the role of spatially explicit models in land-use change ... · the role of spatially explicit...

Agriculture, Ecosystems and Environment 85 (2001) 177–190

The role of spatially explicit models in land-use change research:a case study for cropping patterns in China

Peter H. Verburg∗, A. VeldkampLaboratory of Soil Science and Geology, Department of Environmental Sciences,

Wageningen Agricultural University, PO Box 37, 6700 AA Wageningen, The Netherlands

Abstract

Single research methodologies do not suffice for a complete analysis of land-use change. Instead, a sequence of method-ologies is needed that link up and integrate disciplinary components over a range of spatial and temporal scales. In this paper,a modelling methodology is presented aiming at the analysis of the spatial and temporal dynamics of land use at the regionallevel. The methodology explores the dynamic functioning of land-use systems, which is essential to bridge the gap betweenstudies identifying problems associated with land-use change and studies aiming at understanding and manipulating land-usechange processes. An illustration of the methodology is given for China where we have simulated a scenario of near-future(1991–2010) changes in land-use patterns. The methodology is adapted to include the nested simulation of different crop typesin addition to the simulation of land-cover change. Results are presented for changes in the spatial distribution of cultivatedland and special emphasis is given to shifts in the distribution of different crops. In the northern part of the country a decrease inthe proportion wheat within the cropping system is expected whereas in the southern part the proportion of rice is decreasing.Corn and vegetable crops are expected to become more important within the cropping system in these parts of the country.© 2001 Elsevier Science B.V. All rights reserved.

Keywords: Land-use change; China; Cropping patterns; Scale

1. Introduction

The field of land-use change studies is stronglydivided by scientific discipline, tradition and scaleof analysis. Researchers in the social sciences havea long tradition of studying individual behaviour inhuman–environment interactions at the micro-scale,mostly by narrative approaches. At higher levels ofaggregation, geographers and ecologists have studiedland-use change either by direct observation, usingremote sensing and GIS or have applied the systems/

∗ Corresponding author. Tel.: +31-317-485208;fax: +31-317-482419.E-mail address: [email protected] (P.H. Verburg).

structures perspective to better understand the organi-sation of society and landscapes (Lambin et al., 1999).

Apart from these differences caused by scientifictradition and scale of analysis there are also in-evitably large variations in research approach becauseof differences in research objectives and stakeholdersaddressed. Some of the studies aim at understanding(parts of) the land-use system dynamics by itself,while others aim at intervention in land-use dynamicsby means of land-use planning or the designing ofalternative land-use systems.

Intervention in the dynamics of land-use systemsis impossible without a proper understanding of thedriving factors in these systems and their behaviour.Such a complete analysis of systems as complex asland-use systems, is impossible with a single research

0167-8809/01/$ – see front matter © 2001 Elsevier Science B.V. All rights reserved.PII: S0 1 6 7 -8 8 09 (01 )00184 -0

178 P.H. Verburg, A. Veldkamp / Agriculture, Ecosystems and Environment 85 (2001) 177–190

methodology (Bouma, 1998). Therefore, differentresearch approaches originating from different disci-plinary backgrounds and different scales of analysisshould not be considered in isolation but shouldrather be linked and inter-related following a logi-cal sequence (Levin, 1992; Fresco, 1995; Rindfussand Stern, 1998). This sequence of interconnectedmethodologies for studying land-use change researchproblems can be called a ‘research sequence’ in whichthe different research methodologies (‘tools’) are or-dered according to their spatial scale of analysis andphase of research.

To fit existing research methodologies within such aresearch sequence, methodologies should be designedor adapted so that information and understandingobtained with a single research methodology is com-plementary to other methodologies operating at dif-ferent scales and/or in different phases of research.

This paper discusses a research sequence for land-use change studies and elaborates on the use of aspatially explicit methodology for land-use changemodelling within this research sequence. This metho-dology, the CLUE modelling framework (Veldkampand Fresco, 1996; Verburg et al., 1999a) is designed tobridge the gap between different phases of a land-usechange research sequence. An extension of thismethod, allowing the nested simulation of different

Fig. 1. Research sequence for land-use change research.

crop types in addition to the simulation of land-covertypes, is presented. An illustration of the method isprovided for the exploration of possible shifts in landuse and cropping patterns in China.

2. A research sequence for land-use change studies

This section gives an example of a typical researchsequence which aims to achieve changes in theland-use system by steering specific characteristics ofthe system to avoid or decrease negative impacts ofland-use changes. Fig. 1 illustrates this research chainby providing a logical sequence of the different toolsordered according to their scale of analysis.

