from global economic modelling to household level analyses of...

From global economic modelling to household level analyses of foodsecurity and sustainability: How big is the gap and can we bridge it?

Mark T. van WijkInternational Livestock Research Institute (ILRI), Livestock Systems and the Environment, Nairobi, Kenya

a r t i c l e i n f o

Article history:Available online xxxx

Keywords:Multi-scaleIntegrated assessmentModelsMacro-economyFood securitySustainability

a b s t r a c t

Policy and decision makers have to make difficult choices to improve the food security of local peopleagainst the background of drastic global and local changes. Ex-ante impact assessment using integratedmodels can help them with these decisions. This review analyses the state of affairs of the multi-scalemodelling of policy interventions, with an emphasis on applications in developing countries and livestocksystems. Existing models do not sufficiently capture the complexity of human–environment interactionsacross different scales, and especially the link between landscape and local market levels, and nationaland sub-national level policies and markets is missing. The paper suggests a step wise approach withincreasing data needs to bridge this gap. Improvements need to be made at the description of effectsof the distribution of local markets on price formation and the representation of farm diversity withina landscape. Analyses in contrasting agro-ecological systems are needed to derive generic summary func-tions that can be used as input for macro level model analyses. This is especially pertinent for macro leveldescriptions of crop and livestock production in relation to price developments and of the mosaic of dif-ferent agricultural land use responses in regions with contrasting socio-economic conditions anddevelopments.

! 2014 Elsevier Ltd. All rights reserved.

Introduction

Achieving sustainable food security (i.e. achieving the basicright of people to produce and/or purchase the food they need,while not harming the social and biophysical environment) – in aworld of a growing human population and large scale changes ineconomic development is a major challenge. The way land is usedplays a significant role in the changing global food economy, anddetermines food availability at both macro and micro levels.Livestock plays an important role in achieving food security andis the largest land use sector on earth (Herrero and Thornton,2013). Livestock and its related land use (e.g. grassland, forageproduction) bring both benefits and problems, and is a typicalexample of how agricultural land-use is determined by a multitudeof environmental, economic and socio-cultural conditions(Lotze-Campen, 2008).

The policy environment has profound impacts on the opportu-nities and constraints that affect agricultural land users. Policymakers aiming to improve food security and rural livelihoods inthe developing world in a sustainable manner face many uncer-tainties when exploring the future of food systems (Ericksen

et al., 2009). Policy can play an important role to balance the multi-ple functions of agriculture and support sustainable developmentbut needs adequate information on how different policy optionsaffect the complex issues surrounding food security and sustain-able development. Integrated Assessment Modelling (IAM; IAmodels are defined as models that combine knowledge from multi-ple disciplines, with the aim of shedding light on policy questions(Tol, 2006)) is increasingly seen as a way to explore differentfutures of land use and to support policy-making (Harris, 2002;Rotmans et al., 1990; Rotmans and van Asselt, 1996; Ewert et al.,2009). This paper reviews the application of the multi-scale mod-elling of policy interventions on livestock systems in developingcountries. In the modelling, we emphasise the importance of takinginto account the tribal relationship across food production, povertyreduction, and environmental sustainability.

Rapid development are taking place in model development(Parker et al., 2002), including applications to agriculture (VanCauwenbergh et al., 2007; Van Ittersum et al., 2008). However,most models address specific issues, e.g. assessment of land usechange (Verburg et al., 2008), nitrogen emissions and leaching(Velthof et al., 2007) or food production (Fischer et al., 2005). Manyof these models are also developed to address questions at a spe-cific scale or level, which can vary from a farming system to theglobe (e.g. Ewert et al., 2009).

http://dx.doi.org/10.1016/j.foodpol.2014.10.0030306-9192/! 2014 Elsevier Ltd. All rights reserved.

E-mail address: [email protected]

Food Policy xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

Food Policy

journal homepage: www.elsevier .com/locate / foodpol

Please cite this article in press as: van Wijk, M.T. From global economic modelling to household level analyses of food security and sustainability: How bigis the gap and can we bridge it? Food Policy (2014), http://dx.doi.org/10.1016/j.foodpol.2014.10.003

http://dx.doi.org/10.1016/j.foodpol.2014.10.003

mailto:[email protected]


http://www.sciencedirect.com/science/journal/03069192

http://www.elsevier.com/locate/foodpol


However, for really integrated assessment of land use and foodproduction multi-scale assessments are essential (Rounsevell et al.,2012). The viability of global food production, the maintenance ofecosystems services, and the reduction of poverty, involve increas-ingly complex interactions between land users and their socio-eco-nomic and biophysical environment. To decide where investmentin and new policies for food producing systems are most efficient,ex-ante integrated assessment of the consequences of these invest-ments at livelihood, community, landscape, and regional levelneeds to take place.

In recent years several reviews have been published on thestate-of-the art in IAM and multi-scale assessment (i.e. assess-ments that run across different spatial and temporal extents(scales) and organisational levels (Ewert et al., 2011)). Examplesare IAM in relation to Life Cycle Analyses (LCA) (Creutzig et al.,2012), land use and landscapes (Verburg et al., 2011, 2013a), landuse modelling (Rounsevell et al., 2012; Berger and Troost, 2014)and one review that at least in the title claims to be a review ofIAM and issues surrounding food security (Verburg et al., 2013b).Despite the latter’s title, the focus on food security is still verymuch at landscape or regional scale, ignoring responses of cropand livestock based livelihoods. Despite all these reviews it is stillnot clear how multi-scale assessment can help to improve thedescription of geographical variations in the drivers of agriculturalland use in macro-economic models. Capturing these variations isessential, because they result in a mosaic of different agriculturalland use responses: continued expansion in regions with low pop-ulation densities strongly affecting the functioning of (agro-) pas-toralist communities, intensification in regions with goodconnections to urban markets, and decrease in agricultural landuse near the rapidly expanding major cities. Furthermore, thereviews typically miss a description of how the implications of landuse policies can be quantified at the level where many local landuse decision are made, and where the consequences of a lack offood security are felt, i.e. the farm and household level. This reviewwants to fill this gap, and analyses how macro and micro levelmodels are set up and analyse (livestock related) land use. Besidesgiving an overview of the state of affairs of multi-scale modellingof policy interventions, this paper also identifies key gaps in thecurrent approaches to work truly across multiple integration lev-els, and suggests ways forward to deal with these gaps.

Current state of affair of multi-scale IAM

In the scientific literature modelling efforts have focused onspecific aspects of policy, market and land systems (e.g., Dalgaardet al., 2003; Volk and Ewert, 2011) and there is a clear methodolog-ical divide in describing food production and land use: currentapproaches either use top-down global and continental approaches(e.g. macro-economics and large scale land use modelling (e.g.Zhang et al., 2013; Creutzig et al., 2012) or bottom-up approaches,from farm level upwards (farm household modelling, micro-eco-nomics, agent based models and landscape level land use model-ling) (e.g. Rufino et al., 2011; Parker et al., 2003; Valdivia et al.,2012). This section describes the current state of affairs of bothapproaches, the gap between the approaches, and existingapproaches to bridge the gap.

Top down macro-economic modelling of agricultural land use

Elaborate reviews of macro-economic models can be found inrecent work by Dumollard et al. (2013) and Zhang et al. (2013),so here I will limit myself to a short description of contrastingapproaches. All global and continental models describing landuse and food production have a similar setup (for a schematic

overview see Fig. 1). The globe is divided in a number of regions(currently in most applications this values ranges been 6 and 32regions) and a macro-economic model solves the price equilibriumbased on calculations quantifying regional supply and demand. Theprice formation calculated and the regional estimates for supplyand demand are input to a finer scale land use model thatcalculates how this supply can be generated (Fig. 1). Existingmacro-economic models can be grouped according to severalcharacteristics. At the level of the macro-economic description,two types of models can be distinguished: Computable GeneralEquilibrium (CGE) models, which perform economy-wide analyses(multi-sector analyses) and Partial Equilibrium (PE) models, whichdescribe specific segments of the economy (i.e. only one or a fewsectors). Examples of CGE models are ENVISAGE, FARM, GTAP,GTEM and IMAGE (see Dumollard et al., 2013 for an overview).GLOBIOM (Havlik et al., 2011), IMPACT (Rosegrant et al., 2012;and also Msangi et al., 2014)) and MAgPIE (Lotze-Campen et al.,2008, 2010) are examples of PE models.

