mean = –7.6) (mean = 0.1) Feeding rate Interaction strength Impact on stability (kg ha–1 year –1) (year –1) (%) a b cFigure 2.8 Feeding rates (a), interactions strengths (b) and impacts of the interactions on food web stability (c) arrangedaccording to trophic position in the soil food web of arable fields with conventional agricultural practices at theLovinkhoeve Experimental Farm in The Netherlands. From de Ruiter et al. (1995). Reprinted with permission from AAAS.mostly still for experimental communities. For assess. The study of the effects of climate changefield studies, the life span of the organisms and toxic pollutants on food webs makes it neces-involved is often too long compared with the sary to concentrate in the future on the interplay oflength of the study to unambiguously evaluate internal feedbacks and external forcing. Interest-stability, but recent studies deal with that problem. ing lessons can be learned from the comparison ofInsights gained from the dynamics of specific food the configuration and dynamics of food websweb modules have increased over the last few under different external conditions, e.g. differentdecades. But still the question remains how such temperatures that lead to different primary pro-modules are organized together into complex in- ductivity. Where strong evidence for the existenceteraction webs, and how their interplay changes of trophic cascades initially came from aquatictheir dynamics as observed in isolated modules. systems, terrestrial examples are now known.The relative importance of external forcing versus New studies on the dynamics of complex interac-internal dynamics for causing dynamics in food tion webs have concentrated on the consequenceswebs under natural conditions is still hard to of specific patterning of interaction strengths
  • 50. 36 DYNAMICS a b c 4.5 2.5 4 intraspecific interaction Loop weight (year–1) 2 Minimum degree s of 3.5 3 1.5 2.5 2 1 1.5 1 0.5 0.5 0 0 0 2 4 6 0 2 4 6 0 2 4 6 Loop length Loop length Maximum loop weight (year–1)Figure 2.9 Loop length, loop weight and stability in the soil food web of the short grass prairie of the Central PlainsExperimental Range and randomizations (20) of this matrix. (a) Loop weight versus loop length in the real matrix. (b) Loopweight versus loop length in a randomized matrix (a typical example). Long loops with a relatively small weight (thosewith many bottom-up effects) are not shown because they are not relevant for maximum loop weight. (c) Maximum loopweight and stability of the real matrix (solid diamond) and of 10 randomized matrices (open diamonds). Stability wasmeasured as the value s that leads to a minimum level of intraspecific interaction strength needed for matrix stability. In asensitivity analysis, it was found that variation in the parameter values within intervals between half and twice theobserved value led to only a small variation in stability. From Neutel et al. (2002). Reprinted with permission from AAAS.across the web for the stability of the overall sys- modelling frameworks are required to show thetem. Using classic stability analysis with Lotka– generality of this phenomenon.Volterra interaction terms, soil food webs withlong trophic loops appear to contain relatively Acknowledgementsmany weak links, which increases their stability.Future studies of more webs and with different We thank Dick Visser for drawing Figure 2.2.
  • 51. CHAPTER 3 Modelling the dynamics of complex food webs Ulrich Brose and Jennifer A. Dunne3.1 Introduction network structure, network dynamics and various aspects of stability such as persistence, robustnessThe world is currently facing losses of biodiversity and resilience in complex ‘real-world’ networks isand habitats, species invasions, climate change, one of the greatest current challenges in the naturalgroundwater depletion and other anthropogenic and social sciences, and it represents an excitingperturbations that are resulting in the drastic reor- and dramatically expanding area of cross-disciplin-ganization of many ecosystems. In order to under- ary inquiry (Strogatz 2001).stand, predict and mitigate such reorganizations,which can severely affect ecosystem services(Daily 1997) that humans depend on such as water 3.2 Simple trophic interaction modulessupply and purification and crop pollination, re- and population dynamicssearchers need to broaden their focus from particu- While there are many types of interactions thatlar types of species to whole ecosystems. Ecological species can have with each other, trophic (feeding)network research provides a very compelling interactions are ubiquitous, are central to bothframework for addressing the complexity of species ecological and evolutionary dynamics and are rela-interactions with each other and the environment. tively easily observed, defined, quantified andFor example, the effects of any stressor – abiotic or modelled compared with other kinds of inter-biotic, natural or anthropogenic – on one popula- actions. Within ecology, food web research, thetion can cascade through ecological networks as a study of networks of trophic interactions, repre-result of direct and indirect interactions, potentially sents a long tradition of both empirical and theore-affecting any other population in the same ecosys- tical network analysis (Elton 1933; Lindeman 1942;tem. While much research has focused on direct MacArthur 1955; May 1973; Cohen et al. 1990; Pimmeffects between species or their populations, empir- et al. 1991; see review by Egerton 2007). Ideally,ical and modelling studies have shown that effects food web research seeks to identify, analyse anddue to indirect species interactions can be as impor- model feeding interactions among whole commu-tant as direct effects in driving outcomes (Abrams nities of taxa including plants, bacteria, fungi, in-et al. 1995; Menge 1997; Yodzis 2000). In fact, indi- vertebrates and vertebrates, with feeding linksrect effects can be stronger than the direct effects of representing transfers of biomass via various tro-a stressor, potentially greatly modifying the overall phic interactions including detritivory, herbivory,outcomes for population abundances in the face of predation, cannibalism and parasitism. A great dealextinctions (Ives and Cardinale 2004). Network of food web research, starting in the mid-1970s, hasanalysis and modelling provide approaches for focused on various aspects of how such networksquantifying and assessing both direct and indirect are structured (see review by Dunne 2006), witheffects. In general, determining the interplay among emerging strong empirical and model-based 37
  • 52. 38 DYNAMICSevidence that food webs from different habitatshave similar characteristic topologies (Williamsand Martinez 2000; Camacho et al. 2002; Dunneet al. 2002, 2004; Milo et al. 2002; Stouffer et al.2005, 2006, 2007; see also Chapter 1). However, food webs are inherently dynamicalsystems, since feeding interactions involve variableflows of biomass among species whose populationdensities are changing over time in response todirect and indirect interactions. Because it is diffi-cult to compile detailed, long-term empirical data-sets for dynamics of two or more interactingspecies, much research on species interaction dy-namics relies on modelling. Many modelling stud-ies of trophic dynamics have used analytically Figure 3.1 Structure of food web modules. (a) Tri-trophictractable approaches to explore predator–prey or food chain; (b) omnivory or intra-guild predation module.parasite–host interactions (Yodzis and Innes 1992;Weitz and Levin 2006) or small modules of inter- the intermediate species). Now, omnivory as theacting taxa (McCann et al. 1998; Fussmann and stage between these extremes can stabilize popula-Heber 2002), generally ignoring all but the most tion dynamics by either eliminating chaotic dynam-simple of possible network structures. In natural ics or bounding the minima of the populationecosystems, such interaction dyads or modules are densities away from zero (McCann and Hastingsembedded in diverse, complex food webs, where 1997). However, opposite results can be obtainedmany additional taxa and their direct and indirect with different parameters for population traitseffects can play important roles for both the stabili- (Vandermeer 2006). Generally, omnivory can stabi-ty of focal species and the stability of the broader lize the population dynamics if the tri-trophic foodcommunity. Therefore, it is critically important to chain and the exploitative competition module atask whether knowledge about population dynam- the extremes of the gradient are unstable, whereasics obtained in small interaction modules (Yodzis omnivory should destabilize the system if the mod-and Innes 1992; McCann and Yodzis 1994; McCann ules at the extremes of the gradient are stable (Van-et al. 1998; Fussmann and Heber 2002; Weitz and dermeer 2006). Furthermore, the consequences ofLevin 2006) applies to population dynamics in omnivory in a comparable simple experimentalmore realistically complex food webs. system are highly dependent on nutrient enrich- Analyses of such modules suggest that every ment, since coexistence of both consumers is re-additional feeding link between the species may stricted to intermediate nutrient saturations (Diehlchange the population dynamics dramatically. For and Feissel 2001). Extending the size of the foodinstance, the dynamics of three populations in a tri- web modules beyond three populations, Fussmanntrophic food chain (Fig. 3.1a) can be stabilized or and Heber (2002) demonstrated that the frequencydestabilized by an additional link from the top spe- of chaotic dynamics increases with the number ofcies to the basal species (McCann and Hastings trophic levels, but decreases with other structural1997; Vandermeer 2006). By convention, this mod- properties that cause higher food web complexity.ule (Fig. 3.1b) is termed an omnivory module (or Interestingly, these results indicate that populationintra-guild predation module). Depending on the stability might increase or decrease with food webrelative strength of the links this module represents complexity, depending on which process domi-an intermediate stage between a tri-trophic food nates in a particular food web. These studies havechain (in which the top species does not consume emphasized the critically important roles of net-the basal species) and an exploitative competition work complexity and the distribution of interactionmodule (in which the top species does not consume strengths across feeding links in determining
  • 53. MODELLING THE DYNAMICS OF COMPLEX FOOD WEBS 39population dynamics. However, it remains to be existence by a keystone consumer was first de-seen (1) whether our understanding of population scribed for the starfish Pisaster ochraceus thatdynamics in small modules can be scaled up to preferentially consumes the competitively domi-complex food webs and (2) how global characteris- nant mussel Mytilius californianus, and thereby fa-tics of complex food webs such as connectance (a cilitates the coexistence of a diverse community ofmeasure of link richness – the probability that any basal species that are competitively subordinate totwo species will interact with each other; see Chap- the mussel (Paine 1966, 1974). Thus, the starfish haster 1) affect population dynamics. In the next sec- a positive effect on the biomass density of mosttion, we describe an approach using keystone species in the intertidal food web (i.e. the biomassspecies that addresses the first question. density of most species is higher when the starfish is present than when it is absent). This facilitation by Pisaster was termed a ‘keystone effect’. Similar3.3 Scaling up keystone effects in keystone effects of other consumer species havecomplex food webs subsequently been documented for many other eco-In keystone species modules, a keystone consumer systems (Power et al. 1996), which suggests a broadof a competitively dominant basal species facilitates generality of this phenomenon. However, thethe coexistence of competitively subordinate basal strength of the keystone effect varies dramaticallyspecies (Fig. 3.2). When the keystone species is ex- between years and sites within ecosystems (Paineperimentally removed or goes locally extinct, com- 1980; Menge et al. 1994; Berlow 1999). Analyses ofpetition can lead to extinction of the subordinate keystone modules have shown that keystone effectsbasal species. Such facilitation of basal species co- vary substantially with the presence or absence of peripheral (non-keystone) species and links (Brose et al. 2005). This suggests that keystone effects in Keystone K NK complex food webs (Fig. 3.2b) might be highly con-a Non-keystone text dependent. Competitive dominant Systematic simulation analyses of complex food Competitive subordinate D S webs have revealed surprisingly simple determi- Nutrient most needed Nutrient nants of keystone effects (Brose et al. 2005). Gener- ally, distant effects of the global network structure Strong nutrient uptake or effects of populations that are more than two Weak nutrient uptake degrees (trophic links) separated from the keystone module are buffered by the network structure.b Most likely, the multiple pathways between popu- lations in complex food webs are characterized by effects of different signs that cancel each other out. Thus, effects between two species over pathways longer than two degrees often cancel each other out, and only effects over one or two degrees of separa- K tion that dominate in food webs (Williams et al. 2002) are not balanced by other interaction path- ways of opposite sign and could systematically D S4 vary the strength of the keystone effect. These S1 S2 S3 results suggest that effects within keystone mod- ules that are embedded in complex food webs are affected by other populations within a ‘local inter- action sphere’ of influence, which includes effectsFigure 3.2 Keystone consumer in (a) a small module and of non-keystone species that are separated by one(b) a complex food web. See plate 2. or two links from the keystone module (Brose et al.
  • 54. 40 DYNAMICS2005). If these results generalize to other types of which already had great intuitive appeal as wellmodules such as omnivory modules, population as the weight of history behind it, was thus ac-dynamics in natural food webs may be determined corded a gloss of theoretical rigor (e.g. a ‘formalby local interaction spheres that are larger than the proof’ according to Hutchinson 1959), and took onclassic modules, but smaller than entire complex the patina of conventional wisdom by the latefood webs. 1950s. The concept that complexity implies stability, as a theoretical generality, was explicitly and rigor-3.4 Diversity/complexity–stability ously challenged by the analytical work of Mayrelationships (1972, 1973), a physicist by training and ecologistMuch of the research on food web dynamics in by inclination. He used local stability analyses ofmodel systems with more than two taxa has been randomly assembled community matrices to math-orientated around the classic (May 1972) and ematically demonstrate that network stability de-enduring (McCann 2000) debate on how diversity creases with complexity. In particular, he foundand complexity of communities affect food web that more diverse systems, compared with less di-stability. In the first half of the 20th century, many verse systems, will tend to sharply transition fromecologists believed that natural communities devel- stable to unstable behaviour as the number of spe-op into stable systems through successional dy- cies, the connectance or the average interactionnamics. Aspects of this belief developed into the strength increase beyond a critical value. May’snotion that complex communities are more stable analytical results and his conclusion that ‘in generalthan simple ones (Odum 1953; MacArthur 1955; mathematical models of multispecies communities,Elton 1958; Hutchinson 1959). Some popular exam- complexity tends to beget instability’ (May 2001,ples included the vulnerability of agricultural p. 74) turned earlier ecological ‘intuition’ on itsmonocultures to calamities in contrast to the appar- head and instigated a dramatic shift and refocusingent stability of diverse tropical rainforests, and the of theoretical ecology. His results left many empiri-higher frequency of invasions in simple island com- cal ecologists wondering how the astonishing di-munities compared with more complex mainland versity and complexity they observed in naturalcommunities. It was thought that a community con- communities could persist, even though May him-sisting of species with multiple consumers would self insisted there was no paradox. May framed ahave fewer invasions and pest outbreaks than com- central challenge for ecological research this way:munities of species with fewer consumers. This was ‘In short, there is no comfortable theorem assuringstated in a general theoretical way by MacArthur that increasing diversity and complexity beget en-(1955), who hypothesized that ‘a large number of hanced community stability; rather, as a mathemat-paths through each species is necessary to reduce ical generality, the opposite is true. The task,the effects of overpopulation of one species’. therefore, is to elucidate the devious strategiesMacArthur concluded that ‘stability increases as which make for stability in enduring naturalthe number of links increases’ and that stability is systems’ (May 2001, p. 174).easier to achieve in more diverse assemblages ofspecies, thus linking community stability with 3.5 Stability of complex food webs:both increased trophic links and increased numbers community matricesof species. Other types of theoretical considerationsemerged to support the positive complexity– Despite methodological criticism of May’s method-stability relationship. For example, Elton (1958) ology (e.g. Cohen and Newman 1984) muchargued that simple predator–prey models reveal subsequent work related to food webs was devotedtheir lack of stability in the oscillatory behaviour to finding network structures, species’ strategiesthey exhibit, although he failed to compare them and dynamical characteristics that were consistentwith multispecies models (May 1973). The notion with May’s theorem or that would allow complexthat ‘diversity and complexity beget stability’, communities to be stable or persist. May (1972,
  • 55. MODELLING THE DYNAMICS OF COMPLEX FOOD WEBS 411973) used random matrices of interactions for his teraction strength distributions increased its localanalyses, which yield random network structures stability in comparison with random networks(‘who eats whom’) and random distributions of (Neutel et al. 2002). Together, these studies demon-traits and interaction strengths (‘how much is strate that characteristics of the distribution of linkseaten’) across the populations and links in the net- and interaction strengths within natural food websworks. In natural food webs, however, neither the account for their stability.matrices of interactions nor the distributions of in-teraction strengths are random. Systematic patterns 3.6 Stability of complex food webs:of low assimilation efficiencies, strong self-regula- bioenergetic dynamicstion (negative effects of a species on its ownbiomass) or donor control of biomasses can cause More recent theoretical studies have extended apositive complexity–stability relationships (DeAn- numerical integration approach of ordinary differ-gelis 1975). Subsequent analyses demonstrated ential equations to complex food web modelsthat non-random empirical network structures of (Williams and Martinez 2004; Martinez et al. 2006).natural food webs are more dynamically stable In this approach, the structure of the complexthan random networks (Yodzis 1981), suggesting networks is defined by a set of simple topolog-that natural food webs possess a topology that in- ical models: random, cascade, niche or nested-creases the stability of the population dynamics. hierarchy model food webs (Cohen et al. 1990; Extending the approach of adding empirical re- Williams and Martinez 2000; Cattin et al. 2004).alism to stability analyses, de Ruiter et al. (1995) The dynamics follow a bioenergetic model (Yodzisparameterized May’s general community matrix and Innes 1992) that defines ordinary differentialmodel with empirical food web structures and in- equations of changes in biomass densities for eachteraction strengths among the species. In their em- population. Numerical integration of these differ-pirical data, they found a pattern of strong top- ential equations yields time series of the biomassdown effects of consumers on their resources at evolution of each species, which allows explorationlower trophic levels in food webs and strong bot- of population stability and species persistence.tom-up effects of resources on their consumers at Similar to results from community matrix models,higher trophic levels. Adding empirical interaction non-random network structure increases an aspectstrength patterns to the community matrices stability in these bioenergetic dynamics modelsincreased their local stability in comparison with of complex food webs: the overall persistence ofmatrices with random interaction strength values species (Martinez et al. 2006).(de Ruiter et al. 1995). Most importantly, these re- One key parameter of population dynamic mod-sults demonstrated that natural food web struc- els is the functional response describing the pertures as well as the distribution of interaction capita (per unit biomass of the predator) consump-strengths within those structures contribute to an tion rate of a predator depending on the prey bio-increased local stability of the corresponding com- mass density. Generally the per capita consumptionmunity matrices. Neutel et al. (2002) explained this rate is zero at zero prey density and then increasesfinding with results that showed that weak interac- with increasing prey density. According to thetions are concentrated in long loops. In their analy- shape of this increase classical functional responsesis, a loop is a pathway of interactions from a models are characterized as (1) linear or type I, (2)certain species through the web back to the same hyperbolic or type II, or (3) sigmoid or type IIIspecies, without visiting other species more than functional responses (Fig. 3.3). The linear functionalonce. They defined loop weight as the geometric response was used in classic population dynamicsmean of the interaction strengths in the loop and studies (Lotka 1925; Volterra 1926), but lacks a bio-showed that loop weight decreases with loop logically necessary saturation in consumption ratelength. Again, when applied to the community at high prey density. A generalized non-linearmatrix, this empirically documented pattern of in- saturating functional response model (Real 1977) is
  • 56. 42 DYNAMICS Type I switching and refuge seeking. These types of tro- phic and non-trophic behaviours allow rare or low-Per capita consumption Type II Type III biomass resource species to persist in both natural and model ecosystems, increasing overall commu- nity persistence. In another approach to integrating complex structure and dynamics, and in contrast to most prior dynamical studies, Kondoh (2003, 2006) allowed consumer preferences for resources to Prey density adaptively vary. This adaptation process increases the predator preferences for prey of above-average Figure 3.3 Functional responses: per capita consumption density and decreases preferences for prey of rates of consumers depending on prey density. below-average density. This adaptive foraging model yields positive complexity–stability relation- ships in complex food webs if (1) the fraction of adaptive foragers and (2) the speed of adaptation cB h FðBÞ ¼ ð3:1Þ are sufficiently high, and (3) the number of basal 1 þ cTh B h species does not vary with food web complexity where F is the predator’s consumption rate, B is (Kondoh 2003, 2006). Interestingly, the process of prey biomass density, Th is the handling time that decreasing preferences for prey of low density includes the time a predator needs to catch, kill and causes relaxation of feeding at low prey density, ingest a unit biomass of the prey, c is a constant that which is mechanistically similar to type III func- describes the increase in the attack rate, a, with the tional responses. prey abundance (a$cBhÀ1), and h is the Hill coeffi- cient that varies between 1 (type II functional re- sponse) and 2 (type III functional response). Varying the Hill exponent between 1 and 2 grad- 3.7 Stability of complex food webs: ually converts a type II into a type III functional allometric bioenergetic dynamics response. Williams and Martinez (2004) showed Most species’ traits, T, follow close allometric that in food webs with a type II functional response power-law relationships with their body mass, M: many populations are unstable and prone to extinc- T ¼ aM b ð3:2Þ tion. Slight increases in the Hill exponent can have dramatic effects on stabilizing the dynamics of par- where a and b are constants (often b is approximately ticular species, as well as overall species persistence equal to 3/4) and T can represent the biological rates of (Williams and Martinez 2004). Additionally, Hill respiration, biomass growth or maximum consump- exponents slightly higher than 1 buffer predator– tion (Brown et al. 2004). These relationships can be prey modules and complex food webs against the used to parameterize population dynamic models, destabilizing effects of nutrient enrichment (Rall thus collapsing parts of the multidimensional et al. 2008). One important stabilizing feature of parameter space into a body-mass axis. Moreover, this variation in functional response is the slight natural food webs have a distinct body-mass struc- relaxation of consumption at low resource densi- ture of invertebrate and vertebrate predators being ties, which leads to accelerating consumption rates on geometric average roughly 10 and 100 times, with increasing prey density. This yields a strong respectively, larger than their prey (Brose et al. top-down pressure, which controls the prey popu- 2006a). Implementing theoretically and empirically lation to low equilibrium densities (Oaten and Mur- supported allometric relationships in population doch 1975). The relaxation of feeding at low dynamic models in complex food webs yields popu- resource density is evocative of a variety of well- lation dynamics models that are constrained by documented ecological mechanisms including prey the body-mass structure of the food webs (Brose
  • 57. MODELLING THE DYNAMICS OF COMPLEX FOOD WEBS 43et al. 2006b). In these allometric models, consumer– This result confirms classic food web stabilityresource body-mass ratios define the body mass of a analyses (May 1972) showing negative diversity–consumer relative to the average body mass of its stability relationships in random food webs, inresources. which species are on average equally sized. But it Brose et al. (2006b) varied the consumer–resource also confirms the earlier notion of empirical ecolo-body-mass ratios in allometric food web models gists (Odum 1953; MacArthur 1955; Elton 1958), whosystematically between 10À2 and 106, creating a assumed positive diversity–stability relationships ingradient of food webs with predators that are 100 natural food webs, in which species are on averagetimes smaller than their prey to food webs of pre- 10–100 times larger than their prey. Consideration ofdators that are 106 times larger than their prey (Fig. the body-mass structure of food webs may thus3.4). The persistence of species in the food webs (i.e. reconcile lingering gaps between the perspectivesthe fraction of the initial populations that persisted of theoretical and empirical ecologists on diversity–during the simulations) increased with increasing stability relationships. It is interesting to note that, inbody-mass ratios. Persistence is low when preda- this allometric modelling framework, a differenttors are smaller than or equal in size to their prey, measure of stability, the mean coefficient of variationbut persistence increases steeply with increasing of the species population biomasses in time in per-body-mass ratios (Fig. 3.4a). This increase saturates sistent webs (‘population stability’), decreases withat body-mass ratios of 10 and 100 for invertebrate increasing body-size ratios until inflection pointsand vertebrate predators, respectively (Brose et al. are reached that show the lowest stability, and then2006b), which is highly consistent with the geomet- increases again beyond those points (Brose et al.ric average body-mass ratios found in natural food 2006b). Those inflection points also correspond towebs (Brose et al. 2006a). Moreover, persistence the empirically observed body-size ratios (Brosedecreases with the number of populations in the et al. 2006a; Fig 3.4b). Thus, at intermediate body-food web at low body-mass ratios, whereas it ex- size ratios high species persistence is coupled withhibits a slight increase in persistence at high body- low population stability. Interestingly, this demon-mass ratios (Brose et al. 2006b). strated that an aspect of increased stability of the (a) (b) 1.0 0 –20 Population stability Species persistence (negative mean CV) 0.8 (S persistent / S initial ) –40 2 10 –60 0.6 1 10 –80 0.4 Invertebrates –100 101 Ectotherm vertebrates 102 0.2 –120 –2 0 2 4 –2 0 2 4 log10 consumer–resource body-size ratioFigure 3.4 Impact of consumer–resource body-size ratios on (a) species persistence and (b) population stabilitydepending on the species metabolic types. The inflection points for shifts to high-persistence dynamics are indicated byarrows for both curves, and those inflection points correspond to empirically observed consumer–resource body-sizeratios for invertebrate dominated webs (101, consumers are on average 10 times larger than their resources) andectotherm vertebrate dominated webs (102, consumers are on average 100 times larger than their resources). S, speciesrichness; CV, coefficient of variation. Figure adapted from Brose et al. (2006b).
  • 58. 44 DYNAMICSwhole system (species persistence) is linked to an population dynamics of these local interactionaspect of decreased stability of components of that spheres will be an important scientific challenge.system (population stability). Implementing allometric scaling relationships in models of complex food webs has helped in under- standing the processes that lead to particular dis-3.8 Future directions tributions of interactions and network stability.Our knowledge about the structure and dynamical Further integrations of metabolic allometry withconstraints of complex food webs has greatly im- food web research is likely to elucidate much ofproved over the last decade. The framework of dy- the constraints on structure and dynamics of com-namic food web models offers a great possibility to plex food webs. Here, it remains particularly im-study the consequences of ongoing abiotic and biotic portant to understand the allometric scaling ofeffects on natural ecosystems while including indi- other model parameters such as the functional re-rect effects between species. Future applications of sponses. Moreover, incorporating other types ofthis approach will need to further address how consumer–resource interactions such as parasite–mechanistic knowledge gained in simple food web host links into basic food web models remains amodules can be scaled up to predict patterns and challenge that needs to be addressed. Most likely,processes in complex food webs. Local interaction the framework of dynamic food web models de-spheres identified to influence the dynamics of key- scribed in this chapter will be an important back-stone modules may help in scaling up dynamics to bone for future integrations of these processes intonetworks. Unravelling the factors that determine theoretical community ecology.
  • 59. CHAPTER 4 Community assembly dynamics in space Tadashi Fukami4.1 Introduction Much remains unknown to fully answer these questions, and one major challenge is that speciesSpecies live in a complex web of interactions in interactions can bring about two contrasting typesthe ecological community. What effects do spe- of community dynamics (Fig. 4.1). In theory, strongcies exert on one another, and how strongly? If interactions can make communities either deter-species interactions are mostly strong, how do spe- ministic or historically contingent (Samuels andcies cope with one another and coexist in the same Drake 1997; Belyea and Lancaster 1999; Chasecommunity? In other words, what level of species 2003; Fukami et al. 2005). When deterministic, thediversity and what patterns of species composition effect of species interactions on community struc-should we expect to see if species interactions ture is determined by environmental conditions.strongly affect community structure? These are On the other hand, when historically contingent,some of the fundamental questions that community community structure diverges among localities asecologists seek to answer (Morin 1999). Species pool {1, 2, 3, 4, 5, 6, 7, 8, 9} {1, 2, 3, 4, 5, 6, 7, 8, 9} Immigration history A B C A B C Local community {4, 5, 6} {4, 5, 6} {4, 5, 6} {1, 2, 3} {4, 5, 6} {7, 8, 9} (a) Deterministic assembly (b) Historically contingent assemblyFigure 4.1 (a) Deterministic and (b) historically contingent community assembly. Numbers represent hypotheticalspecies, sets of numbers in brackets represent the species composition of local communities, arrows from the speciespool to local communities represent species immigration, and alphabets represent different immigration histories.Deterministic community assembly refers to situations in which different patches converge to the same speciescomposition regardless of immigration history as long as the communities initially share the same environmentalconditions. Historically contingent community assembly refers to situations in which different patches diverge to containdifferent sets of species if immigration history differs between them, even if the communities initially share the sameenvironmental conditions. Specific species compositions in the figure are arbitrary. Modified from Fukami (2008). Seealso Chase (2003). 45
  • 60. 46 DYNAMICSa result of stochastic variation in the history of spe- deepen our understanding of community structurecies arrivals, even under identical environmental without knowing immigration history. Building onconditions and an identical regional species pool. this framework, this chapter will consider the spa- It is difficult to determine which of these two tial scale of community assembly dynamics as ascenarios happens in natural communities. One potentially important yet relatively overlooked fac-main reason is simply that the immigration history tor that may critically determine the likelihood ofof most communities is unknown. This problem is deterministic versus historically contingent com-apparent in observational studies of community munity assembly. My aim here is not to provide aassembly. First popularized by Connor and Sim- comprehensive review of community assembly re-berloff (1979) in response to Diamond (1975), these search. I will instead use the results of several re-studies use statistical methods called null models to cent studies to highlight ideas that I believe arecompare observed community structures with worthy of further exploration.what would be expected if species interactions didnot exert significant effects on structure (Gotelli 4.2 Determinism and historical2001). The null models are a useful tool for detect- contingency in community assemblying effects of species interactions, but only whenused with caution. An incorrect assumption some- Before considering spatial issues relating to com-times made when using null models is that strong munity assembly, I would first like to clarify what isinteractions should always lead to community meant by determinism and historical contingency.structures that are significantly different from null In this chapter, I define community assembly as theexpectations. Strong interactions, when combined construction and maintenance of local communitieswith variable immigration history, can result in through sequential arrival of potential colonistshistorically contingent community development, from an external species pool (Drake 1991; Warrenwhich can produce apparently random community et al. 2003). As Warren et al. (2003) pointed out,structure. As pointed out by Wilbur and Alford ‘viewed in this way, community assembly empha-(1985) and Drake (1991), species interactions, even sizes changes in the community state rather thanwhen strong, do not necessarily create community embracing all evidence for pattern in communitystructures that are distinguishable from null expec- structure, the broader context in which the termtations that are based on deterministic effects of assembly is sometimes used’.species interactions. This limitation arises largely While community assembly can be historicallybecause most null-model studies use data taken at contingent or deterministic in the absence of speciesonly one point in time. Temporal changes in com- interactions, the focus of this chapter will be com-munity structure, let alone the history of species parison of the two scenarios in their presence. Com-immigration, are usually not considered, simply munity assembly starts with a disturbance, such asbecause such data are rarely available. a fire, flood or hurricane. Because space, nutrients Is it possible at all, then, to deepen our under- and other resources are often abundant in the re-standing of species interactions and community cently disturbed area, competition and other inter-structure without historical information on species specific interactions are unlikely to exert strongimmigration? Studies have recently begun to eval- effects on community structure at this stage. Also,uate possible conditions that make community as- of the potential colonists that can immigrate into thesembly deterministic or historically contingent. For disturbed area, only some will have reached theexample, it has been suggested that the rate of new patch thus far, and which species have arrivednutrient supply determines the extent of historical can be a matter of chance (e.g. Walker et al. 2006). Incontingency (Chase 2003; Steiner and Leibold 2004). this sense, communities are historically contingent,If this is true, then we should be able to calculate at but not as a joint consequence of immigration histo-least how predictable community structure will be, ry and species interactions. Once more time hasbased on nutrient supply rate. These studies indi- passed since disturbance, most potential colonizerscate a potentially promising way in which we can may have arrived, even though species interactions
  • 61. COMMUNITY ASSEMBLY DYNAMICS IN SPACE 47have not yet started to affect community structure references therein). A simple example of priority(e.g. Mouquet et al. 2003). In other words, most effects involves pre-emptive competition, in whichspecies expected to be found there are present. In species that arrive early make resources unavail-this sense, community structure is deterministic, able, by virtue of being there first, to other later-but because this determinism does not involve spe- arriving species that need those resources to sur-cies interactions, it is not what I would like to focus vive and grow (e.g. MacArthur 1972; Sale 1977; Til-on here, either. man 1988). However, priority effects need not Given more time after disturbance, species inter- involve only competition, and can happen via pre-actions will start to influence community compo- dation (e.g. Barkai and McQuaid 1988; Holt andsition more strongly as each species increases in Polis 1997), environmental modification (e.g. Peter-abundance in the patch. These species interactions son 1984; Knowlton 2004) and other types of speciescan make community structure become either de- interactions. Recently, experiments have shownterministic or historically contingent. It is these that not only community assembly over ecologicaltwo contrasting outcomes that are the focus of time, but evolutionary assembly through diversifi-this chapter. According to the deterministic view, cation can also be historically contingent (Fukamithe environmental conditions under which com- et al. 2007).munity assembly happens determine which of the Ever since Lewontin’s early writings (1969),species from the regional pool will remain in much emphasis has been placed on alternative sta-the community as a consequence of species inter- ble states in studying historically contingent assem-actions. In this case, immigration history does not bly. Historical contingency should be consideredinfluence the final species composition of the com- from a broader perspective, however. There aremunity. Such communities are said to follow de- two ways that communities can be historically con-terministic ‘assembly rules’ (Weiher and Keddy tingent even when there is only one final stable1995; Belyea and Lancaster 1999). This idea is root- state to which communities tend over time.ed in Clements’s (1916) climax concept of succes- First, communities can exist in what is called asion. More recently, deterministic assembly rules permanent endcycle. Morton and Law (1997) sug-have been indicated to drive community assembly gested that there are theoretically two types of finalnot just through immigration, but also through states that communities reach. One is called a per-evolutionary diversification (Losos et al. 1998; manent endpoint, and the other a permanent end-Gillespie 2004). cycle. Permanent endpoints consist of subsets of In contrast, if communities are historically con- species from the species pool that are resistant totingent, environmental conditions do not determine invasion by any species that are not members of thea single climax community. Instead, even if two endpoint. When ecologists refer to alternative sta-communities are originally under the same envi- ble states, they are in many cases referring to alter-ronmental conditions, they may contain different native permanent endpoints. In contrast, asets of species if they have different immigration permanent endcycle is ‘the union of the sets ofhistories. Lewontin (1969) is often cited as the first species that occur in a cyclic or more complex se-author to articulate this idea. Here there is more quence of communities’ (Morton and Law 1997).than one final stable state (called alternative stable Each set of species in a permanent endcycle can bestates, multiple stable points, multiple stable equili- invaded by at least one of the other species in the ¨bria, etc.; see Schroder et al. 2005) that communities endcycle, but cannot be invaded by any species notmay approach through assembly; once a communi- in the endcycle (Fig. 4.2). Communities in a perma-ty assumes a stable state, it cannot move to another nent endcycle are contingent on immigration histo-stable state unless heavily disturbed. This phenom- ry, because species composition at a given point inenon is caused by ‘priority effects’, in which early- time depends on the sequence of species invasionarriving species affect, either negatively or positive- as the communities go through the endcycle (Lock-ly, the performance of species that arrive late in wood et al. 1997; Fukami 2004b; Steiner and Leiboldterms of population growth (see Almany 2003 and 2004; Van Nes et al. 2007). This is true even with just
  • 62. 48 DYNAMICS 24 9 4.3 Community assembly and spatial scale {1, 3, 7, 10, 16, 22} Having clarified what is meant by determinism and 1 historical contingency, I would now like to develop 22 the main thesis of this chapter, namely that explicit {1, 3, 7, 10, 16, 24} {3, 7, 9, 10, 16, 22} consideration of spatial scale should help us to 24 better understand the conditions in which commu- 9 nity assembly is deterministic and those in which it {3, 7, 9, 10, 16, 24} is historically contingent. Drawing on recent theo- retical and empirical studies, I will focus on three 1 22 factors relating to spatial scale: (1) patch size, (2) Colonization patch isolation and (3) environmental heterogene- Extinction ity. In discussing these, it will become clear that it is the relative spatial scale of all of these factors si-Figure 4.2 An example of a permanent endcycle. multaneously considered that brings us the closestNumbers represent hypothetical species in a computer to a full understanding of community assemblysimulation. In this simulation, species 1–12 are autotrophs dynamics.and species 13–24 are heterotrophs. Sets of numbers inbrackets represent species composition of a localcommunity, thick arrows represent temporal changes inspecies composition, thin arrows represent species 4.3.1 Patch sizecolonization, and dotted arrows represent local species The local patch is the scale at which communityextinction. In the example shown here, the species poolconsists of 24 species, but only those that participate in assembly occurs (Fig. 4.1). Recent research hasthe endcycle are shown. Modified from Morton and Law suggested that the size of local patches can affect(1997) and Fukami (2008). the degree of historical contingency in community assembly. In their pioneering work, Petraitis and Latham (1999) proposed that historical contingencyone final permanent endcycle, in the absence of leading to alternative stable states occurs only whenalternative final states. patch size exceeds a threshold value. When a newly Second, communities can also exhibit alternative created patch is too small, the species dominant inlong-term transient dynamics. It can take a long and around the patch before disturbance quicklytime, relative to the generation times of the species colonize it from adjacent areas and continue toinvolved, for a community to reach a stable state. In dominate. In this sense, the fate of community as-such cases, communities will be historically contin- sembly in the patch is deterministic. In contrast,gent for a long time on their way to a stable state if when the patch is large, species that are not domi-they follow alternative successional trajectories (Fu- nant in adjacent areas may immigrate from a certainkami 2004a). For example, suppose that species A distance away and subsequently become abundantcompetitively excludes species B regardless of im- before adjacent dominant species take over themigration history. Even so, species B can remain patch. In this situation, the history of species immi-dominant for a long time if it arrives before species gration can influence community membership.A, and before competitive exclusion eventually oc- Thus, this is a historically contingent assembly.curs. This phenomenon is particularly likely when Petratis and Latham’s (1999) idea is mainlydispersal ability and competitive ability are similar derived from their work on rocky intertidal commu-among species (Sale 1977; Knowlton 2004; Fukami nities in the New England region of North America,et al. 2007; Van Geest et al. 2007). For the rest of this where each patch appears to be in either of twochapter, I will consider permanent endcycles and states, algal-dominated or mussel-dominated.long-term transients as well as alternative stable It was suggested that, for a disturbance such as icestates in discussing community assembly. scour to cause a patch to move from algal-dominated
  • 63. COMMUNITY ASSEMBLY DYNAMICS IN SPACE 49to mussel-dominated or vice versa, patch size need- entire islands acting as patches that undergoed to be sufficiently large to prevent nearby domi- community assembly after an island-wide volcanicnants from always driving community assembly. It eruption (e.g. Thornton 1996) and entire ponds act-should be noted here, though, that there is some ing as patches that undergo assembly after droughtdebate about whether these really represent two and subsequent refilling of water (e.g. Chase 2007).alternative stable states (Bertness et al. 2004). But even in patches distant from the species pool, Fukami (2004a) proposed a hypothesis that immigration rates may vary with patch size. Just asseems contradictory to the Petraitis and Latham darts are more likely to hit a larger dartboard, spe-(1999) hypothesis. Experiments showed that com- cies may be more likely to arrive at a larger patch.munity assembly was historically contingent to a Under this target size effect (Lomolino 1990), largergreater extent in smaller rather than larger patches. patches receive more individuals, consequently re-Microbial communities were assembled in the lab- ducing the between-patch difference in the popula-oratory by introducing 16 species of freshwater tion density of early-arriving species. The effect ofprotists and rotifers in four different orders in patch size on historical contingency suggested byeach of four different microcosm sizes. The results Fukami (2004a) may not be as strong then. It is alsoshowed that species diversity was affected more by possible, however, that slower immigration rates inimmigration history in smaller microcosms. This smaller patches make more time available for earlywas explained as follows. Given the same initial immigrants to alter the environment before otherpopulation size, early arriving species can achieve species arrive. This can strengthen priority effectshigh population density more quickly in smaller in smaller patches relative to larger ones, makingpatches. Consequently, resource availability and the difference in the extent of historical contingencyother conditions in patches are more greatly altered between small and large patches more pronounced.by early immigrants in smaller patches, which then The relative importance of these two opposinghas a greater effect on late-arriving species in smal- ways in which patch size affects historical contin-ler patches (see also Orrock and Fletcher 2005). gency requires further investigation. The apparent contradiction between Petraitis and Clearly, the effects of patch size on communityLatham (1999) and Fukami (2004a) stems partly assembly are complex. In particular, it has becomefrom different assumptions made about the source clear that considering the effect of patch size neces-of immigrants. Petraitis and Latham assume that sitates consideration of the areas surrounding theimmigration rates, particularly of species that are patches as well. The following sections will exploredominant near the patches, are higher for smaller surrounding areas a little further. I will first consid-patches. Immigration history itself is, then, more er the degree of patch isolation and then the spatialdeterministic there, resulting in more deterministic scale at which environmental heterogeneity is ob-assembly. On the other hand, Fukami assumes that served relative to the scale of patches.immigration rate and history do not vary withpatch size. In this situation, the inverse relationship 4.3.2 Patch isolationbetween patch size and the rate of increase in pop-ulation density causes larger patches to be more Several studies suggest that community assembly isdeterministic. more sensitive to immigration history when the Which assumption is more realistic? The answer patch is located farther from the species pool. Ro-depends partly on the environment around the binson and Edgemon (1988) assembled microbialpatch. Petraitis and Latham’s assumption would be communities by introducing phytoplankton speciesmore realistic if patches are surrounded by areas into aquatic microcosms in three different orders atthat provide immigrants. Besides rocky intertidal three different rates. Results showed that the effectpatches, forest gaps (e.g. Hubbell 2001) are possible of introduction order on species composition wasexamples. On the other hand, Fukami’s assumption greater when communities were assembled withmay be more realistic if the source of immigrants is lower immigration rates. Because immigration ratedistant from the patches. Examples may include is generally expected to be lower when the distance
  • 64. 50 DYNAMICSfrom the species pool is greater (MacArthur and permanent endcycles was greater when immigra-Wilson 1967), Robinson and Edgemon’s results sug- tion rate is low. As a result, immigration history hasgest that community assembly is historically con- a greater effect on species composition when immi-tingent to a greater extent when the patch is more gration rate is lower (see also Schreiber and Ritten-isolated. house 2004). In a related study, Lockwood et al. (1997) con- These studies all assume that the species poolducted computer simulation of community assem- that provides immigrants exists externally, suchbly using Lotka–Volterra equations modelling that patch community dynamics do not affect thecompetition and predation within patches. Using species pool (Fig. 4.3b). The model of communitytwo immigration rates, they found that immigra- assembly based on this assumption is typically re-tion history influenced species composition under ferred to as the mainland-island model. An alterna-both immigration rates, but that the type of effect tive model has been termed the metacommunitydiffered between the two rates. When immigration model, which describes a collection of multiplerate is low, different immigration histories lead local patches each undergoing community assem-communities to alternative stable states, whereas bly through occasional dispersal of species betweenwhen immigration rate is high, permanent end- the patches (Wilson 1992; Leibold et al. 2004; seecycles occur. The likely reason for this difference Chapter 5). In metacommunities, the species poolhas to do with whether the assembling commu- is internal instead of external, and local patchesnities approach an equilibrium between immigra- serve as the source of immigrants (Fig. 4.3a). Intion events. Low immigration rate allows for this, terms of patch isolation, when patches are moreeventually resulting in a stable state of species com- isolated from one another, the rate of internal dis-position. In contrast, high immigration rate pre- persal is lower (Fig. 4.3a), whereas when patchesvents the community reaching any possible are more isolated from the species pool, the rate ofequilibrium between immigration events. Thus, external dispersal is lower (Fig. 4.3b).high immigration rate maintains species composi- Computer simulations show that higher internaltion in a transient state of change, resulting in per- dispersal (or how isolated patches are to one an-manent endcycles. other) could make community assembly more de- Fukami (2004b) used a similar Lotka–Volterra terministic (Fukami 2005). This theoretical result ismodel to find that community assembly resulted consistent with findings from empirical studiesin permanent endcycles regardless of immigration (e.g. Chase 2003; Cadotte 2006). However, Fukamirate, but that the number of species involved in (2005) also showed that whether this effect of Metacommunity Local community Species pool Species pool Species pool (local patch) (a) Metacommunity model (b) Mainland-island model (c) Unified model Internal dispersal External dispersalFigure 4.3 (a) Metacommunity model, (b) mainland-island model and (c) unified model of community assembly. Arrowsrepresent dispersal between patches within the metacommunity (referred to as internal dispersal). Dashed arrowsrepresent dispersal from the external species pool (referred to as external dispersal). Modified from Fukami (2005).
  • 65. COMMUNITY ASSEMBLY DYNAMICS IN SPACE 51internal dispersal occurs depended on the rate of external species pool that is not influenced byexternal dispersal. Specifically, frequent internal the patches in which historical contingency isdispersal reduces the extent of historical contin- observed.gency if external dispersal is not frequent, but In terms of the spatial scale of environmentalinternal dispersal does not affect historical contin- heterogeneity, the results of Shurin et al. (2004) cangency if external dispersal is frequent. Therefore, be interpreted as follows. Historically contingentthe two dispersal types can reciprocally provide assembly occurs when environmental conditionsthe context in which each affects species diversity. are sufficiently heterogeneous across patchesThese results indicate that in order to understand (Fig. 4.4d) rather than within patches (Fig. 4.4c).historical contingency in community assembly, it Thus, it is the scale of environmental heterogeneityis important, though rarely done, to distinguish relative to the patches in question, rather than itsinternal and external dispersal and to know the absolute scale independent of patch size, thatrelative frequency of the two types of dispersal affects the degree of historical contingency in com-(Fig. 4.3c). munity assembly. The experiment conducted by Drake (1991) pro- vides additional insight into environmental hetero-4.3.3 Scale of environmental heterogeneity geneity and community assembly. Similar in designMany studies, including those discussed above, as- to Robinson and Edgemon (1988) and Fukamisume that local patches share identical environ- (2004a), Drake (1991) assembled aquatic microbialmental conditions. They also assume that the microcosms through sequential introductions ofregion within which local patches are embedded species in various orders using two different sizesis homogeneous across space. Clearly these as- of microcosms. The results showed that, in smallsumptions are not met in many ecological land- patches, the same species dominated the assembledscapes. An interesting question then is how the community regardless of introduction order,scale at which environmental heterogeneity is ob- whereas, in large patches, species introduced earlyserved may influence the degree of historical con- dominated over those introduced late. These resultstingency in community assembly. A study by Shurin et al. (2004) is relevant here.They used a mathematical model to study condi- Local communitytions for coexistence of two competing species at a Metacommunity (local patch)regional scale. The region modelled consists ofmultiple patches that vary in resource supplyratio. In the absence of variation among patchesin resource supply ratio, one of the two speciescompetitively excludes the other. When patches (a) No heterogeneity (b) No heterogeneityvary in the ratio, historical contingency occurs in between or within local patches between or within local patchesterms of which species occupies a given patch.Specifically, in patches where resource supplyratio is intermediate, the species that arrives firstprevents the other from colonizing that patch. Inother patches where the ratio takes more extremevalues, one or the other species dominates. These (c) Heterogeneity (d) Heterogeneitypatches serve as a species pool that provides im- within local patches between local patchesmigrants to patches of intermediate environmental Figure 4.4 (a–d) Spatial scale at which environmentalconditions for historically contingent community heterogeneity is observed. Shading indicates variation inassembly to be realized there (though it involves environmental conditions (e.g. rate of nutrient supply).only two species in the model). Historically con- Heterogeneity is drawn arbitrarily as a gradient. Modifiedtingent assembly occurs only when there is an from Fukami (2008).