2.1. Problem identification phase

Before any in-depth analysis of the land-use dyna-mics can start, a need exists for a detailed identificationand exploration of the problems that may be associatedwith future land-use changes. Most often, problemsare identified by means of rough extrapolations ofcurrent trends or monitoring of changes in the environ-ment. A good example of this type of studies are stud-ies by the World Watch Institute (e.g., Brown, 1995;Brown and Kane, 1994), which contain warnings

P.H. Verburg, A. Veldkamp / Agriculture, Ecosystems and Environment 85 (2001) 177–190 179

for global food shortages as a result of growing fooddemands and deteriorating environmental conditions.In 1995, Lester Brown, President of the World WatchInstitute, published a book called ‘Who Will FeedChina?’ In this rather provoking book he predicts,based on trends and comparisons with other Asiancountries, large shortfalls in grain production in Chinain the near-future. This would be induced by conver-sion of arable land into urban, industrial and horti-cultural use, in combination with increasing demandsfor grain, driven by population growth and changes inconsumption pattern. In another book, published in thesame period, Smil warns for the large extent of envi-ronmental pollution and degradation in China (Smil,1993). These books have made the world aware of thepotential problems China might face in the near-futurewith respect to its food security and the impact thatthese shortfalls in grain production might have for in-ternational trade of grain. Therefore, much researchwas initiated dealing with China’s ability to produceits own grain and how this is influenced by land-usechanges (e.g., Heilig, 1997; Lin, 1998; Alexandratos,1997; Fan and Agcaoili-sombilla, 1997). Althoughmost of this research proved that China’s food prob-lem is not as large as Brown expected, his publicationwas important for putting these issues on the agendaof politicians and scientists. This is exactly the objec-tive of studies in this research phase.

2.2. System description phase

As soon as a potential land-use change problem isidentified and research funds have become available,a need for better insight in the land-use system arises.During this ‘system description phase’ more detailedstatements can be made about the land-use situationby adding a spatial dimension to the research. A typi-cal example of such studies is the identification ofregions and locations with high rates of land-coverchange, so-called ‘hot-spots’, based on remote sensingimages (e.g., Achard et al., 1998; Imbernon, 1999).Apart from these ‘hot-spots’ of land-use change itis also important to obtain more insight into factorsthat cause land-use changes. At the local level this ismostly done by narrative studies revealing causes andincentives of actors of land-use change (e.g., Rudeland Horowitz, 1993; Jones, 1999). Other approachesinclude the agent-based and system approach. The

former seeks to define the general nature and rulesof individual agents behaviour in their daily decisionmaking. Central to this perspective is the significancegiven to human agents in determining land-use de-cisions and the search for generalisations about thisbehaviour. The systems approach in contrast, finds itsbasis in the organisation of society and the characterof nature that establish opportunities and constraintson decision making (Ostrom, 1990). On coarser scalesthis approach is more feasible because it is moredifficult to distinguish individual actors and relationsbetween actors and the diverse environment. At coarsescales, land-use patterns are related to macro-variablesrepresenting proximate drivers of land-use change.Based on these understandings, models can be used toexplore scenarios of near-future land-use change (e.g.,Hall et al., 1995; Veldkamp and Fresco, 1996). Suchspatially explicit models can provide information onwhat might happen if certain policies or other land-usedeterminants change. Such models are needed be-cause land use is the result of complex processesrequiring reflection of non-linearity and spatial andtemporal lags in the analysis. Therefore, these modelsare useful and reproducible tools supplementing ourexisting mental modelling capabilities to make moreinformed decisions (Costanza and Ruth, 1998). Basedon the information and insights obtained in this phaseof the research sequence, scientists, policy makersand other stakeholders can decide if the foreseenland-use dynamics are desirable. Agri-environmentalindicators (Moxey et al., 1998) are a suitable meansto summarise and communicate results and facilitatedecision making on the desirability of the foreseendevelopments. Based upon the understandings ob-tained in this research phase it is possible to effi-ciently design alternative land-use plans focussed onthe appropriate issues and geographical locations.

2.3. Designing phase

Many methodologies for the design of interventionsin present land-use system are available ranging fromthe designing of appropriate policies and subsidiesto landscape designs by architects. At the farm level,prototyping (e.g., Vereijken, 1997) is a way to developfarming systems to meet a set of environmental andsocio-economic objectives. Based on these objectives,established on the basis of the shortcomings of current


farming systems in a region, a theoretical prototypeis designed by linking indicators for these objectivesto farming methods and designing these methods un-til they are ready for testing and implementation. Atcoarser scales, studies are made on optimal land-useconfigurations under a number of constraints. Oneseries of models aimed at designing alternatives forpresent land use are models using linear programmingin order to optimise the land-use configuration andmanagement under a number of agro-technical, foodsecurity, socio-economic and environmental objec-tives (e.g., Bouman and Nieuwenhuyse, 1999; Zanderand Kächele, 1999; Barbier, 1998; Van Ittersum etal., 1998). Results of such a run by a linear program-ming model are characterised by optimised objectivevalues and the associated optimal set of decision vari-ables (agricultural land-use activities: where, whattype of agriculture to which extent). Such results canbe presented in a table or a bar diagram showing theobjective values, and in a map showing the optimumland-use allocation.

2.4. Negotiation and planning phase

Based upon the system description phase whichidentified the driving factors and sensitivities of theland-use system and the designed land-use alternativesand prototypes, a plan can be made to implement theresearch results. During this phase the proper incen-tives and conditions are created, in close collaborationwith the stakeholders involved, that will make thedesigned land-use plans become reality.