Another way to group models is based on the way they describeland use allocation. Three ways can be distinguished. The firstapproach is used by the rule based models (i.e. CGE models andthe IMPACT model) and represents land use at an aggregated spa-tial level (different macro-regions, countries or groups of coun-tries). In CGE models, land is considered as a production factorand the different economic sectors that require land (crop sectors,livestock, forestry) compete on a regional land market. The IMPACTmodel follows a different approach: harvested area for each sectorand each region is expressed as a function of output prices of thesector’s commodity itself but also of competing commodities,which enable the representation of substitution effects. These thenform input to the set of equations IMPACT solves (Rosegrant et al.,2012).

The second approach is used by optimisation models as GLOB-IOM and MAgPIE, in which land use is represented on a spatiallydisaggregated level (Havlik et al., 2011; Lotze-Campen et al.,2008). GLOBIOM and MAgPIE use exogenous agricultural yieldsand production prices that are location specific. As a consequenceland allocation is the only endogenous variable left to reach opti-mality. As yields and production costs are grid-cell specific, thesemodels are able to establish a link between land productivity andland allocation at the local level, although the drivers of the alloca-tion process are the regional food markets (Dumollard et al., 2013).

The third approach is the so-called ‘geographical’ or ‘geostatis-tical’ approach (e.g. Rounsevell et al., 2012). In the past thisapproach was especially applied to regional land use analysesand studies of rural–urban land use connections, but it is nowincreasingly applied in continental or even global land use studies.In models like CLUE (Verburg et al., 1999) and LandSHIFT(Schaldach et al., 2011) spatially explicit land-use patterns are cal-culated using data on land suitability and assumptions on agricul-tural demand. Future land-use change in these models isdetermined by statistical relationships of past trends in land use.The macro-economic models simulate outputs (prices, supplyand demand) that serve as input for the spatially explicit landuse models (e.g. IMAGE, LandSHIFT) and the overall model frame-work can thereby simulate scenarios of developments in land use.

The lowest spatial scale at which the large scale economic landuse models make predictions differs strongly between the models.Many global studies now aim at a 5 arcminute (!10 " 10 km) spa-tial resolution (Verburg et al., 2013b). Models like GLOBIOM, CLUEand LandSHIFT now in general have a resolution of 0.5", whereasCGE models, because of the more comprehensive economic modelthat drives the simulations, land use shifts are calculated at a muchlarger integration level, normally at regional or country level(Zhang et al., 2013). The increasing resolution of the land use mod-els causes problems in the description of land use systems that are

2 M.T. van Wijk / Food Policy xxx (2014) xxx–xxx



highly dynamic in time and space, and in the description of landuse systems that change over larger areas than the smallest resolu-tion of the models. For example, most models use fixed land allo-cation patterns in which it is difficult to capture the dynamics ofpastoralist systems, in which the herders move over large areas.To be able to deal with these systems, generic livestock and fodderproduction values are normally used on a per land area basis (seefor example Msangi et al., 2014, for a detailed discussion), but thisapproach has limited capacity to capture changes, for examplecaused by the increasingly rapid conversion of grassland based sys-tems to mixed crop – livestock systems.

Bottom up approaches to analyse agricultural land use

Bottom-up approaches that take into local information on landuse and management to describe outcomes at regional or land-scape level, can be grouped in roughly three groups: farm house-hold modelling, multi-agent approaches and micro-economicapproaches. In farm household modelling the goal is to model deci-sion making at farm household level. Model outputs are then usedto say something about responses of a community of farmers (e.g.see a recent review of Van Wijk et al., 2014). A standard approachis to classify farm households into farm household types with com-mon characteristics that occur across larger groups of farm house-holds, and apply the model for these farm types to assess theconsequences of certain intervention measures. The responses ofthe individual farm households can thus be scaled up to the totalpopulation of farmers if the total number of farm households isknown together with the relative occurrence of the different farmtypes. This is the simplest way to scale up responses of individualhousehold to a larger community or to landscape level, and itignores interactions between farm households and limitingresources at levels higher than the farm household (for exampleof the availability of communal grassland, e.g. Rufino et al.(2011)). Another approach is to characterise the whole farm popu-lation in a community or landscape, or at least a very large part ofit, and then perform analyses at household level and scale these upto the whole community in a landscape (e.g. Stoorvogel et al.,2004). Also this approach ignores interactions between farmers.

Multi-agent models represent families, farmers or householdmembers as individual entities (agents) explicitly taking intoaccount interactions between these entities. This modelling tech-nique often makes use of the same approaches as dynamic simula-tion models, but where those models typically focus on onehousehold or an average representation of a population of house-

holds, do agent-based models represent multiple instances of indi-vidual households. Agent-based modelling (ABM) has become animportant bottom-up tool that has been used extensively to studycomplexity in theoretical studies, but now is also applied in empir-ical studies (e.g. An, 2012; Berger and Troost, 2014). Several ABMmodels have been developed to study farmers’ behaviour in rela-tion to land use and food production (e.g. Schreinemachers andBerger, 2011), and most of these are spatially explicit by linkingagent to a location on a spatial grid (e.g., An et al., 2005; Bergerand Troost, 2014). ABMs have been coupled to environmentalmodels, for example of soil carbon and water (Schreinemachersand Berger, 2011). In these models the individual agents arelocated at specific positions in the landscape, and their land usedecisions affect the environmental variables for those locations.These changes in the environmental variables are then used asinput for biophysical models that calculate patterns and changesat landscape level. ABMs are capable of incorporating multi-scaleand multi-disciplinary knowledge and agent-based modelling isbelieved to have the potential to facilitate methodologicallydefensible comparisons across case study sites (An, 2012). Forexample, ABM was used to synthesise studies of frontier land usechange in contrasting systems across the globe (Rindfuss et al.,2008). However, further empirical support, especially data aboutthe behaviour of coupled human and environmental systems, isstill considered essential for gaining trust in the simulated resultsof ABMs (An, 2012; Parker et al., 2003; Veldkamp and Verburg,2004).

The calculation of food production and the representation offarm level decision making is normally quite simplified in ABMs,but recently also more complex farm production oriented modelshave used to study land use at farm community level (e.g. Rufinoet al., 2011). The integration of farm production models withinan ABM framework to simulate processes at landscape level seemsa promising approach. Farm household models can give insight infarm level decision making, especially on possibilities for land useintensification. ABMs can calculate what the consequences of thesefarm level decisions are on local and regional land use, for exampleon farm expansion in previously non-agricultural land (e.g., Bergerand Troost, 2014). Pastoralist systems are typically captured atintegrated community and herd level, for example using ABMs(e.g. MacOpiyo et al., 2006) without using the intermediate stepof independent household level analyses. This because manage-ment typically takes place at herd level, and the interactionsbetween cattle, owners and biophysical environment are so strongthat descriptions at individual household level are needed.