  • 66. 52 DYNAMICSappear to contradict those of Fukami (2004a) dis- 4.4 Community assembly and speciescussed above, while being congruent with those of traitsPetraitis and Latham (1999). But neither seems to bethe case. Drake (1991) invoked differences in envi- Ultimately, consequences of patch size, patch isola-ronmental heterogeneity between small and large tion and environmental heterogeneity for commu-patches to explain his results, whereas neither Fu- nity assembly depend on the spatial scale of specieskami (2004a) nor Petraitis and Latham (1999) ex- movement (Cadotte and Fukami 2005). For this rea-plicitly considered environmental heterogeneity. son, it is important to know the dispersal ability ofDrake postulated that environmental heterogeneity the species involved in community assembly inincreased with patch size (light availability was question, in order to address the determinism ver-more variable in larger microcosms owing to sus historical contingency question. Furthermore,increased depth; depth was standardized across the degree of variation in dispersal ability amongpatch size in Fukami (2004a)), and that variation species can also influence historical contingency inamong species in their competitive ability was community assembly. This is because the moresmall when environmental heterogeneity was similar species are in dispersal ability, the moregreat. Communities are thought to be more sensi- stochastic immigration history is expected to be,tive to historical contingency when species are com- which can then lead to less deterministic assembly.petitively more similar (e.g. MacArthur 1972; Smaller variation in competitive ability should alsoHubbell 2001). If this applies to Drake’s micro- result in less deterministic assembly, as prioritycosms, then it explains smaller historical contingen- effects act stronger between competitively morecy in smaller patches. Drake’s (1991) explanation similar species.would need to be tested to be rigorously validated, Furthermore, dispersal ability and competitivebut the suggested potential relationship between ability are thought to sometimes show a trade-off,patch size, environmental heterogeneity, competi- such that species that are good dispersers are poortive relationship and historical contingency re- competitors, and vice versa (e.g. Petraitis et al. 1989;mains novel to this day. Cadotte 2007). In terms of succession, this means that early-successional species are competitively inferior to late-successional species (Petraitis et al. 1989). This trade-off, too, may influence historical4.3.4 Synthesis effects in community assembly. For example, as-In summary, I have considered patch size, patch sembly may be more deterministic when the spe-isolation and the spatial scale of environmental het- cies show a clearer trade-off between these twoerogeneity as three spatial factors influencing the traits. This is because, under a clear trade-off, com-degree of determinism and historical contingency munity assembly is expected to progress predict-in community assembly. These factors do not affect ably to eventually end with a predictable set of late-community assembly independently of one anoth- successional competitive species dominating theer. Instead, their scale and consequently their role community.in community assembly are determined relative to Of course, dispersal ability and competitive abil-those of the others. Through consideration of these ity are just a few of many traits that characterizethree factors, several conditions for historical con- species. There has recently been a renewed interesttingency have emerged. Specifically, community in explaining community dynamics from speciesassembly is hypothesized to be historically contin- traits (e.g. Fukami et al. 2005; McGill et al. 2006;gent to a greater extent when (1) immigration rate is Ackerly and Cornwell 2007). Other traits that canlower, (2) immigration history is more variable and influence community assembly include disturbance(3) the species pool that provides immigrants to tolerance, intrinsic rate of growth and predatorlocal patches undergoing assembly exists more in- avoidance. Including these traits in a frameworkdependently of the community dynamics within for community assembly should enhance our pre-the patches. dictive power. For example, even when there is a
  • 67. COMMUNITY ASSEMBLY DYNAMICS IN SPACE 53clear trade-off between dispersal ability and com- ical contingency and determinism in communitypetitive ability, community assembly can be histor- assembly. Additionally, I have briefly pointed outically contingent in the presence of predators. that the spatial scale of community assembly isRecent experimental work suggests that the timing defined relative to dispersal ability of speciesof predator arrival at the local patch can influence involved. But dispersal ability is often not indepen-the final structure of prey communities under a dent of other traits such as competitive ability, dis-competition–colonization trade-off (Olito and Fu- turbance tolerance and predator avoidance. Explicitkami 2009). consideration of these traits should lead to a better Dispersal ability and other traits may ultimately understanding of the conditions for contingent ver-be determined by the spatial scale of patches that sus deterministic assembly.the species have experienced over evolutionary As discussed in the introduction, much of com-time (Denslow 1980). Patch sizes that species have munity assembly research has traditionally reliedexperienced in the past and those of the present are on null-model approaches using observationalnot necessarily the same. This is particularly true in data. This is because experimental assembly of nat-the presence of anthropogenic disturbance, habitat ural communities is difficult in most situationsfragmentation and exotic species introduction. owing to the large spatial and temporal scalesAnthropogenic disturbance can be evolutionarily involved in this type of work. However, direct ex-novel; habitat fragmentation can create new kinds perimental manipulation of immigration history isof patch size and isolation; and exotic species can necessary in order to rigorously evaluate historicaldiffer from native species in the spatial scale of ¨ effects in community assembly (Schroder et al.patches that they have adapted to, consequently 2005). For this reason, I expect that experimentsdiffering in the way native and exotic species per- will become increasingly important in communityceive spatial scale. How do these anthropogenic assembly research. Experiments have so far beenchanges in the scale of community assembly affect limited mainly to those with microorganisms in thehistorical contingency in assembly? We currently laboratory owing to their logistical advantages, butknow little to answer this question. A better under- we will also need to do more field experiments tostanding of the role of scale in community assembly ensure that the concepts we develop are firmlymay contribute to advancing not only community placed in natural context. Though difficult, fieldecology as a basic science, but also solving applied experiments are feasible by, for example, incorpor-issues regarding the community-level impacts of ating experimental research into ecological restora-species invasions. tion projects (e.g. Fukami et al. 2005; Weiher 2007). In addition, research on community assembly has mainly considered systems in which environ-4.5 Conclusions and prospects mental conditions do not vary considerably exceptI have discussed how the spatial scale at which when pulse disturbance events initiate a new roundcommunity assembly occurs may influence the de- of community assembly. However, environmentalgree to which community assembly dynamics are conditions can of course fluctuate greatly in manydeterministic versus historically contingent. As systems. How do temporal fluctuations affect thespatial factors, I have focused on patch size, patch role of spatial scale in determining the degree ofisolation and the scale at which environmental con- historical contingency in community structure?ditions vary. In combination, these factors are pro- Does the temporal scale of environmental fluctua-posed to jointly affect three elements of community tions relative to that of community assembly affectassembly dynamics: the rate of immigration to local the extent of historical contingency? These ques-communities, the degree to which the species pool tions remain unanswered. My focus here has beenis external to local community dynamics and the spatial scale, but temporal scale should also be ex-extent of variation in immigration history between plored further in future research in relation to com-local communities. I have argued that these three munity assembly, over both ecological andelements will in turn determine the extent of histor- evolutionary time.
  • 68. 54 DYNAMICS This chapter has largely consisted of exploration ture. Historical contingency may well be a mainof ideas rather than evaluation of data. We do not reason behind the absence of such patterns. None-yet have sufficient data to draw general conclu- theless, like other authors (e.g. Long and Karelsions as to how often or to what extent natural 2002; Chase 2003), I believe a good understandingcommunities are governed by historical contin- of the conditions for determinism versus historicalgency. If it turns out in the future that many com- contingency will contribute to building a predic-munities are indeed highly sensitive to historical tive theory of community ecology. Here I haveeffects, then one may question whether communi- sought to provide a first step in this endeavour,ty ecology can be called a science in the first place. with a special focus on the spatial scale of commu-The answer could be no if science was defined as nity assembly dynamics. Many of the ideas pre-discovering general patterns in nature and ex- sented here are only exploratory, and some mayplaining these patterns within a predictive frame- prove wrong. Even so, it is my hope that they servework. In fact, we do know that clear general to stimulate further research on the dynamics ofpatterns are rarely observed in community struc- community assembly.
  • 69. PART IIISPACE AND TIME
  • 70. This page intentionally left blank
  • 71. CHAPTER 5 Increasing spatio-temporal scales: metacommunity ecology Jonathan M. Chase and Janne Bengtsson5.1 Introduction and continents. Furthermore, temporal scales under consideration within the metacommunityCommunity ecology is a dynamic and rapidly context range from seasonal or yearly fluctuationschanging field. While community ecology in the to eons of global climate change and phylogenetic1980s and 1990s focused primarily on how local inertia. Individual species in a metacommunityenvironmental conditions, and species interac- respond to scale differentially. Finally, among spe-tions within those localities, influenced patterns cies in a metacommunity, there are often largeof coexistence and relative species abundances at differences in movements and dispersal, andthe local scale, more recent years have seen anincreasing recognition (or rediscovery) of the im- rates of reproduction and mortality, which dependportant role of space and time (Ricklefs 1987, 2004; on several factors, including behaviour, body sizeRicklefs and Schluter 1993; Hubbell 2001; Leibold and trophic level. Thus, rather than trying to de-et al. 2004; Holyoak et al. 2005). We refer to the fine the range of a metacommunity, we insteadimplicit recognition (and study) of the important focus on the concept of metacommunities as incor-role of spatio-temporal dynamics as metacommu- porating aspects of community ecology at largernity ecology. spatio-temporal scales. These can range from as By analogy with a metapopulation, which refers small a scale as considering the variation of plantto a series of populations interconnected by dis- species that occur on different aspects of a hillsidepersal among patches, a metacommunity refers to slope to as large a scale as considering the varia-a collection of communities of potentially interact- tion in plant species composition across biogeo-ing species that are interconnected by dispersal graphic provinces. When setting out to test a(Leibold et al. 2004; Holyoak et al. 2005). Accord- given metacommunity model, it is important toingly, the simplest model of a metacommunity is recognize the constraints on those models, and toone of several non-interacting species each exist- address their predictions at appropriate spatio-ing in co-occurring metapopulations. In practice, temporal scales. It would, for example, be ratherhowever, the symmetry between metapopulations silly to test the predictive ability of a ‘mass effects’and metacommunities is loose at best. Although metacommunity model, which inherently assumespopulation biologists often make great efforts to spatially heterogeneous environments (e.g. Mou-define the scope of a metapopulation, the scales quet and Loreau 2003), at a scale so small that theunder consideration can be highly variable within landscape is essentially homogeneous.the context of metacommunities. Spatial scales can Throughout this chapter, we will intermingle dis-range from small-scale environmental gradients or cussions of more historical perspectives of meta-patches in which species colonize or go extinct, to community ecology into the contemporarylarge-scale biogeographic studies across provinces perspective. We do this because we feel that in 57
  • 72. 58 SPACE AND TIMEorder to have a deep understanding of the contem- vant at larger scales, and that a perspective basedporary views of metacommunity ecology, it is nec- solely on stochastic processes of colonization andessary to know how the ideas have evolved. In extinction (and speciation) could approximate nat-addition, when possible, we discuss theory and ural patterns (see Chave 2004; Alonso et al. 2006;empirical work hand-in-hand, which we feel is a Holyoak and Loreau 2006). Third, in the context ofmore appropriate way to discuss this information, applied community ecology, spatio-temporal per-rather than keeping them separate. We argue that spectives are often necessary to understand howmetacommunity ecology will most rapidly advance communities are degraded, and how they can bewith an intimate, rather than superficial, connection restored (see Chapter 9). Fourth, statistical analyses,between theory and data. including spatial, multivariate and computational As a first order of business, it is necessary to analyses, have become more sophisticated andrecognize that, although metacommunity ecology powerful (e.g. Clark et al. 2007), and, when appliedas a field has exploded in the past decade or so in combination with theoretical and experimental(Tilman 1997; Leibold et al. 2004; Holyoak et al. information, can provide a deeper understanding2005), its historical roots are much older. For exam- of both the patterns and underlying processes ofple, early experimentalists testing Lotka–Volterra community structure than could have been gainedcompetition and predator–prey models quickly re- previously.cognized that it was exceedingly difficult for com-petitive or predator–prey species pairs to coexist 5.2 The varied theoretical perspectives onwithout some sort of spatio-temporal variation metacommunitiesthat was missing from the simple models (Gause1934; Park 1948, 1954; Huffaker 1958). Insect Recent syntheses have suggested that metacommu-ecologists have long recognized the importance of nity theory can be roughly categorized into fourlarge-scale processes and dispersal in population general conceptual frameworks: Neutral, patchdynamics, and aspects of metapopulation ecology dynamics, species sorting and mass effects (Leiboldwere inherent in the writings of Andrewartha and et al. 2004; Holyoak et al. 2005). The four perspec-Birch (1954). Similarly, during the ‘renaissance’ of tives are not exclusive. For example, mass effectscommunity ecology – the 1960s and 1970s – spatio- and patch dynamics can be viewed as occurringtemporal perspectives were quite common (e.g. along a gradient of organism movement intensityMacArthur and Levins 1964; MacArthur and Wilson and habitat heterogeneity. The relative importance1967; Levins and Culver 1971; MacArthur 1972; of species sorting, patch dynamics and mass effectsHorn and MacArthur 1972; Levin 1974; Slatkin 1974). in metacommunities is a primary issue when exam- Probably as a consequence of the increase in sta- ining the influence of local and regional processestistical and experimental rigor that arose from a on patterns of community composition (seeseries of critiques (e.g. Strong et al. 1984), communi- Chapter 9). Furthermore, the stochastic processesty ecology studies in the 1980s and early 1990s were inherent in the neutral theory are one componentprimarily aimed at local-scale processes, such as the of more complex spatio-temporal niche modelsmechanisms that influence species coexistence and (Chesson 2000; Adler et al. 2007).relative abundances. The recent recognition of the There has been considerable attention directedimportance of spatial processes probably has sever- towards testing empirical data from a variety ofal simultaneous origins. First, while the focus on systems against the predictions of metacommunityexperiments and local-scale mechanisms offers models based on neutral versus niche differencesmuch to community ecology, it is not able to ac- (e.g. Chave 2004; McGill et al. 2006). Unfortunately,count for many of the patterns that are observed in patterns from one type of data are generally notnature, at both local and larger scales (e.g. Ricklefs able to unambiguously separate the different2004). Second, Hubbell’s (2001) ‘neutral’ theory model predictions (Chase 2005; Chase et al. 2005)served to catalyse the field by espousing the contro- (Table 5.1). Although data on the relative abun-versial perspective that local interactions are irrele- dances of species in a community are often used
  • 73. INCREASING SPATIO-TEMPORAL SCALES 59Table 5.1 Summary of predictions from the four metacommunity modelling frameworks (modified fromChase et al. 2005) Model predictionEffect Neutral Patch dynamics Species sorting Mass effectsOverall local Extinction and Extinction and Depends on species Depends on species diversity colonization colonization balance interactions interactions and balance colonization/ extinctionOverall Extinction and Depends on competition– Same as above and Same as above and regional speciation colonization degree of habitat degree of habitat diversity balance trade-off heterogeneity heterogeneityRelative Zero-sum Variable depending on Variable depending Variable depending on species multinomial level of migration and on environmental level of migration abundance (skewed towards degree of interaction conditions rare species)Dispersal Increase First increase, then No effect First increase, then effects on decrease decrease (hump- local (hump-shaped) shaped) diversityDispersal Decrease Decrease No effect Decrease effects on regional diversityDispersal Decrease Global: no effect No effect Global: decrease effects on Local: decrease Local: decrease b-diversityLocal Return immediately Unpredictable Return immediately Return following disturbance successionRegional Random walk Return following Return immediately Return following disturbance succession successionTemporal Variable Variable Static unless Static unless variation: environment environment local changes changesTemporal local Variable Static unless environment Same as above Same as above variation: changes regionalto compare the different model predictions (e.g. The four models make different assumptions re-McGill et al. 2006), several models can predict the garding how species respond to environmental andsame pattern of relative abundances as the neutral spatial gradients, and thus make different predic-model (reviewed in Chase et al. 2005). In addition, tions regarding patterns of community structurestatistically distinguishing between different mod- along those gradients. We do not go into detailels using empirical species-abundance distributions regarding the theoretical reasoning behind each ofis not always straightforward (McGill et al. 2007). the predicted patterns here, but instead simply re-This indicates the limitations of using one pattern, view them in Table 5.1 (for more detail, see Chaseespecially relative abundance patterns, to discern et al. 2005). We start with the simplest model, andamong the validity of the assumptions underlying then move to models using increasingly complexthe different perspectives. assumptions.
  • 74. 60 SPACE AND TIME5.2.1 Neutral case at the patch level. In Levins’ (1969) original metapopulation model, patches are assumed to beAlthough Hubbell’s (2001) construct is the most similar, and species can colonize and go extinctwidely recognized neutral model, there are actually from a patch at defined rates. However, patch dy-many related ‘neutral’ theoretical constructs that namics models of metacommunities go beyond themake no specific assumptions about species traits neutral model by assuming that species may dis-or their responses to the environment (reviewed in play trade-offs in their relative abilities to colonizeChave 2004). The neutral model assumes that spe- and compete, in turn allowing coexistence undercies are neutral with respect to their interspecific some circumstances. There are many ways to depictinteractions as well as the underlying environment. these sorts of colonization–competition and otherThis means that the numbers of individuals and trade-offs, each of which gives slightly differentspecies that occur in any given locality result from sets of predictions (Levins and Culver 1971; Hast-purely stochastic processes (e.g. colonization and ings 1980; Tilman 1994; Hanski and Gyllenbergextinction). MacArthur and Wilson’s (1967) Equi- 1997; Yu and Wilson 2001; Calcagno et al. 2006).librium Theory of Island Biogeography is probablythe most widely recognized neutral model, andsimilar concepts form the basis of Hubbell’s (2001) 5.2.3 Species sortingneutral theory. In the MacArthur and Wilson theo-ry, the number of species in a habitat is solely This framework leaves behind the stochastic pro-determined by the balance between the coloniza- cesses of colonization and extinction inherent in thetion rate of species from a species pool (usually two frameworks described above, and instead fo-mainland) and the local extinction rates of species. cuses on deterministic processes that result fromHubbell’s (2001) model expands on the MacArthur differential responses of species to heterogeneousand Wilson theory in two important ways. First, the environments; i.e. niche differences. Here, a speciesstochastic processes of colonization and extinction will occur in a locality if the abiotic and bioticat the patch level from the MacArthur and Wilson environments are favourable. Trade-offs in toler-theory are transferred to the individual level so that ance to different environmental conditions are typ-they denote birth–death processes, allowing it to ically invoked here to suggest that species that aremake predictions of species’ relative abundances. favoured under some environmental conditionsSecond, rather than defining a species pool (e.g. will be disfavoured under others (Tilman 1982;from a mainland), Hubbell allows the species pool Chase and Leibold 2003).to result from speciation, which is also derivedfrom stochastic processes. While specific details ofthe neutral model and its assumptions have been 5.2.4 Mass effectsheavily debated (e.g. Ricklefs 2003; Etienne et al. This framework combines aspects of patch2007), neither Hubbell (2001) nor others who have dynamics and species sorting by assuming thatdeveloped neutral theories (e.g. Chave 2004) truly species have differential responses to environ-believe that the assumptions of the neutral model mental conditions and differential dispersal,can fully capture the true nature of these commu- colonization and extinction rates (e.g. Mouquetnities. Instead, the neutral models emphasize how and Loreau 2003; Amarasekare et al. 2004). Dis-far one can go by making the simplest assumption persal is assumed to be high enough to influencethat communities are structured by stochastic pro- local dynamics. The coexistence and relative abun-cesses only. dances of the species will depend on rates of dis- persal and extinction, as well as on the source–sink relationship between the different habitats for5.2.2 Patch dynamics each species. The existence of source patches isThis perspective is also borne out of stochastic pro- crucial for a species to persist in a dispersal-drivencesses of colonization and extinction, but in this metacommunity.
  • 75. INCREASING SPATIO-TEMPORAL SCALES 615.3 Metacommunity theory: resolving are operating simultaneously (Tilman 2004; ChaseMacArthur’s paradox 2005, 2007; Gravel et al. 2006; Leibold and McPeek 2006; Adler et al. 2007; Clark et al. 2007). The chal-Each of the four metacommunity perspectives pro- lenge for metacommunity ecologists is to disentan-vides important insights, and they should not be gle the relative importance of the differentviewed as strict alternatives but rather as different processes, and, most importantly, to identify fea-ways to view problems of coexistence and diversi- tures of a given habitat type that would make onety. ‘MacArthur’s paradox’ (Schoener 1989) refers to or the other set of processes exert a stronger influ-the fact that Robert MacArthur played a significant ence on community patterns.role in the simultaneous development of metacom-munity models based on niche differences (Mac-Arthur and Levins 1964) and those based on moreneutral perspectives (MacArthur and Wilson 1967). 5.4 As easy as a, b, g: the importance ofHere, we emphasize that, rather than viewing these scaleperspectives as dichotomous alternatives, they il- Why is it necessary to explicitly consider the role oflustrate different ends of a continuum, and that space in a metacommunity? If spatial processes areeach approach is valid for problems at different not important, then the factors that influence thescales and for different questions. For example, number and relative abundance of coexisting specieswhen the focus is at a relatively coarse scale, it at any given locality should simply ‘add up’ to de-may be an acceptable first approximation to assume scribe those features at the regional scale. If, on thethat all species are equal. This could explain the other hand, there is something inherent in the wayreasonable success that simple ‘neutral’ models that space differentially influences patterns of locallike the MacArthur and Wilson theory and their and regional diversity, then scale (and thus space)derivatives have had for addressing questions at matters (e.g. Whittaker et al. 2001; Chase 2003).coarser scales (Simberloff 1974; Whittaker and Fer- This problem is illustrated by explicitly consider- ´nandez-Palacios 2007), despite the abstraction of ing how scale influences patterns of diversity. Atmuch ecological detail that obviously is important local (small) spatial scales, species diversity (¼ spe-for local coexistence. When the focus is local, or cies richness) is the number of species counted inwhen the questions being addressed are fine- the area defined; known as a-diversity. Becausescaled, these more abstract models often need to there is generally variation in the types of speciesbe expanded to focus specifically on niche differ- (species composition) observed from one locality toences among species and how those differences the next, the turnover of species compositioninfluence patterns of community structure (e.g. among sites is known as b-diversity. Finally, re-Chase and Leibold 2003). gional diversity, known as g-diversity, can be Any theoretical model of metacommunities – be derived by either an additive (g ¼ a þ b (e.g.it based on neutral, patch dynamic, species sorting Lande 1996)) or multiplicative (g ¼ ab (Whittakeror mass effect/dispersal processes – can only be a 1972)) partitioning. This emphasizes how under-cartoon of the processes that are actually occurring standing patterns of b-diversity is critical to under-in that metacommunity. We argue that, although it standing the scalar relationship between local andis instructive to examine patterns observed in natu- regional diversity, which is the domain of meta-ral systems or from experiments in the context of community ecology. b-Diversity can emerge fromthese model predictions (e.g. Table 5.1), this should both deterministic and stochastic processes. First, ifnot be done to test the validity of one model over localities differ environmentally, high b-diversitythe other. Instead, we echo several recent syntheses can emerge from deterministic processes that fa-that have explicitly recognized that, in any meta- vour different species in different environments,community, a variety of processes, including those i.e. species-sorting processes (e.g. Whittaker 1972;related to neutral models (e.g. stochasticity) and Chase and Leibold 2003). Second, even when local-those related to niche models (e.g. determinism), ities have identical environmental conditions, high
  • 76. 62 SPACE AND TIMEb-diversity can emerge from stochastic processes more frequently at regional scales. Chase and(e.g. local extinctions because of demographic sto- Leibold (2002) showed this scale dependence inchasticity; Hubbell 2001), or dispersal limitation surveys of invertebrates and amphibians fromand priority effects, often connected to a patch dy- small fishless ponds in Southwestern Michigan,namics perspective (Chave and Leigh 2002; Condit USA. When considered on a ‘per-pond’ basiset al. 2002; Chase 2003). (local scale), they found a hump-shaped relation- Scale-dependent diversity relationships have ship between productivity and diversity, whereasbeen explicitly observed along gradients of distur- when the same data were considered on a ‘per-bance (Chase 2003, 2007; Ostman et al. 2006) and watershed’ basis (regional scale), a monotonicallygradients of productivity (Mittelbach et al. 2001; increasing relationship between productivity andChase and Leibold 2002; Chase 2003). Scale depen- diversity was observed (Fig. 5.1a).dence will occur whenever responses of a- and g- Such scale dependence is not universal (e.g. Chasediversity do not scale linearly, i.e. b-diversity varies and Ryberg 2004; Harrison et al. 2006a). For example,along the gradient. For example, in a meta-analysis when comparing the ponds discussed above withof the well-studied relationship between productiv- regions where ponds were physically closer andity and species diversity, Mittelbach et al. (2001) dispersal was more likely among ponds, Chase andfound that hump-shaped productivity–diversity Ryberg (2004) found that b-diversity was lower inrelationships tend to emerge more frequently at regions where dispersal was more likely amonglocal scales, whereas monotonically increasing pro- ponds, and no such scale dependence emergedductivity–diversity relationships tended to emerge (Fig. 5.1b). This result could have been due to higher (a) Less Connected Regions 40 40 Local Regional Species Richness 30 30 20 20 10 10 Species Richness 0 0 0 20 40 60 80 100 120 0 20 40 60 80 100 120 (b) More Connected Regions 40 40 Local Regional Species Richness 30 30 20 20 10 10 0 0 0 20 40 60 80 100 120 0 20 40 60 80 100 120 Productivity Productivity (mg cm-2 day-1) (mg cm-2 day-1)Figure 5.1 Scale-dependent productivity–diversity relationships in isolated (a) but not connected (b) watersheds.Redrawn from Chase and Ryberg (2004).
  • 77. INCREASING SPATIO-TEMPORAL SCALES 63dispersal rates among the more closely aligned me- S ¼ CAztacommunities, resulting in a reduced degree of sto- where S is species richness and A is area. When log-chasticity and higher similarity in community transformed, we get the familiar linearized SAR,composition. In Effects of dispersal rates on localcommunity structure (below), we discuss how vari- logS ¼ logC þ z logAation in the proximity of habitats (which is related to where and C and z are curve-fitting parametersthe rates of dispersal) within a metacommunity depicting the intercept (C) and rate of increase (z)might alter patterns of a-, b- and g-diversity. of species with area. Of primary interest here is z, the slope of the log-5.5 Species–area relationships and transformed SAR, indicating the steepness of themetacommunity structure increase in the number of species with increasingAlthough it is conceptually quite useful, partition- area. Although somewhat more complex, varia-ing diversity into a, b and g components is not tion in the parameter z is related to variation inalways straightforward. Theoretically, one can b-diversity (e.g. Connor and McCoy 1979; Rosenz-imagine localities as patches surrounded by inhos- weig 1995; Drakare et al. 2006). For example, SARspitable matrix (e.g. Mouquet and Loreau 2003; are often calculated using a ‘nested’ design (Schei-Amarasekare et al. 2004). In some metacommunities ner 2003), by adding together the area of similarwith patchy structure this assumption is largely sized smaller patches or sampling units into afulfilled, and several such systems have been used larger area. In this case, if the smaller patches areas empirical models of metacommunities. Exam- highly divergent in their community compositionples include water-filled inquilines such as pitch- (higher b-diversity), the increase in species rich-er-plant leaves (e.g. Kneitel and Miller 2003; Gotelli ness with increasing area – the z-value – will beand Ellison 2006), ponds and rockpools (e.g. higher than if the smaller patches are more similarBengtsson 1989, 1991; Chase 2003), islands (e.g. to one another (lower b- diversity). Thus, variationSimberloff and Wilson 1969; Diamond 1975; Scho- in z among sites can provide important informa-ener et al. 2002), host plants for insects (e.g. Zabel tion regarding the partitioning of diversity acrossand Tscharntke 1998; Kreuss and Tscharnkte 2000), scales.moss patches on rocks (Gilbert et al. 1998) and rocky Several factors can create variation in the slopeoutcrops (e.g. Harrison 1997, 1999; Ostman et al. (z) of log-transformed SARs. Even when sampling2007). However, in many terrestrial (forests, grass- methodology is controlled, z varies among com-lands) and aquatic (oceans, large lakes) landscapes, munities (reviewed in Drakare et al. 2006). Some ofthe distinction between where one ‘locality’ ends this variation can be attributed to factors such asand another begins is ambiguous at best (Loreau organism size or trophic position (Holt et al. 1999;2000). In these continuous landscapes, diversity Drakare et al. 2006). However, variation in z canmight be scaled or partitioned more objectively also emerge from differences in metacommunityusing species–area relationships (SARs). structure. For example, z is generally higher on The positive SAR is one of the oldest and best- islands than on continents (Rosenzweig 1995;supported patterns in ecology (Lomolino 2000). ´ Whittaker and Fernandez-Palacios 2007). ThisSARs can be calculated in two ways: (1) using in- might result because islands have lower rates ofcrementally larger areas in which smaller areas are dispersal, and thus higher inter-island differentia-nested or (2) using data on richness and area from tion than similar sized areas of continents thatdistinct separate patches or islands (e.g. Rosenz- have higher rates of dispersal (see Effects of dis-weig 1995). The identity of mechanisms driving persal rates on local communities, below). In addi-variation in the shape of SARs remains at the fore- tion, in a meta-analysis of SARs, Drakare et al.front of research on metacommunity ecology (e.g. (2006) found generally higher z (and thus higherDrakare et al. 2006). The function most frequently b-diversity) at lower latitudes (Fig. 5.2). Thisused to describe the SAR is a power law: indicates the possibility that metacommunity
  • 78. 64 SPACE AND TIME species diversity among animals and plants. 1.0 Among animals, there was a positive, but potential- species–area relationship ly asymptotic or hump-shaped relationship be- Slope (Z ) of the tween the enhancement of local diversity relative 0.6 to the control with increasing dispersal rates, whereas there was generally a positive effect of dispersal on local diversity (most treatment effects 0.2 relative to the control are greater than zero), but this did not seem to vary with the rates of enhanced –0.2 dispersal. 0 20 40 60 80 Higher dispersal through habitat connections Latitude (Њ) does not always increase local diversity. For exam- ple, Hoyle and Gilbert (2004) used an experimentalFigure 5.2 Decline in the slope of the species area curve landscape of moss and its microarthropod inhabi-(z) with decreasing latitude. Redrawn from Drakareet al. (2006). tants similar to that used by Gilbert et al. (1998), but found no positive effects of habitat corridors on species diversity during a year with favourableassembly processes are different between temper- weather conditions. A possible explanation forate and tropical areas. Similarly, Ryberg and Chase this difference is that, during drought years, res-(2007) showed both theoretically and empirically cue effects via corridors may be more important inhow the presence of top predators can reduce z. preventing local extinctions. Similarly, Kneitel andOther community stressors with effects similar to Miller (2003) found strong effects of dispersal onpredators (e.g. increasing local extinction rates) the diversity of a protist community in the absenceshould have similar effects. of predatory mosquitoes, but much weaker dis- persal effects in the presence of mosquitoes. While dispersal’s effects on diversity have5.6 Effects of dispersal rates on local been well-studied at local scales (e.g. reviewedcommunities in Cadotte 2006), its effects on b-diversity areAs noted above, variation in dispersal rates is a less studied. However, both theoretical and em-primary factor that can influence metacommunity pirical results show that rates of dispersal oftenstructure and the partitioning of diversity. It is well result in lower b-diversity (and lower z), homo-known that islands that are closer to the mainland, genizing communities, and potentially ameliorat-and thus have higher rates of immigration, have ing or reversing the local-scale positive effects ofhigher diversity than comparable islands farther dispersal, when diversity is measured at the re-from the mainland, which receive lower rates of gional scale (Harrison 1997, 1999; Amarasekaredispersal (e.g. MacArthur and Wilson 1967). Fur- 2000; Forbes and Chase 2002; Chase 2003; Mou-ther, in a conservation context, experimental data quet and Loreau 2003; Chase and Ryberg 2004;are emerging that show higher levels of diversity in Fukami 2004; Cadotte and Fukami 2005). Finally,habitat patches that are connected by habitat corri- the influence of dispersal on b-diversity anddors than in habitat patches without such corridors scale-dependent patterns of diversity will also(Gilbert et al. 1998; Damschen et al. 2006). depend on the mechanisms that allow species There will often be a limit to the positive effect of to coexist both locally and regionally. Amongdispersal on local diversity, such that diversity can experimental communities treehole-dwellingshow an asymptotic (MacArthur and Wilson 1967) protists, Ostman et al. (2006) found that dispersalor hump-shaped (Mouquet and Loreau 2003) rela- homogenized local communities under benigntionship with increasing dispersal rates. Using conditions (reducing b-diversity) and thus re-meta-analysis, Cadotte (2006) compared standar- duced overall g-diversity. However, whendized dispersal rates with their influence on local drought disturbance was imposed on those
  • 79. INCREASING SPATIO-TEMPORAL SCALES 65communities, dispersal allowed species to recol- chasticity) and those related to niche models (e.g.onize habitats where they were driven extinct, determinism), are operating simultaneously (Chaseincreasing overall b- and g-diversity. 2005, 2007; Gravel et al. 2006; Leibold and McPeek 2006; Adler et al. 2007; Clark et al. 2007). In analogy to population genetics, where genetic drift is only5.7 Local–regional richness relationships part of the equation and is tempered by the impor-It is commonly observed that the relationship be- tance of natural selection, in metacommunity ecol-tween g-diversity and a-diversity – called local– ogy, the ecological drift invoked by neutral theoryregional richness (LRR) relationships – is positive (Hubbell 2001) is only part of the equation, and can(e.g. Ricklefs 1987, 2004; Cornell 1993; Rosenzweig be tempered by deterministic factors, which Chaseand Ziv 1999; Shurin and Srivastava 2005). That is, (2007) referred to as ‘niche selection’.as the number of species in the regional pool in- When niche selection is strong, such as in lowcreases, so does the number of species that coexist productivity or highly disturbed communities, thein any given locality. Originally, the shape of this stochasticity associated with ecological drift shouldrelationship was thought to be able to discern be lower than when niche selection is weakerwhether communities were saturated with species (Booth and Larson 1999; Chase 2003, 2007). Indeed,(in which case the LRR should be asymptotic) or in an experimental study on the assembly of smallwhether they were unsaturated (in which case the pond communities, Chase (2007) found that pat-LRR should be linear) (e.g. Terborgh and Faaborg terns of b-diversity (site-to-site differences in spe-1980; Cornell 1993). However, more recent analyses cies composition) were not different from whathave suggested that LRR relationships can be linear would be expected from purely stochastic (neutral)even when communities are saturated with species, assembly when ponds were permanent. However,or asymptotic even when communities are not when drought was experimentally imposed on one-saturated (reviewed in Shurin and Srivastava 2005). half of those ponds, community structure became Despite the potential problems with inferring more similar between ponds, and patterns ofcausation from LRR patterns, they can provide b-diversity were better predicted by expectationsvaluable information on metacommunity structure. based on models that incorporate species nichesUsing a dataset of several Daphnia species co-occur- (Chase 2007).ring in Baltic Sea rockpools (Bengtsson 1989, 1991), Another way to depict the same idea is to considercombined with a modelling approach, Hugueny the dual mechanisms that can influence coexis-et al. (2007) were able to discern between the patch tence among species in a metacommunity. Co-dynamic (Levins’ metapopulation) and source sink existence results from a balance between stabilizingmodels. Since these models have different under- and equalizing forces (Chesson 2000; Adler et al.lying colonization and competition structures, 2007). Equalizing factors serve to make species re-Hugeney et al. were able to gain a deeper under- sponses to variation in the environment similar, en-standing of the processes structuring this commu- abling them to persist in the same patches, whilenity. Additionally, multivariate analyses exploring stabilizing factors serve to make species responsesthe effects of environmental conditions on patterns to the environment more different, i.e. niche differ-of diversity allow a more direct comparison of how entiation and partitioning (e.g. Chesson 2000). Thethese factors directly and indirectly influence local equalizing factors represent the components of spe-and regional richness when the LRR is explicitly cies coexistence that are neutral, whereas the stabi-considered (Harrison et al. 2006b). lizing factors represent the components of species coexistence that are niche related (Adler et al. 2007). It is easy to see here how both niche and neutral5.8 A synthesis of metacommunity perspectives can (and indeed must) coexist in themodels same conceptual space, since the neutral theory isIn any metacommunity, a variety of processes, in- simply a special case of a more complete theoreticalcluding those related to neutral models (e.g. sto- construct.
  • 80. 66 SPACE AND TIME5.9 Adding food web interactions into the 1.4equation log10 (arthropod sp. 1.3 z = 0.27All of the models discussed above consider species richness)interactions only within a single trophic level. Al- 1.2 z = 0.35though many of the empirical studies above implic-itly considered both predator and prey species intheir analyses, few explicitly considered the role 1.1that food web interactions might play. There isincreasing evidence that not only do predators 1play an important role in structuring communities –1.5 –1 –0.5 0 0.5 Glade size (log10 ha)(reviewed in Chase et al. 2002), but that they areexpected to respond differentially to spatial factors Figure 5.3 Effects of patch area on insect species richnessrelative to prey communities, thus indirectly alter- in Ozark glades with and without insectivorous lizarding metacommunity structure (Holt et al. 1999; Holt predators. Redrawn from Ostman et al. (2007).and Hoopes 2005). Predators often are more sensitive to habitat area Didham 2006; Shulman and Chase 2007). In pond-than prey species (reviewed in Holt and Hoopes dwelling invertebrate communities, Shulman and2005; Ryall and Fahrig 2006), primarily becausepredators typically need larger areas in order to Chase (2007) showed that satellite ponds that weremaintain viable populations. Smaller habitat more isolated from the source pond had lowerpatches are often less likely to maintain large pre- predator–prey species richness ratios relative todators, and thus can have higher diversity and ponds that were closer to the source pond. Thisabundance of prey species than larger habitat resulted primarily because predator richness waspatches (e.g. Terborgh et al. 2001; Gotelli and Ellison highest near the source pond, but decreased with2006; Ostman et al. 2007). Further, because preda- distance, allowing prey species richness to increasetors are more likely to be present in larger habitats, farther from the source.and because predators can strongly influence prey However, predators may also have larger homecommunity structure, this can have a strong influ- areas and be able to utilize a large area of a set ofence on species-area curves (Holt et al. 1999; Holt patches that appear more isolated to prey speciesand Hoopes 2005). For example, on rocky outcrops (Oksanen 1990; Van de Koppel et al. 2005). This may(glades) in the Missouri Ozark mountains, Ostman result in different effects of spatial structure onet al. (2007) found no relationship between habitat predator–prey interactions, as it basically decou-area and insect species richness. However, when ples predators from single prey populations andthe occurrence of the voracious insectivorous lizard may lead to local overexploitation (e.g. van deCrotophytus collaris, primarily only found on larger Koppel et al. 2005). Hence predators may integrateglades, was statistically controlled, significant spe- dynamics over several local ecosystems, which maycies-area relationships emerged, albeit with differ- be similar or of varying quality (see below).ent slopes with and without the predatory lizards(Fig. 5.3). Habitat isolation can also disproportionately in- 5.10 Cross-ecosystem boundariesfluence predators, and thus alter prey metacommu- There is a growing body of evidence suggestingnity structure (reviewed in Holt and Hoopes 2005). that resource subsidies and predators (often termedSeveral empirical studies have shown that more bottom-up and top-down regulation, respectively)isolated habitats have lower rates of predation frequently cross traditionally defined ecosystem(and parasitism), which in turn can alter the struc- boundaries (e.g. Polis et al. 2004). Such fluxes ofture of the entire food web (Zabel and Tscharntke resources and consumers are not generally consid-1998; Kruess and Tscharntke 2000; Watts and ered under traditional metacommunity ecology,
  • 81. INCREASING SPATIO-TEMPORAL SCALES 67since the ecosystems being connected often tran- that occurred in a very different ecosystem a veryscend those types to which ecologists typically re- long distance away.strict themselves. Grassland ecologists rarely study Many species undergo life-history switcheswoodlands. Marine ecologists rarely study fresh- which take them across different types of ecosys-water. Aquatic ecologists rarely study terrestrial tems. For example, many pelagic marine fisheshabitats. Above-ground ecologists seldom study spend their juvenile periods in estuaries, coastalsoil ecology. However, many of the organisms that wetlands or freshwater streams. Additionally,ecologists study readily traverse these boundaries some species undergo much more dramatic onto-seasonally, or across their life span, and do so either genetic niche shifts, and in doing so can shift eco-through behavioural means, such as migration, system types (e.g. aquatic to terrestrial) and trophicthrough life-history shifts (sometimes called onto- roles (e.g. herbivore to carnivore) (reviewed in Wer-genetic niche shifts; Werner and Gilliam 1984), or, ner and Gilliam 1984). These include a large num-like plants, just by growing with roots in the soil ber of insects (e.g. Diptera, Odonata, Megaloptera)and green parts above ground, linking the two sub- and amphibians that have aquatic larval and terres-systems of terrestrial ecology. trial adult stages. In a recent study emphasizing the Organisms frequently move between ecosystem important connections across ecosystem bound-types daily or seasonally for foraging, predator aries, Knight et al. (2005) found cascading effectsavoidance, breeding, overwintering, etc. For exam- of fish in ponds to terrestrial plants through theirple, many birds, ungulates and marine mammals influence of dragonflies with ontogenetic nicheare strong interactors (as both predators and poten- shifts. Specifically, dragonflies are strongly limitedtial competitors) that can migrate over very long by predatory fish in ponds, which translates intodistances. The interactions of these organisms in fewer adult dragonflies near ponds with fish rela-one local community can be the result of processes tive to near ponds without fish. In turn, adult Pollinator Dragonfly Larval Dragonfly Aquatic habitat Terrestrial habitatFigure 5.4 Schematic of the effects of fish in ponds on the adjacent terrestrial food web. Redrawn from Knight et al.(2005).
  • 82. 68 SPACE AND TIMEdragonflies are voracious predators on a variety of broader than previous synthetic treatments of me-terrestrial insects, including species that are impor- tacommunities (Leibold et al. 2004; Holyoak et al.tant pollinators of terrestrial plants. Thus, fish, by 2005). In our view, metacommunity ecology encom-reducing the number of larval and consequently passes the ‘mesoscale’ (sensu Holt 1993) of commu-adult dragonflies near ponds, facilitated higher le- nity ecology, above the level of the localvels of pollination to plants in the surrounding community, but below that of large-scale biogeo-terrestrial community (Fig. 5.4). graphic studies. Thus, there is a rather large spatial The environmental conditions and interactions scale and span of questions that can be addressedthat organisms face in one part of the landscape under the metacommunity umbrella. These in-can strongly influence their interactions in the clude: (1) patterns of metacommunity structure,other. When these sorts of movements are frequent including the relationships between a-, b- and g-and important, we need to expand our view of the diversity; (2) patterns of interspecific interactions,metacommunity. In doing so, we will need to de- relative abundances and coexistence in local com-velop a more general theory that can explicitly ac- munities that are linked through dispersal; (3) SARscount for cross-ecosystem flows of organisms. This and their variation; (4) interactions among speciesmore general theory will have to consider interac- across traditional ecosystem boundaries.tions among organisms that move between ecosys- Metacommunity ecology is a rapidly growingtem types in a landscape context. An organism will field, but one in which considerable new insights,be constrained by the availability of the ecosystem synthesis, theories and empirical studies are needed.type that is most limiting to the population. For We have discussed a number of recent advancesexample, for anadramous fishes, the availability of in metacommunity ecology, including diversity par-pelagic habitat for the adult stage will have little titioning at different spatial scales, the interactioninfluence on abundance if breeding and juvenile between stochastic and deterministic factors, foodhabitats (e.g. streams) are degraded (Allendorf web interactions and cross-ecosystem boundaries.et al. 1997). For many amphibians, the quality of Additionally, although beyond the scope of thethe pond in which larvae will develop is largely current chapter, evolutionary and historical factorsirrelevant in the absence of adequate terrestrial can also play a significant role in the structuring ofhabitat (Semlitsch 1998). We can consider the avail- metacommunities, and this aspect is just beginningability of ecosystem types in a similar manner to the to be considered (e.g. Urban and Skelly 2006;way we consider resources in more traditional con- Kraft et al. 2007). Thus, there is considerably moresumer–resource models (Chase and Leibold 2003). work needed in these and other areas to gain aSpecifically, if one ecosystem type that is needed for better understanding of the processes that maintainan organism to persist is in limited supply, it is that community diversity and structure at differentavailability that will govern the dynamics of the spatial scales. In addition, this understanding willpopulation, and thus the interactions of that organ- be critical in order to develop the concepts and toolsism in the other ecosystem types that it uses. necessary to manage, conserve and restore ecosys- tems in this increasingly human-dominated world (see Chapter 9).5.11 ConclusionsWe have provided a broad overview of metacom- Acknowledgementsmunity ecology. By our definition, metacommunityecology encompasses a large number of spatio-tem- ˚ ¨ We thank Jens Astrom, Wade Ryberg, Herman Ver-poral processes that occur above the level of the hoef and Peter Morin for comments and discus-local community, and thus our treatment is a bit sions that helped us to improve this chapter.