The following paragraphs elaborates on a re-search methodology developed for regional land-usechange analysis during the system description phase.This methodology bridges the gap between researchmethodologies in the problem identification phase andthe designing phase. In most cases design-orientedresearch is started directly after the problem iden-tification. Whenever information on the functioningof the land-use system is used, it is mostly the in-formation attained at the local level. Therefore, theintroduced methodology is developed to generateregional level information. Present applications ofthe method (Veldkamp and Fresco, 1996; De Koninget al., 1999a; Verburg et al., 1999b) are to study theland-use change for land-cover types or large groupsof crops only. However, most of the studies in the

design phase of the research-chain study of theland-use systems was for specific crops or croppingsequences. To match these studies better, the possi-bility to simulate the spatial and temporal dynamicsfor individual crops was included in the methodologyas described in the following paragraph.

3. Methodology

3.1. Modelling framework

In this study, the CLUE modelling framework isused for the simulation of land-use changes (Veldkampand Fresco, 1996; Verburg et al., 1999a). This mod-elling framework has the following characteristics:

• All simulations are made in a spatially explicit wayso that the geographical pattern of land-use changeis resulting. The spatial resolution of the simula-tions is dependent on the extent of the study areaand the resolution of data available for that studyarea. In the present application for China, all simu-lations are made for approximately 9200 grid-cellsof 32×32 km (∼1000 km2). Land use in a grid-cellis represented by its relative land area (expressed asa percentage of the total land area).

• Allocation of land-use changes is based on dyna-mic simulation of competition between differentland-use types. Competitive advantage is based onthe ‘local’ and ‘regional’ suitability of the locationand the national level demand for land-use type re-lated products (e.g., food demand or demand forresidential area).

• The ‘local’ and ‘regional’ suitability for the dif-ferent land-use types is determined by quantifiedrelations between land use and a large number ofexplanatory factors. Explanatory factors includebiophysical and socio-economic factors generallyknown as determinants of the land-use distribu-tion and driving factors of land-use change (Turneret al., 1993). Table 1 lists all variables used in thepresented application for China. Relations betweenland use and the explanatory factors are quantifiedby multiple regression models based on presentland use. It is assumed that these relations are rela-tively stable over the simulation period. The spatialanalysis and resulting regression models for Chinaare described in Verburg and Chen (2000).


Table 1List of land-use types and explanatory factors used

Land-use typesCultivated landHorticultural landForestGrasslandBuilt-up landInland waterUnused land

Crop typesRiceWheatCornOther grainsCash cropsVegetablesOthers

Demographic variablesPopulation densityRural populationUrban populationRural labour forceAgricultural labour force

Socio-economic variablesIlliteracyDistance to nearest cityNet income

Climatic variablesTemperature (minimum, maximum, average, range)No. of months above 10◦CPrecipitation (total, range)No. of wet months (>50 mm precipitation)Percentage of sunshine

Soil-related variablesSoil drainage (three classes)Soil depth (two classes)Soil moisture storage capacity (two classes)Soil texture (three classes)Soil fertility (three classes)Soil suitability for irrigated rice (three classes)Soil suitability for rainfed maize (three classes)

Landform-related variablesMean elevationRange in elevationSlopePhysiography (four classes)Geomorphology (five classes)Distance to nearest river

Fig. 2. Structure of the CLUE modelling framework.

Fig. 2 schematically illustrates the structure of themodelling framework. At the national level, demandsfor different land-use types are determined. Projec-tions for future demands are directly obtained fromother studies that use trend analysis or economicmodels to predict changes in demand for the differentland-use types. In most applications these demands aredirectly converted into areas of the different land-usetypes needed to meet the demands for products fromthese land-use types. Expected developments in yieldlevels and multiple cropping can be taken into ac-count in these calculations. The demands are allocatedon a yearly basis to the different grid-cells in theallocation modules which use the relations betweenland use and the explanatory factors obtained by thespatial empirical analysis. The implementation of theallocation procedure is described in detail in Verburget al. (1999a). Simulations presented in this paper


allocate land-use changes in a nested procedure. First,changes in the different land-cover types are simu-lated. In China, we have subdivided the total land areainto seven categories, including six land-cover types(cultivated land, horticultural land, forest, grassland,built-up land and unused land) and inland water bod-ies. In the simulation these different land-cover typescompete for the total land area available. It is assumedthat the area occupied by inland water bodies is stable.Thereafter, nested within this simulation follows theallocation of different agricultural crops or groups ofcrops. Included are wheat (Triticum spp.), rice (Oryzaspp.), corn (Zea Mays), a group containing othergrains (incl. millet, sorghum (Sorghum spp.), othermiscellaneous grains, tubers and soybeans (Glycinesoya), cash crops (incl. fiber crops, oil crops), tobacco(Nicotina tabacum) and sugar crops), vegetables anda miscellaneous category which includes all minorcrops and green manure crops. In the model thesecrops compete for land within the agricultural area.No direct competition between individual crops andother land-cover types, such as built-up areas andgrassland, is taken into account because it is assumedthat agricultural activities compete with other land-useactivities, indifferent of the actual crops cultivated.The simulations do not account for specific croppingsequences and intercropping systems. At the scale ofanalysis the data represents a mixture of cropping se-quences and systems and does not allow the identifica-tion of individual cropping systems. In the simulationsto be presented it is assumed that the multiple crop-ping index, i.e., the number of times a piece of landis sown to crops during 1 year, as well as the spatialvariability of the cropping index are constant duringthe simulation period. For the allocation of changes inthe different crop types, simulations are only made for