Fig. 1. Generic structure of large scale economic impact assessment tools. In red circles the entry points where incorporating information from small scale bottom upapproaches can improve model reliability. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

M.T. van Wijk / Food Policy xxx (2014) xxx–xxx 3



There are surprisingly few micro-economic studies wheredetailed descriptions of farm level production and decision makingare used to calculate consequences for supply and demand in localor regional markets, and local market price formation is used asinput in farm household level analyses. Up to date only two ofthese multi-scale agricultural models have been developed(Valdivia et al., 2012; Laborte et al., 2007), and both studies showthat these local market feedbacks are affecting individual farmlevel decision making. In a recent study these micro-economicprinciples have also been applied to ABMs (Straatman et al.,2013), opening the possibility to link farm production modelsand ABMs within both an environmental and micro-economicframework to perform local to regional integrated assessmentstudies. However, such applications can only be performed suc-cessfully if model complexity and data needs of such a coupledmodel can be controlled.

From this overview it is clear household level studies analysethe impacts of global change on the functioning of households,and rarely go beyond the local community or landscape level. Atthe moment these results are not used to inform larger scale stud-ies. It is the thesis of this special issue of Food Policy that this kindof micro-level analyses, and the data and approaches that underlieit (see for example Bahta and Malope, 2014), can enrich andimprove macro level models.

The gap

At the moment there is a clear gap between the top down,macro-economic land use models, and the bottom up models usingfarm scale modelling, multi-agent modelling and/or micro-eco-nomic modelling. The large scale models have their own internalmodules to calculate land use and productivity changes at pixelscale, driven by the outputs generated by the large economic mod-els and maps of land suitability and crop and livestock productivity(e.g., JRC, 2011). The price and demand scenarios generated by thelarge scale models are used to drive the small scale models(Herrero et al., 2014; Van Wijk et al., 2014; Claessens et al.,2012), but outputs generated at that lower level do not feed backinto the large scale models: The two approaches are decoupled.

The choice for this decoupling has several logical reasons:building in land user decision making in the large scale modelswould run into problems with model complexity and there is a lackof good data at land user level across large areas of the globe. How-ever, while acknowledging the underlying reasons and needs forusing such simplifications in the representation of land use andlandscapes at the global scale, the implications of such a represen-tation are seldom documented (Verburg et al., 2011). Initially mostglobal scale integrated assessment models were used to study veg-etation dynamics, carbon balances, crop growth and greenhousegas emissions in order to capture important trends in climateand land use, and their feedbacks. However, they are now beingused for an increasing number of indicators, from global food secu-rity to ecosystem service assessments. It could be that for theseanalyses, which are strongly dependent on land use decision mak-ing and the spatial structure in land use and the landscape, the useof such simplified representations of space and land users may bequestionable (e.g., Verburg et al., 2013a; JRC, 2011).

In the past the strong divide between large scale models andbottom approaches was accompanied by a clear spatial scale divideas well (Fig. 2A) (see for example a review by Dowlatabadi (1995)to illustrate the state of the art of IAM in the 1990s). However, thisis no longer true. There, is from a spatial scale point of view, a clearoverlap between the size of the grids now simulated by state-of-the-art IAMs like GLOBIOM (Havlik et al., 2011) and the areascovered by simulations by bottom up approaches (Valdivia et al.,2012; Schreinemachers and Berger, 2011; An et al., 2005). This

means that the spatial scale at which the different models can per-form analyses does not have to be the cause of the strict dividebetween the approaches. The divide is caused by the differentinternal model structures, which at the moment prevents easycommunication between the approaches.

The gap between top-down and bottom up approaches meansthat dynamics and changes occurring at other integration levelsare not included. For example, an expert consultation on large scaleeconomic modelling (JRC, 2011) noted the wide differencesbetween large scale economic land use models regarding assump-tions about yield increases, marginal yield, by-products and pas-ture land effects. Economic models tend to underestimate thelong-term interconnectedness of world production and substitu-tion between crops and forages, because they are calibrated onshort-term annual data (Zhang et al., 2013). The future situationfor developing countries is likely to change given the economicdevelopment surrounding the urban centres, which has strongconsequences for the rural areas that have market connections tothese urban centres. A key weakness in the large scale land usemodels is the allocation of changes in agricultural land use, be itcrop production, fodder production or livestock grazing. Land suit-ability is certainly one driver, and is generally accounted for inthese models, but other socio – economic criteria and political con-straints on land use and land ownership are mostly ignored (JRC,2011). To improve this situation lower scale information is neededto improve the large scale representation of key relationshipsdetermining land use change (JRC, 2011; Rounsevell et al., 2012;Creutzig et al., 2012).

On the other side, bottom up based models are often used inspecific applications with limited geographical extent, thereby lim-iting their usefulness at the larger scales of analyses at which landmanagement and policy plans are developed (Rounsevell et al.,2012). Interactions between the top-down and bottom-up model-ling approaches would lead to a more comprehensive representa-tion of the land system (Rounsevell et al., 2012) and a morerealistic assessment of the consequences of policy options. Toachieve these interactions demands developments at both sidesof the divide. An example in which multi-scale assessment isurgently needed is the debate around land-sharing and land spar-ing, and the potential role of production intensification. Diverselivestock related examples of such ongoing intensification areincreased nutrient use in crop and forage production, improvedintegration of crop – livestock systems, marketing of forage andcrop residues and pastoralist systems turning into agro-pastoralistsystems. Despite considerable research efforts to improve ourunderstanding of agricultural intensification (Bakker et al., 2007;Foley et al., 2005; Lambin et al., 2000; Rudel et al., 2009), it is stillunclear whether intensification can balance the triangle of foodsecurity, poverty and environmental sustainability indicators.Some studies stress the negative environmental impacts of intensi-fication (e.g., IAASTD, 2009), while others highlight the positiveenvironmental effects of potential land-saving through intensivecultivation technologies (e.g. Burney et al., 2010; Phalan et al.,2011). Individual model systems or empirical approaches workingone or a two integration levels can contribute limited informationto the debate surrounding intensification, and multiple scale anal-yses are needed to clarify issues and drivers.

Bridging the gap: current status

At the moment there is only one model framework in whichagricultural household level decision making is integrated intolarge-scale, economic based integrated assessment, and this isthe SEAMLESS framework (Van Ittersum et al., 2008). The SEAM-LESS framework is described in detail in a series of papers (VanIttersum et al., 2008; Ewert et al., 2009), and here I just want to




highlight the most important characteristics of this multi-scaleframework. First of all, four integration levels are explicitly takeninto account: field, farm, region and continental level. For threeof these levels specific models have been used (Ewert et al.,2009), while the regional level is only important for sampling pur-poses, thereby accounting for variations in environmental andsocio-economic conditions across the European continent. Fieldand farm models calculate the supply of agricultural products,and these are used by the continental market model to calculateequilibrium prices. These continental prices are then used as mar-ket prices in the farm models. An important decision made in thedevelopment of the SEAMLESS framework is to not represent localmarkets, market access, road access, local social institutions orlandscapes, thereby making it a framework mainly focusing onproductivity and economics. It also ignores the spatial componentof land use change, which is why it is capable to go across spatialscales and integration levels relatively easily.

No other frameworks exist that represent multiple levels as sys-tematically as the SEAMLESS framework does. Many studies exist

in which scenarios predicted by large scale economic models aredirectly implemented on farm level (e.g., Herrero et al., 2014;Van Wijk et al., 2014; Claessens et al., 2012). In recent projects indeveloping countries regional scenarios of price development aredirectly used to analyse what is happening to ‘typical’ farms usingfarming systems work like Dixon et al. (2001) and Seré andSteinfeld (1996) (e.g. AnimalChange, ‘www.animalchange.eu’).Bottom up approaches that represent the decision making of theland user are not used in the large scale studies, although recentreviews stress that is the way forward for large scale land use mod-elling (e.g., Rounsevell et al., 2012; JRC, 2011). Results fromdetailed simulation studies are only used in large scale studies togenerate maps of crop and livestock productivity (e.g., Havliket al., 2011), which form the basis for the calculations of food sup-ply. Geographical modelling approaches take local level interac-tions implicitly into account through empirical analyses of pastland use change patterns, but adaptive farmer behaviour, for exam-ple based on simulation results at farm household level is nottaken into account.