  • 83. CHAPTER 6 Spatio-temporal structure in soil communities and ecosystem processes Matty P. Berg mechanisms (Berg and Bengtsson 2007). In an6.1 Introduction attempt to answer this question I will use a terres-For some time ecologists have been interested in the trial example, the organic horizon of a coniferousrelationship between the structure of communities forest soil, (1) to quantify community variabilityand the consequences this has for the rate of ecosys- in time and across space and (2) to assess possibletem processes, i.e. the structure–function debate consequences of community variability for an ¨(Vanni 2002; Hattenschwiler et al. 2005), and the important soil process, the degradation of organicrole of community architecture as a stabilizing matter and the subsequent flow of energy andmechanism of communities (McCann et al. 2005). nutrients through soil. First, I will briefly introduceA wealth of evidence exists that links the dynamics soil communities, how feeding links between speciesof populations to rates of ecosystem processes in the community can be visualized in connec-(Jones and Lawton 1995; Wardle 2002 and examples tedness food webs, and how organic matter istherein). The role of invertebrates often relates to structuring these food webs. Second, I will showtheir differential utilization of resources or scarce that soil communities are highly variable in timeelements in the habitat, which can result in inter- and across space, much more variable than currentlyspecific functional dissimilarity (Heemsbergen et al. is appreciated in many empirical and theoretical2004): the more different species are in the functions community and food web studies, and elucidatethey fulfil in ecosystems, the more important they the importance of including community variabilitybecome. However, the enormous diversity in or- in our studies to improve our mechanistic under-ganisms in many communities makes the evalua- standing of how soil organisms regulate soil process-tion of their role in ecosystems a Herculean task. To es. Finally, in order to understand the mechanismsreduce this diversity into smaller components a behind the interplay between community structurefood web approach has been adopted. In this way and soil organic matter quality, I will present a con-the impact of community structure on ecosystem ceptual scheme in which the impact of vertical strat-processes and stability and visa versa can be ana- ification of soil communities on degradation islysed (DeAngelis 1992). further elucidated. Although community structure has been shownto be of great importance in regulating ecosystemprocesses, the question remains how important 6.2 Soil communities, detrital food websspatio-temporal variability in community structure and soil processesis for the regulation of processes, and if we need Soil ecologists have long been fascinated by thethis kind of detail to understand underlying enormous diversity of life in soil. There is a 69
  • 84. 70 SPACE AND TIMEtremendous variety of microbes and fauna in soil, Microbes are fed upon by a number of fungivores,with much variation in body size, feeding speciali- such as springtails, mites and some types of nema-zation, life history strategies, spatial distribution todes, whereas bacteria are fed upon by variousand responses to abiotic factors. Although their protozoan groups, such as amoebae, flagellatescoexistence is still an enigma, patterns and determi- and ciliates and other types of nematodes. Fungi-nants in soil biodiversity are emerging (Wardle vores and bacterivores, in turn, are predated by2002; Bardgett et al. 2005). Factors that seem to predaceous mites, spiders and carabid and staphy-regulate the temporal and spatial patterning of linid beetles (Fig. 6.1). When detritus is added to thesoil biota relate primarily to the heterogeneity of soil there is an increase in microbial biomass, whichtheir environment, especially variability in resource is often associated with an increase in soil faunaquantity and quality and microclimate. This hetero- biomass (Wardle 2002 and references therein).geneity provides possibilities for resource and hab- The coexistence of so many species also raisesitat specialization, or niche partitioning. The the question of whether all these species are ofimportance of detritus for species composition and importance for the functioning of soil ecosystems.abundance is derived from the observation that soil Many experiments show that ecosystem processescommunities are bottom-up or donor controlled, depend greatly on the diversity of the communitywhich implies that resource supply, i.e. detritus, in terms of functional characteristics of the speciesdetermines the structure of the community (Pimm present and the abundance and distribution of1982; Wardle 2002). Detritus is the main resource these species over time and across space (Loreaufor fungi and bacteria and detritivores, such as ¨ et al. 2002; Hattenschwiler et al. 2005; Hooper et al.earthworms, isopods, millipedes and enchytraeids. 2005). Collectively, decomposer organisms are Beetles (o) Springtails Spiders Oribatid mites (g) Oribatid mites Predaceous (b) mites Fungi Prostigmatid mites Nematodes (f) Tardigrades Detritus Enchytraeids Nematodes (o) Amoebae Bacteria Flagellates Ciliates Nematodes (b)Figure 6.1 The connectedness food web of a coniferous forest soil. The boxes symbolize the functional groups oforganisms and the vectors indicate the direction of the flow of C and nutrients through the web. The white boxesindicate the bacteria-dominated energy and nutrient flow and the light grey boxes indicate the fungi-dominated flow.Both flows merge on the level of intermediate and top predators, here in dark grey. The high acidity of the soil preventsthe occurrence of earthworms. g, grazers; b, browsers; f, fungivore; b, bacterivore; o, omnivore. After Berg et al. (2001).
  • 85. SPATIO-TEMPORAL STRUCTURE 71essential for the mineralization of carbon and nu- dead animals and metabolic by-products of micro-trients that are bound to dead organic matter and bial degradation. Input of detritus onto the forestthe provision of resources for primary production. floor is dominated by leaves, flower heads, seeds,To quantify the functional significance of soil or- twigs and bark and is often characterized by aganisms, for instance for nutrient mineralization, seasonal pattern. Although plant species vary inspecies have been assembled in guilds or function- their timing of leaf litter abscission over a year,al groups (Fig. 6.1). The rationale behind collecting most temperate tree species shed their litter in au-species in functional groups, based on similar food tumn. In coniferous forests needle fall is more ortypes, predators, metabolic efficiencies and loca- less continuous over a year, but also peaks in au-tion in the soil profile, is that they may affect tumn. On some occasions the input of plant re-ecosystem processes in a similar way (Moore mains other than leaves can be high. After aet al. 1988). Interestingly, the assumption that severe storm large amounts of branches are blownthese shared characteristics of species reflect simi- from the trees and unusually heavy rains causelar functions is seldom tested. The flow of carbon mast seeding and pulses of primary productionand nutrients through soil largely depends on tro- (Ostfeld and Keesing 2000). Annual fluctuations inphic interactions between these functional groups, the supply of detritus to soil organisms may be aand these trophic relationships can be visualized reason why many species of soil biota show strongin so-called connectedness food webs. A connect- seasonal patterns in activity or abundance (Wardleedness food web shows the architecture of trophic 2002).relationships within soil communities; for exam- In mixed forests where tree species that pro-ple, in Fig. 6.1 a connectedness food web is shown duce litters of different qualities coexist, a patchyfor a pine forest soil in the temperate region (Berg horizontal spatial distribution in detritus qualityet al. 2001). This type of food web indicates only the may occur. Patchy spatial distributions of soilmost dominant feeding relationships and does not biota communities may be correlated with spa-necessarily depict all physiologically possible tro- tial variability in the amount of litter produced.phic interactions between functional groups. For example, in the Amazon forest DicoryniaOften, only those trophic interactions are included quianensis produces litter with a high content offor which we have proof that they matter for car- polyphenolic complexes, whereas Qualea spp.bon and nutrient flows. A food web, therefore, is produce litter with low levels of polyphenols,not more than a reflection of a real community. It but high aluminium levels (Wardle and Lavelledoes not necessarily depict the real organization of 1997). Earthworms are absent from the litter ofthe soil community, because other types than feed- D. quianensis, whereas in litter which accumu-ing interactions are excluded, such as facilitation lates under Qualea endogenic earthworms areor mutualism. More examples of connectedness found, with a high abundance near the treefood webs are given by Hunt et al. (1987; short- trunk where litter input is highest. More evi-grass prairie), de Ruiter et al. (1993; agricultural dence that the spatial positioning of plants is a ¨field) and Schroter et al. (2003; coniferous forests). key determinant of the spatial patterning in soilGiven the importance of detritus, both quantita- biota can be found in Wardle (2002).tively and qualitatively, for the structure of soil The organic compounds that make up detrituscommunities, we need a closer look at the compo- differ greatly in physical attributes and chemicalsition of soil organic matter, the process of organic complexity and, subsequently, in degradability.matter decay and its spatio-temporal distribution The process of organic matter decay is subdividedin soil. into two main phases. In the initial phase the readi- ly decomposable organic compounds, mainly poly- saccharides and non-lignified carbohydrates, are6.3 Soil organic matter degraded. In a subsequent phase the more recalci-Detritus, or soil organic matter, is a collection of trant compounds, such as lignified carbohydrates,various organic compounds, such as plant remains, lignin and organic intermediates, are broken down
  • 86. 72 SPACE AND TIME(Berg and Matzner 1997). The net effect is a relative composition is linked to variability in soil organicincrease in recalcitrant substrates as decomposition matter quality, I present the results from an exper-proceeds, while the amounts of un-decomposable iment that simultaneously describes variability inmaterial accumulate with depth in the organic pro- organic matter degradation and community struc-file. As nitrogen is usually the limiting nutrient in ture over time and across horizontal and verticalthe initial phase of litter decomposition, the C:N space. This study was performed in a first-genera-ratio of organic matter is a general index of the tion, 40-year-old Scots pine, Pinus sylvestris, forest,quality of litter (Gadisch and Giller 1997; Berg and in The Netherlands, planted on a heather field onLaskowski 2006). However, to understand the me- an inland dune of a former sand-drift area (Bergchanisms that regulate the processes of decomposi- et al. 2001). First, I shall describe the variability intion, the type of carbon in organic matter, the community composition over time, then introduceconcentration of other nutrients than nitrogen and, some factors that can explain temporal variabilityespecially, the composition of various secondary in community composition and finally indicateplant compounds are also important. For instance, possible consequences for food web studies.litters with high lignin content, and that are rich inpolyphenols, have a low degradation rate. Lignin isdifficult to decompose by microbial enzymes (Shah 6.4 Variability in time in soil communitiesand Nerud 2002) and can mask cell wall polysac-charides from degradation (Chesson 1997). Poly- Although it has been repeatedly pointed out thatphenols, condensed tannins, terpenes and surface temporal and spatial heterogeneity in soil is crucialwaxes strongly reduce the activity of microbes and for understanding the distribution of soil organ-the palatability of plant residues for detritivores isms and how biodiversity affects key ecosystem ¨(Hattenschwiler and Vitousek 2000). processes (Bengtsson 1994; Ettema and Wardle During degradation, the morphology, biochemis- 2002; Wardle 2002), surprisingly little is knowntry and physical properties of organic matter are about temporal and spatial variability in soil com-continuously modified. These physicochemical munities and detrital food webs (Bengtsson andtransformations result in a decline in substrate Berg 2005). Variability in community compositionquality for microbes and detritivores and are ac- has most often been studied in aquatic ecosystems.companied by succession of soil biota (Ponge 1991; These studies found a considerable temporal andDilly and Irmler 1998). Over time, this older organic spatial variability in the composition of the foodmatter is covered by newly formed detritus with a web of a freshwater pond (Warren 1989), an inter-higher quality, and this stratification in substrate mittent stream (Closs and Lake 1994) and a tidalquality results in shifts in the vertical distribution freshwater river (Findlay et al. 1996). Schoenly andof soil biota (Faber 1991; Ponge 1991; Berg et al. Cohen (1991) analysed a set of aquatic and terres-1998a). In the absence of earthworms detritus is trial food webs for temporal variability in commu-not mixed through the soil and a stratified organic nity composition. Very few species in the collectionhorizon is formed, with a subsequent litter, frag- of food webs occurred on each sampling occasion,mented litter and humus layer on top of parent and most of the species were found only once. Tomaterial. Although in the presence of earthworms examine the temporal and spatial variability in theorganic matter is mixed through the soil, there still structure of a soil community I performed a strati-exists a vertical stratification in organic matter qual- fied litterbag experiment (for details of the design,ity and quantity, but on a larger vertical spatial see Fig. 6.2). Compositional variability, expressedscale. using the Bray–Curtis (BC) similarity index, was These examples imply that the distribution of measured over 2.5 years, for three organic horizons,soil organic matter and soil organisms is often namely litter, fragmented litter and humus (for de-strongly interlinked. To evaluate the importance tails, see Fig. 6.3). From litter to humus the qualityof organic matter for community composition of organic matter for soil biota strongly declines. Asand to assess whether variability in community far as I know, this is the first time that temporal and
  • 87. SPATIO-TEMPORAL STRUCTURE 73 in BCL from 0.73 to 0.39 over 2.5 years). Only the litter horizon shows a significant increase in varia- 1.5–45 m bility over time. This difference in compositional L variability between horizons corresponds with the F N=6 T=1 mass loss of organic matter after 2.5 years, which is H 1.6–3.8 cm significantly higher in litter (mass loss L ¼ 44.2% of initial mass, versus 4.2% and 2.8% for F and H, T=2 respectively). Moreover, mass loss in litter is corre- 8 weeks lated with change in chemical properties of litter, especially C:N ratios (Berg et al. 1998b). This strong- T=3 ly suggests that increase in compositional variabil- ity in litter over time is linked to changes in the quantity and/or quality of organic matter. Organic T = 15 matter turnover is mentioned as one of the most important parts of environmental variability in for- ests (Bengtsson 1994). Within-year variability in community composi- tion is larger than between-year variability in com-Figure 6.2 Stratified litterbag sets to measure organicmatter and functional group dynamics over time and munity composition, but only in litter. Samplesacross space. The litterbags in a set, sized 10 Â 10 cm taken a year apart have significant lower commu-each, were filled with homogenized pine litter (L; mesh nity variability than samples obtained with half-size 3 mm), fragmented pine litter (F; mesh size 1.3 mm) year intervals. This confirms the findings of Bengts-or pine humus (H; mesh size 1 mm on top, 0.2 mm at son and Berg (2005), who showed for a managedbottom) and were 1.0 cm, 2.2 cm and 2.2 cm thick,respectively. Each litterbag set was 5.4 cm thick and pine forest and a virgin spruce forest that within-reflected the thickness and detrital composition of the year variability in the composition of animallocal organic soil horizon in which the sets were functional groups can be as large as between-yearintroduced. The 180 litterbag sets (12 sets, 15 samplings), variability. The seasonal periodic oscillation ineach consisting of three litterbags filled with the three community variability coincides with a similarsuccessive horizons, were placed randomly within a (40 Â50 m) area, and sampled at 8 week intervals over a period seasonal pattern in soil temperature and moistureof 2.5 years. The distance between litterbag sets ranged values (Berg and Verhoef 1998). The annual ampli-between 1.5 m and 45 m. On each sampling occasion six tudes of soil temperature and water content arelitterbag sets were used to extract soil meso- and greater in litter than in underlying horizons.macrofauna and to analyse mass loss and chemical Owing to their short generation times many soilcomposition of organic matter. The remaining six setswere used to extract the microflora and microfauna. See organisms can react relatively fast to short-termBerg and Bengtsson (2007) for details about extraction changes in the environment. This can be of themethods of soil organism and biomass calculation order of some days for organisms at the base ofprocedures. the food web to several months for animals at the top of the food web (Hunt et al. 1987). The highspatial variability in soil community composition within-year variability in the litter community com-has been simultaneously analysed. pared with deeper horizons can be explained by its The composition of the soil community varies higher resource quality. However, within-yearmore over time in the litter horizon (L) than in the variability in community composition is not easilyunderlying fragmented litter (F) and humus hori- explained by changes in organic matter quality,zons (H) (BCL ¼ 0.56 Æ 0.057, BCF ¼ 0.67 Æ 0.030, although input of detritus onto soil shows a season-BCH ¼ 0.68 Æ 0.032). The observed higher average al pattern too. In an additional set of litterbags,variability in litter is caused by a significant de- filled with freshly fallen litter and replaced in thecrease in similarity between samples when the field every 8 weeks to obtain a constant litter qualitytime interval between sampling increases (decrease over time, similar temporal trends in community
  • 88. 74 SPACE AND TIME 0.8 4 FG 48 sp 0.6 BC similarity 0.4 0.2 0 Temp Hor Vert Vert Vert L-F F-H L-HFigure 6.3 Temporal variability, and spatial variability in the micro-arthropod species composition (circles, 48 speciescombined) and the micro-arthropod functional group composition (squares, four functional groups combined(springtails, oribatid browser mites, oribatid grazer mites and predaceous mites)) in which the species were lumped, for aperiod of 2 years. The spiders and omnivorous beetles (no species determined) and prostigmatid mites (dominated for95% by one species) were excluded from these analyses. The observed values for functional groups (squares) are similarto the values observed for the whole community. Compositional similarity clearly declines when the taxonomical level ofresolution (from functional group to species) increases. The Bray–Curtis (BC) similarity index (Legendre and Legendre1998) is used as a measure of community variability. Mean values of BC similarity are shown for the litter horizon overtime (Temp), the litter horizon in horizontal space (Hor) and between the adjacent litter and fragmented litter horizons(Vert) and fragmented litter and humus horizon (Vert) and between the non-adjacent litter and humus horizons (Vert) invertical space. Bars indicate the 95% confidence intervals. Lower BC-similarity values indicate higher variability. Themean biomass values from the six litterbags in each horizon collected at the same time were taken to calculate temporalsimilarity. Hence, the measures of temporal similarity are made on the spatial scale of the whole study plot. For analysesof horizontal spatial similarity, pair-wise similarities were calculated between the micro-arthropod functional group orspecies composition in each of the six litterbag sets sampled at each time. Each horizon was examined separately. Forvertical similarity on the plot scale, the average abundances in each horizon of the six litterbag sets was used to calculatesimilarity between composition in adjacent (L–F and F–H) and non-adjacent (L–H) horizons. For more detail aboutanalyses and for additional analyses in community composition at other levels of taxonomic resolution, see Berg andBengtsson (2007).variability are observed to those in litter that de- Fig. 6.1, are a template for simulation modelsgraded over time. that calculate the amount of C and N processed The relatively high short-term variability of by a functional group or the whole food web.soil fauna communities implies that soil food Using one of these models we derived the min-webs, at least in seemingly stable and homoge- eralization rates for C and N from the trophicneous temperate forests, do not have a fixed interactions among groups of organisms, basedstructure but are instead dynamic. Therefore, on the annual biomass of the functional groupsconclusions from community and food web ana- depicted in Fig. 6.1. The values of C and Nlyses based on yearly averages should be treated mineralization rates indicated by the modelwith caution, especially in ecosystems with a fast were compared with observed C and N lossesturnover of populations and basic resources. For from organic matter in litterbags. The food webexample, connectedness food webs, as shown in model underestimated C mineralization in litter
  • 89. SPATIO-TEMPORAL STRUCTURE 75by a factor of 2, whereas the model predictions 6.5 Variability across horizontal space infor fragmented litter and humus were closer to soil communitiesthe actual C mineralization. The simulated Nmineralization rates resembled the observed Soil organisms are not homogeneously distributedrates in litter, but for fragmented litter and across space, both horizontally and vertically. Ahumus the simulated values accounted for only limitation of studies on spatial variability is thatapproximately 40% and 20% of the actual N they have been done on individual organismmineralization (Berg et al. 2001). However, groups rather than on whole communities. In ourwhen we based our model predictions on the example, the horizontal spatial variability in com-sampling biomasses, not averaged over time, munity composition is rather low and does notand averaged the derived mineralization rates differ between horizons (BCL ¼ 0.69 Æ 0.062, BCFover 1 year, the model significantly overesti- ¼ 0.72 Æ 0.046 and BCH ¼ 0.70 Æ 0.047). Similarly, amated mineralization rates (data not published). low variability in mass loss of organic matter acrossSensitivity analyses of this food web model have space is found for all three subsequent horizons.shown that small changes in the biomass of par- Moreover, a horizontal spatial structure in commu-ticular functional groups, for example basal or- nity composition and mass loss of detritus is notganisms or predators, may have a marked and observed, which means that the level of variabilitydisproportionate effect on the mineralization of does not depend on the distance between samplingnutrients (Hunt et al. 1987; de Ruiter et al. 1993). points. These observations emphasize that organicThe discrepancy between model predictions, matter turnover is an important factor explainingwhen based on temporal variability in functional variability in community composition, and in hori-group biomass versus their annual biomass, zontal space.might relate to the existence of two distinct en- The observed low variability in community com-ergy channels within detrital food webs (Rooney position is not expected, as the horizontal variabil-et al. 2006). One energy channel is based on ity measures are based on single sets of litterbagsbacteria, the other on fungi, and these two chan- rather than on averages of six litterbags (see Fig. 6.3nels differ in both productivity and turnover for details), and horizontal variability in plants,rate. Size differences between fungi and fungi- microbes and animals has been shown to be sub- ˚˚ stantial (Saetre and Baath 2000; Ettema and Wardlevores, on the one side, and bacteria and bacter-ivores, on the other, result in dissimilar 2002; Laverman et al. 2002). A likely explanation ispopulation dynamics, with faster turnover in that the study site lacks ecosystem engineers, suchthe bacteria-dominated energy channel com- as earthworms, because of the acid, non-calcareouspared with the fungi-dominated channel. Aver- soil type. Soil engineers can create small-scaleaging of time in these compartmentalized food variability in the environment by forming casts,web models may account for deviation in C and redistributing organic matter and altering soil tex-N flows found between observed values and ture (for an overview, see Lavelle and Spain 2001).model predictions. It may be more informative Plants have been shown to increase spatial varia-to take temporal variability in food web struc- bility (Klironomos et al. 1999; Wardle 2002). Theture into account when examining C and N dy- possibility of vegetation inducing variability wasnamics, especially when effects of variability are greatly reduced because the litterbags were incu-non-linear. Moreover, identification of which an- bated at a more or less constant distance from equi-imals are of prime importance for flow of C and distantly dispersed trees, in a forest withN might change when the organization of the homogeneous grass–moss undergrowth. The rela-food web is not fixed but dynamic. For a better tively low spatial variability and absence of hori-understanding of how food web structure affects zontal spatial structure may thus be indicativefood web stability and nutrient fluxes through of the absence of factors that create variation,food webs we may need to include more tempo- especially given the short duration of the study –ral details. only 2.5 years. In older, more mature forests with a
  • 90. 76 SPACE AND TIME a b cFigure 6.4 A theoretical, horizontal spatially explicit connectedness food web of three trophic groups (1, microbes; 2,microbivores; 3, predators) with 10 species (indicated with a letter). (a) Distribution of the 10 species (boxes) inhorizontal space, at a specific time. The length of the boxes indicates the amount of horizontal space (home range)occupied by a species. (b) The potential connectedness food web showing ecologically possible feeding interactions(lines) between the 10 species (boxes). (c) Three samples at different points in space (vertical grey arrows) result in threelocal or actual food webs. The actual food web contains a subset of the species and feeding interactions of thepotentially possible food web. Adapted from Brose et al. (2005).high plant diversity and patchy distribution of Nevertheless, horizontal spatial variability indecaying logs a higher horizontal spatial variability community composition is present, and this im-in detritus quality will be found than observed in plies that connectedness food webs can bethis plantation. When soil engineers are also present structured into compartments in which interac-horizontal spatial variability in community compo- tions between species are localized. Compartmen-sition will increase, owing to their direct and indirect talization might strongly influence food webeffects on organic matter quantity and quality. stability properties (Hall and Raffaelli 1997; but
  • 91. SPATIO-TEMPORAL STRUCTURE 77see Morin 1999). This aspect of spatial scale has so (BCF–H ¼ 0.57 Æ 0.085). Moreover, vertical spatialfar been largely ignored in empirical and theoreti- variability in the community composition of non-cal food web studies. In most studies that focus on adjacent horizons is greater than both temporal andhorizontal space trophic interactions are aggre- horizontal variability (Fig. 6.3). Thus, a high pro-gated over broader spatial scales, the scale of a portion of the variation in community structure inforest or even a landscape, and not the scale of soils is likely to be attributed to different soil hor-the organisms (Brose et al. 2005). This results in izons, even if these horizons are only a few centi-potential food web architectures that depict all metres apart.physiological possible feeding links. In reality, Soil temperature and soil moisture content,when sampling on a small spatial scale, not all organic matter texture, quantity and quality arepotential feeding links are realized, because not factors that differ between organic horizons.every species occurs everywhere (Fig. 6.4). The Differences in soil temperature and moisture con-larger the species, the larger its home range, but tent between successive horizons are generallythe lower its numerical abundance and frequency small (Berg and Verhoef 1998), whereas significantin a local community (Cohen et al. 2003; Jennings modifications in organic matter morphology (fromand Mackinson 2003). Similarly, not all small-sized needle to amorphous colloids) are accompanied byorganisms occur locally, owing to environmental a decrease in particle size, organic matter qualityheterogeneity. In Fig. 6.4, three samples at differ- and, subsequently, degradation rates. Decay ratesent locations yield three different local connected- of organic matter decline from litter to humusness food webs. The top predator is not always (mass loss/year: L ¼ 17.6% versus 1.7% and 1.1%present, i.e. the right-hand example in Fig. 6.4c, for F and H, respectively). These morphologicalowing to its low abundance and large home and chemical transitions, which are more pro-range. The spatial dimensions of resource use of nounced the further the horizons are apart, maythe top predator links the other two local webs cause considerable variation in the composition of(Holt 1996; Pokarzhevskii et al. 2003; Hedlund soil communities. When ecosystem engineers areet al. 2004). The more local webs are sampled and present, the scale of vertical gradient in organiccombined, the more the potential food web is ap- matter texture, quantity and quality can changeproached (Brose et al. 2005). Many local versus one from a few centimetres, in coniferous forests, topotential food web and coupling of local webs by decimetres or even metres in other ecosystems,predators are a challenge for future food web re- owing to mixing of detritus through the soil bysearch, especially from a modelling point of view. earthworms.However, the first attempts to deal with these as- The observation of vertical structure in commu-pects of spatial variability have already been made nity composition implies that, even more so than(Kondoh 2003, 2005; Teng and McCann 2004). for horizontal space, food webs can be structured into compartments in which interactions are loca- lized (Fig. 6.5). Most soil food web studies, howev-6.6 Variability across vertical space in soil er, do not consider vertical space, and feeding linkscommunities is high in connectedness food webs are aggregated overThe distribution of groups of soil organisms often the whole organic layer, or over the depth of soilshows a distinct vertical spatial pattern (Faber 1991; core samples. However, on the scale of the organ-Ponge 1991; Berg et al. 1998a). Likewise, in our isms, not all physiologically possible trophic inter-forest example community composition has a dis- actions are realized, as not all species aretinct vertical pattern. The community is significant- everywhere. For instance, the small particle size ofly more variable between non-adjacent organic humus prevents large-bodied animals, such as spi-horizons, the litter–humus comparison (BCL–H ¼ ders, occurring in the humus horizon (Berg et al.0.47 Æ 0.088), than between adjacent horizons, the 1998a), whereas the decline in resource quality withlitter–fragmented litter comparison (BCL–F ¼ 0.62 Æ depth strongly affects succession, hence species0.084) and fragmented litter–humus comparison composition, of microbes (Kendrick and Burges
  • 92. 78 SPACE AND TIME a b cFigure 6.5 A vertical spatially explicit connectedness food web of three trophic groups (1, microbes; 2, microbivores; 3,predators) with 10 species (indicated with a letter). (a) Distribution of the 10 species (boxes) in vertical space. The lengthof the boxes indicates the amount of vertical space occupied by a species. The dashed lines separate the organichorizons. (b) The potential connectedness food web showing ecologically possible feeding interactions (lines) betweenthe 10 species (boxes). (c) Three samples from subsequent organic horizons (L, litter; F, fragmented litter; H, humus)result in three local or actual food webs. The actual food web contains a subset of the species and feeding interactions ofthe potentially possible food web.1962) and fauna (Dilly and Irmler 1998). As a result, differences between local food webs versus onesuccessive organic horizons yield different local aggregated food web is as challenging as forconnectedness food webs (Fig. 6.5). Small-sized horizontal space.organisms, such as microbes and Protozoa, are Berg et al. (2001) have shown, in one of the fewhighly abundant, but their occurrence is vertically food web studies in which vertical space is explicit-localized because of adaptation to particular organ- ly taken into account, that the rates of C and Nic matter qualities, in combination with low dis- mineralization may strongly depend on verticalpersal ability. Fungivores, like springtails and stratification of functional groups and are relatedmites, and small predaceous mites may couple to differences in food web composition between soil ¨ ¨local food webs by vertical migration (Setala and compartments. The observed significant decreaseAarnio 2002), often induced by fluctuations in in C and N losses from litter to humus (50 mgtemperature or soil moisture content (Briones et al. C/g/year in L to 1.5 mg C/g/year in H; 0.9 mg1997), or spatial dimensions in resource use. From a N/g/year in L to 0.05 mg N/g/year in H) corre-food web model perspective the variance in local lates strongly with model simulations of C andvertical food web structure and the structural N mineralization rates for the three successive
  • 93. SPATIO-TEMPORAL STRUCTURE 79horizons (Pearson correlation coefficient for C, r ¼ decade, and large spatial scales, plots to land-0.999; for N, r ¼ 0.996). Moreover, total food web scape, adopted in many studies. In communitybiomass did not explain the observed decrease in and food web studies we have to use the scalesdecay rates of detritus with depth. This indicates at which organisms operate and include theirthe importance of indirect contributions of soil basal resources to fully understand their role infauna, via enhancing the mineralization rates of communities and ecosystems. Time and space are,microbes, sometimes exceeding the direct contribu- however, the same side of the coin. They act si-tion of fauna to C and N mineralization rates. The multaneously on species in the soil communityresults imply that the regulation of ecosystem pro- over time by biochemical and physical changescesses is inextricably linked to the structure of the in resources and across space by burrowingsoil community, and that accounting for vertical organic matter owing to input of fresh litter. Spa-variation in food web structure seems essential for tio-temporal modifications in detrital characteris-understanding energy and nutrient flows in soils. tics as degradation progresses are perceivedEmpirical evidence that micro-stratification of soil differently by bacteria and fungi (Bosatta andorganisms can have specific effects on ecosystem ˚ Agren 1991). Bacteria have shorter generationprocesses is given by Briones and Ineson (2002). times than fungi (Rooney et al. 2006), and reactUsing radiocarbon techniques, they found that differently to disturbances (Orwin et al. 2006).enchytraeids in a blanket bog assimilate carbon Hence, modification in detritus affects the fungi-components of predominantly 5–10 years old. or bacteria-based energy or nutrient channels ofVertical movement of enchytraeids due to abiotic the food web in a different way, and this mayfactors does not affect this as they show similar contribute to the observed high within-year varia-14 C enrichment values at various depths. In re- bility in community composition within organicsponse to warming, however, worms are forced horizons. Moreover, bacteria gain importancedeeper into the soil, where they feed on older from litter to humus, because they benefit fromcarbon components. Similarly, the vertical strati- the increase in resource surface owing to a declinefication of fungivores determines their impact on in particle size (Fig. 6.6). Fungi penetrate wholeecosystem processes (Faber 1991). Surface-living, leaves, a trait that bacteria lack, and do not profitepigeic species affect the colonization of fresh from the decrease in particle size. A decline inleaf litter, and potentially enhance the immobili- particle size and alteration in chemical composi-zation of nutrients. Hemiedaphic species en- tion with soil depth and time may explain thehance the net mineralization and nutrient often observed succession of soil organismsmobilization in fragmented litter. Euedaphic spe- when degradation proceeds, and result in specificcies living in the humus horizon have the poten- local communities in subsequent organic horizonstial to affect plant growth by interference with (Fig. 6.6).mycorrhizal establishment or nutrient uptake by Resource-based modification in food web struc-the roots. Therefore, aggregating organic hori- ture in turn feeds back to organic matter degrada-zons, with their own species compositions and tion, because food web composition strongly affectsspecific impact on processes, is an oversimplifi- the rate of C and N mineralization. Trophic interac-cation that obscures our understanding of the tions within local communities can result in eithermechanisms behind many soil processes. immobilization of nutrients, in litter, or mobiliza- tion of nutrients, in fragmented litter and humus. This shift in the role soil organisms play in the6.7 Spatio-temporal scales of community decomposition processes warrants the subdivisionstudies of the organic horizon in its successive layers, eachThere exists a disparity between the short tempo- representing a different phase in the process ofral scale, days to weeks, and small spatial scale, decomposition, each with a specific soil communitycentimetres to metres, at which most soil organ- composition (Fig. 6.6). However, spatio-temporalisms operate and the long temporal scales, years to changes in resources are not described adequately
  • 94. 80 SPACE AND TIME 400 300 200 100 0 L F H Depth soil Actual food webs Mass loss cm 100 0 1 L 1 75 % remaining mass 2 F Space 50 3 2 4 25 5 H 0 3 6 Parent Time (log scale) materialFigure 6.6 Model for mass loss and rate-regulating factors during degradation of detritus (left; after Berg and Laskowski2006). Phase I curve (equals litter), mass loss is due to loss of water-soluble substances and non-lignified carbohydrates.Mass loss is stimulated by high levels of N, P and S. Trophic interactions result in nutrient immobilization. Phase II curve(equals fragmented litter), mass loss is due to loss of lignified carbohydrates, lignin and lignin-like compounds. Mass lossis stimulated by Mn and suppressed by high levels of N. Trophic interactions result in nutrient mobilization. Phase 3 curve(equals humus), mass loss reaches its limit value; concentrations of lignin and lignin-like compounds are about constant.Trophic interactions result in nutrient mobilization. Over time, which corresponds with across depth, the influence ofclimate on decomposition declines, lignin and N concentrations increase and the degradation rate approaches zero.Particle size strongly declines, resulting in a dominance of bacteria over fungi from litter to humus (bar diagram abovethe model curve). White bars, bacterial biomass; black bars, fungal biomass for litter, fragmented litter and humus,respectively. The y-axis shows biomass C in g C/g dry soil/year (after Berg et al. 2001). Physico-chemical changes indetritus over time, down the soil profile, result in local, horizon-specific food webs (right), often with different speciescompositions (Fig. 6.3). Species with high trophic position and an extensive vertical resource use may couple local foodwebs. The role of soil biota often shifts from the top layer to deeper organic horizons, from immobilization of nutrients inlitter to mobilization of nutrients in fragmented litter and humus. L, litter; F, fragmented litter; H, humus.in many community studies and food models, structural changes in community composition af-while variability in community composition over fect ecosystem processes, and vice versa, we shouldtime and space is often ignored (Berg and Bengts- adopt a more spatio-temporal approach in ourson 2007). To improve our understanding of how studies.
  • 95. PART IVAPPLICATIONS
  • 96. This page intentionally left blank
  • 97. CHAPTER 7 Applications of community ecology approaches in terrestrial ecosystems: local problems, remote causes Wim H. van der Putten7.1 Introduction ecology is that of the consequences of increased habitat fragmentation for the dispersal and survival7.1.1 Issues in applied community ecology of populations and species. Interesting progress inThere are many examples of the applied value of that field of research has been made, for examplecommunity interactions in terrestrial ecosystems. how species from different trophic levels are affect-Enhancing biological control of pests and plagues, ed differently by habitat fragmentation. Habitatfor example, has become a major arena for testing fragmentation could affect trophic control of spe-bottom-up versus top-down control, direct versus cies and, therefore, lead to either outbreaks or ex-indirect defence of plants, and constitutive versus tinctions of species (Tscharntke et al. 2005).induced defence (Karban and Baldwin 1997). This The above-mentioned examples have receivedfield of applied community ecology has developed much attention in the past and there are manyimportant new concepts, such as the concept of overviews and reviews written on these interfacesmultitrophic interactions (Tscharntke and Hawkins of fundamental and applied community ecology.2002). It shows that applied and fundamental re- Here, I want to focus on a current issue: commu-search can go hand in hand fruitfully. Another field nity interactions across system boundaries. Myis that of biodiversity and ecosystem functioning, main point is that applied questions concerningwhich started from the general concern about con- the management of species, communities or eco-sequences of human-induced loss of biodiversity systems at a given place and time may depend onfor the functioning of ecosystems. This led research- processes that take place in adjacent subsystems,ers to question how biodiversity loss may influence or in remote ecosystems, or that have taken placeecosystem processes, their stability and resilience in the past. The soil will play an important role in(Loreau et al. 2001). These ecosystem processes and this chapter, because soil has a strong ‘memory’properties have important consequences for the and legacy effects can cause historical contingencysustainability of human society. However, when (see Chapter 4). Many soil organisms are relativelycompared with biological control, the biodiversity poor dispersers, but they are assumed to exhibitdebate is younger and the concepts are now being considerable redundancy in ecosystem processes.deepened. Studies that went into more complex Nevertheless, redundancy does not apply to allcommunity interactions were exciting and have functions and the response rate of soil commu-made considerable progress (Cardinale et al. 2006). nities to changes may be slower than that ob-A third well-known applied issue in community served above ground. 83
  • 98. 84 APPLICATIONS My message for end-users or stakeholders is that tions for the interpretation of well-controlled plantyour current problem of interest may very well biodiversity experiments, where the intentionaloriginate from historical events, or that the cause sowing of plants promotes spatial and temporalcan be found at your neighbours, or your neigh- stability when compared with naturally colonizedbours’ neighbour’s yard. My message for scientists communities (Bezemer and van der Putten 2007).is that a local focus or a snapshot in time will not be Temporal instability may involve immigration ofsufficient to fully understand community interac- species that introduce novel properties. Immigra-tions that take place here and now. I will argue that tion of species with different traits is a natural pro-applications of community ecology approaches in cess in succession sequences, when grasses andterrestrial ecosystems may require a spatio-tempo- forbs are replaced by shrubs and trees, producingral approach in order to predict, anticipate or solve litter that differs in rates of decomposition. Whileecological problems as a result of natural, or such processes are normal in successional gradi-human-induced, changes in the environment. ents, artificial introduction of species from different continents may change the functioning of entire ecosystems (Vitousek et al. 1987). If such invaders7.1.2 Top-down and bottom-up go hand also reach disproportionate abundance, for whatev-in hand er reasons, incoming species may cause switches inThe long-standing question of what controls the community structure and ecosystem processes.abundance of species was strongly fuelled by the Top-down and bottom-up processes play an impor-debate between Hairston et al. (1960) and Ehrlich tant role in causing invasiveness and in the conse-and Raven (1964). They disagreed about whether quences of invasions. Invasions are among thespecies are controlled by top-down (predator con- major application issues worldwide and under-trolled) or bottom-up (resource controlled) forces. standing why species become invasive, how theseNowadays, however, we realize that top-down invasions can be prevented or how they should beforces may become bottom-up (Moore et al. 2003) managed is an important challenge for communityand that evolution takes place in a multitrophic ecology.selection arena where selection pressures change I will first work out how local community inter-from time to time (van der Putten et al. 2001). Com- actions may be altered by species that move withinmunities are dynamic, and alternating species or across ecosystems. I will argue that local com-abundances may be controlled by a mix of resource munity interactions are influenced by processesand predator controls which, each in their turn, can and interactions between adjacent or distant sys-exert selection pressure on local individuals. Clear- tems. I use the case of plant interactions withly, this is not the end, but only the start of a dyna- above- and belowground organisms and discussmical perspective of communities and community how these organisms may influence each other,interactions. through the shared plant resource. Then, I explain Unlike many simple experimental settings, com- how aboveground–belowground interactions canmunity composition is not constant; but rather spe- be influenced by legacy effects in soil, and howcies move in and out of communities all the time. these delayed influences may affect succession. IThis is observed in lake food webs (Jonsson et al. will focus on ecosystem restoration on post-agri-2005), and in diversity experiments where species- cultural land, which is one of the main applicationdiverse plant communities created by sowing seed issues in the industrialized world. Subsequently, Iwere much more stable in composition than non- will focus on alien exotic plant species, how theysown species-diverse plant communities (Bezemer may influence local community interactions andand van der Putten 2007). These examples illustrate how altering above and belowground bottom-upthat local community interactions can be changed and top-down processes may contribute to inva-by incoming or outgoing species and that immigra- siveness. Finally, I will speculate about the conse-tion and emigration are quite common in natural quences of (human induced) global warmingcommunities. That awareness has major implica- beyond the climate envelope approach, and I will
  • 99. APPLICATIONS OF COMMUNITY ECOLOGY APPROACHES 85conclude with some suggestions for users, stake- large populations of the snow goose cause over-holders and scientists. grazing of shoreline vegetation, transforming the vegetation into bare sediment. Even when the snow goose population is eventually reduced, re-7.2 Community interactions across system covery of the breeding ground vegetation will takeboundaries decades (Abraham et al. 2005). Climate change may have strong local and non-7.2.1 Linkages between adjacent or distant local effects on phenology. Local effects are, forecosystems example, the interruption of prey abundance forAdjacent ecosystems may influence each other in a hole-breeding birds, such as the great tit (Parusvariety of ways. For example, upstream deforesta- major). This forest bird feeds on winter moth,tion may lead to downstream nutrient enrichment which occurs on birch and oak trees. When theyand sedimentation (Ineson et al. 2004). Adjacent are feeding their nestlings, P. major first use winterecosystems also may be mutually influenced by moth caterpillars from birch and then switch toanimal species that have a wide foraging range caterpillars from oak trees. However, climate(van de Koppel et al. 2005). For example, seabirds warming has advanced spring by 10 days overaffect plant productivity, detritus and, consequent- past decades and, while birch trees respond toly, beetles in coastal terrestrial communities (San- warming, oaks do not. As a consequence, the birdschez-Pinero and Polis 2000). Introduced exotic face a gap in prey availability, limiting their abilityspecies may influence these community interac- to raise their chicks (Visser and Holleman 2001).tions. For example, introduced rats in New Zealand Some migratory birds, such as the pied flycatcher,feed on seabirds, reducing the transport of nutri- overwinter in West Africa. As the effects of climateents from sea to land and leading to cascading warming on arrival of spring in North-Westerneffects on belowground organisms and associated Europe are not apparent in West Africa, the Piedecosystem processes (Fukami et al. 2006). flycatchers still arrive at the same dates, but at a Land use changes can have strong and large- later stage of phenology. This reduces preparationscale consequences far from the actual site of time for nest building and egg laying and con-change. The current increasing demand for biofuels strains the ability of the birds to adjust chick feed-to replace fossil fuels in the industrialized world ing to prey availability (Both and Visser 2001).influences production systems in second and thirdcountries. As a consequence, the increasing de- 7.2.2 Linkages between subsystems:mand of biofuels in Europe or the USA will influ- aboveground–belowground interactionsence food production in other parts of the world.This competition for land between food and biofuel Aboveground and belowground compartments ofproduction influences food prices and changes the terrestrial ecosystems traditionally have been con-traditional competition between land needed for sidered separately, but now their interlinkages arefood production and for biodiversity conservation. receiving increased interest (Wardle et al. 2004).Therefore, the desire of human society to counter its Community composition and community process-impact on climate may lead to food limitation in the es in each linked subsystem depend strongly on thethird world. adjacent communities and often there is mutual Local subsidies may also disrupt remote commu- interference. These dependencies can be caused bynity interactions in nature. The increased use of organisms that spend part of their life cycle in thenitrogen fertilizers in the mid-west of the USA is soil and the other part above ground, such as manycorrelated with the increased abundance of mid- root-feeding insects, which mate and dispersecontinent snow geese. Most likely, agricultural above ground. Also, some carnivorous inverte-food subsidy to the geese during their overwinter- brates are supposed to use belowground prey sub-ing in the mid-west indirectly destroys the breeding sidies to survive the absence of aboveground preys,grounds in the Canadian Arctic and sub-Arctic; for example in between two agricultural crops (Bell
  • 100. 86 APPLICATIONSet al. 2008). Plants play a special role in above- benefit. Aboveground–belowground interactionsground–belowground interactions, as they spend also influence plant defence. For example, indirectmost of their life above as well as below the soil defence of plant shoots against caterpillars by para-surface. Plants influence aboveground and below- sitoid wasps is reduced when cabbage root fliesground communities by modifying the abiotic en- (Delia radicum) are present in the soil. The rootvironment, providing structure and resources, and flies reduce the amount of volatiles which normallyby antagonistic interactions involving direct and are released by the shoots and attract parasitoids.indirect defences (Wardle et al. 2004). The presence of root flies enhances the production Traditionally, plant community interactions have of repelling compounds, which distract parasitoidbeen explained primarily by variation in the abiotic wasps (Soler et al. 2007). Aboveground interactionsenvironment. Interspecific competition and above- can also be influenced by changing the entire soilground biotic interactions with vertebrate and in- community composition. For example, in multi-vertebrate herbivores, pathogens, pollinators and species grassland mesocosms, soil nematodes en-other mutualists more recently have been recog- hance the survival and quality of abovegroundnized as influencing plant community structure. mites, resulting in enhanced aphid control by theirThe awareness that plant community interactions natural enemies, aphid parasitoids. The result isare influenced by interactions between plants and that in the presence of nematodes, but not in thesoil organisms is even more recent, whereas interest presence of soil microorganisms, both bottom-upin interactions between plants and below- and and top-down control of aboveground aphids isaboveground multitrophic communities is less intensified (Bezemer et al. 2005). These examplesthan a decade old (van der Putten et al. 2001). show that soil communities can play a key role in Plants interact with soil organisms through exu- the management of production ecosystems; thedates, dead organic material (litter) and through question now is how to translate these findingsliving tissues. Soil organisms can reduce plant into applications.growth (via nutrient immobilization, herbivory,pathogenesis) or enhance plant growth (via symbi- 7.2.3 Consequences for application: findosis, nutrient availability, enhancement through the remote cause of local effectsmineralization) and these processes, and the organ-isms responsible for them, interact (Wardle et al. The existence of linkages across ecosystem or sub-2004). Besides primary metabolites, plants also con- system boundaries has major consequences fortain secondary metabolites, which are involved in management of agricultural ecosystems, as well asplant defence against natural enemies. In fact, a full for understanding and predicting changes in natu-understanding of the ecology of plants cannot be ral ecosystems. The presence and diversity of spe-attained without including their interactions with cies, the abundance of populations, the interactionsboth aboveground and belowground organisms. and organization of communities and the function- Abiotic drivers, such as climate and soil type, ing of ecosystems all may depend on events thatwill have strong predictive power about the com- take place in adjacent populations, communities orposition and dynamics of plant communities on ecosystems. Although conservation tends to be fo-large spatial and temporal scales. However, at cused on red list species or species with iconicsmaller spatial and temporal scales, plant commu- value, conservation requires a more integrated con-nity processes will be much more strongly influ- sideration of community interactions. As many ap-enced by aboveground–belowground community plied problems have remote causes, local solutionsinteractions. For example, aboveground vertebrate will not be effective in the longer term. This appliesherbivores enhance root exudation, which fuels soil to many global human-induced changes, such asdecomposer systems resulting in enhanced miner- land use change, climate change, as well as toalization and increased plant nutrient availability biological invasions. Solutions for local problems(Bardgett et al. 1998). Therefore, vertebrate grazers require insight in processes that exceed local spatialmay enhance primary production to their own and short-term temporal scales. For example,
  • 101. APPLICATIONS OF COMMUNITY ECOLOGY APPROACHES 87counteracting and mitigating impacts of exotic spe- question is whether ecosystems can be restored at all,cies requires insight into how these species are con- because they may have become unalterably influ-trolled in their native range and how these species enced by their history. The soil plays a crucial rolehave evolved during their history of introduction. as the main source of historical legacy effects that mayOn the other hand, local changes, varying from constrain restoration efforts (Fig. 7.1). Soil legacychanged agriculture or industrial discharge to effects cause a major difference between secondarytrade or climate conventions, can have wide and succession, which concerns the transformation fromremote implications. one ecosystem into another (Holtkamp et al. 2008), and primary succession, where ecosystems develop from bare soil (Neutel et al. 2007).7.3 Community interactions and land use One of the main explanations for many exampleschange of unsuccessful ecological restoration is the deple- tion of the soil seed bank and dispersal limitation of7.3.1 Land use change, predictability and plant species from reference sites (Bakker and Be-major drivers of secondary succession rendse 1999). If plant propagules are unlimited,Land use changes are often driven by external factors, land conversion requires a switch from competitionfor example overproduction at the world market, or for light to competition for soil resources (Tilmanopening up of trade barriers, which lead to land 1982). According to the standard view, highestabandonment when prices drop, or to (intensified) plant diversity is obtained at intermediate soil fer-cultivation when prices rise. Land use changes often tility (Al Mufti et al. 1977). This view, albeit devel-go along with altered physical–chemical inputs, such oped by surveying across ecosystems, has providedas soil tillage, fertilization and hydrology measures. the major rationale for restoration in industrializedThese sudden land use changes can have enormous countries, where restoration takes place on relative-impact on species, community interactions and eco- ly nutrient-enriched soils (Marrs 1993). Mowing onsystem processes. Whereas there is extensive experi- wet soils or grazing on relatively dry productiveence with how to bring land into cultivation, soils may help to reduce nutrient availability (Olffecosystem restoration is of a much more recalcitrant and Ritchie 1998). Selective grazing of the dominantnature. Restoration is a long-term process that often forbs and grasses, which have the highest quality,starts with land abandonment and restoring former provides indirect advantage for the rarer, slow-hydrology and nutrient cycles. Simple steering para- growing and poorly competitive plant species. In-meters that enable switching from one state to the troducing aboveground herbivores at mild stockingother (Suding et al. 2004) probably do not exist. One rates is a practice that is often applied in nature Nature Agriculture Root pathway Plant roots Root pathway Bacterial decomposition Bacterial decomposition Exudates and detritus Fungal decomposition Fungal decompositionFigure 7.1 Energy transformation in three belowground pathways: the root–root feeder pathway, the detritus–bacteria–bacterivores pathway and the detritus–fungi–fungivores pathway in agricultural soil and a natural soil.Conversion of land use from agricultural to post-agricultural involves reorganization of the energy channels in the soilfood web according to arrow thickness: the thicker the arrow, the larger the energy flow.
