This study sought to determine if patterns of microbial diversity in canopy soil patches are consistent with patterns of diversity on islands as described by island biogeography theory. The researcher collected soil samples from the ground and canopy of trees in a Costa Rican cloud forest. Microbial morphospecies were identified and diversity metrics were calculated and compared between ground and canopy soil. Canopy soil had significantly lower species richness than ground soil, supporting the idea that canopy patches limit microbial diversity. However, other diversity patterns did not fully match island biogeography predictions, and ground-canopy microbial communities were more similar than expected. This suggests that movement between ground and canopy may not form as strong a barrier to microbial dispersal as the theory predicts.
Microbial Islands: Patterns of Soil Microbe Diversity in Canopy Soil
1. SALTMAN QUARTERLY • VOL 11 sqonline.ucsd.edu sq.ucsd.edu VOL 11 • SALTMAN QUARTERLY37 38
RESEARCH RESEARCH
Introduction
Soil microbial communities are known
to be a significant driving force of global
biogeochemical cycles [1]. However,
compared to that of the well-studied topic
of macroscopic plant and animal diversity,
documentation and understanding of
biodiversity in soil microbial communities
are relatively poor [2]. In order to better
understand how soil microbes fit into
ecosystems, biomes, and the earth system,
studies must be focused on documenting and
characterizing the diversity of soil microbial
communities and its relationship to space
and time. Only once these patterns of soil
microbe diversity are better understood can
action be taken to protect that diversity and
its benefits to humanity.
In the realm of soil biology, canopy soil
has been even less thoroughly studied. The
term “canopy soil” refers to soil that forms
on relatively horizontal surfaces of trees [3].
Airborne organic matter can land on these
surfaces, become stuck, and decompose [3].
For example, falling leaves might land in
the fork of a tree and build up over time.
When this organic matter decomposes, it
forms topsoil, or A horizon soil, in the same
way that O horizon soil (leaf litter) on the
forest floor decomposes to form topsoil [4].
This soil contains microbes and is capable of
supporting the growth of epiphytic plants.
These soil patches are isolated from
the ground soil and from other canopy
soil patches. Microbes may be moved
about on the ground, to the canopy, and
between canopy soil patches by wind [5]
or by animals. Microbial movement to
and between patches is likely limited in
comparison with movement between sites
on the ground. Therefore, canopy soil
patches might be thought of as microbial
islands, and ground soil as a mainland that
is the source of the microbial species that
colonize canopy soil. According to the model
of island biogeography, islands can sustain
fewer species than can a mainland [6]. The
similarity of island communities to mainland
communities decreases as distance from the
mainland increases (Figure 1 left). Also,
because islands are geographically isolated,
movement between them is limited, which
promotes differentiation of community
Microbial Islands: Soil Microbe Diversity in Canopy Soil Patches
Decaying organic matter forms patches of soil in trees, known as “canopy soil”, which are isolated from
the ground. This study sought to determine whether patterns of microbe diversity in canopy soil patches
are consistent with patterns of macrospecies diversity on islands. Via microscopy, I recorded microbial
morphospecies abundance in ground and canopy soil. Canopy soil contained significantly lower alpha
diversity than ground soil and the similarity between the microbial communities of those pairs decreased
as the distance between them increased. Ground-canopy pairs were significantly more similar than pairs
of ground soil samples and pairs of canopy soil samples. I found no significant difference in species
evenness or beta diversity between ground soil and canopy soil. The data indicate that canopy soil
patches limit the number of species that can inhabit them. However, the separation between ground and
canopy soil may not actually form a barrier to microbial movement.
Josh Kenchel
Figure 1 The island biogeography model. (Left) Diagram of species movement between islands and a mainland. (Right) Visualization of canopy soil microbe
diversity following the island model. Following the model of island biogeography, the similarity of a canopy soil microbial community to the ground soil
microbial community should decrease as the height of the canopy soil patch increases.
composition; thus, there is higher species
turnover between islands than on a mainland
[6].
The theory of island biogeography has
been applied before to non-island systems.
For example, the island biogeography
model has been used to conceptualize the
movement of species between patches of
fragmented habitat [7]. In modern usage,
any patch of habitat separated from all
other habitat by non-habitable space might
be termed an “island” of habitat. I posited
that canopy soil patches separated from the
ground soil might be like islands for soil
microbes, because the space in between is
not habitable.