Fig. 2. Previous divide between top down and bottom approaches in relation to spatial scales of analysis (A), and current situation, where there is overlap between theapproaches in terms of spatial extent, while there is still a strict functional divide between the approaches (B).



http://www.animalchange.eu


Bridging the gap: issues, and a way forward

Previous reviews have dealt with several of the issues in multi-scale analyses (e.g., Rounsevell et al., 2012; Verburg et al., 2011,2013a; Van Asselen and Verburg, 2012; Creutzig et al., 2012): forexample the limited data availability of environmental and socio-economic variables, the need for remote sensing data and forimproved representation of ‘soft science’ in the economic models(e.g. effects of social rules, legislation and institutions on landuse change), and the integration of information between differentmodelling approaches (Rounsevell et al., 2012; Creutzig et al.,2012). Here I want to highlight one issue which has not been dealtwith in detail in the previous reviews, and which is a critical bot-tleneck for any scaling exercise that analyses the effects of regionalor national policies on farm households and other land users. Thisis how we should deal with the diversity in land users, markets andinstitutions and how we can represent these key players in landuse in space.

Land users

Agro-ecology, markets and local cultures determine land usepatterns and agricultural management. Farm households differ inresource endowment, production orientation and objectives, eth-nicity, education, past experience and management skills and intheir attitude towards risk. Recognising variability in farm house-holds is key to designing policies to help poor farmers (e.g. Giller,2013), and understanding the main drivers of household diversityand their relationship with livelihood strategies can help to bettertarget agricultural innovation (Klapwijk et al., 2014).

So how to deal with farm household diversity within the settingof integrated assessment? As shown in the SEAMLESS example, oneway is to develop farm typologies, calculate the responses of thesefarm types, and scale these up to the total population of farmers(Ewert et al., 2011). Patterns of farms, farming livelihoods and pro-duction objectives have been used to construct farm typologies in awide range of systems (e.g., Sakané et al., 2013) and these typolo-gies are seen as particularly useful when combined with analysesof different rural household strategies (such as Dorward’s ‘hangingin’, ‘stepping up’ and ‘stepping out’ (Dorward et al., 2008) and dif-ferent ‘rural worlds’ in terms of engagement of rural households inmarkets (Vorley, 2002), see for example Giller (2013)). Also inmany multi-agent models farm typologies are used to deal withthe overall diversity in agricultural decision makers.

However, the use of farm typologies within integrated assess-ment has disadvantages. First, and most fundamental of all, is thefact that the a priori decision on the indicators used to developthe typology to a large extent determines which interventions ortechniques will be adopted according to the modelling analyses.In most cases an indicator representing wealth (e.g. number of cat-tle, size of land, etc.) is the most important factor driving the typol-ogy, and not the way farmers manage their resources. Thisautomatically means that options that improve management effi-ciency of existing management options will be assessed as lessimportant compared to other overall productivity increasingoptions like mineral fertilizer application.

Other approaches make use of the so-called ‘distribution’approach (Fig. 3), in which a population of farmers is treated asthe unit of analysis (Frank et al., 2014). This is a potentially power-ful approach because the analyses of interventions can be based onmore complete information than when farmers are first aggregatedinto typologies (Fig. 3). However, it is not an easy approach toimplement across a landscape or a certain area in land use models.It is difficult to allocate a distribution to a location within a land-scape, whereas allocation of farm types across a landscape is more

straightforward. Any spatial mapping of farming systems or farmhouseholds is a compromise between the attractiveness of show-ing farms or farm types in a spatial grid and the danger of implyingsharp boundaries between neighbouring systems or types. With alarge degree of variation inevitable among types of farm house-holds, there are seldom sharp boundaries between systems(Dixon et al., 2001).

The distribution approach can be useful at integration levels lar-ger than the landscape. For example, in the N2Africa project(www.n2africa.org) farm characterisation surveys were performedalong two axes: agro-ecological potential and market access(Franke et al., 2011). For both axes a ‘high’ and a ‘low’ class weredefined, and farm surveys were executed in four areas that coveredthe possible combinations of the two variables. Such a two-axesclassification can then subsequently be used as the entry pointfor upscaling of farm household level behaviour and responses.Such an approach assumes that production potential and marketaccess are the key drivers of the functioning of farm households,and ignores for example landscape position as a determinant. Toapply these methods across larger regions requires large datasetsof social surveys that characterise land use systems and the deci-sion making in these systems (Rounsevell and Arneth, 2011). Thedevelopment of databases of social surveys from around the worldwould provide an immensely rich resource to be able to derivedescriptions linking land use to the socio-economic and biophysi-cal environment of the decision-maker (Herrero et al., 2014). At themoment no such databases exist and similar efforts like recentlyexecuted successfully in ecology are needed (e.g. Rounsevell andArneth, 2011). Similar work on empirical social data could possiblyallow construction of a generalizable framework of multi-agentmodels which could work over larger areas and thereby informlarge scale models (Rounsevell and Arneth, 2011). A database offarm characterisations could also better inform farm householdmodels, which again could help the search for generalizable rulesof opportunities for successful policy intervention (Van Wijket al., 2014).

Land use data

In recent years the spatial resolution of the physical and socio-economic data that are used as drivers of land change and of dataneeded to assess impacts of land change on environmental indica-tors has increased substantially (Verburg et al., 2013a). See forexample a new global dataset presented by Herrero et al. (2013)

Fig. 3. An example of a farm wealth distribution, and how different technicalinterventions (1–4) and different management interventions (a–b) could be best fitinterventions along this distribution. The farms in this distribution will not onlydiffer in resource endowment, but on a separate axis also in the type ofmanagement they practice. Interventions can therefore be evaluated morethoroughly than with a typology approach, which often uses a wealth-basedcategorisation.



http://www.n2africa.org


in which detailed spatially explicit data are made available on live-stock production, fodder, feed efficiencies and greenhouse gasemissions. Recent advances in the development of such datasetsmay increase the spatial resolution of model predictions even fur-ther (Verburg et al., 2013a). However, it is important to note thatonly increasing the resolution of the land cover and land use datadoes not automatically lead to more accuracy in model predictions.Increasing the resolution of land cover data does not necessarilyallow us to represent those characteristics of landscapes essentialfor its functioning which only become apparent at relatively highspatial resolutions (Verburg et al., 2013a). However, this is noteven the most fundamental problems with spatial explicit analysesof land use, because this problem might be solved in the futurewith even more accurate maps.

The most fundamental issue in spatially explicit analyses ofland use and land and livestock productivity is the basic mismatchbetween pixel and land user: a pixel of a satellite image or a rasterof a GIS raster map is neither a landscape nor a social unit, nor afarm household (e.g., Rindfuss et al., 2004). This puts a fundamen-tal limit on further increasing the resolution of the spatial inputdata needed for running land use models, because at a certain res-olution fundamental mismatches will affect model simulations.This problem is even more pressing for pastoralist systems, inwhich expressing productivity on a per area basis for a specificlocation in space automatically is problematic when the area forwhich the productivity is smaller than the region in which the pas-toralists are moving. See also for example Little et al. (2014) onhow pastoralists of the Horn of Africa negotiate the need for herdmobility (production) under conditions of variable rainfall andgrazing conditions, with the necessity to market animals at fixedmarket locations. Fixed markets can easily be represented in pixelbased framework, but that is less easy for mobile pastoralists.Pastoralists are therefore only represented through fixed produc-tion responses that do not explicitly take into account effects ofmobility, and thereby might lead to over-estimation of the respon-siveness of pastoralists (Little et al., 2014).