Table 2Baseline scenario for CLUE simulation

Land-cover type Area at start of simulation,1991 (million ha)

Area at end of simulation,2010 (million ha)

Cultivated land 132 121Horticultural land 5 9Forest 195 197Grassland 256 256Built-up land 25 29Inland water 34 34Unused land (mainly desert) 296 297

grid-cells, where at least 5% of the land area is usedfor agricultural production in 1991. This effectivelylimits the simulation to about half of the area of China.

3.2. Scenarios

For the simulation of near-future, land-use patternsscenarios need to be formulated. Scenarios can beformulated at the national level, as well as at thesub-national level. At the national level, scenarios in-clude developments of different agricultural demandsthat can be determined on the basis of developmentsof consumption patterns, demographic characteris-tics, land-use policies and export volumes. At thesub-national level, different restrictions towards theallocation of land-use change can be implemented,e.g., the protection of nature reserves or land alloca-tion restrictions in areas susceptible to land degra-dation. Examples of different scenarios and theirimplementation for the land-use situation in Ecuadorare given by De Koning et al. (1999b).

4. Application of the methodology for China

4.1. Scenario formulation

A baseline scenario is formulated to represent themost likely developments in demography and demandfor land-use types and agricultural products in thenear-future, i.e., the period 1991–2010. The demandsfor the different land-use types are based upon an ex-trapolation of trends and estimates of land-use changeat the national level by Smil (1993) for the period1990–2000. The resulting land areas of the differentland-use types are shown in Table 2. It is assumed


that cultivated land will face the largest decrease inarea. About half of the decrease of cultivated land isexpected to originate from increase in built-up area(Smil, 1993) while other losses occur through con-version to horticultural lands, degradation into unusedland and rehabilitation of lands through reforestationor grassland restoration. Changes in demography dur-ing the simulation period are based upon projections ofpopulation growth (US Census Bureau, 1997), urbani-sation (United Nations, 1995), rural labour force (Shenand Spence, 1996) and trends in agricultural labourforce. These population projections are differentiatedthroughout the country based on observed growth ratesbetween 1986 and 1996. Also, income was assumedto grow steadily with growth rates proportional to thegrowth rates observed in the grid-cells between 1986and 1991.

The demand for the different crop types is based ona projection to 2007 by the United States Departmentof Agriculture (USDA, 1998) which provides yearlyprojections for all major crops. No projections aregiven for potato (included in the other grain category),some cash crops, vegetables and crops grouped in themiscellaneous category. The area given to potatoesis relatively small, and is assumed to stay constant.The same assumption is made for the two major cashcrops for which no projections are available, sugarcrops and tobacco. This assumption is consistent with

Fig. 3. Percentage of the harvested area allocated to different (groups of) crops for the baseline scenario.

the expectations of Lin and Colby (1996) who expectthat increasing sugar demand will mainly be fulfilledby imports rather than large extension of the areagrown to sugar crops.

China is expected to become a large vegetableexporter (Crook, 1996) because China will have acomparative advantage in producing vegetables forexport, primarily because of its abundant rural labourresources. As China invests in transportation andstorage infrastructure, and as firms improve grad-ing and packaging standards, China is likely to be-come a fierce competitor on world vegetable markets(Guoqiang, 1997; Lu, 1998). Domestic vegetable con-sumption is also expected to increase due to ratherhigh income elasticities for vegetables, especially inthe rural areas (USDA, 1998; Han and Wahl, 1998).Given these developments we have assumed a dou-bling of the area sown to vegetables between 1991 and2010.

These projections are proportionally scaled to matchthe scenario defined for the changes in total cultivatedarea. Fig. 3 summarises the scenario for the differ-ent crops types. Observed data have been used upto 1997, which can be seen from the fluctuations inrelative harvested area. Between 2007 and 2010 noprojections of USDA were available, a simple con-tinuation of the trends for these years was thereforeassumed.


Fig. 4. Cultivated area in 1991 and changes in cultivated area as simulated for the period 1991–2010.

4.2. Simulation results

The simulations for the baseline scenario are pre-sented in Figs. 4 and 5. Fig. 4 shows the spatialdistribution of cultivated land in 1991, the referenceyear, as derived from our database. The distributionof changes in cultivated area as simulated for theperiod 1991–2010 are indicated in the same figure.Simulations for other land-cover types are not shown.In Fig. 4, a clear pattern of decrease in cultivatedarea throughout China can be observed. Land degra-dation causes large losses of cultivated land in thenorthern part of China, mainly on the Ordos Plateauof inner Mongolia and on the Loess Plateau. Other‘hot spots’ of land-use change can be found in themain agricultural regions of east, central, south andsouthwest China as result of the expansion of thebuilt-up and horticultural area. Especially, the largerarea surrounding Shanghai is expected to face largedecreases in cultivated land.