In this study, I investigated microbial
community diversity in ground soil and
canopy soil in a cloud forest in Costa Rica. The
purpose of the investigation was to determine
whether soil microbe diversity differs between
ground soil and canopy soil, and whether the
patterns of microbe diversity are consistent
withtheapplicationoftheislandbiogeography
model. I defined alpha (local) diversity as the
species richness of each soil patch, gamma
(regional) diversity as the total richness of all
soil patches that I studied, and beta diversity as
the change in species content between patches.
I hypothesized that canopy soil would have
greater beta diversity and gamma diversity,
lower alpha diversity, and lower species
evenness than ground soil. I also expected the
similarity between ground soil and canopy
soil microbial communities to decrease as the
distance between them increased.
Hypothesis: the patterns of microbial
biodiversityincanopysoilpatchesareconsistent
with the model of island biogeography.
Materials and Methods
STUDY SITE — I acquired ground soil
and canopy soil samples from trees in the
cloud forest on the property of the Estación
Biológica Monteverde, Puntarenas, Costa
Rica. The site spanned an area of about 0.25
km2
between the elevations of 1480 m and
1540 m above sea level.
SOIL COLLECTION — I located trees
that had canopy soil located between 0.5
m and 7 m above the ground. Of those
candidate trees, I selected those whose
canopy soil I could safely reach by climbing.
At each selected tree, I collected about 1
g of A horizon canopy soil and 1 g of A
horizon ground soil from the base of the
tree. I also measured the distance from the
ground to the canopy soil patch. I stored
the soil samples in 50 mL plastic bottles. I
recorded the latitude and longitude of each
Figure 2 Change in microbial species similarity as canopy soil patch distance increases. Each point rep-
resents a single site. The index of similarity between the ground sample microbial community and the
canopy sample microbial community for that site is plotted against the distance between the canopy soil
patch and the ground. A negative power model best fits the data, and the trend is statistically significant
(p < 0.001). The maximum value of the index of similarity is one, and the similarity between any two sites
should approach zero as the distance between them increases to infinity. Thus, it stands to reason that a
non-linear model is the best fit for the data.
Table 1 System for identification of morphospecies. I assigned each morphospecies a six-letter code describing (in order) its classification, size, color, shape,
motility, and social aggregation.
2. SALTMAN QUARTERLY • VOL 11 sqonline.ucsd.edu sq.ucsd.edu VOL 11 • SALTMAN QUARTERLY39 40
RESEARCH RESEARCH
tree using a GPS device. In total, I sampled
20 trees, hereafter referred to as “sites”,
between the dates of 10 November 2013 and
29 November 2013.
SLIDE PREPARATION — I stored each
sample at room temperature and analyzed
it within two days of collecting it. Analysis
involved viewing the soil samples through
a compound light microscope. To prepare
slides for viewing, I added 0.2 g of soil to
3.0 mL of distilled water and mixed them
thoroughly. Next, I filtered this mix through
a 0.1 mm filter to remove large soil particles
but leave single-celled organisms in the
suspension. I placed one drop of the filtrate
on the slide for viewing.
MORPHOSPECIES IDENTIFICATION —
I observed one slide from each sample at
400X magnification. Due to limitations of
time and available equipment, I identified
different microbe species only to the
level of morphotype. While not as finely
resolved at the microbial level as other
species concepts such as the molecular
species concept, the morphological species
concept has been used to obtain measures of
biodiversity [8]. Using a transparent grid,
I counted the number of morphospecies
and the number of individuals of each
morphospecies that I observed in a 4 mm2
area. I created a system of classification
of microbial morphospecies (Table 1) in
order to maintain consistency between
samples and facilitate data analysis. I
assigned each morphospecies a six-letter
code describing its visual appearance. For
example, following Table 1, a single 10
µm green rod-shaped bacterium with no
observed motility would be recorded as
BSGRNS. (See Figure 6 for an example of
morphospecies BSGRNS.) Thus dissimilar
morphospecies were assigned unique
codes, but like morphospecies from
different samples were assigned identical
codes, allowing for similarity comparisons
between samples.
STATISTICAL ANALYSIS — For each
sample, I calculated the observed species
richness and the Shannon H’ index for
species evenness [9], as well as evenness
(E), calculated as H’/H’max, which allows
for direct comparison between two samples
with different species richness. For each
site, I calculated the Sørensen-Dice index of
similarity [10] between the morphospecies
composition of the ground soil and canopy
soil. For tests of significance, I treated the
groundsoilsampleandthecanopysoilsample
from each site as a matched pair. I performed
a matched pairs t-test to compare the species
richness of ground soil and canopy soil and
a special t-test to compare the H’ values for
species evenness. I calculated alpha, beta,
and gamma diversities for each canopy and
ground soil.