Farmers, markets and institutions

Answering the question ‘how many are there of which farms,and where are they located’ is pertinent in relation to describingmarket access and landscape level processes. As noted before, pas-toralist communities have their own peculiarities in this respect.To obtain an answer to that question remote sensing, farm charac-terisation questionnaires, and market access and agro-ecologicalinformation need to come together. Two recent examples tryingto link farmers and local markets are Valdivia et al. (2012) andLaborte et al. (2007), and both studies show that local market feed-backs have strong effects on individual farm level decision making.In Valdivia et al. (2012) spatial explicit information on farm loca-tion is used to generate response maps, but how exactly thesemaps were generated and how they are linked to the farm popula-tions analysed within their household model is unclear.

Furthermore, there is little general insight into the role of insti-tutions, land tenure and land governance on land use intensity asmost research is based on case studies that are strongly contextdependent (Verburg et al., 2013b). The fact that a very large partof the land used for food production is privately owned or has beenallocated long term land rights to private persons or companiesmakes the top-down implementation of improved land systemmanagement that better match food production demand and eco-system protection impossible (Verburg et al., 2013b), and againcreates problems for pixel based approaches. IAMs need furtherdevelopment on how ‘soft science’ knowledge can be built in,related to fundamental questions of governance, economic sys-tems, and the assumptions, beliefs, and values underlying human

behaviour. This must involve close integration of social and bio-physical sciences (Reid et al., 2010). Neither bottom up nor topdown approaches are very powerful in this integration at themoment, and there is urgent need to improve model descriptions.

A step wise approach to bridge the gap

Here 6 different approaches are described (see also Fig. 4),increasing in complexity and data needs, to scale up bottom upapproaches so that they can generate information that can usedin the model formulations of large scale economic approaches(see also Fadiga and Katjiuongua, 2014, for how a similar approachcould be used for modelling animal diseases). I assume the bottomapproaches receive region specific information under specific sce-nario conditions from the higher scale models: for example relativechanges compared to a baseline value of prices of products, inputsand labour. This information can be combined with participatoryscenario work to derive development pathways that could encom-pass the regional context and cultural history of the location ofinterest (e.g. Chaudhury et al., 2013), and serve as key drivers ofthe micro-scale models presented in Fig. 4. The micro-scale modelscan then generate information that can improve the descriptions inthe large scale models, focusing there on (1) the descriptions ofcrop and livestock production in relation to price developmentsand in relation to the rapidly developing urban centres; thesedescriptions in this way will take into account the decision makingof land users (e.g., where and when will the switch from farm landextension to farm land intensification take place, and why), and (2)the rules describing the spatial allocation of land use (Fig. 1). Thesetup described here is distinct from the one proposed byRounsevell et al. (2012) which only focuses on multi-agent modelsas the way to go as the bottom up approach (see also Berger andTroost, 2014). Given the fact that application of ABMs across largeregions is still unproven (as stated by Rounsevell et al. (2012)themselves) and that in general there is still a strong lack of datato support such generic development of ABMs across large areas,I aim in this section to describe different approaches along withtheir data needs and the type of information they can generate.

The most data-efficient approach to link farm scale analyses tothe large scale economic outcomes is to use farming descriptionsavailable for the region of interest (for example Dixon et al.(2001) or Seré and Steinfeld (1996)), represent these farming sys-tems by one or more ‘typical’ farm(s), and calculate consequencesof certain scenarios for the food security at farm household level(e.g., Van Wijk et al., 2014). These outcomes can then be taken asrepresentative for the whole region for which the farming systemdescription is supposed to be representative (Fig. 4A). Given thedata available for farm typologies this is an approach that can bereadily operationalized in many regions in the world (Ewertet al., 2011; Herrero et al., 2014). This approach can give improvedinformation on the supply side of the large scale economic models,and but would not necessarily make the large models better intheir predictions of land use change. The approach ignores farmheterogeneity, and can therefore lead to over-prediction of theresponsiveness of farmer population, as only few farms arerepresented.

A second approach could be to disaggregate farming systemsalong the axes market access and agro-ecological potential, twokey drivers of productivity (Fig. 4B). Hereby large farming systemsclusters are broken up and thereby diversity in farms is repre-sented. Farm household characterisation across the whole popula-tion of farmers is necessary (e.g. Franke et al., 2011; Rufino et al.,2013), and the approach therefore demands more information thanthe first approach. The advantage of the approach is that it betterrepresents the differential responses different farmers might have,




but it still ignores interactions between farmers and interactionsbetween farm consumption and production and local market levelsupply and demand.

A third approach would be to specify local markets (identifiedthrough surveys, or by simple rules, e.g., population centres withmore than ‘x’ people), and market access as key drivers of farmlivelihood responsiveness (Fig. 4C). Such a model can be set up

non-spatially explicit (e.g. Laborte et al., 2007) or spatially explicit(e.g. Valdivia et al., 2012). This modelling approach takes feedbacksbetween farmer level decision making and price formation intoaccount. In Valdivia et al. (2012) farmers/farm types are given alocation, and the micro-economic analysis could therefore also beembedded in a landscape level analysis where environmental indi-cators are assessed as well. The approach is already highly data

Fig. 4. Schematic representation of different upscaling methods for representing farm household level decision making at higher scales/integration levels. For explanation seetext.




demanding, but still ignores interactions between farmers, andbetween farmers and their environment.

In the fourth approach interactions between farmers are takeninto account, and these interactions are seen as key drivers of over-all system functioning and behaviour (Fig. 4D). This approach is thebasis of many existing agent based models (e.g., Van Wijk et al.,2014). Important direct buying-and-selling relationships betweenfarmers can be captured in multi-agent models, for example thesale of crop residues of crop farmers as animal feed to livestockowners without the existence of a formal market. While theapproaches represented in Fig. 4A–D are currently available, themore complex setups of Fig. 4E–F are not. The approaches of 4Cand 4D could be combined by giving the local market a specificlocation in the multi-agent interactions (Straatman et al., 2013).The next step would be to bring in the landscape in such a modeldescription, so that the assessment of environmental indicatorscan be improved. Feedbacks between key environmental indicatorslike erosion and nitrogen leaching, which vary strongly across alandscape, and crop, grassland and livestock production can bequantified in such a setup and the longer term sustainability offood production in different locations of a landscape can beassessed. In first instance interactions between farmers could beignored (Fig. 4E), although in a complete description of the land-scape and the local market they will be important to incorporate(Fig. 4F). This full-blown model description will be very data-demanding (they will require farm characterisation data in combi-nation with socio-economic information on institutions, local mar-kets, cultural rules and biophysical information), but can giveimportant information in terms of land management and farmexpansion or farm relocation responses of the local community.These responses, sampled in contrasting sites, could be summa-rised in relation to their drivers, and used for better land use man-agement and land use change descriptions in higher scale models.For example, ABMs have been used in contrasting regions of theworld to generate information about the drivers of land use foresttransition zones (Evans and Kelly, 2008) and especially in ecologymacro-scale patterns have been derived from interactions of heter-ogeneous agents at lower scales (Grimm et al., 2005). Whether thisis a powerful approach for land use modelling at large scales stillhas to be proven though (Rounsevell et al., 2012).

Conclusions

Regional and national policies can provide a framework, guidethe incentive structure and police outcomes that can help toimprove the food security and sustainability of farm based liveli-hoods. Infrastructure development, improvement of value chainsfor agricultural products and innovation systems for the designand promotion of alternative agricultural practices ought toempower small scale farmers to achieve food security, while atleast not endangering ecosystem services (Lopez-Ridaura andGérard, 2012). However, good impact assessment tools that canassess the consequences of regional and national policies acrossintegration levels in developing countries, thereby informing pol-icy makers, are still missing. The existing gap between landscapeand community to sub-national market and policy levels preventstruly integrated assessment of policy options. This paper suggests astep wise approach with increasing data needs to bridge the exist-ing gap with a focus on smallholder farming in developingcountries.