  • 102. 88 APPLICATIONSconservation, although the value of this practice is mechanical soil disturbance (Van der Wal et al.subject to debate among ecologists. 2006). The dynamics of soil pathogens and root Besides reducing and concentrating nutrients, herbivores will change from whole field oscillationherbivores also influence vegetation structure and patterns coinciding with crop presence to finer spa-succession. Aboveground herbivory does not nec- tial and temporal patch dynamics as related to theessarily retard succession in all stages. Above- spatio-temporal dynamics in vegetation composi-ground herbivores may slow down succession in tion. Therefore, the soil community will acquiresome stages, but, initially, they may accelerate suc- (as well as generate) more spatial complexity,cession (Davidson 1993). Aboveground herbivores whereas the food web will develop from a fast to ainteract, thereby facilitating or inhibiting each slower cycling of nutrients and energy (Fig. 7.1).other’s effects. In an exclosure experiment, small Plants interact constantly with other organisms inherbivores (rabbits, voles) did not have consistent soil- and aboveground communities and these feed-effects along a productivity gradient, whereas cattle back effects can drive succession. Feedback interac-increased plant diversity at high-production sites, tions with soil decomposer organisms proceedbut reduced plant diversity at low-production sites indirectly, through root exudation, decaying roots(Bakker et al. 2006). Large vertebrate herbivores, and litter. These interactions most likely take placeespecially cattle, are easy to manage and their ef- at a slower rate than feedback interactions with sym-fects are relatively predictable. Invertebrate herbi- bionts and root pathogens, which have a more directvores can be controlled far less well and and intimate association with plant roots. Feedbacksinteractions between vertebrates and invertebrates between plants and their soil community can behave been rarely explored (Tscharntke 1997). In a assessed by growth experiments and the net effectsstudy focusing on sap-feeding insects, Schmitz et al. range from negative to positive. These effects can(2006) showed that the sap feeders could be con- develop over several months. Negative feedbacktrolled by both bottom-up and top-down forces, leads to coexistence, whereas positive feedback re-which change the rate of succession due to abrupt sults in dominance of plant species (Bever 2003). Theshifts in trophic control. sign of plant–soil feedback varies along successional sequences. In a 35-year-old chronosequence, early successional plant species experienced negative soil7.3.2 Secondary succession from an feedback, mid-successional species had neutral feed-aboveground–belowground perspective back and later successional plant species had posi-When secondary succession begins in post-agricul- tive feedback from the soil community (Kardol et al.tural fields, the soil already contains a well-devel- 2006). This suggests that early succession soil com-oped food web that includes decomposers, munities enhance and later succession soil commu-bioturbators, symbionts, herbivores and pathogens. nities slow the rate of secondary succession.Secondary succession, therefore, requires the trans- Moreover, plant–soil feedbacks of early successionalformation of an arable soil food web into a soil food plant species make them less competitive and alsoweb characteristic of a more natural system that cause a legacy effect to mid-successional plant spe-receives less soil disturbance, no mineral nutrients cies. Such feedbacks can result from soil microorgan-and other fertilizers, no biocide spraying and no isms (Kardol et al. 2007) as well as from soil faunamonocultures of crop species. As a result, the soil (De Deyn et al. 2003).food web will transform from a bacteria-dominated In post-production grassland soils, the soil faunato a more fungi-dominated soil food web, although selectively reduces the abundance of early succes-this process is not necessarily linear over time (Van sional plant species, as well as that of the dominantder Wal et al. 2006). The role of mutualistic sym- plants (De Deyn et al. 2003). While soil fauna en-bionts (especially of mycorrhizal fungi) and of bio- hanced the rate of succession, it increased evennessturbators (earthworms) increases when succession in plant community composition (De Deyn et al.proceeds. The initial changes in community compo- 2003). Nutrient addition increases dominance ofsition can be partly due to the absence of regular fast-growing plant species, but the soil community
  • 103. APPLICATIONS OF COMMUNITY ECOLOGY APPROACHES 89counteracts the effects of nutrient addition (De Enhanced exposure to aboveground and below-Deyn et al. 2004). These inoculation experiments in ground invertebrates did not necessarily increasemesocosms support exclusion experiments in the plant community diversity. We know now thatfield by Brown and Gange (1992) and Schadler ¨ aboveground and belowground biodiversity may,et al. (2004), who used selective biocides to elimi- at least to some extent, interact and also that second-nate belowground and aboveground insects. Early ary succession needs to be perceived from an above-secondary succession slowed down when using soil ground–belowground community perspective. ¨insecticides (Schadler et al. 2004) and accelerated However, this area of research still needs more ex-when using foliar insecticides (Brown and Gange amples for convincing generalizations to be made.1992). Depending on the type of ecosystem, succes-sion may slow down or speed up due to below- and 7.3.3 Consequences for restoration andaboveground activities respectively (Davidson conservation1993; Kardol et al. 2006), but these results showthat ecosystem restoration is profoundly influenced Biodiversity restoration and conservation clearlyby above- and belowground biotic interactions and require an aboveground–belowground approach.not by changes in the abiotic environment alone. Secondary succession is influenced by abiotic con- Aboveground–belowground community interac- ditions, as well as aboveground and belowgroundtions enhance temporal variation in natural commu- community interactions. Probably, these abioticnities (Bardgett et al. 2005). For example, in an and biotic interactions act at different spatial andextensively grazed pasture, cattle facilitated ant ac- temporal scales and there will be hierarchies intivity, resulting in ant mounds with relatively path- controlling effects. For example, climate determinesogen-free soil, which was rapidly colonized by the type of biome, soil type determines which veg-pathogen-sensitive plant species (Blomqvist et al. etation types can occur, and aboveground verte-2000). The result was a spatio-temporal mosaic of brate herbivores determine vegetation structureant mounds and plant populations. Similarly, my- and patch structure in the landscape. At the smal-corrhizal fungi in the plant roots can counteract lest spatial scale, aboveground invertebrate herbi-aboveground insect damage on plants (Gange et al. vores, plant pathogens and plant–soil organism2003) and the individual effects of soil nematodes, interactions ultimately determine the compositionroot-feeding wireworms and aboveground grass- and dynamics of the plant community by influen-hoppers resulted in non-additive effects on plant cing plant competition, performance and abun-community composition (Van Ruijven et al. 2005). dance (Fig. 7.2). Conservation and restoration Factor Scale Climate Continent Soil type Region Hydrology Watershed-field Aboveground Landscape-field biotic interactions Belowground Plot biotic interactionsFigure 7.2 Hierarchies in influences of plant community structure and composition. Climate influences plantcommunities at continental scales, whereas belowground interactions influence plant communities at the smallest (plot,< 1 m2) scale.
  • 104. 90 APPLICATIONSecologists are challenged to integrate all these abi- For example, when species-rich plant communitiesotic and biotic influences in their management de- prevent the establishment of invaders, how docisions. However, there is relatively little known these plant communities initially get and maintainabout how to influence the composition and devel- their high species richness? Moreover, species-richopment of soil communities (Kardol et al. 2008). plant communities not only provide more potentialCommunity ecologists face the enormous challenge competitors of exotic plant species. They also mayof integrating all of these various disciplines to harbour more natural enemies, which increases theimprove current concepts and to develop novel chance that there are local enemies suitable to at-ones, and to facilitate environmental planning and tack invading exotic plants. Therefore, the issue ofconservation decision-making. biotic resistance, which is crucial to predict the long-term development of invasions, needs more multidimensional and multitrophic approaches7.4 Biological invasions than have been taken thus far. The concept of enemy release is supported by7.4.1 Community-related hypotheses that examples of successful control of invasive exoticexplain biological invasions weeds by introduced biological control agents.Biological invasions of exotic species are causing Other studies that support the ER hypothesis havemajor problems worldwide, because of their dis- compared introduced and native plant species inproportional abundance, negative effects on local their amount of natural enemy species. For exam-biodiversity and alterations of ecosystem processes ple, in a review including more than 300 introduced(Williamson 1996). There are many overviews and plant species, Mitchell and Power (2003) showedreviews that discuss biological invasions in depth. that non-native species had fewer abovegroundHere, I will focus on biological invasions in relation pathogen and virus species than comparable nativeto aboveground and belowground community in- plant species. The exotic plants also had fewerteractions. Community ecology theory would pre- pathogens in their new than in their native range.dict that the disproportional abundance of exotic Exotic plant species that were actively dispersed byplants is caused by altered bottom-up and top- humans, for example crops and ornamental plants,down interactions in the novel environment. Unoc- had more pathogen and virus species than othercupied niches can provide exotic species with abun- exotic species. It is not clear why this is; possibledant resources, although some studies argue that usage enhances the chance of exposure to potentialmany invasive species occupy the same niches as enemies, or used exotic species did not pass thethe native species (Scheffer and Van Nes 2006). selection processes as strongly as non-used exoticAlternatively, invasive exotic species may have species. Numbers of enemy species of course arenovel traits, such as the capacity to fix nitrogen not indicative of enemy effects, but these results at(Vitousek et al. 1987) or novel chemical properties least show that exotic plant species are less exposedthat do not have a coevolutionary history with to aboveground pathogen and virus species thanother organisms in the new range (Callaway and natives (Mitchell and Power 2003). There are simi-Ridenour 2004). lar studies on exposure of exotic plants to above- Prominent hypotheses proposed to explain ground insects. They show that exotic invadersbiological invasions from a community perspective have less specialist feeders, but that they still canare biotic resistance (BR) and enemy release (ER) have generalists (Jobin et al. 1996; Memmott et al.(Keane and Crawley 2002). BR was proposed by 2000).Elton (1958), who concluded that species-rich com- All of above-mentioned examples of enemy re-munities are invaded less by exotic species than lease and biotic resistance concern release fromspecies-poor communities. There is evidence for aboveground enemies. Klironomos (2002) showedand against biotic resistance (Stohlgren et al. 1999). that five exotic plants in an old field in Canada exertCertainly, much more work is needed to under- neutral soil feedback, suggesting that these exoticstand when biotic resistance can prevent invasions. plants may have become released from natural
  • 105. APPLICATIONS OF COMMUNITY ECOLOGY APPROACHES 91soil-borne enemies. As soil feedback indicates a net view from the world’s highest peak. Here, I willeffect, it could be that the neutral effect was due to point out some new viewpoints and conclude thatless pathogenic, or to more symbiotic, activity, as biological invasions provide an enormous chal-symbionts may overrule the effects of pathogens. lenge to community ecology both from a funda-However, in the same study, native dominant mental and from an applied perspective.plants also had neutral or even positive feedback A substantial amount of exotic plants contain(Klironomos 2002), so that the exotic plants could chemicals that are novel then native species of thehave shown the same soil feedback as in their na- invaded range (Cappuccino and Arnason 2006).tive range, when they were dominants. Enemy re- These novel chemicals may influence plant–plantlease includes two components. The first is that interactions by allelopathic compounds (Callawayexotic species escape from their native enemies and Ridenour 2004), or they may reduce the inocu-and the second is that the exotic species have less lation potential of root symbionts as arbuscularenemy exposure in their new range. Therefore, such mycorrhizal fungi in invaded habitats (Stinsonstudies need a comparative approach, including et al. 2006). These views contrast with that of theboth the native and non-native ranges (Hierro evolution of increased competitive ability (EICA)et al. 2005). ¨ (Blossey and Notzold 1995), which assumes that Further evidence on ER from soil pathogens defences impose trade-offs with growth whenstems from Prunus serotina (black cherry), which is plants are exposed less to natural enemies. Howev-invasive in Europe and native in the USA. In Eur- er, the novel weapons hypothesis has, thus far, notope, soil feedback was neutral, whereas, in the been widely tested for a range of plant species,USA, the soil feedback was negative (Reinhart whereas EICA is contradicted in a number of stud-et al. 2003). In that example, there is still a possibility ies (Wolfe et al. 2004). Tests of these hypothesesthat symbionts provided an overwhelmingly posi- usually lack the inclusion of negative controls,tive effect, so that the neutral soil feedback effect which would be exotic species that do not becomecould have been due to highly effective symbionts. invaders. These are the unseen majority of non-In a study on Kalahari savanna grasses, however, native species, according to the 10s rule of William-fungi were isolated from a native grass (which ex- son (1996). The problem is, however, species that dopressed negative soil feedback) and from an exotic not become established will not be of use in ecolog-invader (which had neutral soil feedback). The soil ical studies. Therefore, it would be possible to in-fungi from the native species were pathogenic to clude mild invaders in the studies and examinetheir own host, but not to the exotic plant, whereas their means of control. Nevertheless, it would stillsoil fungi from the exotic plant were not pathogenic be possible that these species have not yet reachedto the native and non-native plant species (van der their stage of invasiveness, or that they are alreadyPutten et al. 2007). The exotic species was not exam- over the top of their invasiveness. Including suchined in its native habitat. Therefore, these results all ‘false positives and negatives’ would enhance thepoint to ER from soil pathogens; however, the ulti- objectiveness of ecological studies on causes of in-mate test, specifically including pathogen species vasiveness.instead of treating soil feedback as a ‘black box’, is In the debate on EICA and novel weapons, moststill lacking. studies implicitly assume that plant defences are mainly direct. However, indirect defence, through recruiting the enemies of your enemies, may play7.4.2 Mount Everest or tip of the iceberg? an important role in plant abundance. Loss of theIn spite of the enormous research effort focused on ‘third trophic level’ has rarely, if ever, been consid-biological invasions, most studies still are correla- ered in invasion studies. In contrast, it has beentive (Levine et al. 2003). Moreover, BR and ER have argued that exotic plants may have easy access tomany more dimensions than have been explored symbionts, such as pollinators or mycorrhizal fungithus far. Therefore, it seems as if we have hit the (Richardson et al. 2000). This hypothesis has nottip of the iceberg, rather than getting a panoramic been rigorously tested and the evidence for the
  • 106. 92 APPLICATIONS Mycorrhizal fungi sions provides an enormous task for future re- – search. As soon as invasive species become abundant, it is already often too late for their com- + plete eradication. Introducing biocontrol agentsInvasive plant Native plant – may be helpful for some invasive species, but the – – + risk that biocontrol agents may switch to feeding on Local soil pathogens native plants cannot always be excluded. More- + over, the long-term effectiveness of biocontrol Novel allelochemicals agents could be limited. In a comparison of biocon-Figure 7.3 Three pathways of how invasive plants can trol studies, Burdon and Marshall (1981) concludedindirectly influence native plants: by reducing, or that biocontrol by aboveground insects and patho-suppressing, local symbionts, for example arbuscular gens was least effective against introduced annualmycorrhizal fungi; by enhancing local pathogens or plant species and most effective against introducedparasites; or by excreting allelochemicals that are new to perennial and clonal plants. They concluded that anthe local plants or enemies. The consequences of thesethree effects are that the invasive plant has an indirect annual lifestyle probably enables rapid evolution ofadvantage over native plants, as indicated by the dotted defences against pathogens or herbivores. There-arrow. Arrow thickness indicates the strength of the fore, managers should consider alternatives, sucheffect. as exploring how control agents already present in the invaded range might be used, or cultivated, to switch to and control exotic species. In doing so, thepositive effect of arbuscular mycorrhizal fungi is full array of enemies, including viruses, pathogens,quite contrasting, whereas rigorous inoculation herbivorous invertebrates, both below and abovestudies are lacking (van der Putten et al. 2007). ground, as well as large herbivores could be con- Some exotic plants may accumulate local viruses sidered. There is often little known about whatand pathogens, thereby having an indirect negative controls exotic species in their native range. How-effect on native plant species (Fig. 7.3). For example, ever, at the same time it is unclear whether these, orexotic plants in Californian grasslands accumulate different, controls are required to reduce the abun-local viruses without suffering from them. These dance of the exotic species in their new range.viruses spread into the surrounding native plantcommunities and cause strong negative effects(Malmstrom et al. 2005). Similarly, the invasive 7.5 Discussion, conclusions andtropical shrub Chromolaena in India accumulates perspectivessoil pathogens that have strong negative effects on The promise of ecology as a science lies in devel-surrounding local plants (Mangla et al. 2008). Mod- oping sufficient knowledge to allow us to under-els show that such indirect effects could render stand and predict how individuals, species,exotic plants invasive, even when these plants suf- communities and ecosystems will respond to myr-fer to some extent from mild biotic resistance iad environmental changes. Applied community(Eppinga et al. 2006). Although these invasive spe- ecology should be at the forefront of developingcies make use of local enemies, their effects on such predictive power and testing the value of thenative communities could be quite similar to those predictions. Are there general laws and how wellcases where novel weapons have been proposed to can we predict? How well do relationships devel-cause invasiveness. oped in temperate zones, where most ecological studies are carried out, apply to tropical or boreal regions? These and many other applied issues are7.4.3 Conclusions and consequences for open for discussion. I will focus on a subset ofmanagement these issues, show opportunities for end-usersThe role of community ecologists in preventing, and stakeholders, and point out some areas forcombating or mitigating effects of biological inva- future work in ecology.
  • 107. APPLICATIONS OF COMMUNITY ECOLOGY APPROACHES 93 One main message of community ecology to its conditions to become favourable for new intro-applied end-users and stakeholders is that many duced exotic species. This, together with rangelocal problems can have remote causes. Whether shifts due to climate warming, may elevate thethis concerns local overgrazing, aboveground future incidence of biological invasions. Especiallypests or biological invasions, the problems quite when range shifts release plant species from below-often originate at a different place from where the ground (van Grunsven et al. 2007) or belowgroundproblems occur. Community ecology could pro- and aboveground enemy attack (Engelkes et al.vide help in solving these applied problems by 2008). The same has been shown for higher trophicanalysis of the complex interactions in which level interactions, such as between insects and theirmany species are involved. In some cases, such as enemies (Menendez et al. 2008). These futureecosystem restoration, original key interactions changes will require considerable attention fromneed to be determined and the key players need community ecologists.to be brought into contact. Restoration ecology still The most striking examples of remote causes formakes very little use of community ecological in- local problems are definitely those where speciessights, except the use of large herbivores. Major move over large distances, such as the geese thatomissions concern the involvement of inverte- migrate from mid-west USA to the Arctic zone. Tobrates, of soil communities and of interactions of solve these and to prevent other problems, it isplant communities with above- and belowground crucial to thoroughly analyse all consequences soorganisms. as to not create additional problems while attempt- Biological invasions are a global problem. In- ing to solve an initial one. Community ecologycreasing globalization of the economy, international should be at the forefront in these analyses, giventourism and enhanced emigration–immigration en- the complexities involved. Interestingly, these ap-sure that more and more non-native species will plied questions could also be used as learning op-move to novel areas where they could become in- portunities, because the shortcomings of thevasive. Predictive systems are necessary to forecast predictions will reveal the limits of our knowledge.whether certain species have the potential to be- Therefore, community ecologists should take ancome invasive in their new habitat and, when so, active role in analysing causes of environmentalto identify these species for import limitation. The and biotic change and in forecasting consequencesmain problem is twofold. First, only a minor frac- of new policy and management strategies. Thetion of all introduced species become invasive, so challenge is huge, and includes biobased economy,that the predictions need to detect $ 0.1% of all biofuels, sustainable agriculture, biodiversityspecies as possible invaders. Usually, this is a restoration and conservation, climate warming,range where statistics are uncertain. Second, the and the consequences of genetically modifiedcurrent rapid climate warming may change local organisms.
  • 108. This page intentionally left blank
  • 109. CHAPTER 8 Sea changes: structure and functioning of emerging marine communities J. Emmett Duffy urgent need for rigorous applied community eco-8.1 Introduction logy – we need to understand how complex inter-The earth is in the midst of a global-scale, uncon- actions within real ecological communities playtrolled experiment involving human alteration of out on the large space and timescales relevant toboth the abiotic resources that support ecosystems sustainable management, and how they mediateand the trophic linkages that control their structure stressor impacts on ecosystems and their provisionand function. Inputs of inorganic nutrients have of services to humanity.increased substantially worldwide, causing major Much of our knowledge of the organization andbottom-up shifts in ecosystems (Cloern 2001). dynamics of communities comes from mathemati-Abundances of large vertebrates have been sub- cal theory and controlled experiments. These ap-stantially depleted both on land (Dirzo and Raven proaches have advanced ecology tremendously2003; Cardillo et al. 2005) and at sea (Pauly et al. over the last 50 years or so. But there is also a widely1998; Steele and Schumacher 2000; Jackson et al. recognized trade-off between the clarity and ele-2001; Myers and Worm 2003; Hutchings and gance of theory and experiments and the complexi-Reynolds 2004), resulting in systematically altered ty of the real world (e.g. Carpenter 1996; Oksanenfood web structure (‘trophic skew’; Duffy 2003; 2001). For example, the classical experiments ofByrnes et al. 2007), and exotic invasions are altering community ecology are generally limited to smallcommunity composition locally and homogenizing spatial and temporal scales, and are logisticallycommunities globally (Sax et al. 2005). Finally, prohibitive for the large mobile animals of specialrising CO2 inputs from fossil fuel combustion are conservation interest due to their status as strongcausing climate warming and acidification of the interactors (Soule et al. 2005). Thus, successfullysurface oceans (Orr et al. 2005); these and associated applying principles from academic communitychanges in precipitation and circulation are shifting ecology to real-world management and conserva-species ranges, seasonal cycles and interactions tion remains a daunting challenge. The ultimate test(e.g. Stachowicz et al. 2002b; Voigt et al. 2003; Schiel of ecological theory is its ability to explain, and toet al. 2004; Winder and Schindler 2004; Hays et al. forecast, community responses to environmental2005; Perry et al. 2005; Parmesan 2006). Together, change in the real world (e.g. Clark et al. 2001).these processes are producing novel or ‘emergent In this spirit, I focus here on one major humanecosystems’ (Hobbs et al. 2006) assembled in altered impact, fishing, as a case study to evaluate the suc-habitats from species that may have had little or no cess of fundamental principles of community eco-evolutionary history of interaction. The accelerating logy in understanding impacts of environmentalpace and synergism of these changes create an change in marine ecosystems. I ask whether impacts 95
  • 110. 96 APPLICATIONSof human predation are consistent with expectations system has relatively low impact on the habitatarising from theoretical and experimental communi- (although ‘ghost’ nets can continue to ensnare fish-ty ecology, particularly regarding trophic cascades, es indiscriminately long after they are lost or aban-the influence of biodiversity on ecosystem stability doned); this is in contrast to most human impactsand the nature of regime shifts among alternate on land (Wilcove et al. 1998) and in benthic habitatssemi-stable states. I explore what the impacts of (Watling and Norse 1998), where habitat destruc-human predation on marine communities can tell tion or modification confounds species removalsus about how and whether basic principles in com- with other impacts (Srivastava and Vellend 2005).munity ecology extrapolate to complex ecosystems Third, the major commercial value of fisheriesat large spatial and temporal scales, and thus assess means that there is a large body of detailed datatheir value, if any, for conservation and manage- on which, when and how many fish have beenment. removed from the oceans (FAO 2007). Finally, the growing database on ecological changes within ma- rine protected areas (MPAs) offers important large-8.1.1 Fishing as a global experiment scale experimental controls against which to evalu-in community manipulation ate the effects of fishing (Halpern and Warner 2002;Of the several human impacts on marine systems, Micheli et al. 2004).the strongest and most pervasive is the continuousremoval of large quantities of animal biomass 8.1.2 Physical forcing and the uniquenessthrough fishing. In general, fishery management of marine ecosystemsaims to reduce a fish stock to $50% of its unfishedbiomass in order to maximize productivity. In prac- The ecology of marine communities and their re-tice, many stocks are fished well beyond this target sponses to perturbations are strongly influenced by(Hilborn et al. 2003a; FAO 2007). Modern fisheries, the unique nature of the marine environmentincluding both landings and by-catch, currently (Steele 1985, 1991). The fundamental physical dif-consume 24–35% of global marine primary produc- ference between terrestrial and (pelagic) aquatiction in the continental shelf and major upwelling systems is the greater density of the liquid mediumareas (Pauly and Christensen 1995). Thus, any at- of water (Strathmann 1990), which has three impor-tempt to understand modern marine communities tant consequences for understanding the ecologicalmust reckon with the fact that humans are now the structure and dynamics of terrestrial compareddominant predator throughout the world ocean. with pelagic ecosystems. First, in water, buoyancyThe many direct impacts of human exploitation on allows primary producers (and other organisms) tomarine fish populations and communities are well float and obviates the need for large, expensive,documented (Jennings and Kaiser 1998; Pauly et al. metabolically inert structural tissues required to1998; Jackson et al. 2001; Myers and Worm 2003). compete for light on land. Thus, the dominantBut this intense predation is expected to have ex- marine pelagic autotrophs are microscopic, fast-tensive indirect effects on marine communities as growing and highly nutritious (floating Sargassumwell. Indeed, effects of fishing provide a uniquely accumulations being a conspicuous exception in theuseful case for testing how community models Atlantic gyre). Consequently, compared with land,scale up to real ecosystems, for several reasons. marine ecosystems show higher grazing rates andFirst, fishing impacts have followed similar pat- production: biomass ratios, a much larger faction ofterns in many regions (Jennings and Kaiser 1998; primary production grazed and more efficient con-Pauly et al. 1998; Jackson et al. 2001; Hilborn et al. version of production to herbivore biomass (Steele2003a; Myers and Worm 2003), providing a degree 1991; Cebrian 1999; Shurin et al. 2006). The higherof replication and potential generality. Second, for growth rates, nutritional content and vulnerabilitypelagic fishes specifically, harvesting provides a of marine autotrophs to grazing in turn haverelatively ‘clean’ test of community manipulation important consequences for the structure andin that removal of individuals or species from the functioning of marine communities, including
  • 111. STRUCTURE AND FUNCTIONING OF EMERGING MARINE COMMUNITIES 97stronger top-down control generally and trophic sponses of primary producers within days or evencascades in particular (Shurin et al. 2002), and an hours in the sea. Thus, many marine communities,altered distribution of biomass among trophic levels. especially pelagic communities, tend both to beFor example, pristine coral reefs support (or did more sensitive to disturbances and to reboundhistorically) large populations of apex predators more rapidly after disturbance than terrestrial(Friedlander and DeMartini 2002) but often little ones. Together with strong trophic interactions,visible plant life, in stark contrast to the rain forests this sensitivity should enhance the ability to detectfound above the tide line. In short, top-down control predicted processes and patterns of communityand trophic transfer are more efficient in the sea, and regulation on large spatial and temporal scales.marine biomass pyramids tend to be (or were,primevally) less bottom-heavy than those on land(Odum 1971; Del Giorgio and Gasol 1995). These 8.2 The changing shape of marine foodstrong trophic interactions should enhance the websability to detect predicted responses to food web 8.2.1 Conceptual backgroundalteration in marine systems relative to terrestrialones. Like other optimal foragers, humans generally tar- The second major consequence of water’s density get large and abundant prey preferentially, all elseand buoyancy of biomaterials is the greater impor- being equal. Fishing thus represents not only atance in the sea of advection of materials (inorganic strong, but also a selective press perturbation onnutrients, detritus) and organisms. In certain terres- marine communities, which has been sustainedtrial systems, migrating birds, mammals and even for decades and even millennia in some areasinsects can transport large quantities of materials (Wing and Wing 2001; Barrett et al. 2004; Lotzeover long distances (Polis et al. 1997). Nevertheless, et al. 2005). The responses of marine communitiesconstantly moving currents make marine ecosys- to this strong top-down influence depend on bothtems more open on average than terrestrial sys- ‘vertical’ components (food chain length, omniv-tems. Although pelagic marine communities and ory) and ‘horizontal’ components of biodiversitypopulations are more highly structured than (species or functional group richness and composi-might be expected from the superficially featureless tion within trophic levels), and their interactionsappearance of their habitat, many large predators (Duffy 2002; Duffy et al. 2007; Fig. 8.1). These innevertheless can swim between ocean basins, and turn are mediated by organismal traits. Specifically,larvae of many species can drift for hundreds of focusing on the key related traits of body sizekilometres before settling. These features mean (Woodward et al. 2005), feeding traits that deter-that between-habitat subsidies, source-sink dynam- mine trophic level (Pauly et al. 1998) and life historyics and gene flow tend to be considerably higher, on (Jennings et al. 1998) reveals several apparentlyaverage, in marine communities than on land or consistent patterns in the changing structure andfreshwater. They also suggest that simple models functioning of marine ecosystems, and clarifies theof community structure and dynamics that implic- mechanisms involved.itly assume closed systems may be less likely to An important consequence of the complex pelag-apply in the sea, where metacommunity ap- ic life histories of many marine animals is that theyproaches will probably prove fruitful (see Leibold pass through a large range of body size, and multi-et al. 2004; Chapter 5). ple trophic levels, during their lifetimes. Most new- The third important consequence of water’s den- born marine animals receive no parental care – insity, stemming from both the microscopic size of stark contrast to the birds and mammals that domi-most primary producers and system openness, is nate upper trophic levels on land – and are there-the much closer and more rapid coupling between fore highly vulnerable to predation, starvation andphysical drivers and biological processes in marine abiotic stress. Importantly, the larvae of apex pre-(pelagic) systems than those on land (Steele 1985). dators frequently serve as prey of fishes that theFor example, nutrient loading can produce re- apex predators hunt as adults. Thus, compared
  • 112. 98 APPLICATIONS 3 C1 Top carnivore 01 02 Omnivores ‘Vertical’ biodiversity H1 H2 2 Herbivores 1 P1 P2 P3 P4 P5 P6 Edible plants Inedible plants ‘Horizontal’ biodiversityFigure 8.1 Components of horizontal and vertical diversity in a schematic food web. Vertical diversity includes averagefood chain length and degree of feeding from one (e.g. herbivores) compared with more than one (omnivores, andcannibalistic top carnivore) trophic level. Horizontal diversity includes number of functional groups and degree of feedingspecialization (e.g. species H2) compared with generalism (e.g. H1, O1 and O2). Reproduced with permission from Duffyet al. (2007).with terrestrial food webs, pelagic marine food dance and trophic level of species, with primarywebs tend to be more strongly structured by body producers being both much smaller and moresize than by species, to contain more cannibalistic abundant than predators. These relationships areand omnivorous species (Dunne et al. 2004) and to presumably also characteristic of pelagic marinecontain more ‘loops’. This shifting ontogenetic systems, which are similarly based on microscopicniche identity is well illustrated by stable nitrogen algal producers. In the pelagic system, species rich-isotope data for North Sea fishes, which show that ness also shows a pyramidal distribution with fewthe trophic level of a species, averaged across onto- apex predators and many primary producer speciesgenetic stages, was unrelated to its maximum size, (Cohen et al. 2003; Petchey et al. 2004). Predators, asbut was strongly related to body size at the individ- distinguished from parasites and pathogens, mustual level (Jennings et al. 2001; Fig. 8.2). Despite generally be larger than their prey (Brose et al.such indeterminacy, however, both experiments 2006), and simple allometric theory dictates that(Menge 1995) and observational data (Williams the few larger, less abundant species high in theand Martinez 2004; Thompson et al. 2007) support food web also are slower growing than basalthe reality of discrete trophic levels, at least at lower species.positions in food webs. Moreover, maximum These considerations have several general impli-trophic level in a system is clearly related to the cations for responses of marine communities topresence of species that can attain large size. human influence (Duffy 2002, 2003; Petchey et al. Cohen et al. (2003) showed for the pelagic food 2004). First, the smaller populations and slow pop-web of Tuesday Lake, USA, that there are consistent ulation growth rates of top predators should raiserelationships among body size, numerical abun- their risk of extinction due to demographic and
  • 113. STRUCTURE AND FUNCTIONING OF EMERGING MARINE COMMUNITIES 99(a) blunting of the trophic pyramid due to proportion- ally greater losses of higher level species (Duffy 20 2003). 18 00) 16 8.2.2 Empirical evidence for trophic skew 15N 0/ ( in the ocean 14 Several lines of evidence illustrate that human im- 12 pacts cause predictable changes in the functional structure of marine communities. The importance 10 of life history traits in mediating responses is 4 6 8 10 12 14 16 illustrated by data from the North Sea. Sustained log2 maximum total mass size-selective fishing during the late 20th century(b) shifted the demersal fish community toward 16 increased aggregate growth rate and decreased average age and length at maturity as smaller, 15 faster maturing species gained in relative abun- 14 dance, while larger, slower growing species declined 00) 13 (Jennings et al. 1999a). A similar pattern was found 15N 0/ ( 12 on fished coral reefs (Jennings et al. 1999b). The decline of apex predators in the oceans has 11 taken on iconic status since Pauly et al. (1998) pre- 10 sented evidence for ‘fishing down the food web’, 9 i.e. a worldwide decline in the mean trophic level of 0 5 10 15 fishery landings, which they suggested resulted log2 body mass from sequential depletion of large-bodied preda- tors. Although this pattern results in part from ad-Figure 8.2 Among North Sea fishes, trophic level (as dition of new fisheries at lower trophic levelsindexed by d15N signature) is unrelated to maximum bodymass of the species (a) but closely related to individual (‘fishing through the food web’; Essington et al.body mass (b). Thus, trophic level is a property of 2006), there is abundant evidence that extinctionindividuals, not species, and increases through ontogeny. and depletion of marine animals are consistentlyReproduced with permission from Jennings et al. (2001). biased toward loss of large animals at high trophic levels (Pauly et al. 1998; Jackson et al. 2001; Dulvy et al. 2003; Byrnes et al. 2007), resulting in broad-environmental stochasticity, and may also lower scale declines in both abundance and body size oftheir resilience to demographic perturbation. Sec- marine predators (Baum et al. 2003; Myers andond, harvesting disproportionately targets large an- Worm 2003; Hsieh et al. 2006).imals near the top of marine food chains (Botsford But marine ecosystems are also increasingly af-et al. 1997; Pauly et al. 1998; Jackson et al. 2001). fected by exotic invasions, which could in principleBecause most ecosystems support few species of counteract the loss or depletion of species (Sax andapex predators, this combination of demographic Gaines 2003). Byrnes et al. (2007) explored this pos-vulnerability and targeted human pressure means sibility by synthesizing data from global tallies ofthat marine (and other) ecosystems under human marine extinctions (Dulvy et al. 2003) and marineinfluence are inherently vulnerable to loss of an invasions in four well-documented areas, and as-entire functional group or trophic levels at the top signing each species a trophic level. They foundof the food web (Jackson et al. 2001; Duffy 2002; that the gains and losses of biodiversity did notDobson et al. 2006). The predicted pattern that re- compensate one another functionally but insteadsults is ‘trophic skew’, a vertical compaction and led to consistent directional change in the shape of
  • 114. 100 APPLICATIONS Original trophic Recorded Recorded distribution extinctions invasions –24% –29% –21% +67% –26% Current distribution Projected distribution (after 5.1% turnover) (after 25% turnover) –14.0% –65.1% –5.4% –24.6% +8.6% +50.0% –0.1% –3.3%Figure 8.3 The changing shape of a coastal marine food web, the Wadden Sea, The Netherlands, based on Byrneset al. (2007). Bars represent successive trophic levels: primary producers (dark grey); herbivores, deposit feeders,detritivores and zooplankton (hatched); omnivorous consumers (light grey); and carnivores and parasites (open).food webs (Fig. 8.3). On a global basis, 70% of have a major influence on overall ecosystem struc-documented extinctions were of predators, while ture and functioning, and can shift the systemmost invasions were at lower trophic levels, trun- between alternate semi-stable states (Sala et al. 1998;cating the typical trophic pyramid into a vertically Bakun 2006).compressed food web dominated by filter-feeders In summary, many marine food webs historicallyand scavengers. Thus, invasion by exotic species had a characteristic ‘shape’, with diversity anddid not counteract the trophic skew caused by abundance generally declining, and body size in-predator depletion but in fact exacerbated it. creasing, with height in the food chain. Human One important exception to the general pattern impacts change the shape of marine food websof declining species richness with trophic height, predictably, tending to reduce average food chainwith important implications for community dynam- length and skew communities toward dominanceics, involves so-called ‘wasp-waist’ ecosystems, in by small-bodied, fast-maturing omnivores, detriti-which one or a few species dominate intermediate vores and suspension-feeders. Trophic skew thustrophic levels, such that their particular biological appears characteristic of human-influenced marinecharacteristics control trophic transfer from primary systems, and altered top-down control should be aproduction to upper levels of the food web (Cury central consequence in the sea.et al. 2000; Hunt and McKinnell 2006). The classicexample involves the herbivorous sardines andanchovies that dominate upwelling ecosystems 8.3 Trophic cascades in the seathroughout the world. Typically, wasp-waist has 8.3.1 Conceptual backgroundbeen used to refer to pelagic ecosystems. But thesepelagic systems also bear some functional similari- What are the broader consequences of the system-ties to benthic systems in which the herbivore tro- atic shortening of marine food chains? Hairston,phic level is dominated functionally by a single Smith and Slobodkin (HSS; Hairston et al. 1960)species of sea urchin. In both cases, factors influen- initiated one of the longest running controversiescing abundance of these key intermediate species in ecology with their assertion that trophic-level
  • 115. STRUCTURE AND FUNCTIONING OF EMERGING MARINE COMMUNITIES 101abundance should be controlled alternately by bot- istic shortening of marine food chains under humantom-up and top-down processes as one descends influence, these generalizations imply that a consis-through successive levels of the food chain. The tent consequence of human impacts will involvechain of indirect effects emanating from top preda- cascading indirect effects of reduced predationtors has since been called a trophic cascade (Paine pressure.1980; Carpenter et al. 1985). Despite the obviousgreater complexity of real communities and the 8.3.2 Evidence for trophic cascadesseveral factors that would appear to work against in open marine systemstrophic cascades (Strong 1992; Polis 1999), manycontrolled experiments in marine and other sys- Marine ecosystems are open, with propagules andtems have supported the general HSS hypothesis, apex predators moving over large distances, andshowing that loss of predators indeed often releases are subject to climate forcing and other influencesprey from top-down control, leading in turn to that could attenuate trophic cascades (Jenningsstrong reductions in the prey’s resources (Menge and Kaiser 1998). Perhaps surprisingly, the expec-1995; Shurin et al. 2002). Similar trophic cascades tations from simple theory of alternating bottom-have been documented experimentally in a wide up and top-down control at adjacent trophic levelsvariety of ecosystems (Pace et al. 1999; Borer et al. are nevertheless supported by accumulating evi-2005). Moreover, controlled experiments show that dence that fishing can drive trophic cascades. Thischanges in abundance of apex predators often have evidence includes time-series data from kelp bedscascading impacts throughout ecosystems (Pace (Estes et al. 1998; Davenport and Anderson 2007),et al. 1999; Shurin et al. 2002), and that top-down coral reefs (Dulvy et al. 2004), open ocean planktoninfluences penetrate farther, on average, through (Shiomoto et al. 1997), the demersal communitiesfood chains than do bottom-up influences of nutri- (Worm and Myers 2003) and pelagic communitiesent loading (Borer et al. 2006). Given the character- (Frank et al. 2005, 2007) of continental shelves, anda b c 0.001 0.002 0.003 0.03 Great sharks 100 (% max. count) 0.02 100 Snail mortality 80 (% loss/day1) 75 Otters 60 0.01 40 50 20 25 0.0 0.0 0 0 1970 1990 1970 1990 Urchin biomass 400 (g 0.25/m2) measopredators 0.6 300 Elasmobranch Snail density 200 800 (adults/m2) 0.02 0.4 100 600 0 400 0.01 0.2 60 200 (%/24 hr1) 50 Grazing 40 0 0.0 0.0 30 20 1970 1990 1970 1990 10 0 Spartina biomass 4000 (No. 0.25 m-2) 10 Kelp density 8 200 3000 (g/m3) Scallops 6 2000 150 4 2 1000 100 0 1972 1985 1989 1993 1997 0 50 Tall Short Year zone zone 1970 1990Figure 8.4 Evidence for trophic cascades in disparate coastal marine ecosystems. (a) Killer whales to sea otters to seaurchins to kelps in northeast Pacific kelp beds. Reproduced with permission from Estes et al. (1998). (b) Shell-crushingcrabs and terrapins to snails to cordgrass in west Atlantic salt marshes. Reproduced with permission from Sillimanand Bertness (2002). (c) Great sharks to cownose rays to bay scallops in west Atlantic estuaries. Reproduced withpermission from Myers et al. (2007).
  • 116. 102 APPLICATIONStemperate estuaries (Deason and Smayda 1982; to ‘barrens’ of structure-free coralline algal pave-Myers et al. 2007). These examples involve a wide ments (Sala et al. 1998; Guidetti 2006). Finally, onrange of organisms and environments (Fig. 8.4), rocky reefs of Tasmania, predation by spiny lob-and suggest that systematic depletion of predators sters and fishes on urchins cascades to macroalgaefrom the oceans is producing far-reaching indirect (Shears and Babcock 2003; Pederson and Johnsonimpacts on the structure and functioning of many 2006). In all these systems, loss of top predatorsmarine ecosystems. I divide these examples into shifts a structurally complex, diverse communityrocky bottoms, continental shelves and pelagic dominated by macroalgae to a depauperate ‘ur-systems. chin barren’ dominated by crustose coralline algae and maintained by intense grazing.8.3.2.1 Rocky bottoms But cascades are not limited to urchin-domi-The most famous and dramatic marine trophic nated communities. An intriguingly similar exam-cascade involves the four-level food chain from ple involves small herbivorous crustaceans as thepredatory orcas (killer whales) to sea otters to intermediate link. There have been several reportsgrazing sea urchins to dominant kelps in the of perennial seaweeds such as rockweeds (Fucus)northeast Pacific Ocean (Fig. 8.4a). After fur tra- and giant kelp (Macrocystis) being decimated byders exterminated sea otters on several islands in anomalous outbreaks of grazing amphipod andthe 18th century, their sea urchin prey exploded isopod crustaceans (e.g. Kangas et al. 1982; Haah-and in turn eliminated kelp forests, whereas kelp tela 1984; Tegner and Dayon 1987). For example,beds remained vigorous on islands too remote for ˜ after an El Nino warm-water event destroyed kelpotter harvesting (Estes and Palmisano 1974; Estes communities in California during the early 1980s,and Duggins 1995). Loss of kelps on the otter-free recovering kelp beds were left without their nor-islands in turn led to pervasive ecosystem-level mal assemblage of fishes, and populations of thechanges, including local extinction of several ma- kelp-curler amphipod (Peramphithoe humeralis) ex-rine species associated with the kelp habitat, pos- ploded, devastating the kelps again, probably as asibly reduced abundances of coastal raptors that result of relaxed predator control (Tegner anddepend on fishes, and increased coastal storm Dayon 1987). Mesocosm experiments suggesteddamage stemming from the loss of buffering by that particular species of grazing amphipodskelp forests (Mork 1996). In recent decades, killer might mediate these shifts in dominance by largewhales (orcas) began attacking sea otters, evident- brown algae (Duffy and Hay 2000). Recent fieldly as the pelagic food chains supporting them experiments in California support this hypothesiswithered, and the effects of killer whale predation (Davenport and Anderson 2007), showing that in-cascaded down through sea otters and sea urchins vertebrate-feeding fishes reduced amphipodto reduce kelp again, demonstrating a four-level abundances and grazing impact, and that thesetrophic cascade (Estes et al. 1998). Similarly strong effects in turn cascaded to increase kelp bladecascades involving vertebrate predators, sea urch- growth by 100–300%, with a trend toward alsoins and macroalgae have since been documented reducing kelp mortality. A survey of unmanipu-on rocky bottoms throughout the world. In kelp lated kelp reefs similarly showed that amphipodbeds of the western North Atlantic, archaeological abundance was negatively related to that of fishes.data and time series suggest that overfishing of Intriguingly, many of these communities resemblecod and other groundfish released grazing sea the ‘wasp-waist’ architecture found in some pel-urchins from predatory control, and cascaded agic communities, in which the intermediatedown to decimate kelps (Steneck et al. 2004). In trophic level is dominated by one or a few stronglythe warmer Mediterranean Sea, experiments and interacting species. This hints that low diversitycomparisons of marine protected areas with near- is an important mediator of these strong trophicby fished areas also showed that harvesting of dynamics, as suggested by verbal theorypredatory fishes allowed urchins to proliferate (Strong 1992; Duffy 2002), a point to which I returnand overgraze macroalgae, converting large areas below.