I plotted indices of similarity between sample
pairs against the distance between those pairs
and performed a least-squares regression of
the data. For comparison, I also calculated
the similarity between each unique pair
of ground soil samples and between each
unique pair of canopy soil samples. Using
the GPS data, I calculated the distances
between all of the trees. Then I plotted each
of the similarities between ground samples
and the similarities between canopy samples
against the distances between them. Lastly,
I calculated the mean similarity between
ground-canopy pairs, ground-ground pairs,
and canopy-canopy pairs, and compared
those means with an ANOVA test.
Results
SPECIES RICHNESS — Alpha and beta
diversity values for the ground and canopy
soil samples followed the pattern that was
hypothesized (Table 2). The ground soil
had significantly greater species richness on
average than the canopy soil. Species turnover
was greater for the canopy samples, but the
difference was not statistically significant.
The total richness (gamma diversity) of the
ground soil was greater than that of the
canopy soil.
SPECIES EVENNESS — When the
frequencies for all ground morphospecies
and canopy morphospecies were combined,
the total evenness of canopy soil was greater
than that of ground soil (Figure 5). However,
this difference was not statistically significant.
DISTANCE-DEPENDENT SIMILARITY
— When the index of similarity of each
ground soil sample with its corresponding
Figure 3 Change in canopy microbial species richness as canopy soil patch distance increases. The data
show no significant trend.
canopy soil sample was plotted against the
distance between them, the data yielded
a significant, non-linear, negative trend
(Figure 2). The maximum value of the
index of similarity is one, and the similarity
between any two sites should approach zero
as the distance between them increases to
infinity. Thus, it stands to reason that a non-
linear model is the best fit for the data.
To ensure that the observed trend was not
confounded by a parallel trend of species
richness, I plotted the canopy soil richness
at each site against the canopy soil’s distance
from the ground (Figure 3). In fact, the
data showed a trend of increasing richness
with distance, but the correlation was not
significant. Therefore the observed negative
trend of distance-dependent similarity does
not appear to be the result of confounding
effects by species richness.
When I plotted the indices of similarity
between ground soil samples against the
distance between them, the data yielded
no trend (Figure 4). The analogous plot of
similarity between canopy soil pairs also
yielded a trendless scatter (Figure 4). The
negative relationship between similarity and
distance seen in Figure 2 did not appear over
this larger spatial scale.
Controlling for distance, the difference
between the similarity of canopy soil pairs
and that of ground soil pairs was not
statistically significant. Furthermore, the
similarity between ground-canopy pairs was
significantly greater than that of the other
two pair types (Figure 6).
Discussion
SUPPORTING EVIDENCE — The
patterns of alpha and beta diversity that were
observed in the data support the statement
that canopy soil patches constitute islands of
microbial habitat. Ground soil appears to be
able to support more microbial species than
canopy soil. There may also be greater species
turnover between canopy soil patches than
between ground soil sites. This suggests that
microbial movement between the canopy
and the ground is restricted. However, this
difference in the data is not statistically
significant.
There was also greater gamma diversity in the
ground soil, which suggests that although
species turnover is greater in the canopy, it
is not a hotbed of differentiation. Perhaps
canopy soil is a habitat niche that can support
only a relatively few specialized species. This
is often the case with islands as well, whose
restrictive ranges and habitat diversity limit
the number of species that can survive on
them [6]. However, by comparison, island
chains give rise to greater gamma diversity
through allopatric speciation [11].
I observed the opposite of my hypothesized
difference in species evenness between the
ground and the canopy, but this result was
not significant. It may be the case still that
canopy soil patches tend to be dominated by
a few specialized species. However, it might
also be the case that because there tend to be
fewer species in canopy soil, the distribution
of those species is more even; ground soil
might contain more species, but a greater
proportion of those species might exist at
very low frequencies.
Lastly, and perhaps most convincingly, the
significant trend between canopy soil distance
from the ground and ground-canopy similarity
matches the expected trend. The farther the island
is from the mainland, the less similar the species
composition between the two is. In this case,
the canopy soil patch is the island, the ground
is the mainland, and the space between them is
an ocean that restricts microbial movement. As
shown in Figure 3, this trend is not simply due to
the significantly lower species richness of the soil
islands.Thesimilaritybetweenthegroundsoiland
the canopy soil patch is dependent on the distance
betweenthetwo.