Clear improvements need to be made at the description ofeffects of the distribution of local markets on price formation andthe representation of farm diversity within existing large scalemaps of farming systems (Dixon et al., 2001; Seré and Steinfeld,1996). Especially answering the question ‘how many are there of

which farms, and where are they located’ is pertinent. Here remotesensing, farm characterisation questionnaires, and market accessand agro-ecological information need to come together. Pastoralistsystems especially form a challenge for the land based approachesthat are used in the current models. The modellers’ quest for ‘find-ing homogeneity in heterogeneity’ (Messerli et al., 2009) to be ableto aggregate fine-scale information for up-scaling purposes hasquite some similarities to the one for the holy grail, but it is essen-tial that we develop methods to identify particular land usergroups of which land use and market response are important forpolicy projections. It is easy to drown in model complexity andunrealistic model data demands, although this danger should alsonot result in that we only use simplistic representations of theinherent complex socio-agro-ecological systems (Rounsevell andArneth, 2011). Detailed micro-market and multi-agent studiesare useful in this context if they are used to derive summary func-tions describing the most important effects of feedbacks betweenfarmers’ choices and local markets. Laborte et al. (2007) andValdivia et al. (2012) have shown that these feedbacks matterwhen describing farm level management responses and local mar-ket level price formation. More work on these feedbacks is neededto derive generic summary relations that capture these feedbacksacross a wide range of systems and socio-economic conditions thatcan be used in larger scale studies. Also the income and food secu-rity consequences of interactions between farmers not well quan-tified yet, for example the selling of crop residues as animal feed tolivestock owners without the presence of formal markets. Theseinteractions and feedbacks between farmers are likely to increasethe flexibility and buffer capacity of crop and livestock based live-lihoods to deal with climate and market shocks.

The review did not cover an area like the modelling of regionaland larger scale trade flows, but improved embedding of thesetrade modelling approaches could further improve the descriptionof the scenarios that are used for describing the future develop-mental pathways that act as drivers of change at micro-level. Alsohow market failure and short term government responses (as incontrast to longer policy development) can shape the developmentof the agricultural sector was ignored in this review. It is clear thatby filling the gap between micro and macro level approaches betterinformation can be generated for policy makers regarding the con-sequences of global level changes for local level food security, andregarding the identification of where most efficiently policy invest-ments should take place to achieve larger food security and envi-ronmental impacts. Especially for the livestock sector, as thelargest land use sector in the world with its big role in food produc-tion and its big environmental impact, filling this gap is essential tobetter target policy interventions to increase the benefits of live-stock production while reducing the problems associated with it.

Acknowledgements

This work was funded by CGIAR Research Programs on ClimateChange, Agriculture and Food Security (CCAFS) and on IntegratedSystems for the Humid Tropics (HumidTropics). The paper waspresented at the Mainstreaming Livestock Value Chain Conferenceon 5–6 Nov. 2013, organized by the CGIAR Research Program onPolicies, Institutions, and Markets. I thank Prof. D. Baker and oneanonymous referee for their constructive comments on a previousdraft of the manuscript.

References

An, L., 2012. Modeling human decisions in coupled human and natural systems:review of agent-based models. Ecol. Model. 229, 25–36.

An, L., Linderman, M., Qi, J., Shortridge, A., Liu, J., 2005. Exploring complexity in ahuman–environment system: an agent-based spatial model formultidisciplinary and multiscale. Ann. Assoc. Am. Geogr. 95, 54–79.



http://refhub.elsevier.com/S0306-9192(14)00145-6/h0010






Bahta, S., Malope, P., 2014. Measurement of competitiveness in smallholderlivestock systems and emerging policy advocacy: an application to Botswana.In: Baker, Upton (Eds.), Mainstreaming Livestock Value Chains: Bringing theResearch to Bear on Impact Assessment, Policy Analysis and Advocacy forDevelopment. Food Policy (this issue).

Bakker, M., Govers, G., Jones, R., Rounsevell, M., 2007. The effect of soil erosion onEurope’s crop yields. Ecosystems 10, 1209–1219.

Berger, T., Troost, C., 2014. Agent-based modelling of climate adaptation andmitigation options in agriculture. J. Agric. Econ. 65, 323–348.

Burney, J.A., Davis, S.J., Lobell, D.B., 2010. Greenhouse gas mitigation by agriculturalintensification. Proc. Natl. Acad. Sci. USA 107, 12052–12057.

Chaudhury, M., Vervoort, J., Kristjanson, P., Ericksen, P., Ainslie, A., 2013.Participatory scenarios as a tool to link science and policy on food securityunder climate change in East Africa. Reg. Environ. Change 13, 389–398.

Claessens, L., Antle, J.M., Stoorvogel, J.J., Valdivia, R.O., Thornton, P.K., Herrero, M.,2012. A method for evaluating climate change adaptation strategies for small-scale farmers using survey, experimental and modeled data. Agric. Syst. 111,85–95.

Creutzig, F., Popp, A., Plevin, R., Luderer, G., Minx, J., Edenhofer, O., 2012. Reconcilingtop-down and bottom-up modelling on future bioenergy deployment. Nat.Clim. Change 4. http://dx.doi.org/10.1038/NCLIMATE1416.

Dalgaard, T., Hutchings, N.J., Porter, J.R., 2003. Agroecology, scaling andinterdisciplinarity. Agric. Ecosyst. Environ. 100, 39–51.

Dixon, J., Gulliver, A., Gibbon, D., 2001. Farming Systems and Poverty. ImprovingFarmers’Livelihoods in a Changing World. FAO, Rome and Washington, 407pp.

Dorward, A., Anderson, S., Bernal, Y.N., Vera, E.S., Rushton, J., Pattison, J., Paz, R.,2008. Hanging in, stepping up and stepping out: livelihood aspirations andstrategies of the poor. Dev. Pract. 19, 240–247.

Dowlatabadi, H., 1995. Integrated assessment models of climate change. Anincomplete overview. Energy Policy 23, 289–296.

Dumollard, G., Havlík, P., Herrero, M., 2013. Climate change and land basedactivities: a review of economic models. CGIAR Research Program on ClimateChange, Agriculture and Food Security (CCAFS). Working Paper No. 34, 89pp.

Ericksen, P.J., Ingram, J.S.I., Liverman, D.M., 2009. Food security and globalenvironmental change: emerging challenges. Environ. Sci. Policy 12, 373–377.

Evans, T.P., Kelly, H., 2008. Assessing the transition from deforestation to forestregrowth with an agent-based model of land cover change for south-centralIndiana (USA). Geoforum 39, 819–832.

Ewert, F., van Ittersum, M.K., Bezlepkina, I., Therond, O., Andersen, E., et al., 2009. Amethodology for enhanced flexibility of integrated assessment in agriculture.Environ. Sci. Policy 12, 546–561.

Ewert, F., van Ittersum, M.K., Heckelei, T., Therond, O., Bezlepkina, I., Andersen, E.,2011. Scale changes and model linking methods for integrated assessment ofagri-environmental systems. Agric. Ecosyst. Environ. 142, 6–17.

Fadiga, M., Katjiuongua, H.B., 2014. Issues and strategies in ex-post evaluation ofintervention against animal disease outbreaks and spreads. In: Baker, Upton(Eds.), Mainstreaming Livestock Value Chains: Bringing the Research to Bear onImpact Assessment, Policy Analysis and Advocacy for Development. Food Policy(this issue).

Fischer, G., Shah, M.N., Tubiello, F., van Velhuizen, H., 2005. Socio-economic andclimate change impacts on agriculture: an integrated assessment, 1990–2080.Philos. Trans. Roy. Soc. B: Biol. Sci. 360, 2067–2083.