  • 117. STRUCTURE AND FUNCTIONING OF EMERGING MARINE COMMUNITIES 1038.3.2.2 Continental shelves Chesapeake Bay similarly suggest that cownoseIncreasingly, trophic cascades are being detected rays are now causing severe damage to seagrasseven in large open coastal and oceanic ecosystems. beds as they forage for the infaunal bivalves livingTime-series data from offshore fisheries reveal, for there, some of which are commercially important.example, that the collapse of Atlantic cod (Gadus Similar patterns of observed increases in mesopre-morhua) stocks in recent decades as a result of fish- datory rays and decreases in bivalve populationsing have been accompanied by increases in both have also been observed in the northeast Atlantictheir benthic crustacean prey (Worm and Myers and in Japanese waters (Myers et al. 2007). One2003) and in overlying pelagic communities interesting question arising from such dramatic(Frank et al. 2005). Worm and Myers (2003) ana- changes in community structure is whether andlysed time series from nine cod stocks throughout how predators are sustained after their prey arethe North Atlantic and searched for correlations of driven to such low levels. Evidence consistentboreal shrimp and cod abundance with one another with a trophic cascade has also been found in saltand with water temperature. They found that cod marshes, where areas inaccessible to marine preda-and shrimp abundance showed opposite trends tors, notably blue crabs, have much higher densitiesthrough time at most sites, consistent with a trophic of rasping snails and marsh grasses achieve accord-cascade from humans through cod to benthic ingly lower biomasses; there is concern that declin-shrimp. They were also able to reject an alternative ing populations of heavily fished blue crabs maybottom-up hypothesis that these trends stemmed result in deterioration of salt marshes as a result offrom changing climate: although water tempera- this cascade (Silliman and Bertness 2002).ture affected cod, it had no detectable influence inshrimp abundance. 8.3.2.3 Pelagic systems Clearly such marine trophic cascades are of more Trophic cascades were first documented in fresh-than academic interest. Cod, shrimp, and lobster all water pelagic (Carpenter et al. 1985) and benthicsupport major commercial fisheries, for example. (Power 1990) systems, and are common in pondsAnother sobering case involves the decline of and lakes (Carpenter and Kitchell 1993; Brett andsharks, which have been especially hard hit by fish- Goldman 1996). While they appear less common ining (often as by-catch) throughout the world oceans the more functionally diverse marine pelagic com-(Stevens et al. 2000; Baum et al. 2003). Along the munities (Micheli et al. 1999; Shurin et al. 2002),eastern seaboard of North America, time-series there is growing evidence for trophic cascades indata reveal that precipitous declines of large sharks both estuaries and the open ocean. In estuaries,since 1970 were accompanied by increases of many ctenophores (comb jellies) are voracious predators‘mesopredator’ rays and small sharks, most of of other zooplankton, and often reach very highwhich are eaten almost exclusively by large sharks. densities in summer, where in some systems theyThese patterns suggest that the mesopredators have can crop 20% of standing crustacean zooplanktonbeen released from predation by larger enemies stock per day. A 6 year field study showed general-(Myers et al. 2007), as has been suggested for the ly synchronous but opposite fluctuations of preda-increase of small, mammalian predators on land as tory ctenophores, their copepod prey andwell (Crooks and Soule 1999; Johnson et al. 2007). phytoplankton, particularly the dominant diatomNotable among these marine mesopredators is the species (Deason and Smayda 1982). Integratingcownose ray (Rhinoptera brasiliensis), which feeds abundances of each group by month or season re-largely on bivalve molluscs, and which increased vealed clear evidence of a trophic cascade as cteno-by an order of magnitude over the last three dec- phore ‘blooms’ were followed by decimation ofades. Both survey data and experiments show that herbivorous zooplankton and subsequent phyto-the growing population of cownose rays has in- plankton blooms (Deason and Smayda 1982). Fieldflicted heavy mortality on bay scallops in North observations suggest that such trophic cascadesCarolina (Fig. 8.4c), resulting in a collapse of the also occur in a variety of other marine pelagic sys-century-old fishery for this species. Reports from tems, particularly those dominated by gelatinous
  • 118. 104 APPLICATIONSzooplankton (Verity and Smetacek 1996). The thus seemed to be the exception rather than thestriking example of the Black Sea illustrates both rule in marine pelagic ecosystems. But the storythe complexity of processes driving cascading may be more complex. Detailed analysis of experi-ecosystem shifts and the potentially high stakes ments found that removal of predators in the ma-for human society. Analysis of four decades of rine pelagic frequently did cascade to affecttime-series data revealed patterns consistent phytoplankton biomass, but that the sign of pred-with a cascade through five levels, from predatory ator influence on phytoplankton depends onfishes, to planktivorous fishes, to zooplankton, to food chain length, which in turn depends on cellphytoplankton, to water-column nutrient stocks size and thus taxonomic composition of the domi-(Daskalov 2002; Daskalov et al. 2007). A major nant algae (Stibor et al. 2004). When data fromshift from a clear-water phase supporting abundant three- and four-link experimental food chainslarge fish to a turbid phase began in the early 1970s were averaged, the strong influence of predatorsafter industrial fishing depleted apex predators on phytoplankton was masked. It remains uncer-(Daskalov 2002), a situation strongly reminiscent tain, however, whether these pelagic cascades areof phase shifts in north temperate lakes (Scheffer also common in unmanipulated, open marineand Jeppesen 2007). Because these changes also systems where food chains with three (classical)coincided with eutrophication and the invasion of and four (microbial loop) links operate in parallel.an exotic predatory ctenophore, mass balance mod-els were developed to evaluate the relative impor-tance of bottom-up and top-down mediation of 8.4 Biodiversity and stability of marinethese changes. The model simulations produced ecosystemscascading changes in biomass of lower trophic 8.4.1 Conceptual backgroundlevels quite similar to the observed pattern whenpredators were removed, whereas simulated eutro- Early ecologists, from Darwin (Hector and Hooperphication produced biomass increases across all 2002) to MacArthur (1955) and Elton (1958), be-levels, in contrast to observed patterns (Daskalov lieved that diverse communities were more stable2002). In this system, then, it appears that restora- and better able to resist disturbances than depau-tion of predatory fishes should be at least as effec- perate ones. These ideas have received renewedtive in restoring water quality as reducing nutrient attention as concern about declining biodiversityloading. Negative correlations across trophic levels, has grown (McCann 2000). There are several me-from predatory fishes through zooplankton to phy- chanisms by which biodiversity might increasetoplankton and even down to water-column nutri- stability of community- or ecosystem-level proper-ent stocks, in time series from the North Pacific ties. The most general is simple statistical averaging(Shiomoto et al. 1997) and coastal north Atlantic (also called the ‘portfolio effect’): as long as tempo-Oceans (Frank et al. 2005), suggest that cascading ral fluctuations of co-occurring species are not per-trophic interactions can occur even in open pelagic fectly correlated, variance of their aggregateecosystems. abundance in response to stochastic environmental The question remains whether these patterns are variance will be lower than the average variabilitygeneral. Micheli (1999) conducted a meta-analysis of component species (Tilman et al. 1997; Doak et al.of marine pelagic systems, using data from both 1998). Biodiversity may also stabilize ecosystemmesocosm experiments and time series from properties against perturbations by enhancingunmanipulated systems to ask whether nutrient the system’s ability to absorb a stress withoutloading and predation penetrated through the changing (resistance) or the rapidity with whichfood chain. Although both zooplanktivorous it returns to its original state after perturbationfishes and nutrient loading significantly affect the (resilience), through at least two biological mechan-adjacent trophic level (zooplankton and phyto- isms. First, niche differentiation (functional diver-plankton respectively), these effects attenuated sity) among species increases the probability that atrapidly through the food chain. Trophic cascades least some species will thrive as environmental
  • 119. STRUCTURE AND FUNCTIONING OF EMERGING MARINE COMMUNITIES 105conditions change. Second, and in contrast, func- and Cardinale 2004; Duffy et al. 2005) and to inva-tional redundancy can enhance stability against sion by other native (France and Duffy 2006) andperturbations that cause extinction by reducing non-native species (Stachowicz et al. 1999, 2002a).the probability that extinction removes a function-ally unique species. The last two phenomena are 8.4.2.1 Comparisons through timesometimes combined as the insurance hypothesis Observational evidence from fishery science corro-(Yachi and Loreau 1999). borates predictions (Tilman et al. 1997; Doak et al. In the context of applied marine ecology, a ques- 1998) that diversity can enhance stability of aggre-tion of particular interest is how diversity influ- gate biomass or production (i.e. overall catch), bothences resistance to human predation, i.e. fishing. in the general sense of reducing long-term fluctua-Are more diverse marine food webs more stable in tions and in the specific sense of providing resis-response to such anthropogenic perturbations? The tance to perturbations of fishing and environmentaltheory most relevant to the case of fishing effects in forcing. For example, time series of fish biomassmarine systems involves the role of prey richness in from the North Sea show that aggregate biomassbuffering the community from predator impacts. is less variable than that of individual fish speciesLeibold (1989, 1996) argued that a resource base (Fig. 8.5; Jennings and Kaiser 1998), supporting thewith more species is more likely to contain at least suggestion that diversity enhances general stability.one species that is resistant to consumption and can Diversity can also provide resistance to specificdominate in the presence of a consumer, such that a perturbations. Among the most intriguing cases ismore diverse prey community will maintain higher the link between stock diversity and productivity ofaggregate biomass under predation. Similarly, if Alaskan salmon under decadal-scale climate varia-different prey species are resistant to different pre- tion (Hilborn et al. 2003b). Because salmon return todators, or gear types in the case of fishing, then a the streams or lakes of their birth to spawn, popula-diverse prey community will maintain higher bio- tions are genetically highly structured into distinctmass under fishing pressure. This general argu- genetic populations or ‘runs’ that differ substantial-ment has been extended to suggest that trophic ly in life history, phenology and ecology. Hilborncascades also should be most prevalent in low- et al. (2003b) used historical catch records for Bristoldiversity systems, with one or a few important Bay sockeye salmon (Onchorhyncus nerka) datingspecies at each trophic level (Strong 1992; Duffy back to the 1890s to show that the relative contribu-2002). tions of different populations to total salmon catch differed greatly through time as individual popula- tions responded differently to long-term variation8.4.2 Evidence linking diversity ˇ in climate forcing by the El Nino Southern Oscilla-and stability in marine systems tion (ENSO) and the Pacific Decadal OscillationGrowing empirical evidence suggests that chang- (PDO). The population-specific variability in re-ing horizontal (i.e. within-trophic level) diversity sponse to changing environmental conditions re-can have several important consequences for ma- sulted in an aggregate salmon catch that was morerine food web interactions and ecosystem processes stable through time than was that of any individual(Emmerson and Huxham 2002; Duffy and Sta- population (Hilborn et al. 2003b). This link betweenchowicz 2006; Stachowicz et al. 2007). Experimental stock diversity and stability in response to humanresearch has supported a stabilizing effect of diver- predation is consistent with the proposed impor-sity on community biomass in some competitive tance of niche differentiation as a mechanism byplant assemblages, aquatic microbial food webs which biodiversity can stabilize biomass and pro-and soil microfaunal communities (reviewed by duction (Loreau et al. 2002).McCann 2000; Cottingham et al. 2001; Loreau et al. Similarly, there is considerable evidence that func-2002). There is also some experimental evidence tional diversity of herbivores is important to main-that marine species diversity can enhance trophic taining coral dominance over algae on tropical reefs.level resistance to top-down control (Hillebrand In the Caribbean, overharvesting of herbivorous
  • 120. 106 APPLICATIONS Biomass (tonnes x 106) Sole Plaice 0.15 0.80 0.12 0.60 0.09 0.40 0.06 0.20 0.03 0.00 0.00 1976 1980 1984 1988 1992 1976 1980 1984 1988 1992 Biomass (tonnes x 106) Cod 1.60 Haddock 1.00 0.80 1.20 0.60 0.80 0.40 0.20 0.40 0.00 0.00 1976 1980 1984 1988 1992 1976 1980 1984 1988 1992 Biomass (tonnes x 106) Whiting Saithe 1.20 1.00 1.00 0.80 0.80 0.60 0.60 0.40 0.40 0.20 0.20 0.00 0.00 1976 1980 1984 1988 1992 1976 1980 1984 1988 1992 Norway pout Sandeel Biomass (tonnes x 106) 1.50 5.00 1.20 4.00 0.90 3.00 0.60 2.00 0.20 1.00 0.00 0.00 1976 1980 1984 1988 1992 1976 1980 1984 1988 1992 Biomass (tonnes x 106) Herring All species 5.00 16.0 4.00 12.0 3.00 8.0 2.00 1.00 4.0 0.00 0.0 1976 1980 1984 1988 1992 1976 1980 1984 1988 1992 Year YearFigure 8.5 Biodiversity begets stability. The temporal trend in aggregate fish biomass (bottom right) is substantially lessvariable than the time series for any individual species. Reproduced with permission from Jennings and Kaiser (1998).fishes had little effect on algal biomass initially, per- These observed patterns are consistent with experi-haps because sea urchin grazing compensated for ments and comparisons among protected and fishedthe reduced fish grazing (Hay 1984; Hughes 1994; areas on Kenyan reefs, which showed that sea urch-Jackson et al. 2001). But when sea urchins suffered ins were more abundant on overfished reefs than inmass mortality from disease, algal biomass exploded protected areas, but that experimental reduction of(Carpenter 1990). Macroalgal blooms proliferated urchins allowed large macroalgae to dominate onlyonly after both fishes and sea urchins were reduced. on fished reefs, where herbivorous fishes were
  • 121. STRUCTURE AND FUNCTIONING OF EMERGING MARINE COMMUNITIES 107scarce (McClanahan et al. 1996; see also Hay and nation again comes from a coral reef in Kenya,Taylor 1985). That is, the presence of two effective where experimental reductions of sea urchins al-herbivore groups (fishes and urchins) buffers the lowed algae to proliferate to about twice the levelsystem against algal overgrowth if one group is lost. on fished reefs as on unfished reefs (McClanahan While these examples involve community (or, et al. 1996).more accurately, guild) stability at the expense of A stabilizing role of biodiversity in exploited ma-population fluctuations, experiments show that in- rine communities is also suggested by regionalteractions within diverse natural assemblages may comparisons of trophic dynamics in northwest At-also foster stability at the population level by lantic continental shelf ecosystems (Frank et al.imposing density dependence in demographic 2006). These authors analysed time-series data forrates (Hixon and Carr 1997; Carr et al. 2002). Field several trophic levels, from phytoplankton to har-experiments showed that per capita mortality rates vested fishes, at nine heavily fished sites to assessof recruiting coral-reef fishes in the Bahamas were the strength and direction of trophic control. Corre-strongly density dependent in the presence of mul- lations between adjacent trophic levels variedtiple predators (Hixon and Carr 1997) or predators among sites, being predominantly negative atand competitors together (Carr et al. 2002), whereas higher latitudes, indicating top-down control, butmortality was independent of density when com- positive at lower latitudes, indicating bottom-uppetitors alone, or only one type of predator, was control. Interestingly, the strength and sign of tro-present. These experiments provide an intriguing phic control varied systematically with species rich-contrast with the pattern found in a meta-analysis ness, with stronger top-down control in the moreof predator–prey experiments, in which the pres- depauperate northern sites and bottom-up controlence of predators tended to destabilize temporal at the more diverse southern sites. This weakeningdynamics of their prey (Halpern et al. 2005). It of top-down control with prey species richness iswould be quite interesting to know whether the consistent with results of experiments (Steiner 2001;destabilizing effects of predators in the meta-analy- Hillebrand and Cardinale 2004; Duffy et al. 2005).sis might be an artefact of experiments including However, the link to diversity in the northwestonly one or a few species of predators. Atlantic data is confounded by strong correlations of species richness with latitude and temperature,8.4.2.2 Comparisons across space which are also expected to influence the strength ofSeveral prominent spatial patterns are consistent top-down control through effects on demographicwith the hypothesis that diversity enhances stabili- rates (Frank et al. 2006). Disentangling these influ-ty in marine systems. Across geographic regions, ences is not possible at present.the least stable ecosystems are those low in fish Finally, Worm et al. (2006) conducted a compre-diversity, with a few strong predator–prey links hensive analysis of the links between marine biodi-and little capacity for prey switching (Jennings versity and response to fishing (Fig. 8.6). Theyand Kaiser 1998). For example, the dramatic cas- analysed relationships between species richnesscades emanating from sea otters to kelp in the rela- and fishery production for the world’s 64 Largetively low-diversity community of Alaska were not Marine Ecosystems (www.fishbase.org). Regionsobserved in the more diverse kelp beds of southern with naturally low fish diversity supported lowerCalifornia (Dayton et al. 1998), even though both average fishery productivity, and had more fre-regions lost sea otters by the early 19th century. A quent ‘collapses’ (strong reductions in fisherypossible explanation is that, in southern California, yield) and lower resilience (degree of recoveryspiny lobsters and sheephead, which also feed on after overfishing) than naturally species-rich sys-urchins, compensated for the reduced impacts of tems. Worm et al. (2006) suggested that the greatersea otter predation; indeed, when lobsters and resilience of more diverse ecosystems may be ex-sheephead were heavily exploited in the 1950s, plained by the greater ability of fishers to switchkelps did decline in southern California (Dayton among target species in diverse ecosystems; whenet al. 1998). Evidence consistent with such an expla- abundance of a species declines to a low level, it is
  • 122. 108 APPLICATIONS 80 Productivity in catch (%) 100 (a) r = –0.57 (b) r = 0.57 Collapsed taxa (%) p < 0.0001 70 p < 0.0001 80 60 50 60 40 40 30 20 20 10 0 0 1 2 3 4 1 2 3 4 (log) Species richness (log) Species richness 40 (c) r = 0.48 150 (d) r = 0.57 Fished taxa richness p = 0.0001 Per cent recovery 120 p < 0.0001 30 20 90 10 60 0 30 –10 0 1 2 3 4 1 2 3 4 (log) Species richness (log) Species richness 3500 (f) r = 0.73 Average catch (Gg/yr) 0.9 (e) Coefficient of variation r = –0.30 3000 p < 0.0001 0.8 p = 0.0291 2500 0.7 2000 0.6 1500 0.5 1000 0.4 500 0.3 0 0 50 100 150 200 0 50 100 150 200 Fished taxa richness Fished taxa richnessFigure 8.6 Correlations between biodiversity (fish species richness) and the productivity and resilience of fishery catchesacross 64 large marine ecosystems. Reproduced with permission from Worm et al. (2006).more profitable for fishers to target other species, Atlantic, where longline fisheries for billfish show awhich provides the overfished species with a refuge pattern of sequential depletion of species (Myersand allows them to recover. This explanation is and Worm 2003), with decline of blue marlin inconsistent with the negative relationship between the 1960s accompanied by a rise in catch of sailfish,fished taxa richness and interannual variation in which then declined in turn as swordfish catchescatch (Worm et al. 2006, see also Fig. 8.5). increased through the late 1970s and 1980s. The result was that total billfish catch remained relative-8.4.2.3 Mechanisms ly stable through time despite boom and bust pat-What is the mechanism for these putative effects terns in the catch of individual species. Similarof biodiversity on ecosystem resilience? One likely patterns have been observed in demersal ecosys-candidate is functional compensation (or functional tems, using both catch data (Myers and Worm‘redundancy’) among species within a guild or 2003) and fisheries-independent data (Shackellfunctional group, such that decline of one species and Frank 2007), in which declines of targeted fishis compensated by increase in another species with species were accompanied by compensatorysimilar functional characteristics, e.g. through increases in other groups. These patterns of speciesrelaxed competition. Evidence potentially consis- turnover are very similar to that predicted bytent with this mechanism comes from the tropical theory when a consumer imposes mortality on
  • 123. STRUCTURE AND FUNCTIONING OF EMERGING MARINE COMMUNITIES 109prey species differing in competitive ability and modelling (de Ruiter et al. 2005). A principal chal-edibility (Leibold 1996). lenge is the paucity of data on interaction strengths to parameterize models realistically (Wootton and Emmerson 2005). Most existing food webs are ‘con-8.5 Interaction strengths and dynamic nectance webs’ or ‘energy webs’, based on qualita-stability in marine food webs tive trophic links between species or on patterns of energy flow, respectively. Paine (1980; see also Raf-8.5.1 Conceptual background faelli and Hall 1996) showed that links in the foodPredicting the cascading indirect effects of predator web that are important energetically are often notdepletion in food webs depends both on the topo- the same links that are functionally (dynamically)logical structure of webs, which the previous section important, i.e. that have strong impacts on the struc-showed are changing systematically, and on the ture and organization of the community and, bydistribution of interaction strengths. One approach inference, on ecosystem processes. It is the latter,to the challenge of scaling up ecological processes to functional food web structure that is of primaryreal ecosystems uses empirical data on topology interest in understanding a system’s stability andand interaction strengths to explore dynamic re- response to perturbations (Berlow et al. 2004; Woot-sponses to perturbations through simulation and ton and Emmerson 2005). a c Number of species 25 20 15 10 5 0 0.000 –0.005 –0.010 –0.015 Interaction strength b 1.0 % of interactions 0.8 0.6 0.4 0.2 0 1 2 3 Interaction strengthFigure 8.7 Skewed distributions of interaction strengths in (a) 45 herbivore species feeding on sporelings of giant kelp inCalifornia (reproduced with permission from Sala and Graham 2002), (b) three bird species feeding on 23 species ofintertidal animals in Washington, USA (reproduced with permission from Wootton 1997) and (c) fishes feeding ondiverse prey in a large Caribbean coral reef food web (reproduced with permission from Bascompte et al. 2005). Notethat, in all food webs, only a small percentage of interactions have strong effects (thickness of arrows is proportional tointeraction strength in (c)).
  • 124. 110 APPLICATIONS The small but growing number of communities that the preponderance of weak interactions inin which interaction strength has been estimated many food webs is unlikely to protect them fromsuggests that strong skew in interaction strength the specific sorts of impacts imposed by fishing.may be a general feature of communities, withmany links between species having negligible ef-fects on the dynamics of either party, and a few 8.6 Alternate stable states and regimelinks having very strong impacts (reviewed by shifts in marine ecosystemsWootton and Emmerson 2005, Fig. 8.7). This has 8.6.1 Conceptual backgroundimportant implications for community stability, asskew in interaction strength has been shown to An ecological phenomenon of growing concern inconfer stability on trophic networks in theory understanding marine ecosystem dynamics and(McCann et al. 1998; Emmerson and Yearsley 2004) their implications for society is the phenomenonand in simulation studies of empirical food webs of regime shifts. Regime shifts can be defined as(Emmerson and Raffaelli 2004). relatively rapid transitions between distinct and relatively long-lasting, semi-stable states of a sys- tem (Knowlton 2004; Steele 2004). The potential8.5.2 Empirical evidence existence of alternate stable states has been a sub-Simulation-based approaches to exploring food web ject of keen interest and controversy in ecology forstability have generally measured stability as resis- decades (Lewontin 1969; Sutherland 1974). In re-tance to small perturbations imposed on species ran- cent years, regime shifts have gained new promi-domly. A critical question for applied ecology is how nence in the context of conservation andinteraction strength of a species covaries with its management (Scheffer et al. 2001). The term hasvulnerability to real perturbations such as fishing, been used in two ways. The first definition is phe-and how those perturbations ripple through the nomenological, and refers to multi-year periods ofweb. Soberingly, analysis of a diverse Caribbean relative stability in time series of observationalreef food web found that, while combinations of data that are separated by abrupt shifts to intervalsstrong interactions capable of generating trophic cas- that fluctuate around a different mean. The sec-cades were quite rare, human impacts fell dispropor- ond, stricter definition involves the dynamics oftionately on species involved in such interaction the system and refers to the existence of two orcombinations, primarily because large- bodied spe- more semi-stable states, or alternate attractors, incies are both strong interactors and disproportion- an ecosystem. Regime shifts in the phenomenolog-ately targets of human impacts (Bascompte et al. ical sense have been described in a number of2005). Thus, in contrast to the stabilizing effects of marine ecosystems, and appear often to track de-many weak interactions predicted by theory in di- cadal-scale climate variation (e.g. Overland et al.verse communities such as reefs, the targeted har- 2006).vesting of large predators can have important Recent syntheses (Collie et al. 2004) have definedcascading impacts because fishing imposes strong, three types of regime shifts, which are actuallypersistent and non-random perturbations. points along a continuum (Fig. 8.8a): (1) a smooth There is reason to expect that the results of Bas- regime shift, defined by a quasi-linear relationshipcompte et al. (2005) may be common in that strong between a forcing variable and a response variable,interactors are typically the larger species in a com- (2) an abrupt regime shift, in which the relationshipmunity (although see Sala and Graham 2002 for an is non-linear and (3) a discontinuous regime shift,exception), and at least for direct harvesting such as in which the relationship is not only non-linear butfishing, large species are also those targeted prefer- the trajectory of the response variable differs whenentially. Thus, it is likely that there will often be the forcing variable is declining compared withpositive covariance between a species’ interaction when it is increasing. The latter situation results instrength and its vulnerability to fishing. Although two possible states of the response variable at athese conclusions are preliminary, this suggests given value of the forcing variable and is also
  • 125. STRUCTURE AND FUNCTIONING OF EMERGING MARINE COMMUNITIES 111 a b c 15 400 Thousand tonnes Biomass Catch Frequency 300 10 Abundance 200 5 100 0 0 Smooth 1940 1960 1980 2000 0 100 200 300 400 Ex Year Biomass ter na e lF ur or d e ct cin ru Abrupt St g al 80 rn 80 te 1931–1965 in 1966–2000 60 60 Catch (kt) Catch (kt) Discontinuous 40 40 20 20 0 0 0.0 0.2 0.4 0.6 0.0 0.2 0.4 0.6 Fishing mortality Fishing mortality f g h 2 0 75 80 2 97 1.5 77 85 1.5 78 Phytoplankton Planktivorous fish 99 Zooplankton 85 83 1 –0.5 1 2000 2000 82 88 89 91 0.5 0.5 79 81 2000 94 97 0 –1 0 98 93 89 94 83 –0.5 –0.5 90 96 –1 91 –1.5 –1 87 –1 0 1 2 –1 0 1 2 –1.5 –1 –0.5 0 0.5 Fishing mortality Gelatinous plankton ZooplanktonFigure 8.8 (a) A conceptual model of the three forms of regime shift, and empirical evidence consistent withdiscontinuous regime shifts in (b–e) Georges Bank haddock and (f–h) the Black Sea pelagic ecosystem. In Georges Bankhaddock, the regime shift is illustrated by (b) a discrete shift from higher to lower abundance after ~1965, (c) a bimodaldistribution of biomass across the time series, (d) a different functional relationship between fishing mortality and catchin the two time periods and (e) hysteresis in this functional relationship in simulations of a model fit to empirical data.Reproduced with permission from Collie et al. (2004). In the Black Sea, the regime shift is similarly supported by (f–h)changing functional relationships through time between predators and prey. Reproduced with permission fromDaskalov et al. (2007).referred to as hysteresis (Fig. 8.8a). Scheffer and order of colonization during succession), shift to-Carpenter (2003) described a series of criteria for wards a distinctly different stable state after a pulseidentifying regime shifts, none of which alone diag- perturbation and hysteresis, i.e. change of the eco-noses a regime shift but which, together, constitute system along different pathways when the forcingstrong evidence. In field data these include abrupt variable is increased compared with when it is de-shifts in time series, a bi- or multimodal frequency creased. Two important questions for applied ecol-distribution of states in a time series, and dual (or ogy are whether rapid shifts between relativelymultiple) relationships between ecosystem state long-lasting states of an ecosystem can be forcedand a forcing variable. Experimental evidence in- by gradual changes in conditions, and whethercludes dependence of final state on initial state (e.g. these shifts are reversible.
  • 126. 112 APPLICATIONS8.6.2 Empirical evidence for regime shifts hysteresis characteristic of a discontinuous regimein marine ecosystems shift (Fig. 8.8f–h). Models confirm that overexploi- tation can trigger such shifts between alternateAt the population level, Collie et al.(2004) applied states (Daskalov 2002; Collie et al. 2004).Scheffer and Carpenter’s (2003) criteria to test for Are such regime shifts common in marine eco-regime shifts in Georges Bank haddock, based on systems and, if so, how do they relate to forcingage-structured abundance time series for the period mechanisms? Feng et al. (2006) explored this ques-1931–2000. Although some of these criteria are dif- tion using dynamic simulation of mass-balanceficult or impossible to evaluate in open marine eco- models (Ecopath with Ecosim; Christensen andsystems (e.g. whether the system goes to different Walters 2004) applied to 24 marine ecosystems.states after a perturbation or under different start- The models imposed a simulated perturbation ofing conditions), others can be addressed with avail- 10 years of intensified fishing, then relaxed fishingable time-series data (Fig. 8.8b–e). First (Fig. 8.8b), to the initial level and followed the system’s long-they demonstrated a discrete step in the average term (70 years) trajectory, asking whether each sys-stock biomass, which dropped abruptly after a tem returned to its original state or equilibrated tospike in catch due to influx of foreign fishing fleets an ‘alternative attractor’. Six scenarios consideredin the early 1960s, and remained low for most of the fishing on top predators compared with intermedi-remaining century. Second (Fig. 8.8c), the distribu- ate levels (wasp-waist system) under bottom-up,tion of biomass values was bimodal. Third (Fig. mixed or top-down control. The simulations8.8d), catch showed a different functional relation- showed that, under top-down or mixed control,ship to fishing mortality before and after the shift in 11–28% of the ecosystems showed alternate attrac-the 1960s. These three observations support the tors, i.e. shifted into a new regime that persistedexistence of two distinct regimes during the time after fishing pressure was relaxed, whereas none ofseries. Moreover, simulations of a population the ecosystems showed alternate attractors undermodel fit to empirical data showed that catch fol- bottom-up control (Feng et al. 2006). These modellowed different trajectories when fishing mortality results, together with a few well-documented em-was increased versus decreased, i.e. the system ex- pirical examples such as the Black Sea, suggest thathibited hysteresis (Fig. 8.8e), a key piece of evidence intense fishing pressure can produce shifts to newfor a discontinuous regime shift between alternate ecosystem states that are difficult to reverse, sup-semi-stable states (Collie et al. 2004). porting suspicions that such regime shifts may be at An ecosystem-wide regime shift, evidently least partly responsible for the failure of manyforced in part by overfishing, has also been docu- heavily fished species to rebound even decadesmented in the Black Sea (Daskalov et al. 2007). Here after fishing moratoria were enacted (Hutchingsstrong fishing pressure on top predators caused and Reynolds 2004). If such regime shifts aretheir decline and eventual collapse in the 1970s, indeed common responses to top-down perturba-which was accompanied by cascading changes in tions, they have serious implications for mana-lower trophic levels, leading to phytoplankton gement of natural ecosystems.blooms and nutrient depletion; a subsequentchange in the focus of fisheries to smaller plankti-vorous fishes such as sprat and anchovy (‘fishing 8.6.2.1 Mechanismsdown the food web’; Pauly et al. 1998) then led to a Several mechanisms potentially can produce re-subsequent collapse of these planktivores and gime shifts between alternate attractors or semi-corresponding increases in the jellyfish that com- stable states (Collie et al. 2004; Folke et al. 2004;pete with them (Daskalov 2002). Plotting the time Knowlton 2004). At the population level, the mosttrajectories of the various trophic groups shows general mechanism involves the Allee effect, i.e.that, for several interactions, the relationships be- depensation or positive density dependence attween consumer and prey abundances differed in low population sizes (Knowlton 1992), althoughearly compared with later years, suggestive of the this process requires some additional factor to
  • 127. STRUCTURE AND FUNCTIONING OF EMERGING MARINE COMMUNITIES 113prevent extinction at low population density. Alter- Regime shifts may also result from effects ofnative stable states of a population can also result organisms on the environment. In benthic systemsfrom size dependence of vital rates such as recruit- in particular, ecosystem engineers or other speciesment, growth and fecundity (Botsford 1981). For may modify the environment such that it becomessimple predator–prey systems, the range in types less hospitable to species characteristics of the al-of regime shifts can be generated from the same ternative regime. Certain taxa of infaunal inverte-model simply by changing parameters, particularly brates, for example, are both more tolerant ofthe ratio between prey carrying capacity and mobile sediment resuspension and more active inpredator half-saturation constant (K/D), the rela- resuspending it; these may prevent establishmenttionships between maximal predation rate and of species that would otherwise dominate in stableprey growth rate, and the minimum timescale sedimentary environments (Peterson 1984; vanfor shifts (Collie et al. 2004). Alternate attractors Nes et al. 2007). Similarly, in lakes and probablymay also arise from idiosyncrasies of species beha- also in estuaries, clear-water phases are main-viour. For example, pelagic ecosystems often show tained in part by growth of benthic macrophytes,rapid shifts between decadal-scale states domi- which bind sediment, preventing its resuspension;nated by different planktivorous fish species, such when macrophytes are lost for whatever reason,as sardine and anchovy. At high population density the mobility of both sediments and sediment-such fishes are usually found in ‘pure’, monospecif- bound nutrients foster turbidity in the water col-ic schools, but, when reduced to low density by umn, which resists re-establishment of benthicfishing or other processes, their strong schooling macrophytes.inclination causes them to join schools of other Finally, a link between changing biodiversity andspecies, which may place them into conditions regime shifts, though not rigorously studied, is sug-that are poor for feeding and reproduction, driving gested by several lines of evidence. First, rapidtheir population further toward decline (Cury et al. transitions in ecosystem state appear to be better2000). documented in relatively low-diversity systems, in- Regime shifts in predator–prey interactions may cluding temperate lakes, the Black Sea (Daskalovalso be mediated by the ontogenetic shifts in trophic et al. 2007) and the North Atlantic (Frank et al. 2007).level (Fig. 8.2) common in both benthic (Barkai and In particular, wasp-waist ecosystems appear espe-McQuaid 1988) and pelagic (Swain and Sinclair cially prone to rapid, pronounced ‘regime shifts’2000; Bakun 2006) ecosystems. In many such sys- between alternate semi-stable community states,tems, the dominant species in one of the alternative and this vulnerability has been attributed in partstates feeds on early life history stages of the alter- to the low diversity and low resilience of this inter-native dominant, generating an unstable feedback mediate trophic level. Second, experiments showloop that prevents the alternative dominant from that invasion of marine communities by exotic spe-gaining abundance. A potential example of this cies, which can trigger irreversible shifts in ecosys-phenomenon involves the collapse of cod in the tem structure and function, is generally moreBaltic Sea, which was accompanied by a shift to frequent in communities of low diversity (Stacho-dominance by the cod’s ‘wasp-waist’ prey, plankti- wicz et al. 1999, 2002a).vorous herring and sprat (Bakun 2006). Cod havefailed to rebound from their initial collapse 8.7 Emerging questions in emergingthroughout much of the north Atlantic, despite se- marine ecosystemsvere fishing restrictions, probably in part becauseabundant herring and sprat feed heavily on cod The ocean of the 21st century is changing at rateseggs and larvae. Thus, for pelagic marine food and in directions never before seen in human histo-webs, the ontogenetic size structuring of trophic ry. The causes involve both abiotic changes – in-interactions may be a key factor in mediating com- cluding eutrophication, habitat alteration, and,monly reported regime shifts between alternate sta- increasingly, climate warming and acidification –ble states (Bakun 2006). and direct alteration of community structure
  • 128. 114 APPLICATIONSvia intense predation by humans and invasion Finally, it is becoming increasingly clear that evo-of non-indigenous species. The studies reviewed lutionary change often occurs on similar timescaleshere suggest that these emerging marine ecosys- to ecological interactions among species, and can betems show several recurrent patterns that raise critical to understanding the dynamics of those in-the following intriguing questions for future re- teractions (Thompson 1998; Hairston et al. 2005).search. This is especially true of human predation on marine The decline in large animals through overhar- fishes, which generally targets larger, more econom-vesting constitutes an important loss of functional ically valuable individuals, and accordingly has pro-diversity, which is eroding complex food webs into duced declines in average body size in manytopologically simpler, and probably more strongly exploited marine fish species over recent decadeslinked, food chains. This simplification may lead to (Hsieh et al. 2006). If length and age at maturity arefundamental change in community and ecosystem at least partially heritable, the resultant size- anddynamics with several potential consequences. Are age-selective mortality means that rapidly maturingthese simpler food chains indeed less stable gener- genotypes will be favoured under fishing mortalityally and, specifically, more vulnerable to dramatic (Law 2000) and this truncation in size structure willregime shifts than naturally diverse communities produce not only ecological ramifications through(Folke et al. 2004)? the ecosystem but also evolutionary change. In par- There is some evidence that intense exploitation ticular, length and age at sexual maturity are key lifemay change not only the strength of interactions history traits affecting fitness. Controlled experi-but also the mode of control, from bottom-up in ments in both laboratory (Conover and Munchlightly exploited, naturally diverse systems to top- 2002) and field (Reznick and Ghalambor 2005) con-down in heavily exploited systems, as in the north- firm that size-selective mortality can produce sub-west Atlantic (Frank et al. 2006). Indeed, this re- stantial genetically based changes in age and size atview suggests that trophic cascades may be more maturity within a few generations. Data from com-common in marine systems than concluded previ- mercially exploited fishes also indicate that age andously (Jennings and Kaiser 1998). The different size at maturity have substantial heritabilities, andconclusions may stem in part from newer, fine- many stocks indeed have shown predicted declinesscale data (e.g. Frank et al. 2006); but it is also in age and size at maturity over recent decadesconceivable that marine ecosystems have changed (Hutchings and Baum 2005). A critical question foreven in the last decade toward states more vulner- future research is how much of the change in lifeable to perturbation. Does heavy exploitation gen- histories of wild fish stocks results from evolutionerally shift marine communities towards stronger versus other factors such as release from competi-top-down control? tion, and whether this evolution reinforces the hys- Accelerating invasions of non-indigenous species teresis between exploited and unexploited states ofare changing marine communities worldwide (Ruiz marine ecosystems, as suggested for cod-dominatedet al. 2000). Do these emerging communities interact ecosystems (Olsen et al. 2004; de Roos et al. 2006).in different ways as a result of the lack of sharedevolutionary history among species? For example, Acknowledgmentsdo marine consumer–prey interactions involvingnon-indigenous species differ systematically from I am grateful to James Douglass, Peter Morin andthose involving only native species, as they do on Herman Verhoef for comments that improved theland where non-indigenous consumers promote manuscript, and to the National Science Founda-‘invasional meltdown’ (Parker et al. 2006)? tion for support (OCE-0623874).
  • 129. CHAPTER 9 Applied (meta)community ecology: diversity and ecosystem services at the intersection of local and regional processes Janne Bengtsson range from biological control, regulation of wildlife9.1 Introduction and fisheries to nutrient cycling and crop produc-The most pressing ecological questions today con- tion. On the other hand, there is a long traditioncern the management of natural resources. At the among ecologists to separate ecosystems into natu-heart of natural resources lies biodiversity. The ral versus disturbed ones, in which the latter werecommon biodiversity in ecosystems provides disturbed mainly by human activities (Worstergoods and services for society, such as food (e.g. 1994). At times, it was even questioned whethercrops, grazing animals, fish), materials for build- such human-dominated ecosystems had any ‘realings and other human artefacts; bioenergy; and ecology’, and, if so, if it was the same ecology as thatprocesses such as plant or animal production, for undisturbed ecosystems. Hence, studies of com-biological control and pollination that result in munities had to be made in pristine systems wherethose goods (Daily 1997). Ecosystems are also uti- the unnatural humans played a negligible role.lized to treat the waste products from human activ- Traces of such notions can still be found withinities (e.g. Folke et al. 1997). Behind most of these conservation biology, especially in the USA, whenecosystem services are interactions between species emphasis is put on ‘wilderness’ and ‘naturalness’.or functional groups, which are – no matter wheth- However, the emergence of recent major issues iner we are conscious of them or not – modified by ecology, such as biodiversity conservation, ecosys-human activities. The ultimate test for the validity tem services and climate change, show that this isof community ecology, as outlined in the previous an unrealistic approach.chapters, is whether it can provide useful knowl- In this chapter, I argue that community ecologyedge for managing ecological interactions and ought to understand ecological processes that arebiological resources in a sustainable way. crucial for providing goods and services to society. Humans are now considered to be the dominat- Applied questions can both provide large-scaleing ecological and evolutionary force on Earth tests of general ecological theories and lead to cru-(Vitousek et al. 1997; Palumbi 2001). Despite this, cial insights into basic ecological theory. Using thein the past, community ecology had an awkward theoretical framework of spatial population andrelationship with human-dominated landscapes. community dynamics, now more trendily referredOn one hand, applied questions have been crucial to as metacommunity dynamics (Chapter 5), I startfor theory and empirical studies in both population with a short background in classical patch andand ecosystems ecology. Well-known examples niche theory. The juxtaposition of local versus 115
  • 130. 116 APPLICATIONSregional dynamic processes in metacommunity the- an indicator of species being present to do the workory is then used to suggest guidelines for manage- in ecosystems (Bengtsson 1998). Management ofment of biodiversity and ecosystem services in ecosystem services in heterogeneous landscapes ul-human-dominated landscapes. As the literature on timately relies on understanding communities ofthese subjects is increasing rapidly, I do not attempt interacting populations.to cover the recent literature. Instead, I use someexamples to illustrate how applied and basic ques-tions in ecology can gain from being studied in 9.2 A theoretical backgroundtandem in the same managed system. My examples 9.2.1 A simplified historical narrativemainly derive from studies on biodiversity andecosystem services in terrestrial agricultural land- Island biogeography and metapopulation theoryscapes. I use the terms ecosystem functioning and emerged during the 1960s almost simultaneouslyecosystem processes to refer to any process carried and actually within the same intellectual environ-out in ecosystems by organisms maximizing fitness, ment (Worster 1994). Many ecologists seemed tofor example biomass production or decomposition, regard them as essentially the same, but from sev-while ecosystem services are the subset of ecosystem eral aspects and especially for conservation theseprocesses that benefit humans (or society). Diversity theories offer different views of the world. A majoris used synonymously with species richness, unless conceptual difference is that island biogeographystated otherwise. (MacArthur and Wilson 1967) saw the world as Basic in biodiversity and ecosystem functioning mainlands sending off migrants to islands, an ex-research is the role of diversity. It is important to treme source–sink situation with respect to coloni-remember that diversity in itself does not carry out zation–extinction dynamics (but see Schoenerany process. Ecosystem processes are carried out by 1976). On the other hand, in metapopulation theoryspecies (populations and individuals); diversity is (Levins 1969) patches are colonized from other Island biogeography Metapopulation dynamics Important Diversity Single – few species characteristics External mainland species pool Internal pool of colonizers Islands sinks Patches equal (can be relaxed) Stochastic extinctions Stochastic extinctions View of the world Basic equation Diversity For single species solutions S* = IM/(E+I) P* = 1 – e/m For single species For competitors: e = e0 + eij Equilibrium proportion occupied islands P* = m/(e+m) Importance for Historically large but Large management in practice useless Diversity and species because islands depend persistence dependent completely on the mainland on isolation and regional commonnessFigure 9.1 Contrasting views on the world in the basic theories of island biogeography and metapopulation dynamics.S, species number; M, mainland pool species number; E and I, number of extinct and immigrating species per unit time(in island biogeography model); e and m, extinction and migration rates (in metapopulation model); P*, equilibriumproportion of occupied islands/patches. See plate 3.