Fig. 5 presents the results for the different crop typesor group of crops. Results for the group of miscella-neous crops have been omitted. The maps represent therelative share of the considered crop within the totallycropped area, i.e., a figure of 40% for rice means that atthat location 40% of all sown area is occupied by rice.

Rice is mainly grown in southern China, where hightemperatures and high precipitation favour the culti-vation practices. From the map some rice far up northin the North Eastern plain and in the Beijing-Tianjin

area can be distinguished. This rice cultivation doesnot follow the climatic subdivision of grain crop cul-tivation. In these areas special varieties are grown thatcan tolerate the cool climate. Yields are however low.Reasons for growing rice in these areas which are moreproductive for other grains, are mainly historic andbecause of consumer preference for the high qualityvarieties from this area. Because of the large demandfor these ‘good tasting’ varieties, prices are high andtherefore the cultivation is profitable in spite of thelow yields (Ren, 1991). Simulation results show thatthe decrease in demand for rice area is causing a de-crease in the share of rice within the cropping systemthroughout the whole rice growing area. Because nodistinction is made between rice varieties grown in thenorthern part of the country and in the southern part ofthe country, the impact that increasing incomes mighthave on the demand for the ‘northern varieties’ is notwell represented. Increasing consumer preference forthe northern rice varieties might keep rice productionin the northern part of China more important than an-ticipated by the model.

Wheat is traditionally grown in the valleys of theYellow river and the large North China floodplainof the same river. The increasing cultivation of cornand cash crops in these areas will decrease the im-portance of wheat in the cropping system, especiallyin the southern part of the wheat growing area. In alarge part of the wheat growing area labour-intensivecropping systems combining winter-wheat cultivation


Fig. 5. Percentage of the harvested area allocated to different (groups of) crops for 1991 and simulation results for 2010.


Fig. 5 (Continued).


and cash crop cultivation (e.g., partial intercroppingof winter-wheat and cotton) are found. Increasingcash crop cultivation for the domestic and interna-tional market will make cash crop cultivation the mostprofitable part of the cropping system.

Corn is mainly grown in three regions, namelythe spring-sowing maize belt in the north; thesummer-sowing maize belt in the plains of the Yellowriver; and the southern maize belt in the upland ar-eas of the provinces Sichuan, Guizhou, Guangxi andYunnan. In 1994, 64% of all produced maize wasused for animal feeds, while 20% was being used fordirect consumption. The remaining 16% was used forseed, processing, other uses and wastage (Guoqiang,1997). Very fast growth of the livestock populationis responsible for the fast growth in demand forcorn (Simpson et al., 1994; Verburg and van Keulen,1999). Simulations show large increases in the shareof corn production in the southern part of China, alsoin areas where corn is presently almost not cultivated.This increase in corn production in these areas seemsrather realistic as in these areas a large number of pigfattening farms is found. Results for cash crops showthat in a large part of the agricultural area of Chinacash crops are going to make up a substantial part ofthe cropping system. Of course, large differences inclimatic conditions and resource endowments causelarge regional differences in the type of cash crops.In the scenario large increases were defined for thevegetable area. Most of these increases are allocatedin the southern part of China. Recent developments,however, show increases in vegetable area in otherparts of eastern China, mainly around the major urbancentres (Guoqiang, 1997). The model might thereforeoverestimate increases in the southern part becauseof the historic occurrence of vegetables in this partof China, which determines the spatial relation be-tween land use and its explanatory factors used in themodel. Fastly growing vegetable producing zones,as designated by the Chinese Government, include(in correspondence to the simulation results) the sixsouthern provinces, where vegetables are grown forconsumption in the northern provinces; the winter andspring vegetable zone in the floodplain of the Yellowriver (not well represented in the simulation results);and the low season vegetable zone in north China.

A validation of the performance of the CLUEmodel has been made for a number of countries (De

Koning et al., 1999a; Verburg et al., 1999c; Kok et al.,2001) and resulted in confidence in the simulation ofland-use change dynamics. Unfortunately, validationof the application for China has not been possible dueto the absence of an independent, comparable data seton land use for China. Therefore, all interpretationsof modelling results have to be made carefully. Cau-tion with interpretations of modelling results wouldstill hold when the model for China would have beenvalidated, because high levels of uncertainty are com-mon in the modelling of complex systems integratinghuman–environment interactions.