Foley, J.A., DeFries, R., Asner, G.P., Barford, C., Bonan, G., Carpenter, S.R., Chapin, F.S.,Coe, M.T., Daily, G.C., Gibbs, H.K., Helkowski, J.H., Holloway, T., Howard, E.A.,Kucharik, C.J., Monfreda, C., Patz, J.A., Prentice, I.C., Ramankutty, N., Snyder, P.K.,2005. Global consequences of land use. Science 309, 570–574.

Frank, A.C., van den Brand, G.J., Giller, K.E., 2014. Which farmers benefit most fromsustainable intensification? An ex-ante impact assessment of expanding grainlegume production in Malawi. Eur. J. Agron. 58, 28–38.

Franke, A.C., Rufino, M.C., Farrow, A., 2011. Characterisation of the Impact Zones andMandate Areas in the N2Africa Project. Wageningen University, Milestonereference number 1.4.1, 50pp.

Giller, K.E., 2013. Guest editorial: can we define the term ‘‘farming systems?’’ Aquestion of scale. Out. Agric. 42, 149–153.

Grimm, V., Revilla, E., Berger, U., Jeltsch, F., Mooij, W., Raisback, S., Thulke, H.,Weiner, J., Weigand, T., DeAngelis, D., 2005. Pattern-oriented modeling of agent-based complex systems: lessons from ecology. Science 310, 987–991.

Harris, G., 2002. Integrated assessment and modelling—science for sustainability.In: Costanza, R., Joergensen, S.E. (Eds.), Understanding and SolvingEnvironmental Problems in the 21st Century. Elsevier, pp. 5–17.

Havlík, P., Schneider, U.A., Schmid, E., Böttcher, H., Fritz, S., Skalsky, R., Aoki, K., Cara,S.D., Kindermann, G., Kraxner, F., Leduc, S., McCallum, I., Mosnier, A., Sauer, T.,Obersteiner, M., 2011. Global land - use implications of first and secondgeneration biofuel targets. Energy Policy 39, 5690–5702.

Herrero, M., Thornton, P.K., 2013. Livestock and global change: emerging issues forsustainable food systems. PNAS 110, 20878–20881.

Herrero, M., Havlík, P., Valin, H., Notenbaert, A., Rufino, M.C., Thornton, P.K.,Blümmel, M., Weiss, F., Grace, D., Obersteiner, M., 2013. Biomass use,production, feed efficiencies, and greenhouse gas emissions from globallivestock systems. PNAS 110, 20888–20893.

Herrero, M., Thornton, P.K., Bernues, A., Baltenweck, I., Vervoort, J., van de Steeg, J.,Makokha, S., van Wijk, M.T., Karanja, S., Rufino, M.C., Staal, S.J., 2014. Integratinglivestock feeds and production systems into agricultural multi-market models.Glob. Environ. Change. 24, 165–182.

IAASTD, 2009. Agriculture at a Crossroads. International Assessment of AgriculturalKnowledge, Science and Technology for Development (IAASTD), Global Report.Island Press, Washington, DC.

Joint Research Centre (JRC), 2011. Critical Issues in Estimating ILUC Emissions.Outcomes of an Expert Consultation, November 9–10, 2010, Ispra, Italy.European Commission, Publications Office of the European Union,Luxembourg. <http://publications.jrc.ec.europa.eu/repository/bitstream/111111111/22908/2/reqno_jrc64429_cirtical%20issues%20in%20estimating%20iluc%20emissions %20print%20version.pdf>.

Klapwijk, C.J., van Wijk, M.T., Rosenstock, T.S., van Asten, P.J., Thornton, P.K., Giller,K.E., 2014. Review – analysis of trade-offs in agricultural systems: current stateand way forward. Curr. Opin. Environ. Sustain. 6, 110–115.

Laborte, A.G., Van Ittersum, M.K., Van den Berg, M.M., 2007. Multi-scale analysis ofagricultural development: a modelling approach for Ilocos Norte, Philippines.Agric. Syst. 94, 862–873.

Lambin, E.F., Rounsevell, M.D.A., Geist, H.J., 2000. Are agricultural land-use modelsable to predict changes in land-use intensity? Agric. Ecosyst. Environ. 82, 321–331.

Little, P.D., Dejene Negassa Debsu, Waktole Tiki, 2014. How pastoralists perceiveand respond to market opportunities: the case of the horn of Africa. In: Baker,Upton (Eds.), Mainstreaming Livestock Value Chains: Bringing the Research toBear on Impact Assessment, Policy Analysis and Advocacy for Development.Food Policy (this issue).

Lopez-Ridaura, S., Gérard, B., 2012. Adaptation to climate change for cereal basedsmall scale farming systems. Options at different levels, from plant to region. In:Third International Agronomy Congress, Agriculture Diversification, ClimateChange Management and Livelihoods, November 26–30, New Delhi, India.

Lotze-Campen, H., 2008. The role of modelling tools in Integrated SustainabilityAssessment (ISA). Int J Innov Sust Dev 317, 70–92.

Lotze-Campen, H., Müller, C., Bondeau, A., Jachner, A., Popp, A., Lucht, W., 2008. Fooddemand, productivity growth and the spatial distribution of land and water use:a global modeling approach. Agric. Econom. 39, 325–338.

Lotze-Campen, H., Popp, A., Beringer, T., Müller, C., Bondeau, A., Rost, S., Lucht, W.,2010. Scenarios of global bioenergy production: the trade-offs betweenagricultural expansion, intensification and trade. Ecol. Model. 221, 2188–2196.

MacOpiyo, L., Stuth, J., Wu, J. Use of GIS and agent-based modeling to simulatepastoralist mobility in the rangelands of East Africa. In: GSDI-9 ConferenceProceedings, 6–10 November 2006, Santiago, Chile.

Messerli, P., Heinimann, A., Epprecht, M., 2009. Finding homogeneity inheterogeneity – a new approach to quantifying landscape mosaics developedfor the lao pdr. Hum. Ecol. 37, 291–304.

Msangi, S.D. Enahoro, Herrero, M., Magnan, N., Havlik, P., Notenbaert, A., Nelgen, S.,2014. Integrating livestock feeds and production systems into agriculturalmulti-market models: The example of IMPACT. In: Baker, Upton (Eds.),Mainstreaming Livestock Value Chains: Bringing the Research to Bear onImpact Assessment, Policy Analysis and Advocacy for Development. Food Policy(this issue).

Parker, P., Letcher, R., Jakeman, A., et al., 2002. Progress in integrated assessmentand modelling. Environ. Modell. Softw. 17, 209–217.

Parker, D.C., Manson, S.M., Janssen, M.A., Hoffmann, M.J., Deadman, P., 2003. Multi-agent systems for the simulation of land-use and land-cover change: a review.Ann. Assoc. Am. Geogr. 93, 314–337.

Phalan, B., Balmford, A., Green, R.E., Scharlemann, J.P.W., 2011. Minimising the harmto biodiversity of producing more food globally. Food Policy 36, S62–S71.

Reid, W.V., Chen, D., Goldfarb, L., Hackmann, H., Lee, Y.T., Mokhele, K., Ostrom, E.,Raivio, K., Rockström, J., Schellnhuber, H.J., Whyte, A., 2010. Earth systemscience for global sustainability: grand challenges. Science 330, 916–917.

Rindfuss, R.R., Walsh, S.J., Turner, B.L., Fox, J., Mishra, V., Hanson, S., 2004.Developing a science of land change: challenges and methodological issues.Proc. Natl. Acad. Sci. USA 101, 13976–13981.