  • 131. DIVERSITY AND ECOSYSTEM SERVICES 117patches in the region, not from a mainland (Fig. conditions allowing survival) and a species’ impact9.1). Island biogeography was important in early on the environment (withdrawing resources butconservation biology, despite the conceptual prob- also incorporating ecological engineering andlem that islands (patches) require a mainland to other aspects of niche construction; Odling-Smeesustain their diversity, which makes it practically et al. 2003) are both major components of the nicheuseless to apply in reserve design. Metapopulation (Chase and Leibold 2002). (2) The idea that, whenmodels in which the number of colonizers depends environments vary over time or in space, stability ofon how many patches are inhabited in the region, ecosystem functioning is ensured by species withand how isolated those patches are, provide a more similar effects on ecosystems while at the same timerelevant model of the world, especially for conser- showing a diversity of responses to environmentalvation of species in fragmented landscapes. The variation (response diversity, Elmqvist et al. 2003;general acceptance of the metapopulation concept insurance hypothesis, Loreau et al. 2003). (3) Thefor single species was crucial for the later develop- neutral community theory (Bell 2001; Hubbellment of metacommunity theory. 2001) was important by emphasizing the role of Although both island biogeography and meta- dispersal limitation for community composition,population theory could be formulated to incorpo- despite dismissing niches as being important forrate species interactions – in both cases by allowing species coexistence (building on earlier notions ininterspecific competition to increase extinction rates especially plant ecology).(e.g. Hanski and Ranta 1983) – their basic formula- Niche theory, as well as ecology in general, re-tions and later interpretations, especially in conser- garded the local habitat to be of prime importance.vation biology, modelled the world as essentially This was especially so when dealing with conserva-consisting of single species with independent dy- tion of single species or diversity. Local habitatnamics. This limited the usefulness of metapopula- patches were the objects to be managed becausetion theory for interacting communities. Only after they were threatened by human activities. Habitatsother areas of ecology, such as food webs (Polis et al. of high value for conservation were set aside as1997), had provided a spatial perspective on species reserves or land-owners given payments to manageinteractions and community structure, the time be- them in an economically less rational way, often bycame ripe for an emerging metacommunity theory traditional practices. Metapopulation and land-that linked landscape ecology, food webs, metapo- scape ecology theory allowed researchers and man-pulation dynamics and ecosystem functioning. agers to see effects of landscape and patch In the 1960s, niche theory for local communities configuration on species, but the incorporation ofwas also developed by MacArthur and Levins (see this insight into management was slow. However,Chapter 5). For example, the theory of limiting sim- the rise of the corridor concept – despite the con-ilarity was used to explain the limits to local diver- troversies surrounding it – opened up for a recog-sity (see Chase and Leibold 2002). Classical niche nition of the importance of landscape context, intheory, despite its falling popularity, forms an im- this case connectivity, for managing diversity inportant background for community ecology, and is individual patches.important for the discussion of why diversity might In addition, insect ecologists had long recognizedaffect ecosystem functioning. The complementarity the importance of larger-scale processes and dis-hypothesis (Loreau et al. 2001), which predicts in- persal in population dynamics. In fact, already increases in various aspects of ecosystem functioning the 1950s, metapopulation ideas were developed byas diversity increases, is based on niche differences Andrewartha and Birch (1954). Migration betweenin resource utilization between species. However, crop fields and other habitats, both among pestsnot until the last decade had niche theory devel- and among natural predators, was identified earlyoped to be integrated into a metacommunity frame- as important for conservation biological controlwork. Important ideas in this process were as (Barbosa 1998).follows: (1) The recognition that a species’ environ- From these sources emerged the modern view ofmental requirements (resources and environmental local habitat patches interacting via dispersal in a
  • 132. 118 APPLICATIONSlandscape consisting of habitat patches of varying Leibold et al. (2004; see also Chapter 5) offered fourquality, and a matrix of varying but lower quality simplified perspectives of how metacommunity the-between patches. Crucial for this view being accept- ory might view the world, i.e. neutral, patch dynamics,ed was the accumulation of observations that pat- species sorting and dispersal-driven metacommunitiesterns in local diversity and local interactions could (also termed mass effects, but in this author’s view thisnot be fully explained by local processes and local term is ambiguous and should be replaced to empha-environmental conditions (e.g. Ricklefs and Schlu- size the main role of dispersal for the dynamics of thister 1993; Polis et al. 1997; Holyoak et al. 2005). This type of communities). Each of these offer interestingled a new generation of ecologists to design studies insights into how diversity, species composition andin which the effects of local conditions and regional ecosystem services are affected by local (sorting) anddynamics or landscape composition could be sepa- regional (dispersal) processes, but, as noted in Chaserated more rigorously. Many such studies tested and Bengtsson (Chapter 5), the four perspectives arevery applied questions. Some examples are the im- not exclusive. The basic processes and managementportance of the amount of natural habitats in the implications of the four metacommunity perspec-surrounding landscape for parasitoids on pest in- tives are summarized in Table 9.1.sects in crops (Thies and Tscharntke 1999), and the The metacommunity view of the world is mainlyeffects of different agricultural practices and land- one of discrete patches and a matrix between patchesscape heterogeneity on insect diversity (Weibull that is non-habitable environment. The quality of theet al. 2000). In both these as well as many other matrix in this view mainly affects dispersal (migra-cases, the effects of landscape were significant and tion rates). However, as pointed out by Vandermeersometimes larger than effects of local management. and Perfecto (2007) the matrix is often habitat for many species in agricultural and forest systems. The effect of a matrix that is habitable or even similar in quality to patches is most easily included in the9.2.2 Implications of metacommunity dispersal-driven (mass effects) perspective. This view oftheory the matrix highlights the points of Oksanen (1990)Allowing theoretical predictions to guide manage- and Holt (2002) that the relative productivity ofment is risky. Many applied questions in ecology patches or of the matrix compared with patches isrequire both knowledge of ecological theory and of great importance when examining the effects ofintimate understanding of the natural history of spatial heterogeneity. Spill-over from more produc-the system under study. Furthermore, for many tive habitats or patches can greatly affect speciesapplied questions the system that needs to be un- interactions. It may complicate management, but isderstood includes not only the ecosystem, but also important for the delivery of ecosystem services suchmany aspects of humans as part of linked social– as biological control and pollination (see below).ecological systems, including sociology, economy, Some general insights for ecosystem managementpolitics and the conflicting interests of those advo- can be deduced from these perspectives (e.g. Mouil-cating different ideological views of the world (see, lot 2007), although for particular systems and ques-for example, Lawton 2007). Nonetheless, it is valid tions predictions have to be much more specific. Into examine what the emerging metacommunity the- the following, unless otherwise stated, communitiesories might predict about the management of diver- are assumed to be at one trophic level, structured bysity and ecosystem services in human-dominated competition for space or other resources, and therelandscapes (see, for example, Bengtsson et al. 2003 is a trade-off between competitive ability and traitsin the context of reserves). An analogy from evolu- associated with dispersal.tionary biology is useful: Predictions on evolution- The species sorting perspective implies thatary changes in pests or pathogens in response to management of local conditions and local habitathuman actions are of interest, although we know heterogeneity is most important to maintain diversi-that decisions in society do not fully take these into ty in a patch (Table 9.1). Because species sorting allowsaccount. the most efficient competitors to dominate patches,
  • 133. Table 9.1 Implications of different perspectives on metacommunity dynamics (Leibold et al. 2004) for management ofbiodiversity and ecosystem services (inspired by Mouillot 2007)Perspective Ecological processes Management implicationsSpecies sorting Local processes determine diversity. Proper local management most Coexistence by niche separation, local important. heterogeneity creates niches leading to Local diversity: maintain local conditions higher diversity. by management. Disturbances decrease local diversity.Low dispersal Regional diversity: maintain diversity of into patches, but some dispersal needed for local conditions on regional scale; manage communities to track environmental changes. patches differently when appropriate. Sorting of most efficient competitors results in high Maintain some connections between ecosystem functioning patches to allow environmental tracking. Local management to maintain diversity increases resource use and local ecosystem servicesPatch dynamics Metapopulation dynamics, patches fairly equal. Local and regional diversity and species Extinctions can be stochastic, caused by occurrence maintained by higher competitors or predators, disturbance-driven or connectivity between patches, and deterministic. Colonization often distance maintaining disturbance regimes. dependent, dispersal limitation may be Local diversity maintained by important. maintaining the regional species pool. Regional coexistence and diversity depends on Ecosystem services depend on dispersal trade-offs, e.g. between dispersal and of functionally dominant species or competitive ability. sets of species with similar effects on Stochastic extinctions lead to lower ecosystem functioning functioning Low dispersal rates and low regional occurrence result in lower average ecosystem services with a high variability among patchesDispersal-driven Island biogeography roots. Maintain dispersal but do not metacommunities Source-sink dynamics, patches of varying quality. homogenize region. (mass effects) Dispersal from outside local patches, i.e. other Important to identify and manage (1) between source patches or matrix, has major effects on source patches. Promote diversity of patches, or local dynamics and composition (diversity). conditions regionally to maintain (2) from matrix to Local diversity may increase with increasing sources for different species. patches dispersal of new and less competitive species into Ecosystem services depend on dispersal patches, but then decrease as immigration of from sources, but high regional good dispersers dominate dynamics, which also dispersal may prevent local sorting of leads to homogenization of regional diversity efficient species. High-quality matrix important if dispersal from matrix enhances ecosystem functioningNeutral Acts on longer time scales. Short-term management implications Individuals of different species are similar in fitness, less clear. competitive ability, etc., but are not necessarily Maintain regional dynamics and habitat equal. diversity. Balance between ecological drift and speciation Do not change fitness relations maintains diversity drastically (e.g. many forestry practices predominantly decrease fitness of old-growth species, thus decreasing diversity)The perspectives are patch-based, and the uninhabitable matrix affects only migration rates. However, in manymanaged systems the matrix is often habitat for many species and is most easily included in the dispersal-driven (masseffects) metacommunity perspective.Note: Perspectives are not exclusive (see text).
  • 134. 120 APPLICATIONSecosystem functioning increases with local diversity, ditions vary over time, dispersal may still be impor-and thus by proper local management. tant to maintain ecosystem services (Norberg et al. In the patch dynamics perspective, which is based 2001; Loreau et al. 2003). If local diversity and eco-on classical metapopulation theory, local commu- system functioning also depend on the matrix, thenities are assembled from the regional species pool matrix has to be managed in an appropriate way.and then subjected to local sorting (Bengtsson et al. Management implications of the neutral perspec-2003). At least moderate dispersal between patches tive are less obvious (Table 9.1). This is mainlyin the region is required for the maintenance of because it concerns the long-term evolutionary dy-local and regional diversity, and to ensure reliabili- namics of large-scale systems, while its short-termty of ecosystem services (Table 9.1). applied ecological consequences are unclear. When dynamics in local patches are largely de- Some simple rules of thumb for managementtermined by dispersal from source patches or ma- can be suggested by combining these perspectivestrix habitats (dispersal-driven metacommunities), local (Box 9.1a). First, it is not enough to manage habitatsdiversity may first increase and then decrease as patch-wise. Instead whole landscapes must beimmigration of good dispersers increases (Mouquet managed as networks or mosaics (Bengtsson et al.and Loreau 2003). Regional diversity is maintained 2003; Lindenmayer et al. 2008). Second, this regionalby patches with different conditions having source perspective means that a diversity of conditionscommunities with different composition. Dispersal and management strategies should be maintainedmay prevent adaptation of the local community to in a region, and that a certain degree of connect-local conditions, which may decrease ecosystem ions between habitat patches is needed to managefunctioning (Table 9.1). However, when local con- diversity and ecosystem services properly (Box Box 9.1. Metacommunity theory, landscape management and land use intensification a Rules of thumb for landscape management b Summary of predictions from following from one or several metacommunity metacommunity theory on the effects of land perspectives use intensification • Maintain local conditions by management 1 Landscape heterogeneity and higher con- (species sorting) nectivity in the landscape result in higher • Manage not only single patches but whole diversity and a larger regional species pool. landscapes (patch dynamics, dispersal-driven, This increases the possibility for local ecosys- neutral) tem services to be maintained. • Maintain diversity of local conditions in 2 The intensity of dispersal influences the ef- region (species sorting, dispersal-driven, fect of landscape context on metacommu- neutral) nity dynamics. • Maintain connections between patches 3 Homogenization of landscapes and land use without homogenizing the landscape intensification create a weedy world. (species sorting, patch dynamics, 4 Intensification of land use will affect trophic dispersal-driven, neutral) structure and strength of local interactions. • Maintain disturbance regimes close to The effect will vary depending on relative natural (patch dynamics, neutral) patch productivities and the importance of the matrix. 5 Homogeneous landscapes have lower diversity and biomass of many functional groups and thus less efficient ecosystem ser- vices.
  • 135. DIVERSITY AND ECOSYSTEM SERVICES 1219.1a). Third, natural or close to natural distur- ing by dispersal ability will take place, becausebance regimes and landscape mosaics should be species that are poor dispersers but, presumably,maintained, as most species are likely to be good competitors will not be able to persist (Neewell adapted to these over evolutionary and eco- and May 1992). As intensification of land use pro-logical time (Bengtsson et al. 2003; Lindenmayer ceeds, species with poor dispersal will approachet al. 2008). remnant metapopulations in which extinctions are not balanced by colonizations (Eriksson 1996). Long-lived species such as plants may persist for a9.2.3 Metacommunities in human- long time, but finally go extinct. The end result willdominated landscapes: effects of habitat be scattered patches with similar community com-loss and fragmentation position, dominated by the best dispersers in theFrom an applied perspective, a crucial question con- region or those common in the intensively managedcerns the effects of land use changes under the differ- matrix. That is, the metacommunity will again ap-ent metacommunity perspectives. The intensification pear to be dispersal driven, but with low local andof agriculture and forestry during the last century regional diversity and impaired ecosystem services.has resulted in an unprecedented loss of natural Alternatively, assume an originally heterogeneoushabitats and increased distances between the remain- and mosaic landscape, in which niche differencesing ones (fragmentation). In addition, management and environmental variation result in a high diversi-intensification has resulted in an increase in the area ty of species. This metacommunity will be closer to acovered by monocultures of certain crops and tree species sorting perspective, and to variable degreesspecies in large patches. At the same time, intensifi- patch dynamics. When such landscapes are fragmen-cation has resulted in an increased disturbance fre- ted and dispersal is made more difficult, the resultquency and intensity, to a larger extent in agriculture will first be sorted remnant metacommunities inthan in forestry. Note that a similar drastic increase in patches of varying quality and composition. Furtherdisturbance frequency also has taken place on most land use intensification may have different conse-fishing grounds. The high frequency of disturbance quences. In remnant patches of sufficient qualityis expected to favour certain species only. Traits such and size, species adapted to local conditions andas rapid exploitation and high dispersal ability may with high competitive ability can remain for a longmake these specific species more likely to persist in time, and the species sorting perspective prevails. ¨such landscapes (Tremlova and Munzbergova 2007). Alternatively, if local extinction rates are higher, What are the expected consequences of these pro- there will be a shift from competitive dominants tocesses on different types of metacommunities? To species sorted by dispersal, and the dispersal-domi-discuss this question, we first need an idea of what nated perspective will be more applicable.kinds of metacommunity dynamics were dominant A third scenario concerns disturbance-driven mo-in the landscape before the onset of land use changes. saic landscapes, where disturbances, e.g. fire orThen a short scenario related to the effects of frag- wind-throws, are major causes of local extinctionsmentation and habitat loss will be discussed. (e.g. Bengtsson et al. 2003). At moderate to high dis- Assume a fairly homogeneous landscape in turbance frequencies the role of species sorting willwhich communities can mainly be described as be reduced, and communities will be dominatedclose to the neutral or dispersal-driven metacommu- by fast resource trackers with high dispersal ability.nity types, because dispersal distances between Depending on disturbance frequency such land-patches are negligible. Although species may be scapes will originally appear to be dominated bydispersal limited, the short distances between suit- dispersal or patch dynamics. However, some speciesable patches allows persistence of many species and may also persist in these landscapes by escaping ina high local and regional diversity. If this landscape time rather than space, e.g. by having dormant stagesis broken up into smaller fragments by human ac- waiting for the next disturbance. Human activitiestivities, the situation will first approach the patch altering disturbances and increasing fragmentationdynamics perspective. Locally and regionally, sort- will increase the role of dispersal for species
  • 136. 122 APPLICATIONS Local Regional (a) (c) Local Regional diversity diversity (and β-diversity) Good competitors can disperse Mass effects of to all suitable patches good dispersers decrease diversity Note: Effect on β-diversity Local diversity depends on variation increased by Only good dispersers among patches and mass effects persist (assuming local on dominance of extinctions occur) dispersal vs. sorting (b) (d) Local “ecosystem Average “ecosystem High connectivity functioning” functioning” in region allows local sorting Only competitive extinctions of best competitors (species sorting) Decline if dispersal dominates over sorting Stochastic extinctions Good dispersers and environmental assumed to be less variation added efficient (trade-off) Landscape connectivity Landscape connectivity Dispersal to patch Low HighFigure 9.2 Relationships between landscape connectivity and (a) local diversity, (b) local ecosystem functioning,(c) regional diversity and (d) regional average ecosystem functioning. Suggested relationships are based onmetacommunity theory but may not always hold (see text).persistence, for many species by first approaching a decrease diversity. This is because species in thepatch dynamics perspective, but as land use intensi- region then will behave as one population (patchyfication proceeds the end result will be low-diversity populations, sensu Harrison 1991), decreasing bothcommunities of r-selected good dispersers. local and, if dispersal dominates over sorting, re- Taken together, human alterations of habitats gional diversity. A larger regional species pool in-and disturbance regimes will likely result in sub- creases the possibility for local ecosystem services tostantial decreases in habitat and landscape quality be maintained. Niche theory suggests that increasedfor many species. However, for some species, the diversity increases utilization of available resources,landscape created by humans is of exceptionally at least at low levels of diversity. Also, a diversehigh quality. These species are admittedly a minor- species pool increases the probability that at leastity but often abundant and obnoxious ones, with some species survive more variable environmentalgood dispersal ability and either generalistic re- conditions (Elmqvist et al. 2003; Loreau et al. 2003).quirements or being specialists on monocultures. Second, dispersal ability and intensity influence Some general expectations from metacommu- the effect of landscape context (connectivity) onnity theory on the effects of land use intensification metacommunity dynamics. It is predicted that loware summarized in Box 9.1b and Fig. 9.2. dispersal increases the importance of local sorting First, in human-dominated landscapes, increased in patches, while at intermediate dispersal landscapeheterogeneity and connectivity is likely to lead to factors are more important. At high dispersal, localhigher diversity and a larger regional species pool conditions in patches may again seem more impor-(Benton et al. 2003). However, high dispersal and tant, because then sorting can be expressed in allhigh connectivity may homogenize landscapes and patches.
  • 137. DIVERSITY AND ECOSYSTEM SERVICES 123 Third, homogenization of landscapes and duration (spatial and temporal scale) of phenome-increased disturbances associated with land use na, and it is a general problem that conclusionsintensification ‘select’ species with particular traits about a system depend on the scale at which wesuch as high dispersal ability, generalistic resource view it. For example, there have been thousands ofand habitat requirements, and ability to rapidly ex- laboratory and small plot experiments in appliedploit abundant resources. In essence, homogeneous ecology. Often, such small-scale studies have beenlandscapes are creating a weedy world (this term was used to guide management at larger scales, as inoriginally coined by Carl Folke; note the similarity forestry where the majority of experiments havewith classical r-selected species (MacArthur and been made on plots smaller than a hectare (100 ÂWilson 1967), such as ruderals and rats). 100 m) monitored over much less than a forest Fourth, because different organisms are likely to generation. Still it is assumed that they inform usdiffer in their sensitivity to habitat loss and increased about the large-scale and long-term effects of vari-isolation, and in their response to intensification, dif- ous forestry practices, despite the fact that this as-ferent landscapes will contain different sets of species sumption often is unlikely to hold. Management(metacommunities) and the strength of trophic inter- advice based on small-scale studies is risky.actions in patches will vary between landscapes. For However, management of real ecosystems isexample, if predators have lower population sizes or carried out on large scales, and different manage-lower dispersal rates than their prey, then increased ment practices can provide information about theisolation, smaller habitat patches and increased inten- dynamics of ecosystems on larger scales. Whilesification is predicted to decrease ratios of predators small-scale experimental model systems can be in-to prey, resulting in less predator control of prey, e.g. structive, any advice on, for example, corridorspests. However, predators may spill over from more or land use should be relevant at the scales atproductive matrix habitats to small patches or to in- which decisions about land use are made. In agri-tensively cultivated areas, resulting in higher preda- cultural landscapes such decisions are made at thetion. Thus, the mechanisms for different levels of scale of single fields to farms or even landscapespredation need to be understood in a spatial context with many farms. These decisions are made by(Oksanen 1990; Polis et al. 1997; van de Koppel et al. farmers, advisors and land use planners based on2005; Chapter 5) socioeconomic factors, naturally given landscape Finally, as a consequence of the third and fourth and soil features, but also on individual feelingspoints above, homogeneous landscapes are likely and prevailing ideologies. This creates variation into show a lower diversity and biomass of many func- management that can be used to examine the effectstional groups, except those deliberately favoured by of various land use practices in a pseudo-experi-agriculture or forestry through planting (note that mental way.planting is not the case in fisheries, which exacer- An example of this is how different farming sys-bates these effects). Changes in diversity and species tems affect biodiversity and ecosystem services.composition are often associated with changes in size The original question was if organic farming –distributions, which in turn may entail less efficient a system characterized by using no pesticidesecosystem services (Bengtsson 1998). and no inorganic fertilizers and consequently usually more perennial semi-permanent grasslands with nitrogen-fixing plants – had higher biodiversi-9.3 A selection of empirical studies ty than conventional farming systems. Clearly, plot experiments on the scale of a few metres are9.3.1 Applied questions allow almost useless for answering such a question. Itexperimental studies on management should be answered at the management scalescales of fields or preferably farms. However, at thisA major issue in ecology is on which scales larger scale, landscape structure, such as thephenomena and processes should be studied (e.g. amount of natural habitat or the diversity of habi-Peterson and Parker 1998). Scale refers to the size or tats, is important for species richness of many
  • 138. 124 APPLICATIONSorganisms from plants to insects and birds (Steffan- ogy will require more studies of how organisms dis-Dewenter et al. 2002; Benton et al. 2003). How can perse and how they utilize the matrix between focalwe distinguish farming system effects from such patches.landscape effects? A straightforward way is to account for the 9.3.2 Biodiversity in human-dominatedeffect of landscape by examining farms (or fields) landscapes: local or landscapesituated in different landscapes, but making sure management?that organic and conventional farms are evenlydistributed over the landscapes. The simplest way The intensification of agriculture in recent decades isis to choose matched pairs of organic/conventional a good example of how management practices affectfarms along a gradient of landscape heterogeneity, whole landscapes as well as the quality of habitatand then analysing the results with an analysis of patches. How these qualities affect organisms andcovariance or similar models. This approach was their interactions are exactly the kind of appliedused by Weibull et al. (2000) and has since then been question that metacommunity theories should try toextensively used by a variety of authors in Ger- explain. In the UK, the major changes in agriculturalmany, the UK and Sweden (Roschewitz et al. 2005; landscapes took place in the 1970s as it joined the ¨Fuller et al. 2005; Rundlof and Smith 2006). With European Union (Chamberlain et al. 2000), similar tothis design, it is also possible to examine whether many other Western European countries.there are interactions between the farming system It is well established that modern agriculture (andand landscape structure. For example, organic forestry, e.g. Bengtsson et al. 2003) has resultedfarming might have a large effect in homogeneous in declines in biodiversity (Donald et al. 2001;agricultural landscapes but not in small-scale mo- Tscharntke et al. 2005; Biesmeijer et al. 2006). In agri-saic landscapes. cultural landscapes the reasons are well known: Sim- However, in order to extract the variables that plified crop rotations, farm specialization, increasedbest describe the habitat and landscape factors use of pesticides and herbicides, heavy fertilizationthat organisms are likely to respond to, the biology and loss of marginal and natural habitats, singly orand ecology of the organisms under study must be together, have affected various organisms negative-understood. Here, we often encounter an imbalance ly. For organisms depending on traditionally man-in our knowledge: Most field biologists working on aged habitats such as semi-natural grasslands, theiran organism group have an intuitive feeling for habitats nowadays occur as isolated patches in a seawhich local habitat factors influence the presence of crops or forest.of species. On the other hand, understanding re- To reverse this trend, intuitively we would sug-gional influences on species distributions require gest that diversity could be restored by increasingthat we also have a good idea about the factors the amount of seminatural habitats and landscapeaffecting dispersal and movements of organisms heterogeneity, and decreasing the use of pesticidesin the landscape. Such knowledge is more difficult and other intrants. This has led to suggestions thatto obtain and for many organisms quite scanty. organic farming practices would result in a higherHence, while the local factors included in an analy- farmland biodiversity, because organic farming issis often capture what organisms really respond to, supposed to have, among other things, a higherregional scale variables are often merely informed crop diversity, more complex crop rotations withguesses, and determined by the availability and res- more semi-permanent fields like leys and pastures,olution of the landscape maps used to measure, for and not using pesticides, herbicides and inorganicexample, proportion of habitats, landscape heteroge- fertilizers. Organic farms are also more likely to beneity and edge zones. Thus, almost by default region- mixed farms with both animal and crop produc-al influences will contribute less to the explained tion. However, all these features of organic farmingvariation in species distributions and diversity than may not be realized on single farms, although pes-the more precisely known local habitat factors. In ticides, herbicides and inorganic fertilizers alwaysview of this, major advances in metacommunity ecol- are prohibited. In many landscapes organic farmers
  • 139. DIVERSITY AND ECOSYSTEM SERVICES 125 Table 9.2 Organism mobility in relation to contributions of landscape, farming system and local habitat type to (A) species richness (diversity) and (B) variation in species composition on 16 farms in mosaic agricultural landscapes in east central Sweden Mobility: PlantsÆ CarabidsÆ Butterflies A. Diversity Landscape (þ) r ¼ 0.47 (þ) r ¼ 0.58 (þ) r ¼ 0.52 Farming system ns (þ) P ¼ 0.02 ns Habitat þ þ þ B. Species composition Landscape Æ 0% % 5% % 10% Habitat % 10% % 5% > 10% Model r2 15% 15% 20% Mobility is assumed to increase from plants to carabids to butterflies, based on data on adult movements. Habitat types were pasture, semi-permanent ley and field margin. The ¨ table summarizes results in Weibull et al. (2003) and Weibull and Ostman (2003). For diversity (A), only habitat was significant in the full model, but effects of landscape (highest correlations with landscape variables shown) and farming system (carabid diver- sity lower on organic farms) were found in other analyses (see Weibull et al. 2003). þ, P < 0.05; ns, not significant. In (B) the figures refer to the proportion of the variation in species composition explained by canonical correspondence analysis (CCA) axes related to landscape or habitat (Weibull ¨ and Ostman 2003) and interpreted by this author. Model r2 is the total variation explained by the CCA model.may for economic reasons specialize on a single or a margins or spiders, although landscape heteroge-few crops. This is especially a risk as contracts with neity at the farm scale influenced these groups pos-large retailer chains such as Wal-Mart or Sainsbury ¨ itively (Weibull et al. 2000, 2003, Oberg et al. 2007).become more common. In addition, local habitat type was more important The expected positive effects of organic farming for diversity than landscape factors (Table 9.2).on biodiversity are one of the reasons it has been The importance of local habitat and landscape forincluded in agri-environmental schemes. An im- species composition appeared to vary between theportant applied question is then whether organic three organism groups (plants, carabids and butter-farming really delivers a higher biodiversity. For flies) in relation to mobility (Table 9.2). The propor-more basic questions, the advantage of using organ- tion of explained variation accounted for byic farming is that it is a farm-scale manipulation of landscape measurements increased from almostlandscape structure. Organic farming mainly af- zero for plants to almost 50% in butterflies, whichfects how fields are managed, and, for organisms in general are the most mobile of the three groups.in natural habitat patches, it may mainly increase Other studies have also found that responses to thethe quality of the matrix (Vandermeer and Perfecto landscape may vary with differences in mobility2007). This is likely to have effects on metacommu- (Steffan-Dewenter et al. 2002; Tscharntke et al. 2005nity dynamics and trophic interactions. for an overview). This is in accordance with meta- Our studies were conducted in a mosaic agricul- community theory. The importance of the land-tural landscape in central Sweden, where fields are scape, either other patches or the matrix, shouldinterspersed with areas of glacial till and bedrock increase with organism mobility and dispersalwhich are not used for crops but often contain semi- rates (second expectation, above; Fig. 9.2a).natural pastures or small forests. We did not find Such patterns may also be found for the sameany positive effects of organic farming on diversity group in different landscapes, as suggested by aamong butterflies, carabid beetles, plants in field study of plants in field buffer zones in Finland
  • 140. 126 APPLICATIONS(Ma 2006). In the most connected landscape, dispers- scapes, but this is also where farming is presentlyal effects seemed most important and resulted in a less profitable and farmland risks being abandonedlower regional diversity and more similar local with current policies. Any scheme that resultscommunities, associated with the dominance of a in continued farming in these areas will counternumber of competitive species which were argued decreases in biodiversity and also have other socio-to disperse more easily in this landscape. In the economic benefits, and organic farming is oneleast connected landscape, species such as dande- such scheme. In the most homogeneous landscapes,lions (Taraxacum spp.) and Aegopodium podagraria the increase in landscape quality that organic farm-were more common, supporting the view that frag- ing entails will mainly favour a few generalisticmentation of habitats may result in local commu- species that are quite common in other parts ofnities being more dominated by weedier species the human-dominated landscape, such as villageswith high dispersal ability. The landscape with in- and urban areas. The fact that the effect appearstermediate connectivity had the highest regional large is that the most intensively used land-diversity, resulting in the predicted hump-shaped scapes are so extremely depauperated that any in-pattern of regional diversity in relation to connec- crease in diversity will appear substantial. Fortivity (see Fig. 9.2c). example, in many agricultural areas in The Nether- Our results showing that organic farming had no lands, regional bumble bee diversity is in essenceeffect on diversity were at odds with generally held only four species (Kleijn and Langevelde 2006),notions, and also with the majority of other studies whereas more than 10 species are regularly found(analysed by Bengtsson et al. 2005). Why was Sweden in agricultural and urban areas in southern and cen-different? The large variation in effect size among ¨f tral Sweden (Rundlo 2007; Andersson et al. 2007;studies in the literature, coupled with closer inspec- J. Risberg, J. Bengtsson and B. Cederberg, unpub-tion of individual results, suggested that the effect of lished data).farming system might be larger in more intensively In addition, there appears to be a positive effectmanaged and more homogeneous landscapes than of the amount of organic farming in the landscape.in small-scaled mosaic landscapes with many semi- ¨ Rundlof et al. (2008) found that for butterflies therenatural habitats. This is in essence a test of the hy- was a higher local diversity in landscapes with apothesis that at low dispersal the importance of local high proportion of organic farming within a 1 kmsorting increases, while at intermediate dispersal the radius, irrespective of farming system. This indi-landscape is more important (see above). If this is cates that, along these long gradients from homo-true, an interaction between landscape and farming geneous to mosaic landscapes, the landscape effectsystem is expected in studies conducted along a suf- on diversity can be substantial.ficiently long gradient of landscape heterogeneity. An important message for managers emergingExactly this was found independently in southern from these studies is that landscape effects on di- ¨fSweden for butterflies (Rundlo and Smith 2006) versity or species composition of plants and insects ¨fand bees (Rundlo 2007), and in central Germany often may be found on the scale of management, infor arable weeds (Roschewitz et al. 2005) and bees agricultural landscapes usually individual farms.(Holzschuh et al. 2007). Such results are highly infor- In our studies in Sweden, landscape variablesmative from a basic metacommunity perspective, but measured at the 25 km2 scale were not significant,if they are common they also pose problems when whereas heterogeneity at the farm and multiple-implementing ameliorative measures such as agri- field scales was (approx. 0.5À2 km2) (Weibull et al.environmental schemes: Should organic farming be 2000). This means that the choices of individualencouraged only in homogeneous and intensively landowners often can have a large influence onmanaged landscapes which have lower regional di- diversity, and identifies the farmers as importantversity, because this is where the largest gains in decision-makers for conservation. However, therediversity can be achieved? may be more immediate gains for farmers manag- In fact this may not be the case. The highest biodi- ing their land for higher diversity, to which we willversity is usually found in heterogeneous land- now turn.
  • 141. DIVERSITY AND ECOSYSTEM SERVICES 1279.3.3 Local and regional effects on Pollinators often rely on semi-natural and less dis-ecosystem services turbed habitats, but still use flowering crops for nectar and pollen. However, biological control ofThe species most likely to profit from organic farming pests in crops provides an even better example ofand other agri-environmental schemes decreasing how metacommunity theories concerning speciesagricultural intensity can be termed ‘the common interactions in spatially structured habitats natural-biodiversity’ (I owe this term to Anki Weibull, and ly interact with very applied questions. There ishave found it very useful to distinguish these species also a direct economic benefit of biological controlfrom the red-listed species that often – and rightly so – for farmers.are the major concern in conservation biology). These Two types of biological control are discussed inspecies are not rare, but may be declining. They are the literature. One is control of pests by specializedcommon enough to have direct effects on ecosystem predators. In this case the trick is to maintain bothservices, such as predators controlling pests, insects predators and pests so that control continues overpollinating fruits and vegetables, or earthworms time, either naturally or by continuous addition ofmaintaining soil fertility. The common biodiversity predators. This has been quite successful in, forprovides farmers, landowners and society with a example, greenhouses. The second is the control ofnumber of ecosystem goods and services whose va- pests by generalist predators, which do not dependlues usually are unknown and thus never enter the on, but readily feed on, the pest. The latter typeeconomic calculations that constrain how ecosystems often relies on predators being present in the fieldsare managed. Clearly, a useful community ecology when potential pests emerge or colonize. Theseshould provide a better understanding of ‘the ecology predators usually depend on habitats surroundingof ecosystem services’. the fields for persistence. Early examples were po- Many ecosystem services provide interesting lyphagous insects such as carabids feeding on cere-cases of natural habitats interacting with crops, al aphids (Ekbom et al. 1992) and parasitoids onwith the quality of the matrix of agricultural fields rape seed beetles (Thies and Tscharntke 1999), butbetween natural habitats playing an important role. Crop field Edge zones, ley Generalist predator guild Intra-guild Movement predation Spiders Spiders Carabids Carabids Biological control Context Pest insects Insect Aphids Plant larvae etc. material Collembola Collembola (detritivores) (detritivores) Interactions = mechanismsFigure 9.3 The biological control of a pest in a crop, for example aphids, is only a part of the interactions that thepredators (and pest) are involved in, and proper management requires that all interactions in the food web are managed.In this case, including the edge habitat results in a basic food web question in a spatial context, incorporating predator–prey dynamics and spatial subsidies.
  • 142. 128 APPLICATIONSspiders have also been studied in this context biological control will be affected by the produc- ¨(Schmidt and Tscharntke 2005; Oberg et al. 2007). tivity of predators in the second habitat (field mar- The basic system is shown in Fig. 9.3. In crop gins) and their movements, which can depend onfields, aphids on cereals can be serious pests, but distance but also on the abundance of other prey inthey are preyed upon by predators which require the crop.more permanent habitats like field margins for sur- For control of pests such as aphids with multiplevival and overwintering. The predators also feed on generations during a season, it is crucial that theother prey, both in the field and in the margins. predators are present in the fields when aphidsPresented in this way, this clearly is a system in colonize (Ekbom et al. 1992). Aphid abundance inwhich the basic theories for trophic interactions in one year does not depend on particular fields andspatially structured habitats developed by Holt their margins; their dynamics are on a much larger(2002), Oksanen (1990) and Polis et al. (1997) are scale. Hence aphids can be considered as ‘donorapplicable. It is a metacommunity problem where controlled’ and we cannot expect the predators inthe strength of trophic interactions depends on local a field to regulate aphids, only keep their numbersand regional (landscape) factors. The strength of below damage thresholds. (a) (b) 0.08 600 Organic Aphid establishment 5 Number of aphid days Without predation 2 Conventional 0.07 500 With predation (winged aphids/tiller) 5 3 400 0.06 1 Heterogeneous 300 landscapes 0.05 4 200 1 0.04 3 4 100 2 0.03 0 100 200 300 400 500 600 1 Conv 1 Org 2 Conv 2 Org 3 Conv 3 Org 4 Conv 4 Org 5 Conv 5 Org Perimeter/area ratio (c) 1 Predation rate on aphids r = –0.47, n=8, ns 0.8 0.6 0.4 0.2 –4 –2 0 2 4 6 Diversity (residuals from landscape–carabid diversity regression)Figure 9.4 Effects of organic farming and landscape heterogeneity on aphid establishment in spring cereal fields, andthe total number of aphid days (an indicator of yield loss) on 10 farms in central Sweden in 1999. Landscapeheterogeneity was measured as the perimeter-to-area ratio, high values indicating mosaic landscapes. Organic farms areindicated with closed symbols, conventional farms with open symbols. (a) Aphid establishment during the earlycolonization phase. (b) The number of aphid days in areas within (hatched+black) and outside (black) predator ¨exclosures. Effects of organic farming and landscape were both significant (Ostman et al. 2001). (c) Predation rates onaphids glued on paper and placed in the field in relation to diversity. Diversity was measured as the residuals from the ¨regression of carabid diversity (y-axis) versus landscape heterogeneity (x-axis). Extracted from data in Ostman et al.(2001) and Weibull et al. (2003).
  • 143. DIVERSITY AND ECOSYSTEM SERVICES 129 We examined the biological control of the bird- farmers rely on natural predators for pest control,cherry oat aphid Rhopalosiphum padi by the guild of they represent a substantial value for farmers, indi-generalist predators, and could by exclusion experi- cating that landscape management is an importantments roughly estimate the gain in cereal yields re- tool for achieving economically sustainable farms.sulting from predation on aphids and thus the However, when attempting to estimate economiceconomic value of predation in one particular year value, we encounter different perceptions of econ- ¨in one particular region (e.g. Ostman et al. 2001, 2003). omists and ecologists. For an economist, the eco-The results are summarized in Figs 9.4 and 9.5. nomic value on the market is the marginal increase First, predators had a larger impact on aphid po- in yield values under the present rate of servicepulations in heterogeneous landscapes and on or- delivery (Fig. 9.5a). This value usually levels offganic farms. In both cases, aphid numbers early in when the service is available in abundance, imply-the season were lower (Fig. 9.4a). This could be at- ing that ecosystem services are worth less whentributed to a higher predation rate on aphids (Ost- ¨ they function well. For ecologists, the value of aman et al. 2001), resulting in lower numbers of aphids service such as predation is the total effect of pre-during the growing season. Further, exclusion of dators, which we measured as the effect of exclud-predators resulted in a much higher abundance of ing all (or most) predators (Fig. 9.5b). This meansaphids (Fig. 9.4b). Because there is a direct correspon- that economists and community ecologists maydence between the number of aphid days on the crop perceive the value of an ecological service quiteand yield loss, predation clearly affects the yield a differently. ¨farmer gets (Ostman et al. 2003). However, there wasno relation between carabid diversity and predationrates (Fig. 9.4c). This suggests that either the abun- 9.3.4 What have we learned in the contextdance of particular species rather than predator di- of metacommunity ecology?versity plays a role for predation, or predator First, metacommunity theories will be importantdiversity in this area (15–35 spp.) was sufficiently tools in analyses of interactions between specieshigh for the relation between diversity and ecosys- living in adjacent habitats, such as crop fields,tem function to saturate towards an asymptote. field margins and semi-natural habitats. These si- Finally, using the price of cereals we could estimate tuations are common and crucial to understand forthe value of natural predators for farmers in 1999 as applied questions such as biological control andapproximately  40 or $37 per hectare. We do not pollination.know whether this particular year and these land- Second, the ecological processes involved in main-scapes are representative, and a generalization clearly taining biodiversity and ecosystem services in agri-requires more studies. Nonetheless, because organic cultural landscapes often act on larger scales than (a) (b) Marginal value = market price 40 Value of product % yield increase 35 (e.g. yield) 30 25 20 c Perceived ecological o 15 value of biocontrol 10 5 0 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Rate of ecosystem service Predation rateFigure 9.5 Different perceptions of the value of ecosystem services such as biological control in ecology and economy ¨(a). In (b) the perceived ecological value of natural predators for biological control is indicated. Data from Ostman et al.(2001, 2003).
  • 144. 130 APPLICATIONSindividual fields, but for many organisms such as questions, despite the fact that full control over theinsects still on scales that are possible to manage by treatments is not achieved. However, this is balancedindividual decision-makers such as farmers, i.e. on by an increased realism and applicability of the re-whole farms or multiple field-matrix mosaics. This sults in management and policy, and the possibilitymay be different from many birds and game animals, of applying theoretical frameworks on larger scalesand from forestry, which may require management rather than extrapolating from small-scale modelof larger landscapes to include the relevant distur- systems whose relevance for management often isbance regimes (see Bengtsson et al. 2003). questioned. Third, these examples illustrate that metacommu-nity theories can be useful because they address the Acknowledgementsinterplay between local and regional dynamics atscales that are relevant to managers. These theories ¨ ˚ ¨ ¨ I thank Tomas Part, Jens Astrom, Erik Ockingercan thus be used to guide management strategies for ´ and Henrik Andren for comments on variousconservation and ecosystem services in human- parts of the manuscript, my previous and presentdominated landscapes. At the same time, applied PhD students and collaborators for their intellectualstudies can provide tests of metacommunity theory input, and the organizers of the Community Ecolo-on the roles of landscape and local factors for diver- gy course in 2005 in The Netherlands, where thesity and ecosystem functioning, and of theory on two lectures on which this chapter is based weretrophic interactions in spatially heterogeneous eco- held. I have been funded by the Swedish Researchsystems. Variation in management may provide the Council and FORMAS for the research reportedtreatments needed to more rigorously examine these here.
  • 145. CHAPTER 10 Community ecology and management of salt marshes Jan P. Bakker, Dries P.J. Kuijper and Julia Stahl elevation than the adjacent depressions. Erosion10.1 Introduction protection measures, coastal defence or agriculturalSalt marshes are ecosystems at the edge of land and purposes play no critical role. They occur in sandysea. They are influenced by tidal movement. It is the back-barrier conditions on islands such as Mellum,interaction of the vegetation and sediment trapped Spiekeroog (Germany), eastern parts of Amelandfrom inundating water that creates a salt marsh. and Schiermonnikoog (The Netherlands). OnCurrently, there are about 176 000 ha of salt marsh the other hand, semi-naturally developed saltaround the Baltic and Atlantic coasts of Europe. For marshes either have an extensive wide-stretchedthe Wadden Sea the area of the salt marshes can be natural creek system but are affected by measuressubdivided into $13 000 ha of salt marshes on to enhance livestock grazing (e.g. back-barrier con-the barrier islands and $26 000 ha of salt marshes ditions at the peninsula of Skallingen (Denmark)along the mainland coast (Bakker et al. 2005a). Back- or feature a salt marsh within sedimentation fieldsbarrier marshes develop at the lee side of the sand with a man-made drainage system by ditchesdune system of barrier islands in front of the main- and are grazed by livestock or left fallow after pre-land coast, where foreland marshes develop. vious grazing (e.g. artificial marshes along the Salt marshes are considered to represent one of mainland coast of The Netherlands, Germany andthe few pristine ecosystems in North-West Europe. Denmark; Bakker et al. 2005a).That may be true for some marshes, others are Abiotic conditions on salt marshes are related todistinctly influenced by humans (Davy et al. 2009). the inundation period and frequency dependingThe role of salt marshes along the coast has been on an elevation gradient running from the uppertransformed from primarily coastal protection marsh at the foot of a dune at the back-barriertasks to a combination of the former with nature marshes, or the foot of the seawall along the main-conservation interest. Large areas are nowadays land coast to the intertidal flats. This elevationalassigned to nature reserves or national parks. gradient also influences the rate of sedimentation,These designations initiated critical debates on nat- which is the main driver of plant succession.uralness and suitable management of marshes The rate of sediment input on salt marshes variesand concern especially the need and intensity of from < 5 mm/year on sandy back-barrier marsheslivestock grazing (Bakker et al. 2003a). to up to 20 mm/year on marshes in sedimentation Naturally developed salt marshes feature a self- fields (Bakker et al. 2002). This results in a distinctstimulated development and geomorphological zonation of plant communities (Bakker et al. 2002),condition and growth that are not affected by invertebrate communities (Andresen et al. 1990),humans. They show a natural drainage system avian herbivores (Stahl et al. 2002) and mammalswith meandering creeks and levees with higher (D.P.J. Kuijper unpublished data). 131
  • 146. 132 APPLICATIONS In this chapter we will discuss the naturalness iment is higher at the low marsh than at the higherof salt marshes and their plant cover and the inter- marsh. Apart from the zonation from low to highaction of the vegetation with abiotic conditions, marsh, the thickness of the sediment layer changessuch as sediment and nutrient input, and with biotic over time from a young marsh to an older marsh.conditions, such as wild herbivores and livestock. The back-barrier salt marsh of the Dutch islandWe will particularly address the long-term dynam- of Schiermonnikoog shows such a successional pat-ics of salt-marsh communities. We will demonstrate tern. The eastern part of the island gradually ex-to what extent the findings of small-scale experi- tends further eastward. Hence, a chronosequencements on individual salt marshes can be generalized representing vegetation succession (De Leeuw et al.to add to our understanding of community ecology 1993; Olff et al. 1997) has established with veryof salt marshes, and how this knowledge can be young marsh (from 0 years onwards) at the farapplied for management purposes. east and older marshes (up to 150 years) more to the west (Fig. 10.1). Increasing age of the marsh coincides with a thicker layer of sediment resulting10.2 Natural salt marsh: the back-barrier from tidal inundation. Thus, the eastern part ofmodel including a productivity gradient Schiermonnikoog features a matrix of two phenom-Barrier islands in the Wadden Sea feature sandy ena: zonation and succession. While walking frombeaches along the North Sea and silty salt marshes east to west at high or low elevation levels, succes-along the Wadden Sea. Sedimentation of fine sus- sion of the higher and lower marsh can be studied,pended material (silt or clay) can take place in respectively. With the sediment, organic matter in-the shelter of dunes. The geomorphological condi- cluding nitrogen is imported. The nitrogen pooltions of the sandy subsoil show a gradual slope of the top 50 cm of the soil, i.e. the rooting depthfrom the foot of the dunes towards the intertidal of most plant species, is positively related to theflats. As the period of inundation is longer and the thickness of the sediment plus underlying sandyfrequency higher at low elevation, the input of sed- soil. By comparing various back-barrier systems(a) <1809–1848 1874–1894 1913–1955 1964 1974 1986 (b) 1993 1500 Surface area (ha) 1000 500 1km 1800 1900 2000Figure 10.1 (a) The development history of the eastern part of the Dutch Wadden Sea island of Schiermonnikoog. Thedifferent shadings represent different age classes on the basis of maps and aerial photographs. (b) Development of thesize of the vegetated marsh and dune area on the eastern part of Schiermonnikoog from 1989 onwards. After Van derWal et al. (2000b).