5. Discussion

The methodology presented in this paper fits wellwithin the land-use change research sequence pre-sented in Fig. 1. The identified niche of the method-ology is regional with respect to the scale of studyand belongs to the ‘system description phase’ of theresearch sequence. It directly links up with studiesin earlier phases of the research sequence: trends inland demand, projected within studies that belongto the ‘problem identification phase’ of the researchchain, are direct input for the simulations throughthe demand module. Resulting shifts in croppingpatterns are useful inputs for the ‘design phase’.Studies in the ‘design phase’ commonly result instatic realisations of optimised land-use configura-tions (obtained by linear programming models) thatcan only be realised in a distant future (Fig. 6). Thepresented methodology in this paper results in tra-jectories of land-use change that are expected to takeplace more or less autonomous under the conditionsdefined in a scenario. Comparing these developmentswith the designed (optimised) land-use configurationshelps to indicate locations and conditions that con-strain the implementation of the designed land-usealternatives. Evaluation of model runs for differentscenarios, including different land-use policies, in-dicate which conditions lead to the desired land-usechange trajectories. Furthermore, this confrontationwith near-future developments might lead to a morerealistic definition of the objectives of the linearprogramming and prototyping models. Identified‘hot-spots’ of land-use change can also help to focusin-depth research belonging to the design phase to the


Fig. 6. Schematic representation of the differences and linkages between studies in the system description phase and the design phase.

areas and land-use systems that are facing the highestdynamics.

The nested simulations for different crop types area new feature within the CLUE modelling frame-work. In the previous applications, only land-covertypes (Veldkamp and Fresco, 1996; De Koning et al.,1999a; Verburg et al., 1999b) or a limited numberof very dominant crop types (Kok et al., 2001) weresimulated. The simulation of individual crops makesit easier to communicate, discuss and compare resultswith economist and agronomist, who mostly focus onindividual commodities or crops instead of land-usetypes. Economic models also often include changes intechnology and productivity of agricultural land-usesystems (Jones and O’Neill, 1992). In the presentversion of CLUE, changes in these conditions areonly included in the scenarios defined at the nationallevel (demand module). Including the dynamic, spa-tially explicit simulation of productivity change intothe modelling framework will certainly improve thecomparison and linkage of the modelling results withother modelling approaches. A preliminary analysis ofproductivity change in China is provided by Verburget al. (2000).

Apart from linking up different stages of researchit is also needed to link up different scales of analysis.

The presented method uses a multi-scale approach forland-use allocation (Verburg et al., 1999a). Most de-tailed units of analysis in China, however, have a sizeof 32 × 32 km, which is a fine resolution when com-pared to studies for China on the national or provinciallevel, but very coarse as compared to local studies atthe farming system or village level. This means that lo-cal driving factors, bottom-up effects and interactionsare not well represented in this study. On the otherhand, most local level studies disregard the impact ofprocesses and feed backs occurring at higher levelsof scale. Therefore, higher-scale systems, as analysedin this paper, should be represented in local scalestudies. That should at least guarantee that importantchanges in the context of the situation are taken intoaccount and that the system does not simply exportits problems to its neighbours (Musters et al., 1998).Analysis and understanding of complex systems oversuch a wide range of scales is still problematic and anissue of discussion (Gibson et al., 2000; Müller, 1992;Caldwell and Fernandez, 1998). The presented methodis a starting point for such multi-scale research be-cause it already covers a relatively wide range ofscales. The further understanding of scalar dynamicsis only possible after an integration of the presented(macro-level) approach with local (micro-level)


studies. Such integration also means that different re-search philosophies need to be combined. Micro-levelapproaches have their strength in the explanation of theprocesses leading to differences in behaviour. Straight-forward upscaling of such explanations is impossibledue to scaling properties, including non-linearity(Rastetter et al., 1992) and emergent properties (e.g.,collective action, Ostrom, 1990; Gibson et al., 2000).Macro approaches are often criticised for their deduc-tive approach leading to empirical relations betweenland use and its driving factors of which the causalitycannot be guaranteed. Therefore, an improved linkageof micro- and macro-studies will not only improve therange of scales studied but will, at the same time, im-prove the understanding of the processes and patternstaking place at different levels of analysis (Mertenset al., 2000). This type of understanding of the scalardynamics of land-use change is also urgently neededto better link up with studies in the design and im-plementation phases. When stakeholders or plannerscannot steer specific characteristics of the land-usesystem at one scale level, it does not necessarilymean that they cannot be steered at other scale lev-els. Therefore, awareness of the scalar dynamics ofland-use change will improve land-use planning.

6. Conclusions

• The CLUE methodology can provide important in-formation needed for land-use planning in additionto insights obtained by existing research metho-dologies.

• The inclusion of crop-specific simulations withinthe CLUE methodology enables a linkage of thismethodology with crop suitability analysis and lin-ear programming studies that aim at the design ofalternative land-use configurations.

• The multi-scale approach used is useful for studyingcomplex systems such as the land-use system.

Acknowledgements

The research described in this paper is financed bythe Netherlands Organisation for Scientific Research(NWO/NOP-II). All persons and institutes who kindlymade their data available for this analysis are acknow-ledged.

References

Achard, F., Eva, H., Glinni, A., Mayaux, P., Richards, T., Stibig,H.J., 1998. Identification of deforestation hot spot areas inthe humid tropics. TREES Publication Series B. ResearchReport No. 4. Joint Research Centre, European Commission,Luxembourg.

