Van wijk - HHreview - Modeling Workshop - Amsterdam_2012-04-23
A review of farm household models with a focus on food security, climate change adaptation, risk management and mitigationM.T. van Wijk, M.C. Rufino, D. Enahoro, D. Parsons, S. Silvestri, R.O. Valdivia, M. Herrero
Introduction• Insight in farm functioning is important both from an agricultural, social and from an environmental perspective• Modeling farm and household level decision making and its consequences is no sinecure• Farming systems across the world are highly complex and diverse and tools that address their behaviour are very diverse• No review available that focuses on climate change and adaptation as a specific model application area.
IntroductionThe specific goals of this review are:• To present a comprehensive overview of farm and household level models and to analyse trends in the use of modelling techniques in publications in peer reviewed scientific journals• To analyse how (combinations of) different approaches and techniques are used or can be used to study adaptation of farm systems to changes in the biophysical and socio- economic environment.• To identify models and modelling techniques that can be further developed to improve their representation of adaptation of farm/households in response to environmental change.
Approach• SCOPUS search engine• Central search terms: – ‘Model’ AND ‘Farm’ OR ‘agriculture’ OR ‘household’
Search termsTarget search terms1. ‘Livestock’ OR ‘poultry’ OR ‘cattle’ OR ‘pig’ OR ‘dairy’ OR ‘beef’ OR ‘sheep’ OR ‘goat’ OR ‘small ruminant’2. ‘Fisheries’ OR ‘aquaculture’3. ‘Crop’ OR ‘horticulture’ OR ‘tree’ OR ‘grass’4. ‘Soil’ OR ‘landscape’ OR ‘land use’5. ‘Water’ OR ‘hydrology’ OR ‘nutrient’6. ‘Ecosystem’
Search termsDomains of application terms1. ‘Adaptation’ OR ‘mitigation’2. ‘Smallholder’ OR ‘peasant’ OR ‘small-scale’ OR ‘commercial’3. ‘Productivity’ OR ‘yield’4. ‘Production’ OR ‘consumption’5. ‘Biodiversity’ OR ‘wildlife’ OR ‘conservation’6. ‘Emission’ OR ‘pollution’ OR ‘leaching’ OR ‘loading’ OR ‘runoff’ OR ‘erosion’7. ‘Profit’ OR ‘income’ OR ‘utility’
Results• Close to 16000 articles• Based on the abstract we selected 2,528 articles for further reading.• Of those articles, only 480 were selected for detailed evaluation because they included explicitly the farm or household level.• We selected 126 models (presented in 160 papers) that included farm or household level decision making with climate as input – 24 MP models – 36 MP combined with simulation models – 52 simulation models – 14 agent based models
Number of publications 60 Total number of publications Publications presenting new model 50 Models using more than 1 techniqueNumber of publications per year 40 30 20 10 0 1980 1985 1990 1995 2000 2005 2010 Year
Re-use of models 1 0,8New models relative to total 0,6 0,4 0,2 All models Mathematical programming Simulation models 0 1990 1995 2000 2005 2010 Year
Name of model Reference Components included Soil Crop Livestock HouseholdMP modelsFSRM (Dake et al. 2005) X Crop coefficients used which X Gross margin and variance in gross margin can be varied stochastically are evaluatedMUDAS (Kingwell et al. 1993) X Simple description X Production as technical X Simple description X Optimises income through tactical responses coefficients to seasonal weather.MP models combined with simulation modelsIMPACT-HROM (Zingore et al. 2009, Waithaka et X Soil model of X APSIM is used to estimate X RUMINANT is used to X Net income is maximised while also al. 2006) APSIM crop production estimate livestock indicators as food security and food self production sufficiency are calculated at household levelSAPWAT-LP (Grove and Oosthuizen 2010) X Soil water balance X Crop yield deter-mined by X Farm gross margin optimised evapora-tion reduction(dynamic) Simulation models (Shepherd and Soule 1998) X Simple soil model X Crop growth model per X Simple livestock X Cash and food balance at household level season production modelAPS-FARM (Rodriguez et al. 2011, Power et X Soil model of X APSIM crop growth model X Farm level production and economics are al. 2011) APSIM evaluatedTOA (Claessens et al. 2010, Stoorvogel X Soil models X Crop production model X Livestock production X Trade offs between socio-economic and et al. 2004) included included included environmental indicators assessedAgent based modelsPALM (Matthews and Pilbeam 2005) X Century model X DSSAT model included X Agent level evaluation of food production included and incomeMPMAS (Schreinemachers and Berger 2011) X Can be included in X Simple crop growth model X Livestock model can X Agent level evaluation of income, food framework included be included production and possible other indicators
Spatially Dynamic / Time-step Climate as input Feed-backs Inputs Decision making Regions of applicationName of model explicit Multi-period/ referenceMP modelsFSRM No No - Implicitly in No Prices, stochastic Trade off gross margin and New Zealand analysis production levels variance in gross marginMUDAS No No - Yes No Prices, 9 climate season Optimisation through LP Australia typesMP models combined with simulation modelsIMPACT-HROM No Yes, simulation 1 day, Yes, daily input Yes, soil Prices, climate, production Optimisation through LP Zimbabwe, Kenya models are optimisatio for APSIM orientation n longerSAPWAT-LP No Yes, SAPWAT 1 day Yes, daily Yes, soil and Prices, climate, risk Non-linear optimisation South Africa water aversion of farmer(dynamic) Simulation models(Shepherd and No Yes 1 year Yes, yearly Yes, livestock Climate, prices Rule based KenyaSoule 1998) values and soilAPS-FARM No Yes 1 day Yes, daily input Yes, soil Climate, prices, setup Rule based AustraliaTOA Yes Yes, at least the 1 day / Yes, daily input Yes, soil Soil, Climate, prices, Maximisation of net returns Andes, simulation multiple Management (econometric simulations) Kenya, Senegal, models years Netherlands, USA, PanamaAgent based modelsPALM No Yes 1 day Yes, daily input Agents and Climate, prices Rule based Nepal soilMPMAS Yes Yes 1 day / 1 Yes Agents and Climate, prices LP optimisation Chile, Germany, Ghana, year soil Thailand, Uganda, Vietnam