  • 147. COMMUNITY ECOLOGY AND MANAGEMENT OF SALT MARSHES 133in the Wadden Sea (Schiermonnikoog, The Nether- The productivity gradient (chronosequence) onlands; Terschelling, The Netherlands; Skallingen, Schiermonnikoog is accompanied by plant speciesDenmark), this appeared to be a general phenome- replacement. The unproductive lower salt marshnon (Olff et al. 1997; Van Wijnen and Bakker 1997). is dominated by Salicornia spp., Puccinellia maritima, Soil nitrogen is a limiting factor for plant produc- Plantago maritima and Limonium vulgare, whereastion in salt-marsh systems (see overview in Davy et al. the oldest stages are dominated by Atriplex portula-2009). As the nitrogen availability is positively related coides. The unproductive higher marsh featuresto the nitrogen pool (Bakker et al. 2005b), the plant Puccinellia maritima and Festuca rubra followed byproductivity increases with a growing thickness of Artemisia maritima and, eventually, Elymus athericusclay layer (Van de Koppel et al. 1996). In other (Elytrigia atherica) at the productive marsh (Olff etwords the chronosequence of increasing thickness of al. 1997; Van der Wal et al. 2000a). Both at the lowsediment represents a productivity gradient. and high salt marsh, succession eventually features a tall canopy of Atriplex portulacoides or Elymus athericus, respectively. Recently, it was noticed that Elymus athericus spread into lower elevation at10.3 Effects of plants on herbivores older marshes (Olff et al. 1997). These tall plant(bottom-up control) species outcompete other species by light intercep-Along the productivity gradient the density of tion (Huisman et al. 1999; Van der Wal et al. 2000a),the wild herbivores such as different species with subsequent decline in plant species richnessof Arctic geese (e.g. brent goose Branta bernicla berni- (Bakker et al. 2003b).cla, barnacle goose Branta leucopsis), brown hares Herbivores are evicted by plant succession.(Lepus europaeus) and rabbits (Oryctolagus cuniculus) Goose numbers were estimated at young, interme-initially increases to an optimum at intermediate pro- diate and older parts of the salt marsh on Schier-ductivity, but declines at sites with high productivity monnikoog between 1971 and 1997 (Fig. 10.2). In(Van de Koppel et al. 1996). According to theory, at the late 1970s brent goose numbers were high in thesites with low productivity, plant biomass is too low old marsh. However, goose numbers declined sig-to support a herbivore population, and plant growth nificantly in the following 20 years (Van der Wal etwill be regulated by bottom-up effects such as nutri- al. 2000b). This decrease is not related to a decreaseent availability (Oksanen and Oksanen 2000). With in size of the area. On the contrary, the surface areaincreasing productivity a shift from bottom-up to top- increased over the years as a result of sedimenta-down effects is expected to occur. Top-down regula- tion (Fig. 10.1). Goose numbers increased in thetion of plant biomass occurs at sites of intermediate intermediate aged salt marsh followed by a slightlevels of productivity, and herbivore population will but significant decrease towards 1997. Develop-be top-down regulated by carnivores at high produc- ment of new young marsh in the east led to ativity (Oksanen and Oksanen 2000). However, in the further eastward movement and an increase ofabsence of carnivores (e.g. one of our study systems, goose abundance (Van der Wal et al. 2000b). TheSchiermonnikoog) bottom-up effects remain to play decrease in number of brent geese at the olderan important role even at highly productive sites. marsh coincided with a change in vegetation com-Herbivore density can decrease even in the absence position. In 1977, when goose abundance wasof carnivores. Intake rate of geese levels off or declines still high, the clonal shrub Atriplex portulacoideswith biomass above a certain threshold (Van der was lacking. Since then, the Atriplex communityGraaf et al. 2006). Forage quality declines at sites has spread into the lower elevation salt marsh,of high biomass and tall canopy (Van der Wal et al. and this coincided with the observed decline in2000a; Kuijper and Bakker 2005) featuring a decreas- goose numbers. Part of the Limonium communitying leaf–stem ratio. This bottom-up control of herbi- was transformed into the Atriplex portulacoides com-vore density at high productivity sites is referred to as munity. The open Limonium vulgare communitythe ‘quality threshold hypothesis’ (Van de Koppel et harbours the preferred goose food plants such asal. 1996; Olff et al. 1997; Huisman et al. 1999). Puccinellia maritima, Festuca rubra and Triglochin
  • 148. 134 APPLICATIONS (a) Comparably, on the high elevation salt marsh, foraging patch choice and spatial distribution of 2000 Old b c brown hares is influenced by the ongoing vegeta- 1600 Schiermonnikoog a tion succession. The tall-growing plants Elymus 1200 athericus and Artemisia maritima are invading at 800 these sites with short vegetation consisting of the preferred food plant for hares, Festuca rubra (Kuij- 400 per et al. 2008). The increasing abundance of theseMaximum number of geese per day 0 tall-growing plants, which are not preferred as food (b) plants, reduces the grazing intensity of hares. As a 1000 Intermediate aged result, hare numbers decrease with increasing salt- 800 marsh age; hence, they are also evicted by vegeta- 600 tion succession (Kuijper and Bakker 2008). 400 200 10.4 Effects of intermediate-sized 0 herbivores on plants (top-down control) 0 (c) Are small herbivores only a victim of plant succes- 1000 Young sion? Studies on American salt marshes show that small- to medium-sized herbivores can regulate 800 plant biomass. For instance, grazing by insects 600 (Bertness and Shumway 1992), crabs (Bortolous 400 and Iribarne 1999), snails (Silliman et al. 2005) and 200 greater snow goose (Chen caerulescens atlantica) 10 (Smith and Odum 1983) can regulate plant biomass 0 1974 76 78 80 82 84 86 88 90 92 94 96 in Spartina-dominated marshes. The effects of lesser snow goose (Chen caerulescens caerulescens) on sub-Figure 10.2 The number of brent geese in (a) old, (b) arctic marshes along the Hudson Bay, Canada, areintermediate and (c) young parts of the salt marsh of another example (Jefferies et al. 2006). But what isSchiermonnikoog between 1974 and 1997. Maximumnumbers of geese counted per day ± SE is given for all known about the effects of intermediate-sized her-years separately. The location of the three parts of the bivores in European salt-marsh systems?marsh is indicated in the inset in (a). The absence of barsindicates no data, unless stated otherwise. The sizes ofthe study areas were 58.2 ha, 39.8 ha and 78.9 ha in 10.4.1 Experimental evidence1977, and 97.4 ha, 37.5 ha and 79.2 ha in 1996 for theold, intermediate and young marsh, respectively. After Theory predicts the effects of herbivory to changeVan der Wal et al. (2000b). along a productivity gradient. The strongest top- down effects are predicted at sites of intermediatemaritima, which were replaced by non-preferred productivity (Oksanen et al. 1981). At the back-bar-species such as Artemisia maritima, Atriplex portula- rier salt marsh on Schiermonnikoog the wild browncoides and Limonium vulgare itself (Van der Wal et al. hares occur year round, whereas brent and barnacle2000b). However, the losses of the Limonium vulgare geese are spring-staging visitors on their way tocommunity were compensated for by an increase in arctic breeding grounds (Stahl 2001). Althoughthis community in newly developed parts of the rabbits are also found at the salt marsh, theirsalt marsh at the east. We observed that ongoing grazing pressure is more than a factor of 10 lowerplant succession pushed the geese eastward and than that of hares and geese, and they mainly for-geese had to follow the changing vegetation or, in age along the foot of the dunes high on the marshother words, ‘vegetation succession evicted spring- (Kuijper and Bakker 2005). Hence, their role on saltstaging geese’ (Van der Wal et al. 2000b). marshes is expected to be low. Exclosures were
  • 149. COMMUNITY ECOLOGY AND MANAGEMENT OF SALT MARSHES 135established at four sites along the chronosequence 2.0on the island of Schiermonnikoog. The sites were aestablished 1, 8, 20 and 30 years previously, when 20 yearsexperiments started. At each site, four exclosures 1.5 1yearwere established in autumn 1994: two in the high 30 years Axis 2marsh and two in the low marsh. Every exclosureplot included three treatments. ‘controls’ were free- 1.0ly accessible to geese and hares. ‘goose exclosures’kept geese out and allowed hares to enter freely. 8 years 0.5‘full exclosures’ excluded both geese and hares(Kuijper and Bakker 2005). Dropping counts de-monstrated that different herbivores were success- 0.0fully excluded in the treatments. The vegetation 0 1 2 3 4 5was monitored from 1995 to 2001. 2.0 b10.4.2 Effects of herbivores at high marsh 30 years 1.5Multivariate analyses revealed that full exclosures 20 years Axis 2in the 1-, 20- and 30-year-old marshes showed over-all a different shift in plant species composition 1.0 1year 8 yearscompared with goose exclosures and control plots(Kuijper and Bakker 2005), whereas the goose ex- 0.5closures did not differ from the control plots. How-ever, these changes in cover of individual plantspecies did not show consistent responses to treat- 0.0ments. To study the effects on vegetation species 0 1 2 3 4 5composition, detrended correspondence analysis Axis 1(DCA) was used. This analysis orders a data setand plots data points that are most similar close Figure 10.3 The change of position of the centroids of the quadrats in the ordination diagram between 1995together in a diagram. DCA can be used to show and 2001 is shown by arrows on (a) the high marsh andgraphically how the plant community structure, (b) the low marsh on Schiermonnikoog. Note thetaking the changing abundances of all plant species differences in scale between (a) and (b). Treatments areinto account, is changing in response the different indicated by thick lines (full exclosure), thin lines (control)treatments. First, when all vegetation releves ´ and dashed lines (goose exclosure). Sites of different ages are indicated by different symbols: closed circle, 1 yearwere ordered in the DCA, typically early succes- old; open circle, 8 year old; closed triangle, 20 year old;sional species such as Elymus farctus, Parapholis and open triangle, 30 year old marsh. After Kuijper andstrigosa and Ammophila arenaria were located at the Bakker (2005).left-hand side of the diagram. The typically latesuccessional species Elymus athericus was at theright-hand side, and Festuca rubra and intermediate each treatment, indicative of the averages of treat-successional species were in the middle of the dia- ments, at each site revealed different starting posi-gram (Fig. 10.3a). The ordination showed an order- tions in the diagram. This resulted from theing of plant communities typical of young marshes different species composition at the establishment(left in Fig. 10.3) to older marshes (right in Fig. 10.3). of the exclosures. The centroids of all treatmentsSecond, the positions of all exclosures (and con- (control, goose exclosure and full exclosure) at thetrols) at the start and at the end of the experiment youngest sites moved in the direction of increasingwere included in these diagrams to show the cover of Festuca rubra, whereas all other centroidschanges in plant community. The centroids of moved towards increased cover of Elymus athericus.
  • 150. 136 APPLICATIONSThe magnitude of this change did not show consis- species. Moreover, all sides converge to the sametent differences between treatments, as indicated by point, indicating that all sites started to resemblethe lengths of the arrows. Only the full exclosures in each other in species composition. The largestthe 20 year-old marsh showed larger changes than changes in plant species occurred in the full exclo-the other treatments. This shows that, at all but the sure at the 1 yr-old marsh; this is indicated by theyoungest sites, Elymus athericus increased at the largest vector (Fig. 10.3), which describes the changeexpense of Festuca rubra, and, despite of the signifi- in community composition. Here, the largest in-cant treatment effects, no clear pattern in the direc- crease in vegetational cover of late successional spe-tion of plant communities could be detected. cies occurred in the absence of herbivores (Kuijper and Bakker 2005). These experiments revealed that grazing by in-10.4.3 Low marsh termediate-sized herbivores retards vegetation suc-The effects of the treatments on the species compo- cession and that these top-down effects are mostsition in the low marsh were more pronounced than pronounced at low, young salt marshes. The openthose in the high marsh. On the low marsh, the full vegetation in the young unproductive marshes of-exclosures showed a different shift in species com- fers the opportunity for late successional species toposition in the 1-, 8- and 30-yr-old marshes, but not become established as long as selective grazing byin the 20-year-old marsh where highest herbivore herbivores is absent. Once late successional speciesdensity occurred (Kuijper and Bakker 2005). Over- have established, they will spread more rapidly inall, the full exclosures explained most of the varia- the absence of herbivores, indicating that establish-tion in the shift in species composition. Typical ment is actually the limiting factor in this invasionplant species that increased in cover inside full ex- and herbivory can retard further spread. In theclosures at the 1-year-old marsh were Atriplex por- absence of herbivores, late successional speciestulacoides and Festuca rubra, whereas Salicornia spp. can directly invade, during the ‘window of oppor-increased the least compared with the other treat- tunity’ in young marshes, and will dominate thements. Also, the typically late successional species vegetation at an earlier stage. Hence, the top-Elymus athericus had become established, whereas it down effects of the herbivores combined with thecould not be found in the area surrounding the bottom-up effects of the vegetation can retard veg-exclosures at the young marsh site. At the 8-year- etation succession in these salt-marsh systems forold marsh, Festuca rubra increased most in the full several decades (Kuijper et al. 2004).exclosures at the expense of Puccinellia maritima. At A second conclusion is that small migratorythis site the goose exclosures showed a shift in herbivores such as geese alone do not show aspecies composition that was intermediate between long-lasting impact on the vegetation, but the com-the full and control treatments. At the oldest site (30 bination with hares is essential to retard succession.years) two typically late successional species It was argued that the hare is the most important ofincreased in cover and dominated the vegetation: these two herbivores in determining the effects.in one full exclosure Elymus athericus, in the other First, migratory geese use the salt marsh in springexclosure Atriplex portulacoides. before peak productivity periods of most plant spe- In the DCA diagram, typically early successional cies. This allows plants to recover from gooseplant species such as Salicornia spp., Spartina anglica, grazing once the geese have left the salt marsh.Spergularia maritima and Suaeda maritima were in the Second, in spring, hares and geese have a stronglybottom left-hand corner whereas late successional overlapping diet, namely early successional plantspecies such as Elymus athericus and Atriplex portu- species such as Festuca rubra, Puccinellia maritima,lacoides were in the upper right-hand corner (Fig. Triglochin maritima and Plantago maritima. Howev-10.3b). All treatments showed a similar direction in er, in winter, hares eat late successional woodythe shift of species composition, i.e. they moved in species which are sensitive to grazing, such as Atri-the direction of increasing cover of late successional plex portulacoides, Artemisia maritima and Elymusspecies and decreasing cover of early successional athericus (Van der Wal et al. 2000c).
  • 151. COMMUNITY ECOLOGY AND MANAGEMENT OF SALT MARSHES 137 During undisturbed succession at the high occurred at a similar elevation with higher covermarsh in temperate European marshes, the low- on Schiermonnikoog than on Rottumerplaat andstatured species Festuca rubra eventually will be Mellum (Fig. 10.4). Plantago maritima wasreplaced by the tall-growing grass Elymus ather- rarely found on Rottumerplaat and Mellum. Festucaicus (Leendertse et al. 1997). Both species were rubra, a preferred food plant for both geese andaffected when herbivores were excluded, indicating hares, occurred over a large part of the elevationlocal effects of grazing by intermediate-sized herbi- gradient on Schiermonnikoog, but was found atvores, because the herbivores are not able to only a small part of the gradient on Rottumerplaatprevent the increase of Elymus athericus at the and Mellum (Kuijper and Bakker 2003). In contrast,high marsh (Kuijper and Bakker 2005). The main the typically late successional species Atriplexreason for this may be that Elymus athericus is not portulacoides dominated the lower elevations onpreferred by any herbivore (Prop and Deerenberg both Rottumerplaat and Mellum, whereas it had1991; Van der Wal et al. 2000a; Kuijper et al. 2008), low cover on Schiermonnikoog (Fig. 10.4). Elymusand grazing pressure drops dramatically once athericus, a characteristic late successional species ofthis species dominates the vegetation (Kuijper the high marsh, occurred with higher cover at bothet al. 2008). low and high elevation on Rottumerplaat and Mel- lum compared with that on Schiermonnikoog. At the upper part of the elevation gradient on10.5 Large-scale effects of an Rottumerplaat and Mellum a monoculture ofintermediate herbivore on salt-marsh Elymus athericus, covering 100%, was found. In con-vegetation trast, on Schiermonnikoog, Elymus cover didThe small-scale exclosure experiments and studies not reach values higher than 70% (Kuijper andon individual plants on the salt marsh on Schier- Bakker 2003).monnikoog revealed that plant species replacement It can be concluded that the small-scale exclosureis retarded by herbivory. The effects of hare grazing experiments on Schiermonnikoog are not applica-especially were dominant and were most pro- ble only to understanding the local effects ofnounced in young salt marshes (Kuijper and Bakker grazing, but can also be extrapolated to a larger2005). Grazing by hares retarded succession by scale. Intermediate-sized herbivores affect the com-more than 25 years (Van der Wal et al. 2000c). This munity structure of large-scale salt-marsh systemsimplies succession should proceed fast when hares on the back-barrier Wadden Sea islands.are not present at the initiation of salt-marsh devel-opment. Hence, late successional species should 10.6 Interaction of herbivory anddominate at an earlier stage of development com- competitionpared with salt marshes that developed in the pres-ence of hares. This idea was tested by comparing Apart from experiments focusing on the level of thethe hare-grazed salt marsh on Schiermonnikoog entire vegetation, detailed experiments with indi-with those of two Wadden Sea islands without vidual plant species may reveal which mechanismshares, namely Rottumerplaat (The Netherlands) play a role in plant species replacement along theand Mellum (Germany). productivity gradient. In addition to plant–plant On all three islands, sites were selected where competition, plants have to deal with changing le-salt-marsh development had started in the early vels of herbivory. The small highly herbivore-pre-1970s. Transects of 1000 m running from the foot ferred Triglochin maritima is hardly present at theof a dune towards the intertidal flats were matched very young and old marshes, but is very abundantfor surface elevation with respect to the level of at intermediate-aged marshes. Competition andmean high tide and sediment thickness (Kuijper grazing are closely linked: when grazing pressureand Bakker 2003). Early to mid-successional plant is relaxed, competition with neighbouring plants isspecies Puccinellia maritima and Plantago maritima, intensified. Grazing is shown to influence thesewhich are the preferred food plant of geese, competitive interactions between plants, acting
  • 152. 138 APPLICATIONS (a) 60 Festuca rubra 40 (b) 20 Atriplex portulacoides 100 0 80 4 Plantago maritim a 60Cover (%) 3 40 2 20 1 0 100 Elymus athericus 0 2.5 Puccinellia maritima 80 2.0 60 1.5 40 1.0 0.5 20 0.0 0 0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140 Surface elevation (cm +MHT) Surface elevation (cm +MHT)Figure 10.4 Average cover of (a) important food plant species for hares and geese, Festuca rubra, Plantago maritimaand Puccinellia maritima, and of (b) typically late successional, unpalatable species, Atriplex portulacoides andElymus athericus, at different marsh surface elevation (cm + mean high tide (MHT)) for Schiermonnikoog (islandwith hares), Rottumerplaat and Mellum (islands without hares). Closed circles, Schiermonnikoog; closed triangles,Rottumerplaat; open circles, Mellum. After Kuijper and Bakker (2003).both directly on the target plant and indirectly younger marsh and increasing competition forthrough its neighbours. The significance of compe- light in the older marsh (Van der Wal et al. 2000a).tition and herbivory largely depends on plant stat- Adult plants of Elymus athericus are tall and noture relative to the neighbouring vegetation. preferred by any of the herbivores. However, experi-Although establishment of Triglochin maritima starts ments in which grazing and competition were ma-from seed, the high grazing pressure at younger nipulated along the productivity gradient show thatmarshes determines its abundance in the sward. herbivory negatively affects the survival of seedlingsHowever, at productive old marshes this small-sta- (being a good food source) in the unproductive sites.tured plant is outcompeted by tall-growing late At the productive sites, plant competition becomessuccessional species. The distribution of T. maritima an overruling factor. When seedlings grow in naturalis ‘sandwiched’ between intense grazing in the vegetation, the increased competition prevents any
  • 153. COMMUNITY ECOLOGY AND MANAGEMENT OF SALT MARSHES 139increase in biomass, whereas in the absence of com- geese was completely matched by increased harepetition the plant can grow fast because of high nu- grazing, while for Puccinellia only part of the surplustrient availability along the productivity gradient. was grazed. Apparently, competition for food be-Even though Elymus athericus is an unpalatable supe- tween hares and brent geese also occurs and playsrior competitor as an adult plant at highly productive a role in the habitat use of hares (Van der Walsites, in its seedling phase its growth is strongly et al. 1998).reduced by herbivory at unproductive stages and Competitive and facilitative interactions betweencompetition with neighbouring plants at the produc- geese (barnacle and brent geese) (Stahl 2001) andtive stages (Kuijper et al. 2004). geese and hares were studied on Schiermonnikoog (Stahl et al. 2006). Biomass (through temporary exclosures) and quality (by fertilizer application)10.7 Competition and facilitation of grass swards were manipulated and the foragingbetween herbivores preferences of the herbivores were recorded. Cap- tive barnacle geese were used to set the stage for10.7.1 Short-term competition and a choice experiment with captive brent geese, asfacilitation between hares and geese the latter species normally exploits the vegetationFor a large part of the year hares and geese forage on ‘on the heels’ of the former. Brent geese preferredthe same food plants, hence competitive interactions to forage on vegetation previously grazed by bar-may also occur. Exclusion of brent geese at scales nacle geese, probably reacting to enhanced qualityranging from 30 m2 to 1 ha at the salt marsh on of the regrowth, in spite of the higher biomassSchiermonnikoog enhanced the level of utilization of the ungrazed swards (Stahl 2001). In anotherby hares in both Festuca rubra- and Puccinellia mari- experiment with captive barnacle geese, it wastima-dominated marshes. The more geese were ex- demonstrated that grazing affected the swardcluded from a site, the stronger the increase of characteristics significantly: the proportion of deadhare grazing pressure. When geese were excluded, biomass in the vegetation was reduced, and thethe ‘original’ decrease in Festuca consumption by production of additional axillary tillers increased 80 Goose No cattle Cattle grazed Hare Rabbit Consumption (gDW/m2/year1) 60 40 20 0 5 15 25 35 60 90 110 130 190 Age of the salt marsh (years)Figure 10.5 Total plant consumption of geese, hares and rabbits in salt marshes of different ages at Schiermonnikoog.Cattle grazing occurs only at the older marshes. Consumption was calculated on the basis of total droppings weightby multiplying the cumulative amount of droppings during 1 year (1999–2000) by the droppings weight per species.Subsequently, consumption was calculated from: total faecal mass/(1 – DE). Digestive efficiency (DE) for hares andgeese was obtained from literature (Van der Wal et al. 1998). After Kuijper (2004).
  • 154. 140 APPLICATIONS(Van der Graaf et al. 2005). Both barnacle and Although hares can retard vegetation successionbrent geese selected plots with plants that have a for several decades (Van der Wal et al. 2000c;high nitrogen content. Barnacle geese avoided Kuijper and Bakker 2005), they eventually lose con-plots with high biomass. Geese mainly selected trol in the higher ranges of the productivity gradi-plots that have been previously grazed by either ent. Large herbivores, such as livestock, are neededgeese or hares within the same season. Grazing to set back the successional clock. Indeed, at theby both geese and hares leads to an increased older cattle-grazed salt marsh in the chronose-quality of the sward. Under these circumstances, quence on Schiermonnikoog, grazing pressure ofherbivores profit from the increased tissue quality hares and geese increases again compared withas a result of an elevated rate of nutrient in- the ungrazed older marsh (Kuijper 2004; Fig. 10.5).take. However, when the forage resource is used An experiment with exclosures on the cattle-grazedjointly by more than one herbivore species, a shift marsh revealed that after 30 years of cessation oftowards less preferred plots by one species may cattle grazing no hares grazed inside the exclosurestake place. Hares prefer the combination of high when the cover of tall plants, such as Elymus ather-biomass with high plant quality in the absence of icus, was > 30%. Thus, clear facilitative effects ofgeese (Stahl et al. 2006). Van der Wal et al. (1998) cattle on the feeding opportunities of hares weresuggested that large flocks of socially foraging found (Kuijper et al. 2008). This finding is in con-geese rapidly deplete preferred salt-marsh sites in trast to studies from other areas that reported onlyspring and evict hares to alternative less favourable competitive interaction between hares and live-foraging sites. stock (Hulbert and Andersen 2001; Smith et al. 2004). The contrasting conclusions of these studies may be the result of the timescale of the experi- ments. Facilitative effects between cattle and hares10.7.2 Long-term facilitation between on Schiermonnikoog were observed only whenherbivores looking at the long-term effects, including the effectThe previous section showed that the cover of spe- of cattle on the competitive replacement of plantcies that are selected as food plant by both geese species. Only when species replacement didand hares, such as Puccinellia maritima, Plantago occur in the absence of cattle was an effect on themaritima and Festuca rubra, is higher at hare-grazed abundance of hares observed. In contrast, in aislands, whereas the cover of unpreferred plants, short-term experiment on Schiermonnikoog insuch as Atriplex portulacoides and Elymus athericus, is which cattle were excluded for 5 years, plant bio-lower. Hare grazing may thus facilitate food supply mass increased inside the exclosure, but the periodfor geese (Kuijper and Bakker 2003). This idea was was too short for plant species replacement totested experimentally at the salt marsh on Schier- occur. In this short-term study no effect on themonnikoog. The woody shrub Atriplex portulacoides abundance of hares was detected (Kuijper et al.is unpalatable for geese. It can overgrow the pre- 2008). This suggests that at a short timescale noferred food plant Pucinellia maritima. When Atriplex effect of cattle grazing on hare abundance is appar-portulacoides was removed, goose grazing, ex- ent, whereas at a longer timescale facilitation occurspressed as the number of droppings found, was (Kuijper et al. 2008).higher than in the control plots. In contrast, goose It can be concluded that competition betweengrazing declined when Atriplex portulacoides indivi- different species of herbivores occurs only in theduals were planted in a Pucinellia maritima sward short term, i.e. within one spring season. In the(Van der Wal et al. 2000c). Knowing that hares long-term, facilitation plays an important role. Atforage on Atriplex portulacoides during winter, this the salt marsh on Schiermonnikoog, barnacle geeseexperiment clearly demonstrated the effect of facilitate for brent geese within one season, haresgrazing facilitation by hares for geese. facilitate for geese for several decades, and
  • 155. COMMUNITY ECOLOGY AND MANAGEMENT OF SALT MARSHES 141ultimately cattle facilitate for hares and geese, when community, 27 years after the cessation of grazinghares have lost control of the vegetation. (Bakker et al. 2003a). Taking into account the aforementioned natural succession without livestock grazing, it is likely that10.8 Exclusion of large herbivores: effects the oldest part of the salt marsh with a thick layer ofon plants clay in most sites will eventually be covered by the Elymus athericus community at both the high and10.8.1 Natural marshes the low salt marsh. That is exactly what happensThe effects of large herbivores on salt marshes is after the long-term exclusion of livestock. The ces-restricted to that of livestock. In fact, livestock sation of livestock grazing produces two main con-grazing is the most common land use of North- clusions. Initially, the vegetation transforms into aWest European salt marshes (Bakker et al. 2005a; ‘flower garden’ as many existing species have theDavy et al. 2009). Hence, the obvious way to study opportunity to flower during the first few yearsthe effects of livestock grazing is to establish exclo- before tall species become dominant and replacesures. In 1973 at the oldest part of the chronose- the present plant community with another one.quence on Schiermonnikoog (> 150 years) that was Eventually, most plant communities are replacedalways cattle grazed, two exclosures were estab- by the Elymus athericus community at the saltlished, one at the higher and one at the lower marsh on Schiermonnikoog. Another part of themarsh. At the higher marsh Elymus athericus was salt marsh on Schiermonnikoog was abandoned inalready present in the grazed area. The Elymus 1958 for cattle grazing and grazed anew from 1972athericus community established at the expense of onwards. Permanent plots in exclosures revealedthe Juncus maritimus community within five years that different plant communities converged intoafter the cessation of grazing. The deposition of the Elymus athericus community after various peri-driftline material initiated temporary spots with ods of cessation of grazing: the Juncus maritimusthe annual Atriplex prostrata, but within two years community, the Plantago maritima/Limonium vulgarethese were taken over again by the Elymus athericus community and the Artemisia maritima communitycommunity. This community also spreads at the after 30 years and the Juncus gerardi communitytransition to the low dune, but only gradually, and after 35 years. The only exception was the Festucaafter 27 years remnants of the Festuca rubra commu- rubra/Armeria maritima community, which was notnity with Armeria maritima were still present. It replaced 35 years after cessation of livestock grazingseems that Elymus athericus is also spreading in the (Van Wijnen et al. 1997). Perhaps the combination of agrazed area, but this is mainly due to the fact that thin layer of sediment (low nutrient pool) at this highthe tall Juncus maritimus is not preferred by cattle elevation site and evapotranspiration during dryand protects Elymus athericus from grazing, thus summer periods with subsequent high soil salinityacting as a ‘natural’ exclosure (Bakker et al. 2003a). have until now prevented replacement. At the lower marsh Elymus athericus was lacking ¨ The natural marsh of Suderhafen (Germany) de-in the grazed area at the start of the experiment. The veloped in the shelter of the former salt-marsh is-Artemisia maritima community dominated within land of Nordstrand after 1925. The site was hardlyfive years in the relatively higher parts inside the grazed before 1968, and not at all since 1971. Re-exclosure. It took 12 years before the first clone of peated vegetation mapping in 1968 and 1995 re-Elymus athericus found its window of opportunity vealed an expansion of the Elymus athericusand became established. After 22 years the Elymus community at the expense of the Festuca rubra com-athericus community expanded. The initially bare munity, and of the Atriplex portulacoides communitysoil at the lowest places became covered by the at the expense of the Puccinellia maritima communi-Plantago maritima/Limonium vulgare community ty (Bakker et al. 2003a).after about ten years, after which the Atriplex portu- Combining permanent plot data from experimen-lacoides community took over after 22 years. The last tally ungrazed sites on the back-barrier marshes onhas locally been replaced by the Elymus athericus Schiermonnikoog (The Netherlands), Terschelling
  • 156. 142 APPLICATIONS 10.8.2 Artificial salt marshes 36 38 44 38 38 74 38 90 38 88 88 38 88 30 30 58 40 18 66 18 18 62 18 18 58 18 10 n= 16 P < 0.05 There are experimentally ungrazed plots in artifi- 14 cial marshes in Dollard Bay, The Netherlands. InSpecies richness (2 x 2m2) 12 these brackish, highly productive marshes the ex- clusion of cattle resulted in the increase of Elymus 10 repens within six years, mainly at the expense of 8 Puccinellia maritima. Species richness was higher in 6 grazed than in excluded plots (Esselink et al. 2002). When salt marshes are broad enough, a gradient in 4 grazing intensity emerges. Cattle and sheep tend to 2 Grazed concentrate near the seawall, where fresh drinking Ungrazed 0 water is available. Hence a reduction in grazing is 0 5 10 15 20 25 found at the seaward site of salt marshes resulting Length of experimental period (years) in a taller canopy. Indeed, gradients of increasing canopy height towards the marsh edge were re-Figure 10.6 The development of plant species richness ported in the Dollard (Esselink et al. 2000), the Leyover time in paired livestock-grazed and ungrazed ¨ Bucht (Andresen et al. 1990) and Sonke-Nissen-permanent plots in the Wadden Sea. Sample sizes (n) peryear since the start of the treatment are indicated in the Koog (Germany) (Kiehl et al. 1996).top of the diagram. After a period of three years the No controlled large-scale grazing experimentsdifferences between grazing treatments were significant, have been established along the Dutch mainlandindicated by the arrow with text P < 0.05. Observations coast with artificial marshes. However, three goodfor two sites started one year before the treatment was examples can be found along the German coast. Theestablished. After Bos et al. (2002). first site is located at Friedrichskoog in Lower Sax- ony. It developed after 1854 and was long-term(The Netherlands) and Skallingen (Denmark) re- sheep grazed. The experiment was established invealed that the convergence to the Elymus athericus 1988 to study the effects of different stocking ratescommunity after the exclusion of livestock grazing is on soil and vegetation (Kiehl et al. 1996; Kiehl 1997).a general phenomenon (Bos et al. 2002). The stocking rate was expressed in sheep-units, i.e. Not only the diversity of plant communities de- adult sheep including their lambs (1 sheep-unitclined after the cessation of livestock grazing. The equals 2.8 sheep). The area was heavily grazed ‘asspecies richness within plant communities in a golf course’ by 3.4 sheep-units/ha at the end ofpaired permanent plots in experimentally un- the grazing season. The control area with 3.4 sheep-grazed and control plots also decreased significant- units/ha was compared with 1.5 and 1.0 sheep-ly after five years (Fig. 10.6). (Data have been units/ha and cessation of grazing. At the start of thecombined from the back-barrier marshes on Schier- experiment this salt marsh harboured mainly Festucamonnikoog, Terschelling and Skallingen (Bos et al. rubra community, at the lower marsh Puccinellia mar-2002; Bakker et al. 2003a).) These permanent plots itima community, and at the intertidal flats Spartinaalso revealed that out of 30 frequently occurring anglica and Salicornia spp. communities were foundplant species only four had a significantly higher (Kiehl 1997). The vegetation revealed a relativelyoccurrence at ungrazed than at grazed marshes, small coverage of the Elymus athericus communitynamely Artemisia maritima, Atriplex portulacoides, At- after the cessation of grazing 11 years after the startriplex prostrata and Elymus athericus. Three species of the experiment (Bakker et al. 2003a).were indifferent, namely Festuca rubra, Juncus mar- Apart from the above large-scale patterns, theitimus and Lotus corniculatus. All remaining 23 spe- Friedrichskoog experiment also revealed differentcies had a significantly higher occurrence at grazed micropatterns in the vegetation with the variousthan at salt-marsh sites excluded for more than 20 stocking rates seven years after the start of theyears (Bos et al. 2002). experiment. The micropatterns were formed by a
  • 157. COMMUNITY ECOLOGY AND MANAGEMENT OF SALT MARSHES 143mosaic of short and tall Festuca rubra stands on a both the higher and the lower marsh, but to a lesserscale of square decimetres in transects of 2 m  degree than at the abandoned area.10 m. In the most intensively grazed and the aban- Both artificial and the natural back-barrier saltdoned paddocks, no micropattern was found. The marshes tend to transform into a dominance of thevegetation in the transects was homogeneously Elymus athericus community after the cessation ofshort or tall, respectively. However, micropatterns livestock grazing within 10–30 years, as could beoccurred in the three intermediately grazed pad- predicted from succession without livestockdocks with the highest spatial diversity in the 1.5 grazing. However, a correlation between the num-sheep-units/ha (Berg et al. 1997). ber of years of exclusion of livestock grazing and ¨ The second site is located at Sonke-Nissen-Koog the spreading of Elymus athericus is not alwaysin Schleswig Holstein. It developed after 1924 and found. The salt-marsh sites that do not follow thiswas long-term sheep-grazed. The experimental rule seem to have a low sediment (nitrogen) inputtreatments were established at the same time and ¨ (Schroder et al. 2002). In these sites exclusion ofhad the same layout as at the Friedrichskoog site. At livestock grazing did not result in a dominance ofthe start of the experiment this salt marsh mainly Elymus athericus within 30 years. A complicationharboured the Puccinellia maritima community with may be that because of the low sediment inputlocally some Festuca rubra and Elymus athericus com- these sites are building a sedimentation deficit duemunities, and with Spartina anglica and Salicornia to continuous sea-level rise, and hence are becom-spp. near the intertidal flats (Kiehl 1997). The ing wetter. This may be an extra factor preventingmarsh showed a large coverage of the Elymus ather- the establishment of Elymus. Another conclusion isicus community after the cessation of grazing. The that grazing with low stocking rates cannot preventcommunity covered smaller areas at the lower the spread of Elymus athericus, but only retards thestocking rates, 11 years after the start of the experi- spread. In contrast to intensive grazing and noment (Bakker et al. 2003a). grazing at all, intermediate grazing can create The third site is in the Ley Bucht (Germany). The small-scale patterns in the vegetation.site was cattle-grazed since its formation after 1950.The site was established as an experiment in 1980. 10.9 Exclusion of large herbivores: effectsThe area with 2 cattle/ha was compared with areas on invertebrateswith stocking rates of 1 and 0.5 cattle/ha and cessa-tion of grazing. The zonation included Elymus re- On the natural mainland salt marsh in Mont Saint-pens/Elymus athericus and Festuca rubra communities Michel Bay (France), the invasive species Elymusclose to the seawall, the Agrostis stolonifera commu- athericus outcompetes Atriplex portulacoides. Apartnity at the transition, the Puccinellia maritima com- from changes in plant communities, this results inmunity at the lower marsh, and Spartina anglica and changes in invertebrate communities, particularlySalicornia spp. communities near the intertidal flats. spiders. The invasion of Elymus athericus led to anEight years after the cessation of grazing, the Ely- increase in the overall species richness. Causes maymus athericus community covered large areas at the be the formation of a dense, tall sward which allowshigher salt marsh and one spot at the lower marsh. colonization of web-spinning species such as Ar-It hardly occurred at the other grazing regimes giope bruennichi, Neoscona adianta and Larinioides cor-(Bakker et al. 2003a). The Elymus athericus commu- nutus. The building of a deep litter layer favoursnity quickly spread over both the higher and the nocturnal wanderers (Gnaphosids, Clubionids),lower marsh, and covered nearly the entire gradi- ambush hunters (Thomisids) and litter-sensitiveent 20 years after the cessation of grazing, at the sheet-weavers. Non-coastal species such as theexpense of the Festuca rubra and the Agrostis stolo- ground-living nocturnal Pachygnatha degeeri andnifera communities, and the Puccinellia maritima the halophilic sheet-web spinning Arctosa fulvoli-community, respectively. Also the 0.5 cattle/ha re- neata increased. However, the dominant halophilicgime revealed a spread of the Elymus athericus com- species Pardosa purbeckensis was strongly negativelymunity 15 years after the start of the experiment at ´ affected by the invasion of Elymus athericus (Petillon
  • 158. 144 APPLICATIONSet al. 2005). Some halophilic ground beetle species the abandoned site, partly as a result of immigra-were more abundant in grazed than in abandoned tion from adjacent grassland. But the main reasonsites and vice versa. In general, no effect of manage- was that many species are damaged by grazingment on species richness was found for ground (Irmler and Heydemann 1986). The authors espe-beetles. Generally, spiders seem to be more depen- cially stress the damaging effects of grazing ondent on vegetation and litter structure than ground many arthropod communities. ´beetles (Petillon et al. 2008). Two and three years after the start of the afore- The aforementioned experiment in the artificial mentioned grazing experiment in Friedrichskoogmarsh of the Ley Bucht aimed to study the effect of ¨ and Sonke-Nissen-Koog, invertebrates were moni-various stocking rates on the invertebrate fauna tored. Mainly herbivorous and flower visitors were(Andresen et al. 1990). For invertebrates, it may positively affected by cessation of sheep grazingnot only be the plant species composition that is and the resulting flowering of Aster tripolium andimportant. Non-flowering Asters were found only Plantago maritima. A minor part of the herbivorousat the higher salt marsh within the highest stocking fauna profits from enhanced plant growth in mod-rate. The canopy height of the understorey was erately grazed sites. Typical soil dwellers benefithigher in the abandoned site than in the grazed from grazing owing to greater amounts of baresites (Andresen et al. 1990). In the third year of soil (Meyer et al. 1995).cessation of grazing, positive effects for several in- In general, the community structure changesvertebrate groups were recorded for Collembola, from a dominance of detritivores to a dominanceAranea, Amphipoda, Coleoptera and Diptera. This of herbivores after the cessation of sheep grazing,was attributed to the accumulation of litter, increase and after the cessation of cattle grazing. The num-of flowering plants and hence availability of pollen ber of species and individuals increases shortlyand nectar and therefore higher aboveground bio- after the cessation of grazing, but after a longermass for leaf- and stem-dwelling species (Irmler period of cessation of grazing typical halophilicand Heydemann 1986). Erigone longipalpis, a halo- species may decrease. For the time being it is notphilic species, is the most important spider species possible to discuss top-down or bottom-up con-in the Puccinellia maritima community. Other spe- cepts with respect to the interaction between vege-cies occur mainly in the Festuca rubra community tation and invertebrates. Food web studies couldand cannot be considered halophilic, namely Oe- help in this discussion and are currently being car-dothorax retusus, Pardosa agrestis and Pachygnata ried out.clerki. Whereas Erigone longipalpis still occurred inhigh abundance in the Festuca rubra community atthe start of the experiment in 1980, it has since 1982 10.10 Exclusion of large herbivores:moved to the lower Puccinellia maritima and Salicor- effects on birdsnia spp. communities in the abandoned site. The 10.10.1 Migrating birdsother species spread into the lower salt marsh atdifferent rates (Andresen et al. 1990). A distinct In order to evaluate the importance of livestockzonation of the invertebrate communities was ob- grazing for habitat use by geese in the Waddenserved in the first three years of the experiment. The Sea, a large-scale inventory was made. Sixty-threecommunity diversity was highest in the abandoned transects were established, subdivided over 38site, since communities of the higher marsh spread sites. Only those sites with a stable and clearlyinto the lower marsh. In 1988, however, the com- defined management regime for at least six preced-munity of the higher marsh had spread over the ing years were included. Management was subdi-entire elevation gradient, and completely replaced vided into ‘long-term ungrazed’(> 10 years), ‘short-the communities of the lower marsh in the aban- term ungrazed’ (6–10 years), ‘lightly grazed’ (lowdoned site. Hence, eventually the community di- stocking rates, i.e. 4.5 sheep/ha or 1 cow/ha),versity was lowest at the abandoned site. and ‘intensively grazed’ (i.e. with high stockingHowever, the number of species became highest at rate). Only marshes with sufficiently large surface
  • 159. COMMUNITY ECOLOGY AND MANAGEMENT OF SALT MARSHES 145area (> 5 ha), large enough for a flock of geese to comparative study in the entire Wadden Sea, as forland on, were included. The sites were distributed migrating birds, has not been carried out so far. Weover the entire Danish (n ¼ 11), German (n ¼ 17) derive our knowledge from a small number of stud-and Dutch (n ¼ 10) Wadden Sea. Twenty-two sites ies describing differences in ungrazed and differ-harboured transects with at least two different ently grazed marshes. At the natural marsh ongrazing regimes under similar abiotic conditions. Schiermonnikoog, including some low dunes, theSeventeen sites with paired transects were visited breeding population was monitored in 1973 andtwice, once in April and once in May 1999. The 1978. The 83 ha of marsh ungrazed since 1958 har-transects on back-barrier marshes were, with one boured maximally 31 species with in total 850–1000exception, visited only by brent geese, whereas breeding pairs, the 77 ha continuously cattle-grazedmost transects on artificial marshes along the main- marsh hosted maximally 25 species with in totalland coast were utilized by both brent and barnacle 550–600 breeding pairs. In 1978, the grazed marshgeese. For each management regime at each site, harboured 133 breeding territories for oystercatch-one transect was established perpendicular to the er, 10 for lapwing and 71 for redshank, whereas theseawall and the coastline, along the entire extent of grazed marsh harboured 85, five and 48 territories,the marsh. Hence, transects were variable in length, respectively (Van Dijk and Bakker 1980).ranging from 100 m to 1000 m, and included high- Studies on the relationship between managementmarsh, mid-marsh and lower marsh sections. and vegetation, and the occurrence of breedingTwenty plots of 4 m2 were sampled per transect, birds have been summarized by Koffijberg (inand the accumulated number of goose droppings press). Most studies have been carried out on artifi-were counted and the plant community was as- ¨ cial marshes in Germany (Halterlein 1998; Eskildsensessed (Bos et al. 2005). ¨ et al. 2000; Halterlein et al. 2003; Oltmanns 2003; The communities of Elymus athericus, Artemisia Schrader 2003; Thyen and Exo 2003, 2005; Thyenmaritima and Atriplex portulacoides had a significant- 2005). They reveal a trend that relaxation of former-ly taller canopy, but a lower goose dropping densi- ly heavily grazing regimes results in an increase inty than the communities of Agrostis stolonifera, species richness, particularly due to a species groupFestuca rubra and Puccinellia maritima. Dropping shift from waders, gulls and terns towards ducksdensity at the transect level declined with decreas- and songbirds. Another trend is the decrease ofing livestock grazing regime. However, only the avocet (Recurvirostra avosetta), great ringed ploverlong-term ungrazed regime combined for barrier (Charadrius hiaticula), Kentish plover (Charadriusmarshes and artificial marshes had significantly alexandrinus), common tern (Sterna hirundo) andlower dropping densities than the other regimes Arctic tern (Sterna paradisaea) after the cessation of(Fig. 10.7). These results are valid for May, the end grazing and subsequent vegetation succession. Aof the staging period for both goose species. In problem in these studies is that the results representApril, goose-dropping densities at the transect snapshots, describing ‘pioneer situations’ a fewlevel did not differ between grazing regimes. years after transition of management, and do notThere were no significant differences in dropping include the long-term effects of cessation of grazing.densities by geese between transects grazed by For some species more detailed information issheep or cattle (Bos et al. 2005). We conclude that available. Increased grazing negatively affects thethe long-term exclusion of livestock on salt marshes number of redshanks. This was attributed to thewill result in a decline in utilization of these areas destructive effects of trampling of nests and hatchl-by spring-staging geese. ings, whereas changes in the vegetation composi- tion were considered less important (Schultz 1987). However, in salt marshes in Great Britain the occur-10.10.2 Breeding birds rence of redshank densities were positively relatedThe effects of excluding livestock grazing on breed- to the extent of the Elymus athericus community.ing birds cannot be studied in small-scale exclosure This relation could be explained by the variationexperiments as for plants and invertebrates. Also a in vegetation structure. Cattle-grazed plots, with
  • 160. 146 APPLICATIONS n= 26 21 25 13 were partly replaced by bare soil and secondary pioneer community of Salicornia spp. and Suaeda 10 a maritima, which was, however, attributed to in-Goose dropping density (no./m2) creasing numbers of spring-staging barnacle geese 8 a and not to decreased cattle grazing (Esselink 2000). We have to conclude that the effects of cessation 6 ab of livestock grazing on breeding birds need further study. From the results so far, we suppose an initi- 4 ally positive, but in the long term negative, effect. 2 b 10.11 Ageing of salt marshes and 0 implications for management Intensively Lightly Short-term Long-term grazed grazed ungrazed ungrazed As long as the area of salt marshes increases, Grazing regime marshes will feature the successional series of pio- neer, young and older mature marshes. When theseFigure 10.7 Average goose grazing pressure at the extension processes stabilize eventually, only ma-transect level in relation to livestock grazing regime for all ture marshes will be found. This happens at back-transects that were paired within the same site. Bars thatdo not share the same letter differ significantly (P < 0.05). barrier marshes that do not expand. It also happensAfter Bos et al. (2005). along the mainland coast where the present area is maintained, and no further expansion into the in-Elymus athericus covering up to 30%, supported the tertidal flats takes place. In the past, it was econom-most structurally diverse vegetation and the high- ically feasible to embank marshes, and start newest breeding densities. In contrast, ungrazed plots sedimentation fields (Esselink 2000). Nowadays, itof similar habitat contained tall, uniform vegetation is no longer economically feasible for many farmersand supported significantly lower breeding densi- to graze livestock at the marshes. The combinationties (Norris et al. 1997). The period of abandonment of decrease in the pioneer zone, and hence matura-was not indicated. However, a survey on 77 salt- tion of the marshes, and abandonment of livestockmarsh sites in Great Britain revealed that breeding grazing results in the encroachment of Elymus ather-redshank densities were lowest on heavily grazed icus on artificial marshes (Dijkema 2007).marshes and tended to be highest on lightly or What are the implications for management (oftenungrazed marshes (Norris et al. 1998). Redshanks livestock grazing) in view of these ageing processesbreeding on salt marshes partly feed on nearby of salt marshes? According to the ‘wilderness con-intertidal flats and build their nests hidden among cept’ (a contradiction in itself for an artificialvegetation of intermediate height, avoiding areas marsh), the solution with respect to the questionwith low cover or with very tall vegetation ‘to graze or not to graze’ (Bakker et al. 2003a) is(Cramp and Simmons 1983). In the Dollard (The easy: the management option will be ‘no grazing’.Netherlands) cattle-grazed salt marsh, densities of This will undoubtedly result in a loss of biodiversi-redshanks were approximately two breeding pairs ty at the local scale. However, at the scale of theper hectare at a grazing regime of $200 cattle-days/ entire Wadden Sea, it should be a preferred optionha in 1984, and decreased to less than one breeding for the marshes that have never been grazed bypair per hectare in 1998. Within the same period livestock such as the eastern parts of Terschellingcattle grazing was reduced to $50 animal-days/ha. (The Netherlands), Schiermonnikoog (The Nether-The redshanks preferentially breed in the Elytrigia lands), Ameland (The Netherlands) and Spiekeroogrepens community, and in the less preferred short- (Germany). In the long run, these areas will dem-grass stands with Festuca rubra, Agrostis stolonifera onstrate whether there is a world beyond Elymusand Puccinellia maritima. Especially the latter stands athericus.