Alexandratos, N., 1997. China’s consumption of cereals and thecapacity of the rest of the world to increase exports. FoodPolicy 22 (3), 253–267.

Barbier, B., 1998. Induced innovation and land degradation: resultsfrom a bioeconomic model of a village in West Africa. Agric.Econ. 19, 15–25.

Bouma, J., 1998. Introduction. In: Stoorvogel, J.J., Bouma, J.,Bowen, W.T. (Eds.), Proceedings of an International Workshopon Information Technology as a Tool to Assess Land UseOptions in Space and Time, Lima, September 28–October 4,1997. Quantitative Approaches in Systems Analysis No. 16,Wageningen.

Bouman, B.A.M., Nieuwenhuyse, A., 1999. Exploring options forsustainable beef cattle ranching in the humid tropics: a casestudy for the Atlantic Zone of Costa Rica. Agric. Syst. 59,145–161.

Brown, L.R., 1995. Who Will Feed China? Wake-up Call for aSmall Planet. W.W. Norton, New York.

Brown, L.R., Kane, H., 1994. Full House: Reassessing the Earth’sPopulation Carrying Capacity. W.W. Norton, New York.

Caldwell, R.M., Fernandez, A.A.J., 1998. A generic model ofhierarchy for systems analysis and simulation. Agric. Syst.57 (2), 197–225.

Costanza, R., Ruth, M., 1998. Using dynamic modeling to scopeenvironmental problems and build consensus. Environ. Mgmt.22 (2), 183–195.

Crook, F.W., 1996. The development of China’s vegetablemarkets. International Agriculture and Trade Report, China.Doc. RS-96-2. ERS–USDA, Washington, DC.

De Koning, G.H.J., Verburg, P.H., Veldkamp, A., Fresco, L.O.,1999a. Multi-scale modelling of land use change dynamics inEcuador. Agric. Syst. 61 (2), 77–93.

De Koning, G.H.J., Veldkamp, A., Fresco, L.O., 1999b. Exploringchanges in Ecuadorian land use for food production and theireffects on natural resources. J. Environ. Mgmt. 57, 221–237.

Fan, S., Agcaoili-sombilla, M., 1997. Why projections on China’sfuture food supply and demand differ? Aust. J. Agric. Res.Econ. 41 (2), 169–190.

Fresco, L.O., 1995. Agro-ecological knowledge at different scales.In: Bouma, J., Kuyvenhoven, A., Bouman, B.A.M., Luyten,J.C. (Eds.), Eco-regional Approaches for Sustainable Land Useand Food Production. Proceedings of the symposium at ISNAR,The Hague, December 1994. Kluwer Academic Publishers,Dordrecht, The Netherlands.

Gibson, C., Ostrom, E., Ahn, T.-K., 2000. The concept of scaleand the human dimensions of global change: a survey. Ecol.Econ. 32 (2), 217–239.

Guoqiang, C., 1997. Market prospects for upland crops in China.Working Paper Series No. 24. CGPRT Centre, Bogor.


Hall, C.A.S., Tian, H., Qi, Y., Pontius, G., Cornell, J., 1995. Mode-lling spatial and temporal patterns of tropical land use change.J. Biogeog. 22, 753–757.

Han, T., Wahl, T.I., 1998. China’s rural household demand forfruit and vegetables. J. Agric. Appl. Econ. 30 (1), 141–150.

Heilig, G.K., 1997. Anthropogenic factors in land-use change inChina. Popul. Dev. Rev. 23 (1), 139–168.

Imbernon, J., 1999. Changes in agricultural practice and landscapeover a 60-year period in North Lampung, Sumatra. Agric.,Ecosyst. Environ. 76, 61–66.

Jones, S., 1999. From meta-narratives to flexible frameworks: anactor level analysis of land degradation in highland Tanzania.Global Environ. Change 9, 211–219.

Jones, D.W., O’Neill, R.V., 1992. Endogenous environmentaldegradation and land conservation: agricultural land use in aregion. Ecol. Econ. 6, 79–101.

Kok, K., Farrow, A., Veldkamp, A., Verburg, P.H., 2001. A methodand application of multi-scale validation in spatial land usemodels. Agric. Ecosyst. Environ. 85, 223–238.

Lambin, E.F., Baulies, X., Bockstael, N., Fischer, G., Krug, T.,Leemans, R., Moran, E.F., Rindfuss, R.R., Sato, Y., Skole,D., Turner, B.L., Vogel, C., 1999. Land-use and land-coverchange (LUCC): implementation strategy. A core projectof the International Geosphere–Biosphere Programme andthe International Human Dimensions Programme on GlobalEnvironmental Change. IGBP Report No. 48 and IHDP ReportNo. 10. IGBP, Stockholm, 125 pp.

Levin, S.A., 1992. The problem of pattern and scale in ecology.Ecology 73 (6), 1943–1967.

Lin, J.Y., 1998. How Did China Feed Itself in the Past? HowWill China Feed Itself in the Future? Second DistinguishedEconomist Lecture. CIMMYT, Mexico, DF.