Rindfuss, R.R., Entwisle, B., Walsh, S.J., An, L., Badenoch, N., Brown, D.G., Deadman,P., Evans, T.P., Fox, J., Geoghegan, J., Gutmann, M., Kelly, M., Linderman, M., Liu,J., Malanson, G.P., Mena, C.F., Messina, J.P., Moran, E.F., Parker, D.C., Parton, W.,Prasartkul, P., Robinson, D.T., Sawangdee, Y., Vanwey, L.K., Verburg, P.H., 2008.Land use change: complexity and comparisons. J Land Use Sci 3, 1–10.

Rosegrant, M.W., the IMPACT Development Team, 2012. International Model forPolicy Analysis of Agriculture Commodities and Trade (IMPACT). ModelDocumentation.

Rotmans, J., van Asselt, M., 1996. Integrated assessment: growing child on its way tomaturity. An editorial essay. Clim. Change 34, 327–336.

Rotmans, J., Deboois, H., Swart, R.J., 1990. An integrated model for the assessment ofthe greenhouse-effect—the Dutch Approach. Clim. Change 16, 331–356.

Rounsevell, M.D.A., Arneth, A., 2011. Representing human behaviour and decisionalprocesses in land system models as an integral component of the earth system.Glob. Environ. Change 21, 840–843.

Rounsevell, M.D.A., Pedroli, B., Erb, K.-H., Gramberger, M., Busck, A.G., Haberl, H.,Kristensen, S., Kuemmerle, T., Lavorel, S., Lindner, M., Lotze-Campen, H.,Metzger, M.J., Murray-Rust, D., Popp, A., Pérez-Soba, M., Reenberg, A.,Vadineanu, A., Verburg, P.H., Wolfslehner, B., 2012. Challenges for landsystem science. Land Use Policy 29, 899–910.

Rudel, T.K., Schneider, L., Uriarte, M., Turner, B.L., DeFries, R., Lawrence, D.,Geoghegan, J., Hecht, S., Ickowitz, A., Lambin, E.F., Birkenholtz, T., Baptista, S.,Grau, R., 2009. Agricultural Intensification and Changes in Cultivated Areas,1970–2005. PNAS106, pp. 20675–20680.
















http://dx.doi.org/10.1038/NCLIMATE1416
























































http://publications.jrc.ec.europa.eu/repository/bitstream/111111111/22908/2/reqno_jrc64429_cirtical%20issues%20in%20estimating%20%20iluc%20emissions%20%20print%20version.pdf






















































Rufino, M.C., Dury, J., Tittonell, P., van Wijk, M.T., Herrero, M., Zingore, S., Mapfumo,P., Giller, K.E., 2011. Competing use of organic resources, village-levelinteractions between farm types and climate variability in a communal areaof NE Zimbabwe. Agric. Syst. 104, 175–190.

Rufino, M.C., Thornton, P.K., Ng’ang’a, S.K., Mutie, I., Jones, P.G., van Wijk, M.T.,Herrero, M., 2013. Livestock transitions in agro-pastoralist systems of EastAfrica: Impacts on food security and poverty. Agric. Ecosyst. Environ. 179, 215–230.

Sakané, N., Becker, M., Langensiepen, M., van Wijk, M.T., 2013. Typology ofsmallholder production systems in small East-African wetlands. Wetlands 33,101–116.

Schaldach, R., Alcamo, J., Koch, J., Kölking, C., Lapola, D.M., Schüngel, J., Priess, J.A.,2011. An integrated approach to modelling land-use change on continental andglobal scales. Environ. Modell. Softw. 26, 1041–1051.

Schreinemachers, P., Berger, T., 2011. An agent-based simulation model of humane–environment interactions in agricultural systems. Environ. Modell. Softw. 26,845–859.

Seré, C., Steinfeld, H., 1996. World Livestock Production Systems: Current Status,Issues and Trends. FAO Animal Production and Health Paper 127. FAO, Rome,Italy, 82pp.

Stoorvogel, J.J., Antle, J.M., Crissman, C.C., 2004. Trade-off analysis in the NorthernAndes to study the dynamics in agricultural land use. J. Environ. Manage. 72,23–33.

Straatman, B., Marceau, D.J., White, R., 2013. A generic framework for a combinedagent-based market and production model. Comput. Econ. 41, 425–445.

Tol, R.S.J., 2006. Integrated Assessment Modelling. Working Papers: FNU-102,Research Unit Sustainability and Global Change. Hamburg University, Hamburg,Germany.

Valdivia, R.O., Antle, J.M., Stoorvogel, J.J., 2012. Coupling the tradeoff analysis modelwith a market equilibrium model to analyze economic and environmentaloutcomes of agricultural production systems. Agric. Syst. 110, 17–29.

Van Asselen, S., Verburg, P.H., 2012. A land system representation for globalassessments and land-use modeling. Glob. Change Biol. 18, 3125–3148.

Van Cauwenbergh, N., Biala, K., Bielders, C., Brouckaert, V., Franchois, L., GarciaCidad, V., Hermy, M., Mathijs, E., Muys, B., Reijnders, J., Sauvenier, X., Valckx, J.,Vanclooster, M., Van der Veken, B., Wauters, E., Peeters, A., 2007. SAFE—a

hierarchical framework for assessing the sustainability of agricultural systems.Agric. Ecosyst. Environ. 120, 229–242.

Van Ittersum, M.K., Ewert, F., Heckelei, T., Wery, J., Alkan Olsson, J., Andersen, E.,Bezlepkina, I., Brouwer, F., Donatelli, M., Flichman, G., Olsson, L., Rizzoli, A.E.,van der Wal, T., Wien, J.E., Wolf, J., 2008. Integrated assessment of agriculturalsystems—a component-based framework for the European Union (SEAMLESS).Agric. Syst. 96, 150–165.

Van Wijk, M.T., Rufino, M.C., Enahoro, D., Parsons, D., Silvestri, S., Valdivia, R.O.,Herrero, M., 2014. Farm household modelling and its role in designing climate-resilient agricultural systems. Glob. Food Security. 3, 77–84.

Velthof, G.L., Oudendag, D.A., Oenema, O., 2007. Development and Application ofthe Integrated Nitrogen Model MITERRAEUROPE. Alterra Report. Alterra,Wageningen, p. 102.

Verburg, P.H., de Koning, G.H.J., Kok, K., Veldkamp, A., Bouma, J., 1999. A spatialexplicit allocation procedure for modelling the pattern of land use change basedupon actual land use. Ecol. Model. 116 (1), 45–61.

Verburg, P., Eickhout, B., van Meijl, H., 2008. A multi-scale, multi-model approachfor analyzing the future dynamics of European land use. Ann. Reg. Sci. 42, 57–77.

Verburg, P.H., Neumann, K., Nol, L.A., 2011. Challenges in using land use and landcover data for global change studies. Glob. Change Biol. 17, 974–989.

Verburg, P.H., van Asselen, S., van der Zanden, E.H., Stehfest, E., 2013a. Therepresentation of landscapes in global scale assessments of environmentalchange. Landscape Ecol. 28, 1067–1080.

Verburg, P.H., Mertz, O., Erb, K.-H., Haberl, H., Wu, W., 2013b. Land system changeand food security: towards multi-scale land system solutions. Curr. Opin.Environ. Sustain. 5, 1–9.

Volk, M., Ewert, F., 2011. Scaling methods in integrated assessment of agriculturalsystems—State-of-the-art and future directions. Agric. Ecosyst. Environ. 142, 1–5.

Vorley, B., 2002. Sustaining Agriculture: Policy, Governance, and the Future ofFamily-Based Farming. IIED, London, pp. 1–196.

Zhang, W., Yu, E.A., Rozelle, S., Yang, J., Msang, S., 2013. The impact of biofuelgrowth on agriculture: why is the range of estimates so wide? Food Policy 38,227–239.





































































from global economic modelling to household level analyses of...

Documents