  • 161. COMMUNITY ECOLOGY AND MANAGEMENT OF SALT MARSHES 147 According to the ‘biodiversity concept’ the answer after a period of intensive livestock grazing. Theto the question ‘to graze or not to graze’ will be: result will be flowering of the plants and the possi-define the biodiversity target at a distinct scale, and bility of replenishing the soil seed bank. Flowersdecide to what extent livestock grazing as a manage- and taller stems will attract invertebrates, whichment tool may help to reach the biodiversity target. can be the prey items for breeding birds. Before It is known that no grazing results in a low diver- Elymus athericus invades, the intensive grazing re-sity for plants and less favourable feeding conditions gime should be re-installed. The results of such afor hares and spring-staging geese. High-intensity rotational grazing regime have not been monitored solivestock grazing is a good option for spring-staging far. Salt-marsh communities and their managementgeese. Low-intensity grazing renders a pattern of will profit from large-scale and long-term experi-intensively grazed short swards and lightly or no- ments in which the interactions of plants, inverte-grazed taller patches of vegetation. The difference brates and birds are studied.with respect to the options no grazing or intensive In summary, in order to have a full displaygrazing seems the patchiness and the spatial scale. of salt-marsh communities, including many speciesHowever, our knowledge of the consequences of of plants, vertebrates and invertebrates, the bestsuch a mosaic for the diversity of breeding birds management option is to have variety in the struc-and invertebrates is fragmentary. ture of the vegetation. This can be achieved by Another option with respect to grazing is rota- variation in grazing management, both in spacetional grazing. Livestock grazing can be abandoned and time.
  • 162. This page intentionally left blank
  • 163. PART VFUTURE DIRECTIONS
  • 164. This page intentionally left blank
  • 165. CHAPTER 11 Evolutionary processes in community ecology Jacintha Ellers evolutionary perspective unnecessary. However,11.1 Introduction evolutionary biology has produced compelling evi-Community ecology examines the distribution, dence that strong selection pressures and fast (co-)abundance, demography and interactions among evolution are commonplace in nature (Thompsonpopulations of coexisting species at a particular 2005). Hence, local adaptation of populations tosite or in a specific area. In past decades, it has community context may significantly affect com-been extremely successful in describing and ex- munity functioning. A growing number of commu-plaining patterns of species diversity, food web nity ecologists have come to realize the value of anstructure, invasion, etc. However, at the same time evolutionary perspective for addressing ecologicalcommunity ecology has been under fire for lack of questions.two important attributes. First, community ecologyis short of general, mechanistic principles leading toquantitative predictions (Lawton 1999). Mechanis- 11.1.1 Bridging the gap betweentic approaches in ecology try to functionally link evolutionary biology and communitytraits of individuals to higher level processes such ecologyas multispecies interactions and community struc-ture. Yet in an effort to reduce the inherent com- Several recent papers have explored the possibili-plexity of biological communities, most studies ties for a synthesis between community ecologyhave grouped species according to their trophic and evolutionary biology (Agrawal et al. 2007; John-positions while disregarding differences in traits son and Stinchcombe 2007). Evolutionary biologywithin these functional groups. Such simplification studies how ecological factors regulate genetic var-of community structure facilitates the study of eco- iation and evolution, and can thus provide commu-logical community properties, but it overlooks the nity ecology with a more mechanistic, quantitativespecies-specific contribution to multiple species in- insight into how genetic and phenotypic diversityteractions in the community. Analysis of functional can shape community processes. On the othertraits is one way to quantitatively predict the im- hand, most evolutionary studies perform experi-pact of local species loss or biological invasions on ments with the single species in isolation, or onlyecosystems. Second, community ecology has al- in direct interaction with another species such asways assumed homogeneous populations that are studies on coevolution. Including community com-impervious to evolutionary change, thereby exclud- position as an ecological context for evolution maying any evolutionary community dynamics result- increase our understanding of the maintenance ofing from selection on traits of individuals. For a genetic and phenotypic variation in the field.long time, it was assumed that the relatively slow However, part of the difficulty in bringing to-timescale of evolutionary changes rendered an gether the community perspective with the 151
  • 166. 152 FUTURE DIRECTIONSevolutionary perspective is that the yard-sticks munity properties result from variation andused to measure community performance differ selection at the level of individual organisms. Sub-from those relevant to evolutionary biology. Com- sequently, I will address how variation in one spe-munity properties most often used include produc- cies mediates interaction at higher trophic levels ortivity, carbon storage, nutrient acquisition or at lower trophic levels. I will focus specifically ondecomposition rate, all of which are related to com- two aspects of evolutionary community ecology: (1)munity functioning. Others reflect community how genetic diversity and phenotypic plasticity ofproperties desirable for nature management such single species influence the composition of commu-as species diversity, community stability and resil- nities and (2) how species diversity and composi-ience. Although such properties can certainly de- tion in a community influence the genetic diversityrive from species composition and individual and realized phenotype of single species.performance, they are not the primary targets of This chapter will not go into research on macro-natural selection. Rather, from an evolutionary evolutionary timescales such as community phylo-point of view, they are by-products of the competi- genetics and co-diversification of communities.tive and mutualistic interactions between indivi- Even though evolutionary history can moderateduals under selection to enhance their own fitness. community assembly, I will focus on the shorterThis discrepancy in views has led community biol- timescale interactions as these are the most impor-ogists to discard individual variation and selection tant ones to cope with the ever-changing environ-in favour of a more phenomenological approach, mental dynamics.whereas it has caused evolutionary biologists toturn away from community ecology because of its 11.2 Evolutionary biology: mechanismsdescriptive nature. The value of a synthesis be- for genetic and phenotypic changetween community ecology and evolutionary biolo-gy will hence critically depend on the potential for Evolutionary ecologists study the cause and effectcommon standards. of variation in fitness-related traits of individuals In this chapter, I aim to outline a newly emerging such as longevity, fecundity and feeding rate infield of research: evolutionary community ecology. their natural environment. They lean heavily onI will argue that considering variation in traits of the robust framework provided by quantitativeindividuals and species is essential to understand and population genetics for quantitative predictionand predict community functioning and composi- of evolutionary change. The two main topics thattion, and, vice versa, that community composition can be identified are maintenance of genetic varia-is a key component of the selective forces determin- tion and the population benefits of variation. Ining genetic and phenotypic variation at the individ- other words, how does natural selection act onual level. Bridging the gaps between community phenotypes while still preserving genetic diversity,and evolutionary ecology can be mutually benefi- and does variation in phenotype affect performancecial. In fact, we may even need an integrative ap- of an individual and its associated population? Un-proach in order to face fast changing environmental derstanding these processes at the population levelconditions such as global warming and urbaniza- may allow us to extrapolate the consequences oftion, which pose ecological as well as evolutionary genetic diversity to community level and may illus-challenges. trate its effect on community properties. I will first review the main principles regardinggenetic and phenotypic variation from evolution- 11.2.1 Benefits and maintenance of geneticary biology, including maintenance of genetic di- diversity at the population levelversity, measuring genetic variation andphenotypic plasticity. I will emphasize parallels The population-level consequences of genetic vari-between genetic diversity and species diversity, ation have been the domain of evolutionary biolo-and population and community ecology, respec- gists for decades. Genetic variation can betively. I will show as a proof of principle that com- generated through sexual reproduction and the
  • 167. EVOLUTIONARY PROCESSES IN COMMUNITY ECOLOGY 153associated recombination during meiosis. Alterna- shoot number and more biomass than geneticallytively, mutations, insertions or deletions can change impoverished communities or monocultures ofthe genetic make-up of individuals. The majority of genotypes (Fig. 11.1). They were able to attributemutations and new genetic combinations are be- the effect of genotypic diversity to positive interac-lieved to be deleterious, either because they render tions between genotypes, particularly facilitation.an enzyme or promoter region non-functional or Some genotypes that performed poorly in monocul-because they break up co-adapted gene complexes. ture had a proportionally strongly reduced mortal-Natural selection will quickly purge such variation ity in genotypic mixtures (Reusch et al. 2005). Otherfrom the population. But even without selection a studies, specifically testing the advantage of geneticsignificant proportion of genetic variation is lost complementarity using clonal diversity, also foundeach generation through the random effect of ge- effects. Semlitsch et al. (1997) found significant dif-netic drift. Especially in small populations, random ferences in the life history traits among differentdrift can quickly reduce genetic variation. Only if clonal groups in Rana esculenta and an increasedgenetic variation has a clear adaptive value will it proportion of metamorphosed larvae in clonal mix-be maintained in the population, because the fitness tures of frogs compared with a clone reared alone.benefits associated with genetic diversity in a pop- In another study, genetically diverse groups ofulation will decrease extinction risk. clones were demonstrated to be better invaders Two principal mechanisms of reducing extinc- than genetically uniform groups of invaders. Thetion risk have been put forward. First, the tangled better performance of the genetically diverse groupbank hypothesis proposes that individuals with of invaders was attributed to competitive releasedifferent genotypes may each use a slightly differ- experienced by individuals in genetically diverseent niche and therefore together are able to extract populations (Tagg et al. 2005).more food from their environment than genetically Experimental evidence for improved populationidentical individuals (Bell 1991; Barrett et al. 2005). performance by reduced infection risk in genetical-In addition, different genotypes may compete less ly diverse populations is rare. Schmid (1994) foundbecause they explore different microniches (Anto- evidence that genetic diversity can influence mil-novics 1978). The second hypothesis to explain the dew infection levels in Solidago altissima. Infectionadvantage of genetic diversity is the Red Queen rate affected individual performance withhypothesis, which argues that genetic diversity re- increased height and biomass of less infectedduces the risk of infection or attack by natural ene- plants, but mean plant performance per plot wasmies (Jaenike 1978). Empirical studies have not correlated with genetic diversity levels. The fewprovided ample support for the benefits of genetic studies so far indeed provide evidence for geneticdiversity within populations, including positive complementarity as a significant mechanism main-correlations with increased resistance to distur- taining genetic variation.bance by grazing (Hughes and Stachowicz 2004),increased productivity and foraging rate (Mattila 70 Leaf shoots (0.25 m2)and Seeley 2007) and enhanced settling success 60 One genotype Unmanipulated *(Gamfeldt et al. 2005). Three genotypes * 40 Six genotypes However, few studies have actually determined **the underlying mechanism of fitness enhance- 20ments. The tangled bank hypothesis assumes thatdifferent genotypes show positive complementari- May June July August Septemberty, because microniche differentiation causes themto compete less with other genotypes than with Figure 11.1 Comparison of mean leaf shoot density (SE)their own genotype. In a study by Reusch et al. of Zostera marina among one-, three- and six-genotype treatments after climate perturbation, and natural shoot(2005) on a coastal community dominated by the density at the experimental site. From Reusch et al.seagrass species Zostera marina, experimental plots (2005). Copyright 2005 National Academy of Sciences,with a high genotypic diversity produced a higher USA.
  • 168. 154 FUTURE DIRECTIONS11.2.2 The source and nature of genetic studies to determine the mean correlation betweenvariation molecular and quantitative measures of genetic variation. Although molecular measures are com-Before discussing any further the possible benefits monly used as a proxy for quantitative genetic var-of genetic diversity, it is worth considering the iation, the observed correlation between the twoquestion of what is genetic variation? In many stud- was weak and not significant (r ¼ 0.217). This anal-ies on the effect of genetic diversity on population ysis shows the risk in using molecular measures ofor community functioning, genetic variation is genetic diversity to predict population performancetaken to be the number of genotypes present. How- or other population properties, unless there is di-ever, no measure of the degree of genetic dissimi- rect evidence that the markers used are indicativelarity is given. Although chances are negligible that of the underlying quantitative genetic variationtwo genotypes are exactly identical, that still leaves (Merila and Crnokrak 2001). In all other cases quan-a wide range of genetic dissimilarities possible. Ge- titative variation is better measured directly.nome-wide screening methods such as amplifiedfragment length polymorphisms (AFLPs) or theuse of large numbers of microsatellites or single 11.2.3 The relationship between geneticnucleotide polymorphisms (SNPs) may give a and phenotypic diversitymore or less reliable measure of overall geneticdiversity because they encompass the entire ge- Another reason to apply caution in the extrapola-nome. Even then, overall genetic diversity in itself tion from genetic diversity to population perfor-might not be sufficient to enhance population evo- mance is the existence of phenotypic plasticity.lutionary potential. Genetic variation matters only Phenotypic plasticity is the property of a givenif it translates into phenotypic variation upon genotype to produce different physiological orwhich natural selection can act. If the rise of molec- morphological phenotypes in response to changingular biology has taught us one thing, it is that the environmental conditions. Differences in pheno-relationship between genetic and phenotypic diver- typic plasticity can be quantified by measuring asity is not a straightforward one. Previously, the reaction norm (Via et al. 1995), which describes themost important distinction in genetic variation change in a trait across environments (Fig. 11.2a).was between coding variation and non-coding var- Reaction norms with a steep slope indicate strongiation such as neutral markers. Genetic variation in trait sensitivity to the environmental factor, i.e.coding regions leads to amino acid substitutions in strong phenotypic plasticity, whereas a flat reactionfunctional proteins, and therefore can contribute to norm denotes weak phenotypic plasticity.the relative fitness of the genotype in the popula- Phenotypic plasticity may account for much oftion. However, the long-held assumption that non- the phenotypic variation in populations and com-coding or neutral markers are not part of the pro- munities and is thought to play an important role incess of transcription and translation, and hence are adaptation to spatio-temporal heterogeneity in en-not subject to natural selection, has been challenged vironmental conditions. Induced phenotypic re-by advances in molecular biology. A much more sponses are a successful conditional strategy tocomplex picture of the molecular genetic organiza- cope with fluctuating conditions such as tempera-tion structure now recognizes variation in coding ture, particularly when there are costs involved inregions, promoter regions, transcription factor the induction of the response. For example, thebinding sites, regulatory genes, etc. induction of heat shock proteins protects the organ- The neutral molecular markers often used to ism from damage through misfolded proteins duequantify the genetic variability of populations are to heat shock exposure. However, there are sub-at best poorly related to variation in quantitative stantial energetic and metabolic costs involved,traits (Merila and Crnokrak 2001; Reed and Frank- due to repression of standard cell activity after ex-ham 2001; McKay and Latta 2002). Reed and Frank- posure to heat shock (Krebs and Holbrook 2001).ham (2001) carried out a meta-analysis across 71 Constitutive expression of this heat shock response
  • 169. EVOLUTIONARY PROCESSES IN COMMUNITY ECOLOGY 155 (a) conditions, however, often reveals hidden variation for such traits (Fig. 11.2b). For example, now that global change causes extreme temperatures to be- come more common, genetic variation hitherto un- exposed to selection will become apparent and important for survival. Overwhelming evidence now exists that the degree of phenotypic plasticity is genetically determined (Scheiner and Lyman Performance 1989; Loeschke et al. 1999) and that natural selection can lead to local adaptation in plasticity levels (b) (Liefting and Ellers 2008). There is a growing ap- preciation of the potential of phenotypic plasticity to modify species interactions. 11.3 Proof of principle: community properties result from genetic identity and selection at the level of individual organisms Environment How do population genetic processes extend to the community level? Population genetic theory couldFigure 11.2 (a) Schematic representation of different equally well apply to communities if genes havereaction norms, showing continuous reaction norms thatdiffer in the degree of sensitivity of the trait (slope of the extended phenotypes (Dawkins 1982; Whithamreaction norm) and in the maximum performance (height et al. 2003). Genes with extended phenotypes affectof the peaks), but not in optimal value of the not only the individual carrying the genes but alsoenvironmental variable. (b) Canalized reaction norms the performance of associated species in the com-show the same phenotype under average environmental munity. For example, we can think of genetic dif-conditions but deviate in performance at either extremeof the range of environmental conditions. ferences in secondary chemical compounds that affect plant defensive capability against foliar her- bivores (Havill and Raffa 2000; Harvey et al 2003).is therefore undesirable. Environmental conditions Although interspecific differences in allelochemicaltriggering phenotypic plasticity explicitly include composition of host plants have long been recog-the biotic environment as well. Well-known exam- nized as an important factor structuring commu-ples of phenotypic plasticity induced by species nities (Dungey et al. 2000; Sznajder and Harveyinteractions are the production of defensive plant 2003; Wimp et al. 2005), the potential effect of intra-compounds in response to herbivory and the induc- specific differences on community composition andtion of anti-predator defences. functioning has been acknowledged only relatively Genotypes may differ in their response to envi- recently. The performance and abundance of herbi-ronmental change, i.e. have different levels of phe- vores can be expected to be influenced in a similarnotypic plasticity. In other words, genotypes that way by intraspecific variation in host plants,have similar phenotypes under one environmental provided that the magnitude of genetic differencescondition may diverge in their performance when is large enough to detect extended phenotype ef-the environment changes, or vice versa. A common fects. Other examples of traits with extended phe-phenomenon causing seemingly monomorphic po- notypes are induction of morphological anti-pulations is canalization: genotypes have all predator defences in many amphibians, or varia-evolved to identical optimal trait values under the tion in thermal tolerance of algae in coral commu-most common environmental condition (Schlicht- nities which is related to bleaching mortalitying and Pigliucci 1998). Exposure to more extreme (Fabricius et al. 2004).
  • 170. 156 FUTURE DIRECTIONS The extended consequences of intraspecific genetic herbivore’s diet on growth and development of itsdifferences have received ample attention over re- primary parasitoid and secondary hyperparasitoid.cent years. Several recent studies show that plant Hyperparasitoid body mass and survival were sig-genotype significantly affects various community nificantly larger if the herbivore had fed on plantsproperties, such as disease infection (Roscher et al. containing lower concentrations of glucosinolates, a2007), arthropod abundance and diversity (Stiling chemical defence compound. Hence, phenotypicand Rossi 1996; Johnson and Agrawal 2005; Craw- plasticity caused by the nutritional differences inford et al. 2007) and decomposition rate (Madritch herbivore diet may affect performance across sev-et al. 2006). For instance, genetic identity of the even- eral trophic levels.ing primrose (Oenothera biennis) explained more than40% of the variation in arthropod diversity (Johnson 11.4 Effects of genetic and phenotypicand Agrawal 2005). Genotypic differences accounted diversity on community composition andfor more variation in arthropod community structure species diversitythan did environmental variation, indicating the rel-ative importance of genetic factors. Not only species In recent years more and more researchers haveinteractions are influenced by plant genotype. Ma- identified the influence of genetic diversity on com-dritch et al. (2006) showed that community properties munity functioning as an emerging frontier in ecol-such as decomposition and nutrient release from ogy (Whitham et al. 2003; Agrawal et al. 2007). Yet itaspen litter are determined by genotype identity. is not sufficient to merely observe a correlationLitter from different aspen genotypes differed signif- between genetic diversity and community proper-icantly in carbon and nitrogen release, most likely ties. To contribute to the understanding of generalbecause litter chemistries varied across genotypes mechanistic principles of community ecology, stud-(Madritch et al. 2006). ies need to assess the relative importance of ecolog- Most studies have focused on genetic differences ical and genetic factors, and address the mechanismin plants because plant diversity is thought to shape underlying these correlations.the composition and dynamics of animal commu- Individual organisms in communities can con-nities. Only some studies have looked at the effect tribute in two ways to the genetic diversity of theof genetic differences at other trophic levels than community. First, each species represents a discreteprimary producers. For instance, plant productivity genetic entity, and individuals belonging to differ-itself is to a large extent controlled by association ent species cause an associated rise in genetic diver-with arbuscular mycorrhizal fungi (AMF). Experi- sity within the community. The effect of speciesments with genetically different isolates of the ar- diversity on community stability and function hasbuscular mycorrhizal fungus Glomus intraradices been discussed extensively elsewhere (Hooper et al.show that AMF genotype can affect plant growth 2005), so I will touch upon this only briefly. Aresponse and the extension of the fungal mycelium consensus is emerging that, if plant species richness(Munkvold et al. 2004; Koch et al. 2006). The effect of increases, community productivity as well as stabil-genetic identity of Glomus intraradices on plant ity increases, although this effect levels off at rela-growth ranged from enhanced plant growth to no tively low species numbers. In the context of thisgrowth benefit or even a reduced growth, also de- chapter, I will mainly focus on the other way indi-pending on the environmental conditions (Koch et vidual organisms may contribute to communityal. 2006). Associations with different individuals in genetic diversity – if individuals represent differentthe AMF population will present plants with either genotypes within species.costs or benefits, potentially mediating the outcomeof competition between plants. 11.4.1 Effects of genetic diversity on Although most studies address the importance of community functioninggenotype identity, effects of different phenotypesmay also mediate community processes. Harvey Composition and functioning of communities iset al. (2003) examined the effect of differences in a thought to be strongly shaped by plant species
  • 171. EVOLUTIONARY PROCESSES IN COMMUNITY ECOLOGY 157diversity, hence the effect of genetic variation in the fects rates of decomposition and nutrient availabili-dominant plant species is the best studied. Since ty. Soil processes, in particular, appear to beplant genotypes can exploit slightly different re- influenced primarily by the functional characteris-sources, analogous to population-level effects of tics of dominant species rather than by the numberwithin-species genetic diversity, the prediction is of species present (Heemsbergen et al. 2004). As athat genetic diversity enhances productivity or re- final remark, genetic diversity is mostly defined assource efficiency of the plant, thereby increasing the the number of genotypes included in the communi-total availability of resources in the community. ty. The actual extent of genetic differentiation be-The most straightforward experimental approach tween genotypes, however, is not known, nor is theis to manipulate genotypic diversity in the commu- level of phenotypic differentiation or functional dis-nity and compare indicators of community perfor- similarity among genotypes included. Given themance, such as plant productivity, carbon storage, weakness of the relation between genetic diversitynutrient acquisition or decomposition rate. A flurry and variation in quantitative traits, the present con-of experimental work has now shown that clusion that genotypic diversity has communityincreased plant genotypic diversity can explain a benefits through complementarity still greatlysignificant proportion of the community properties, lacks detailed understanding.including an increased net primary productivity(Crutsinger et al. 2006), higher fruit production 11.4.2 Diversity begets diversity?(Johnson et al. 2006), higher resistance to invasion(De Meester et al. 2007) and accelerated decomposi- Can within-species variation also enhance speciestion rate (Madritch et al. 2006). Particularly in spe- diversity of the associated animal communities?cies-poor communities the effects of within-species Different genotypes of the dominant plant speciesdiversity have close parallels to benefits at popula- may favour different species in competitive andtion level. trophic interactions, leading to a mosaic of spatially It is unlikely that the positive effects of plant genet- varying selection pressures with a distinct set ofic diversity on community functioning are solely due associated arthropod species. This is known as theto sampling or selection effects. The sampling effect ‘diversity begets diversity’ hypothesis (Whittakercan be observed if diverse mixtures have a higher 1975; Vellend and Geber 2005). On the other hand,chance of containing highly productive genotypes genetic diversity enhances productivity and re-(Huston 1997). The positive effect of genotypic diver- source efficiency, such that the competitive strengthsity was not due to greater chances of obtaining of genetically diverse species increases and maymixtures with more productive genotypes in diverse lead to competitive exclusion of other species fromcommunities (Crutsinger et al. 2006). Similar to the the community (Vellend and Geber 2005). Al-benefits of genetic diversity at population level, the though such a hypothesis would predict that genet-complementarity principle seems to be the major ic diversity lowers community species diversity, itdeterminant of the increased fitness. Complementar- does not account for the positive effect of the in-ity may indicate that mixed genotypes facilitate one crease in available energy on the number of herbi-another (Hector et al. 1999; Mulder et al. 2001), or vores and predator species that can be sustained,niche differentiation among different genotypes the so-called ‘more individuals hypothesis’ (Srivas-causes the available resources to be used more tava and Lawton 1998). The relative strength ofcompletely (Loreau and Hector 2001). species exclusion versus energy availability deter- A major limitation of the former studies is that mines the sign of the relationship between geneticnearly all have focused on plant genetic diversity in diversity and species richness.aboveground communities. A badly needed next In fact, experimental studies present overwhelm-step is to test the generality of the effect of within- ing evidence that plant genotypic diversity is posi-species genetic diversity, by including higher tro- tively correlated with arthropod abundancephic levels such as herbivores. Also, little is known (Reusch et al. 2005; Crawford et al 2007) and arthro-about how genetic diversity of soil organisms af- pod diversity (Dungey et al. 2000; Johnson and
  • 172. 158 FUTURE DIRECTIONS 70 invasions, the main drawback of the competitive exclusion hypothesis is that it considers genetic 60 Total arthropod richness variation only in the keystone species of a commu- 50 nity. Although a number of invasibility studies sup- port the hypothesis (De Meester et al. 2007), it 40 appears invalid to explain species coexistence if all 30 species in a community show genetic variation. Booth and Grime (2003) found enhanced coexis- 20 tence of competing plant species in the most genet- 10 ically diverse communities. To a large extent this effect could be attributed to genotypic differences 0 1 3 6 12 in the initial population, but, especially in geneti- No. of plant genotypes cally impoverished communities, genotype by en- vironment interactions determined the structure ofFigure 11.3 The relationship between genotypic diversity of the resulting plant communities (Whitlock et al.Solidago altissima and total arthropod species richness in 2007) Given the paucity of empirical studies ad-one-, three-, six- and 12-genotype treatments. Circles dressing this issue, a provisional conclusion shouldindicate plot-level observations, and horizontal lines indicatetreatment means. Squares indicate the number of arthropod be that genetic diversity within species reduces thespecies predicted by simple additive models. Error bars rate at which species diversity declines.indicate 95% confidence interval. From Crutsinger et al.(2006). Reprinted with permission from AAAS. 11.4.3 Phenotypic diversity is also important for community diversity andAgrawal 2005, Wimp et al. 2005; Crutsinger et al. composition2006; Johnson et al. 2006). Experimentally manipu-lated plots with 12 randomly selected genotypes of Although some studies have been able to attributeSolidago altissima contained on average 27% more the positive effects of increased genetic diversity toarthropod species than monocultures (Crutsinger facilitation (Reusch et al. 2005; Crutsinger et al.et al. 2006). Both herbivorous and predatory species 2006), it has remained unresolved whether andbenefited significantly from increased plant geno- how non-additive effects among genotypes can betypic diversity, even more strongly than predicted predicted. The incidence of facilitative or inhibitoryby the cumulative species richness from each geno- interactions has been hypothesized to depend ontype grown in monocultures (Fig. 11.3; Crutsinger functional dissimilarity, defined as the degree ofet al. 2006). In a more detailed study on the same study dissimilarity in traits that affect community pro-system, the 12-genotype plots were also shown to cesses. This hypothesis is confirmed only at speciescontain 80% more galls than the one-genotype plots. level. Microcosm experiments with soil ecosystemsSince galls are a preferred habitat for a community and aquatic grazer communities demonstrated thatof secondary users and their predators, Solidago al- the effects of community composition on key eco-tissima genotypic diversity alters community struc- system processes can be predicted by measuringture of its associated arthropods through the the functional dissimilarity in the effect of individ-increased abundance of galls (Crawford et al. 2007). ual species on community processes. (Heemsber- As described above, the competitive exclusion gen et al. 2004; Wojdak and Mittelbach 2007).hypothesis predicts that genetically diverse popu- Facilitative interactions occurred in species mix-lations will have increased resource efficiency and tures with high functional dissimilarity or lowtherefore outcompete populations of genetically niche overlap, independent of species number. Ifimpoverished species. This hypothesis addresses the findings at species level have general applica-the effect of genetic diversity within a trophic bility, in genetically diverse populations the occur-level rather than across it. Because it was developed rence of facilitative or inhibitory interactions wouldto explain the resistance of communities to species depend on the degree of phenotypic dissimilarity
  • 173. EVOLUTIONARY PROCESSES IN COMMUNITY ECOLOGY 159 Trait-mediated interactions Genetic and phenotypic trait (a) (b) diversity (species 1) Phenotypic plasticity Selection A Prey feeding ratePlant growth rate Environmental Interspecific interactions B conditions (e.g. predation, competition) A Selection B Phenotypic plasticity Genetic and phenotypic trait diversity (species 2) Nutrient availability Absent Present Predator Figure 11.5 The three-way interaction between two species and their abiotic environment (G Â G Â E).Figure 11.4 The role of phenotypic plasticity in trait- Environmental conditions induce selection or phenotypicmediated interactions illustrated by two hypothetical plasticity in traits in component species, but notexamples. (a) A switch in competitive dominance under necessarily of the same strength. Performance traits ofincreased nutrient availability. Plant species A and B have species may also be mediated by direct interactions sucha differential growth rate response to nutrient availability as competition or physical interactions. For illustrationalindicated by the steepness of the reaction norm. Under purposes, only two species are shown here but anylow nutrient availability, species B has the highest growth number can be included.rate, but species A is better able to profit from highnutrient availability. Hence competitive dominanceswitches when nutrient availability is high. (b) Predator terspecific interactions transforms the standard ge-presence changes competitive interactions of two notype by environment (G Â E) interaction into aspecies. Species B has a higher feeding rate when three-way interaction between two species and thepredators are absent, but is also more sensitive topredation than species A. Therefore, when predators are environment (G Â G Â E; Fig. 11.5). Trait-mediatedpresent, species B has to lower its feeding rate to avoid interactions appear to be common in nature; forpredation, while species A can continue to feed at the instance, the presence of predators may influencesame rate and now outcompetes species B. foraging behaviour of prey and decrease prey growth rate (Peacor and Werner 2000; Prasad and Snyder 2006). Although identifying how each spe-among genotypes. However, these experiments cies alters its traits in the presence of others may behave yet to be performed. a daunting task, Relyea and Yurewicz (2002) show Some researchers refute a mechanistic approach that this approach yields accurate qualitative pre-to community ecology on the basis of the inherent dictions on the outcome of mesocosm experiments.complexity of multiple species interactions, and More simply measured species traits such as bodystate that, owing to non-additivity of interactions, size do not influence community system function-the dynamics of communities cannot be predicted ing in the long run, and seem to be merely effects offrom the individual characteristics of the compo- initial experimental conditions (Long and Morinnent species (Werner 1992; Sih et al. 1998). Howev- 2005). Including the details of such phenotypicallyer, these early studies did not take the effect of plastic responses in theoretical models can affectphenotypic plasticity into account. If interacting community dynamics and stabilize tritrophic sys-species differ in their response to environmental tems (Bolker et al. 2003; Verschoor et al. 2004).conditions, environmental change can alter thesign and magnitude of interactions among species 11.4.4 Phenotypic plasticity and invasive(Fig. 11.4), for which Abrams (1995) coined the term success‘trait-mediated interactions’. In addition to trait-mediated interactions, interacting species may di- Phenotypically plastic responses have also beenrectly modify each others’ gene expression or phe- implied in the context of invasion success. Invasivenotype by chemical communication, as in the case species can dramatically change community com-of induced defences. Condition dependence of in- position owing to their negative effects on
  • 174. 160 FUTURE DIRECTIONSindigenous biodiversity (Mack et al. 2000). Pheno- the ‘diversity begets diversity’ hypothesis proposestypic plasticity may increase invasive success if an that species diversity can act as a source of diversi-individual’s plastic response to environmental fying selection on single species, because each in-change allows it to expand ecological niche breadth teracting species represents a different competitive(Baker 1965; Richards et al. 2006). Invasive species environment. If different genotypes of the focalare therefore predicted to be more plastic than their species have differential competitive responses,native community members. Richards et al. (2006) a positive effect of community diversity on within-describe three ways in which phenotypic plasticity species genetic diversity is predicted. In fact, suchmay contribute to invasive success: a jack of all fine-scale environmental variation in selection pres-trades, able to maintain fitness in unfavourable en- sures suggests the occurrence of local co-adaptationvironments; a master of some, able to profit in fa- between neighbours. Empirical evidence of suchvourable environments; and a jack-and-master, interactions is found for Taraxacum officinale geno-which combines both. Several studies have demon- types, which exhibited differential response in rootstrated that the shape of the reaction norm differs biomass, total biomass and leaf area depending onbetween native and invasive species; however, the the species identity of the competitor (Plantagoform of changes and the type of traits showing major, Poa pratensis or Trifolium pratense). Some gen-differences in plasticity levels seem to be idiosyn- otypes performed best with a specific competitor,cratic (Chown et al. 2007; Muth and Pigliucci 2007). but others behaved like generalists with nearly equal performance with all competitors (Vavrek 1998). Similar results, at an even more detailed11.5 Effect of community composition on level, demonstrate that performance depends onthe genetic and phenotypic diversity of the genetic identity of both interacting speciessingle species (Fridley et al. 2007). These studies provide strongSo far, I have discussed only the possible conse- support for the diversity begets diversity hypothe-quences of individual diversity on higher level pro- sis, and give good reason for more empirical em-cesses, and the benefits of adopting a more phasis on this topic. Although it is obviously notevolutionary perspective for community ecologists. feasible to include such detailed genetic analysis asThe opposite side of this integrative approach is a standard protocol in community ecology, it willobviously that evolutionary biologists should be- increase our appreciation of the relevance of geneticcome aware of the community context in which diversity in multispecies assemblages and the im-evolution takes place (Fig. 11.5). Performance of portance of spatially fluctuating biotic environ-individual species is typically mediated by inter- ments for the maintenance of genetic variation.specific interactions, and the identity of interacting The alternative hypothesis on the effect of com-species can determine individual response to munity composition on genetic diversity of singlechanging conditions (Davis et al. 1998; Jiang and species proposes that a diverse community con-Morin 2004). Natural selection experiments often strains the ability of the focal species to exploitleave a large proportion of evolutionary change different niches in the environment. Therefore,unexplained, perhaps because fine-scale environ- community diversity should act as a source of sta-mental variation due to identity of competitors is bilizing selection, reducing the variation withinnot considered. Several experiments have now re- species. To my knowledge, no studies have explic-vealed that plant response to competition is itly measured the degree of genetic variation of amediated by interspecific and intraspecific varia- focal species in communities with manipulatedtion in community composition (e.g. Vavrek 1998; community composition. We may tentatively rejectCallaway and Aschehoug 2000, Fridley et al. 2007). this hypothesis because in experimental commu- To predict how community diversity may affect nities with a fixed number of species with eithergenetic diversity of single species, we can basically high or low genotypic diversity, each species re-just reverse the causation in the two hypotheses tains more genotypes when reared with highly di-discussed in the previous section. In this context verse heterospecifics (Whitlock et al. 2007). This is
  • 175. EVOLUTIONARY PROCESSES IN COMMUNITY ECOLOGY 161opposite to the prediction of the hypothesis, but ever, we are only at the beginning of the difficultevidently more work is needed to confirm these process of integrating evolutionary principles withresults. the community level. There is an obvious need for a A specific prediction derived from this hypothe- greater number of field studies to show the gener-sis is that an increase in community diversity ality of the results discussed above, especially fromthrough invasion by a new species will lead to a the part of the evolutionary biologists. Also, thereduction of genetic variation in the native species range of taxa involved in these studies needs to bewithin that community. So far, changes in genetic expanded, because the majority of studies concernvariation as a result of community invasion have plants. Because for some communities measuringbeen assessed in only one study system. Hild and genetic and phenotypic diversity for all species willco-workers studied genetic variability in three na- be a prohibitively large undertaking, a carefultive grass species after invasion of Russian knap- choice of study systems as well as inventive use ofweed (Acroptilon repens) and found a significant molecular tools is necessary.reduction in genetic variability after invasion in If an integrated approach succeeds in addingone native species, but not in the two others (Mea- significant explanatory and predictive power, itlor et al. 2004). Also, genetic similarity of both grass may develop into a new discipline of evolutionaryspecies was smaller between invaded and non-in- community ecology. A true integrative approach tovaded communities than within community type, evolutionary community ecology can address aindicating that there is natural selection in response much wider range of issues than just the reciprocalto community invasion (Mealor and Hild 2006). effects of genetic and species diversity that are cur-Consistent with this finding, an increasing number rently the main focus. Looking from an evolution-of studies shows that natives do show phenotypic ary perspective, a crucial question is if, and to whatchanges after exotics have been introduced (re- extent, genetically diverse species persist longer inviewed in Strauss et al. 2006; Strayer et al. 2006) a community than genetically uniform species. Thisand these studies may provide indirect evidence is particularly relevant if one considers plant spe-on changes in genetic diversity of native species. cies that can propagate clonally, or animals thatAdaptation to exotics occurs often in the form of reproduce parthenogenetically. Another essentialshifts in single traits due to directional selection, question is the relationship between communitypotentially decreasing genetic variation in these composition and evolutionary change of speciestraits. On the other hand, in the case of plant inva- within communities. To what extent can communitysions, native phytophagous insects can even be ex- composition drive selection on traits and can diver-posed to divergent selection if the induced gent selection caused by differences in communityevolutionary change involves adaptation to a composition perhaps initiate ecological speciation?novel host plant, which would enhance genetic On the other hand, multispecies interactions invol-variation. At present, we lack sufficiently detailed ved in diffuse coevolution within communities maystudies to draw any clear conclusions on this hy- constrain the possibilities for evolutionary change ofpothesis. trait values, but favour the development of plastic phenotypes. A multitude of key questions can also be ad-11.6 Future directions dressed from a community ecological point ofThe empirical evidence presented in this chapter view. One question I would regard as crucial isoverwhelmingly supports the reciprocal relation- whether species at particular trophic positions in aship between community ecology and evolutionary community are more likely to be genetically di-biology. Genetic diversity and phenotypic plasticity verse? If genetic diversity is maintained by fine-of component species can shape community struc- scale environmental variation in interspecific inter-ture and composition. Similarly, interspecific inter- actions, the degree of connectiveness of species mayactions and the species composition of communities govern their genetic diversity to a large extent. Acan influence evolutionary change of species. How- second major question is the relationship between
  • 176. 162 FUTURE DIRECTIONSphenotypic plasticity and community diversity. It is clear to most of us that evolutionary processesPhenotypic plasticity often allows individuals to in community ecology can no longer be ignored.maintain fitness homeostasis over a much broader Whether the emerging field of evolutionary commu-range of conditions. Can phenotypically plastic spe- nity ecology will reach its full and exciting potentialcies obtain a broader niche in the community than will depend on the willingness of researchers to lookspecies that have a fixed phenotype? If so, does beyond the boundaries of their own discipline. Weplasticity give them a competitive advantage over need to stop seeing the inclusion of genetic diversitycompeting species and is the effect of phenotypic or community context as a dictated nuisance, butplasticity equivalent to the effect of intraspecific rather explore its usefulness in obtaining a better un-genetic diversity? derstanding of pure and applied problems in biology.
  • 177. Chapter 12 Emergence of complex food web structure in community evolution models Nicolas Loeuille and Michel Loreau12.1 A difficult choice between dynamics that food webs are small worlds (Martinez et al.and complexity? 1999; Montoya et al. 2006), that they are built in compartments (Pimm 1979; Krause et al. 2003) andA food web is defined as the set of species linked by that they contain many loops (Polis 1991; Neuteltrophic interactions in a given ecological community. et al. 2002), a lot of omnivores (Polis 1991), etc.As such, it contains only a subset of the many possi-ble types of ecological interactions and it is a very · finding simple models that would be able to re- produce these features; models such as the Cascadesimplified representation of natural communities. In model (Cohen et al. 1990; Solow and Beet 1998), thespite of this simplification, food webs appear to be Niche model (Williams and Martinez 2000) and thehighly complex networks, if only because any natural Nested Hierarchy model (Cattin et al. 2004) havesystem contains several hundreds of species, most of been relatively successful in reproducing some ofthem preying upon or being preyed upon by many the patterns observed in these binary data sets.others (e.g. Polis 1991). Many food web data sets arenow available (Baird and Ulanowicz 1989; Warren Binary approaches to food webs have been used to1989; Hall and Raffaelli 1991; Martinez 1991; Polis draw conclusions about community structure (e.g.1991; Goldwasser and Roughgarden 1993). food web stability: Pimm 1979; Krause et al. 2003) or It is possible to divide such data sets into two conservation issues (fragility of food webs to spe-broad categories. The first category will be called cies removal: Dunne et al. 2002). In spite of these‘binary’ data sets. Binary data sets simply list spe- results, drawing conclusions from binary data sets,cies in the food web and the trophic interactions or from models that are built on them, to broadamong these species. They do not contain any in- ecological issues has proved to be very controver-formation in terms of species abundances or trophic sial. Binary approaches have a number of short-interaction strength. Food web theory that deals comings:with binary data sets is primarily interested in: · descriptors used in binary approaches are highly· comparing food web networks with other types dependent on species lumping (Solow and Beetof networks such as protein, genetic, social, neuro- 1998) and on the resolution of the data set (Wine-nal and communication networks (Barabasi and miller 1990; Martinez 1991)Albert 1999; Amaral and Ottino 2004; Milo et al. · properties measured on binary data sets do not2004; Grimm et al. 2005; Proulx et al. 2005) describe the ecological properties of the community· from this comparison, determining properties satisfactorily; for example, Paine (1980) criticizedthat are specific to food webs as compared with the use of connectance as derived from these dataother types of networks – for example, the fact sets 163
  • 178. 164 FUTURE DIRECTIONS· binary approaches generally describe foods webs One possible solution is the use of communityat a given time, while food webs prove to be highly assembly models, in which species are drawn fromvariable in time (Paine 1988) a predetermined regional pool (Post and Pimm· because they do not measure species abundances 1983; Taylor 1988; Morton and Law 1997; Steinerand interaction strength (Cohen et al. 1993a; Berlow and Leibold 2004). This type of model has providedet al. 2004), they are unable to deal with conserva- useful information on the conditions for the main-tion issues (mostly based on species abundance) or tenance of large, stable communities. An obviousfunctional aspects of ecosystems (such as energy limitation of these models is that, even when theand nutrient fluxes). pool of species is large, it is unable to account for novelties that arise through evolution, and that areAn obvious alternative to these high-diversity, static potentially infinite. This shortcoming has been ad-approaches is to describe the dynamics and evolu- dressed by the recent development of evolutionarytion of species in small food web modules. This ap- food web models (Caldarelli et al. 1998; Drossel et al.proach has been recently reviewed extensively by 2001; Christensen et al. 2002; Anderson and JensenFussmann et al. (2007). Theoretical studies that follow 2005; Loeuille and Loreau 2005; Ito and Ikegamithis approach often consider coevolution of two spe- 2006; Rossberg et al. 2006).cies (e.g. Levin and Udovic 1977; Saloniemi 1993; The Webworld model (Caldarelli et al. 1998;Abrams and Matsuda 1997; Loeuille et al. 2002; Der- Drossel et al. 2001), for example, is based on acole et al. 2006) or evolution of food web modules that large number of traits that may mutate. Traits maycontain a restricted number of species (Vermeij 1987; be present or absent; thus, species are coded byAbrams 1991, 1993; Abrams and Chen 2002; Yamau- vectors of 0s and 1s of a predefined length so thatchi and Yamamura 2005). These models provide in- the set of species is still finite. An alternative is toteresting insights into species coexistence (Yamauchi base evolutionary models on a few key traitsand Yamamura 2005) the strength of bottom-up or (Loeuille and Loreau 2005; Ito and Ikegami 2006),top-down controls (Loeuille and Loreau 2004), the among which there are trade-offs that are eitherconditions for the maintenance of intra-guild preda- known or inferred from physiological or morpho-tion or omnivory (Krivan and Eisner 2003), the con- logical constraints. An obvious candidate in theditions for the stability of food web modules case of trophic interactions is body size. Body size(Abrams and Matsuda 1997; Loeuille et al. 2002; has been suggested to play an important role in theYamauchi and Yamamura 2005; Dercole et al. 2006), structure of food webs (Cohen et al. 1993b, 2003;etc. It is unclear, though, how such mechanisms Neubert et al. 2000; Jennings et al. 2002a, b; Wood-derived from a small number of species may be ex- ward and Hildrew 2002; Emmerson and Raffaellitended to natural ecosystems that are much more 2004; Williams et al. 2004, Crumrine 2005). Confir-speciose and complex. mation of this importance has come from measures Thus, theory is abundant either when dealing showing the tight relationship between the relativewith large systems but without dynamics or quan- difference in body size between predators and preytitative information, or when dealing with small and the strength of their interaction (King 2002;dynamical systems in which populations and inter- Jennings et al. 2002b; Emmerson and Raffaelliactions are explicitly described. The remaining 2004). Body size has also been shown to be ofchallenge is to develop frameworks that are able importance for many other life-history traits (Klei-to deal with dynamical systems that contain a ¨ ber 1961; Peters 1983; Bystrom et al. 2004; Jetz et al.large number of species and that are able to account 2004; Savage et al. 2004; Reich et al. 2006).satisfactorily for the binary and quantitative aspects In this chapter, we summarize some of the prop-of food webs. This is a long-standing issue since erties of a community evolution model that is en-Polis (1991) already stressed 16 years ago that tirely based on the evolution of body size. Thetheory (Pimm 1982; Pimm and Rice 1987; Cohen model shows that, starting with only one morphet al. 1990) was insufficient to tackle the complexity characterized by its body size, it is possible to ob-of natural systems. tain stable, complex food webs out of repeated
  • 179. EMERGENCE OF COMPLEX FOOD WEB STRUCTURE 165adaptive radiations. The results of the model are Community evolution models are in some waysthen compared with those of other evolutionary very close to classical community assembly models,food web models to give an overview of possible since these also contain invasion and selection pro-uses of community evolutionary approaches in cesses. The main difference between the two typescommunity ecology. of models lies in the details of the invasion process. In community assembly models, species are intro- duced from an existing regional species pool (e.g.12.2 Community evolution models: Post and Pimm 1983; Taylor 1988; Morton and Lawmechanisms, predictions and possible 1997; Steiner and Leibold 2004). For this reason, thetests introduced species do not have to be functionally similar to species already present in the commu-Community evolution models let entire commu- nity. Trade-offs between species traits are generallynities emerge from the basic evolutionary processes not considered. The timing of the invasion is notof mutation and selection. These models start constrained, and the diversity of the local specieswith one or a small number of species, and assemblage is bounded by the total number of spe-new morphs emerge out of repeated mutations. cies present in the regional species pool. By con-When a mutant is introduced in the system, the trast, community evolution models have harsherselection process comes into play to determine