Lin, W., Colby, H., 1996. China brings volatility to the world sugarmarket. China, Situation and Outlook Series. Report WRS-96-2.ERS–USDA.

Lu, F., 1998. Grain versus food: a hidden issue in China’s foodpolicy debate. World Dev. 26 (9), 1641–1652.

Mertens, B., Sunderlin, W.D., Ndoye, O., Lambin, E.F., 2000.Impact of macroeconomic change on deforestation in SouthCameroon: integration of household survey and remotely senseddata. World Dev. 28 (6), 983–999.

Moxey, A., Witby, M., Lowe, P., 1998. Agri-environmental indi-cators: issues and choices. Land Use Policy 15 (4), 265–269.

Müller, F., 1992. Hierarchical approaches to ecosystem theory.Ecol. Model. 63, 215–242.

Musters, C.J.M., de Graaf, H.J., ter Keurs, W.J., 1998. Definingsocio-environmental systems for sustainable development. Ecol.Econ. 26, 243–258.

Ostrom, E., 1990. Governing the Commons: The Evolution ofInstitutions for Collective Action. Cambridge University Press,Cambridge.

Rastetter, E.B., King, A.W., Cosby, B.J., Hornberger, G.M.,O’Neill, R.V., Hobbie, J.E., 1992. Aggregating fine-scaleecological knowledge to model coarser-scale attributes ofecosystems. Ecol. Appl. 2 (1), 55–70.

Ren, C., 1991. The components of agriculture. In: Guohua, X.,Peel, L.J. (Eds.), The Agriculture of China. Oxford UniversityPress, Oxford, UK.

Rindfuss, R.R., Stern, P.C., 1998. Linking remote sensing andsocial science: the need and the challenges. In: Liverman, D.,Moran, E.F., Rindfuss, R.R., Stern, P.C. (Eds.), People andPixels, Linking Remote Sensing and Social Science. NationalAcademy Press, Washington, DC.

Rudel, T.K., Horowitz, B., 1993. Tropical Deforestation. SmallFarmers and Land Clearing in the Ecuadorian Amazon.Colombia University Press, New York.

Shen, J., Spence, N.A., 1996. Modelling urban–rural populationgrowth in China. Environ. Plann. A 28, 1417–1444.

Simpson, J.R., Cheng, X., Miyazaki, A., 1994. China’s Live-stock and Related Agriculture. Projections to 2025. CABInternational, Wallingford.

Smil, V., 1993. China’s Environmental Crisis: An Inquiry into theLimits of National Development, M.E. Sharpe, New York.

Turner II, B.L., Moss, R.H., Skole, D., 1993. Relating land useand global land-cover change: a proposal for an IGBP–HDPcore project. IGBP Report No. 24/HDP Report No. 5.

United Nations, 1995. World urbanization prospects: the 1994revision. United Nations Department for Economic and SocialInformation and Policy Analysis, Population Division, NewYork.

US Census Bureau, 1997. International database. http://www.census.gov.

USDA, 1998. International agricultural baseline projections to2007. Market and trade. Agricultural Economic Report No. 767.United States Department of Agriculture, Economics Division,Economic Research Service.

Van Ittersum, M.K., Rabbinge, R., Van Latesteijn, H.C., 1998.Exploratory land use studies and their role in strategic policymaking. Agric. Syst. 58, 309–330.

Veldkamp, A., Fresco, L.O., 1996. CLUE-CR: an integratedmulti-scale model to simulate land use change scenarios inCosta Rica. Ecol. Model. 91, 231–248.

Verburg, P.H., Chen, Y.Q., 2000. Multi-scale characterization ofland-use patterns in China. Ecosystems 3, 369–385.

Verburg, P.H., van Keulen, H., 1999. Exploring changes in thespatial distribution of livestock in China. Agric. Syst. 62 (1),51–67.

Verburg, P.H., de Koning, G.H.J., Kok, K., Veldkamp, A., Bouma,J., 1999a. A spatial explicit allocation procedure for modellingthe pattern of land use change based upon actual land use.Ecol. Model. 116, 45–61.

Verburg, P.H., Veldkamp, A., Fresco, L.O., 1999b. Simulation ofchanges in the spatial pattern of land use in China. Appl. Geogr.19, 211–233.

Verburg, P.H., Veldkamp, A., Bouma, J., 1999c. Land use changeunder conditions of high population pressure: the case of Java.Global Environ. Change 9 (4), 303–312.

Verburg, P.H., Chen, Y.Q., Veldkamp, A., 2000. Spatialexplorations of land-use change and grain production in China.Agric. Ecosyst. Environ. 82, 333–354.

Vereijken, P., 1997. A methodological way of prototypingintegrated and ecological arable farming systems (I/EAFS) ininteraction with pilot farms. Eur. J. Agron. 7, 235–250.

Zander, P., Kächele, H., 1999. Modelling multiple objectivesof land use for sustainable development. Agric. Syst. 59 (3),311–325.

the role of spatially explicit models in land-use change ... · the role of spatially explicit...

Documents