Name of model Economic performance Food self-sufficiency Food securityMP modelsFSRM X Gross margin and variance in gross margin optimised along trade off curveMUDAS X Income maximisationMP models combined with simulation modelsIMPACT-HROM X Net farm income is X Is explicitly analysed X Purchased food is taken into maximised account, stored food notSAPWAT-LP X Farm gross margin is optimised for different risk aversion values(dynamic) Simulation models(Shepherd and Soule 1998) X Farm profit is calculated X Is assessed X Food purchased included, not food storageAPS-FARM X Annual operating returns are X Could be assessed, not a focus of the calculated studyTOA X Income maximisation within X Can be used for this trade off settingAgent based modelsPALM X Farm income and food X Is assessed X Is assessed through food purchase, production are evaluated not through food storageMPMAS X Farm income and food X Can be quantified X Can be quantified production is evaluated
Climate variability and change Risk Mitigation AdaptationName of modelMP modelsFSRM X Is assessed through random X Price and climate related risks are assessed X Different prices and production coefficients will lead to variations in yield levels through random variations different optimal management and trade offsMUDAS X 9 season types are X Climate risk assessed, no assessment of X Tactical decisions are adapted in relation to different represented price risks seasonsMP models combined with simulation modelsIMPACT-HROM X Climate will affect crop X Risks related to prices and climate could be X Soil carbon and methane X Changes in prices and climate will lead to shifts in optimal production analysed, not in these studies however emissions from cattle managementSAPWAT-LP X Climate determines X Drought risk is assessed versus a risk X Changes in climate, prices and risk aversion lead to variability in crop yields aversion factor which can differ between different optimal management farmers(dynamic) Simulation models(Shepherd and X Climate affects production X Climate and market related risks could be X Soil carbon could be X Could be assessed through what-if scenarios for theSoule 1998) assessed on yearly basis assessed decision rulesAPS-FARM X Climate has effects on crop X Climate and price related risk can be X Soil carbon could be X Could be implemented through what-if scenarios for the production assessed assessed management rulesTOA X Climate has effects on X Price and climate related risks can be X Soil carbon can be X Trade offs and management options will change yield and other indicators assessed (production risk, environmental assessed, no part of study depending on climate and prices and farm configuration. risk)Agent based modelsPALM X Affects crop production X Climate and market related risks can be X Soil carbon could be X Agent behaviour can change depending on conditions; assessed assessed could also be assessed through what-if scenarios for the decision rulesMPMAS X Climate has effects on crop X Climate and market related risks can be X Different GHG indicators X In agent behaviour optimisation changes in prices and production assessed can calculated climate will lead to other optimal behaviour
Main findings• Although many models use climate as an input, few were used to study climate change adaptation or mitigation at farm level.• Few studies performed detailed risk analyses;• The limited number of studies focusing on risk defined it either as the failure of supplying enough food for the family or as the lack of cash resulting in bankruptcy.• There is a wide range of modelling techniques available to address specific questions
Main findings• A range of integrated crop – livestock models available that also perform economic analyses• Recent developments show that new modelling frameworks combine the strengths of different modelling techniques• Promising mixtures of methodologies include – mathematical programming for farm level decision making – dynamic simulation for the production components and – agent based modelling for the spread of information and technologies between farmers.
Main findings• The terms adaptation and vulnerability are well defined in literature• But still need specific and wide-spread implementation in farm systems level research and• Definition at a scale that is relevant to the farm.
Main findings• The appropriate incorporation of model and input uncertainty is important for climate related applications and has only been done in few studies.
Main findings• Modeling decision making remains a difficult issue – Optimisation has its limitations – ‘biodecision models’: decision making with ‘if … then…’ model constructions: seems powerful for limited number of factors/rules
OverallThere are enough techniques for integratedassessments of farm systems in relation toclimate change, adaptation and mitigation, butthey are scattered:they have not yet been combined in a way thatis meaningful to decision makers at farmhousehold level.