COUNTEREVIDENCE — The data
show that this downward trend of similarity
disappears at the larger spatial scale in Figure
4. The data suggest no more than a mean
index of similarity around 0.2, independent
of distance. It may be the case that within any
plot of cloud forest of the size of my study
site – a few hundred meters in diameter – the
true index of similarity of the soil microbe
communities at any two sites is no less than
0.2. It does appear that the data in Figure 2
approach a potential asymptote of similarity
of 0.2 as distance increases. However, this
does not explain why the mean similarity
between canopy samples is not significantly
different than the mean similarity between
ground samples, or why the mean similarity
between ground-canopy pairs is significantly
greater than the other two. If microbial
movement to the canopy is in fact restricted
in comparison to movement between sites
on the ground, then ground-canopy pair
Figure 4 Similarity between ground samples and between canopy samples vs. distance. For each of the
190 unique pairwise combinations of sites, the indices of similarity between the two ground samples
and between the two canopy samples are plotted against the distance between the sites. The data show
no significant trends.
Table 2 . Diversity comparison between ground and canopy samples. For both ground and canopy, n = 20.
3. SALTMAN QUARTERLY • VOL 11 sqonline.ucsd.edu sq.ucsd.edu VOL 11 • SALTMAN QUARTERLY41 42
RESEARCH RESEARCH
similarity should be significantly lower than
the similarity between ground samples,
and the similarity between canopy samples
should be lower still. Therefore, the data
suggest that movement between the ground
and the canopy is in fact no more restricted
than movement between any two sites on the
ground. Transport by wind, birds, or other
animals might be approximately equal on the
ground, between the ground and the canopy,
and between canopy patches.
RESOLVING THE CONTRADICTION —
This result may be due to the difference
in spatial scale of the height of trees and
the distance between trees. The canopy
soil patches I sampled were not truly in
the canopy, but rather much closer to the
ground. A negative trend between height
and similarity was resolved at this smaller
spatial scale, but disappeared at the larger
spatial scale. Some site pairs were less than
10 m apart; those data alone also do not
show any trend. However, I measured
these distances using a GPS device that
is only accurate to within a 10 m radius.
Thus, the answer to resolving the apparent
contradiction in the data may lie in using
more comparable spatial scales. To return to
the analogy of island biogeography, the index
of similarity between Isla del Coco, Costa
Rica and an island in the Santa Barbara
Channel, California is likely similar to that
between mainland Costa Rica and mainland
California. Likewise, the indices of similarity
of those islands to their respective mainlands
are almost certainly greater than both the
island-island and the mainland-mainland
indices of similarity.
Anotherpossible–andnotmutuallyexclusive
– explanation of the apparent contradiction
in the data also involves the concept of scale.
Classification by morphotype is relatively
easy and inexpensive, but when applied to
microbes it has limited accuracy. I classified
morphotypes according to motility and
clumping behavior, which may be dependent
on temperature or other factors [12]. Also,
two identical microbial morphospecies might
very well be different species. For example,
consider the morphospecies mentioned in
the description of Table 1 and shown in
Figure 6, BSGRNS. Visually similar or even
identical green, rod-shaped bacteria (likely
cyanobacteria) can be found in nearly every
body of water and probably every soil in
the world, but DNA analysis would show
that they are in fact many different species.
It is impossible for me to know how many
different species of BSGRNS, or other
common morphotypes, I actually observed.
That knowledge may have affected the results.
There are ways to easily identify all the
species in a soil sample to the molecular level.
Recent advances in molecular technology
have led to the development of high-
throughput DNA sequencing techniques
[13]. The emerging field of “metagenomics”
involves the use of these techniques to
identify all of the microbial species present
in an environmental sample at the molecular
level [13]. Metagenomic techniques might
be used to resolve soil microbe community
composition to a finer level and improve the
accuracy of the measures of diversity and
similarity that I used in this study.
Therefore, for future attempts to study this
system, I can suggest two improvements.
The first is to change the spatial scales of the
study so that they are comparable. Sample
the very tops of trees in order to increase the
distance of the island from the mainland.
The index of similarity of a soil patch 50
m high can more accurately be compared
to the index of similarity of ground soil
samples from the bases of two trees 200 m
apart. Controlled experiments to determine
the method of transport of microbes on the
ground and to canopy soil patches should
also be performed. Second, identify species
to a finer level than morphotype. This might
involve classical microbiology techniques
such as pure culture and dichotomous
tests, but I suggest more high-throughput
methods such as metagenomics. The finer
the resolution is, the more accurate the data
will be.
Conclusion
The model of island biogeography does
not appear to perfectly fit the dynamics of
microbial diversity in canopy soil patches,
but neither is it entirely invalid. The data
do support the conclusion that like islands,
canopy soil constitutes patches of microbial
habitat, limited in size and perhaps
resources, which limit the number of species
that can inhabit them. However, it cannot
be concluded that microbial movement
between the ground and the canopy is any
less frequent than movement between sites
on the ground. Perhaps future studies will
resolve whether this is the case.
Acknowledgements
I would like to thank Sofía Arce Flores
and Federico Chinchilla for their guidance
and support in planning, executing, and
presenting this project. I am grateful to the
Leitón Bello family for housing and taking
care of me; this project would not have
been possible without them. El Institúto
Monteverde and UCEAP provided the
necessary funding and materials for this
project. Thank you to Morgan Boyles and
Phebe Meyers for their help with GPS
and GIS, to Denisse Ruiz for proofreading
my Spanish, and to Shohei Burns for
accompanying me on one of my more
vertigo inducing climbs. Lastly, I would like
to thank each of my EAP student peers and
my professors for continuously inspiring me
to do great things, for making me proud to
be a biologist, and for making this work all
the more enjoyable through their company
and counsel.
References
1. Fitter, A. H. Darkness visible: reflections on
underground ecology. Journal of Ecology 93,
231-243 (2005).
2. Fierer, N. et al. Cross-biome metagenomic
analyses of soil microbial communities and their
functional attributes. PNAS 109, 21390-21395
(2012).
3. Orlovich, D. A. Piracy in the high trees:
ectomycorrhizal fungi from an aerial “canopy
soil” microhabitat. Mycologia 105, 52-60 (2013).
4. Killham, K. Soil Ecology. Cambridge
Figure 5 Comparison of microbial species evenness between ground and canopy soil. Values shown are
the total species evenness of all ground and all canopy samples, respectively.
University Press, Cambridge, UK, 1994.
5. Griffin, D. W., Kellogg, A. C., and Shinn,
E. A., Dust in the wind: Long range transport
of dust in the atmosphere and its implications
for global public and ecosystem health. Global
Change and Human Health 2.1, 20-33 (2001).
6. MacArthur, R. H. The Theory of Island
Biogeography. Princeton University Press,
Princeton, NJ, 1967.
7. Andrén, H. Effects of habitat fragmentation on
birds and mammals in landscapes with different
proportions of suitable habitat: a review. Oikos
71, 355-366 (1994).
8. Nübel, U., Garcia-Pichel, F., Kühl, M., and
Muyzer, G. Quantifying microbial diversity:
Morphotypes, 16S rRNA genes, and carotenoids
of oxygenic phototrophs in microbial mats.
Applied and Environmental Microbiology 65,
422-430 (1999).
9. Magurran, A. E. Measuring Biological
Diversity. Blackwell Publishing, Malden, MA,
2004.
10. Wolda, H. Similarity indices, sample size and
diversity. Oecologia 50, 296-302 (1981).
11. Itow, S. Species diversity of mainland-and
island forests in the Pacific area. Vegetatio 7, 193-
200 (1988).
12. Paster, E. and Ryu, W. S. The thermal impulse
response of Escherichia coli. PNAS 105, 5373-
5377 (2008).
13. Handelsman, J. Metagenomics: Application
of genomics to uncultured microorganisms. ASM
Microbiology and Molecular Biology Reviews
68, 669-685 (2004).
WRITTEN BY JOSH KENCHEL
Josh is a Biochemistry & Cell Biology and
Ecology, Behavior, & Evolution double major
from Revelle College. He will graduate in 2014.
Figure 7 Example of morphospecies identification. Here an example of the morphospecies mentioned
in the description of Figure 1, BSGRNS, is shown. The width of the field of view shown is approximately
400 m.
Figure 6 Comparison of mean indices of similarity of ground-canopy, ground-ground, and canopy-
canopy pairs. “Ground-canopy pairs” displays the mean of the indices of similarity plotted in Figure
5. “Between ground samples” and “Between canopy samples” display the means of the indices of
similarity plotted in Figure 7. Error bars show +/- 1 standard error. The three means are significantly
different (ANOVA, F= 17.04, df = 399, p < 0.0001). However, a matched pairs t-test demonstrated
no significant difference between the indices of similarity between ground pairs and between canopy
pairs.