Chapter 18

DYNAMIC SIMULATION OF LAND-USE

CHANGE TRAJECTORIES WITH THE CLUE-s
MODEL

P.H. Verburg and K.P. Overmars

Department of Environmental Sciences, Wageningen University, The Netherlands

Abstract: The CLUE (Conversion of Land Use and its Effects) model is one of the most
widely applied models with approximately 30 applications in different regions
of the globe focusing on a wide range of land-use change trajectories including
agricultural intensification, deforestation, land abandonment and urbanisation.
The model is a tool to better understand the processes that determine changes
in the spatial pattern of land use and to explore possible future changes in land
use at the regional scale. This chapter describes the functioning of the model
and illustrates the potential of the model for scenario-based simulation of land-
use change trajectories with two case studies, one which is a rural landscape in
the eastern part of the Netherlands and one which is a strongly urbanized
watershed surrounding Kuala Lumpur in Malaysia.

Key words: Land use; modelling; competition; Achterhoek; allocation; cellular automata;
Kuala Lumpur.

1. INTRODUCTION

The CLUE (Conversion of Land Use and its Effects) modelling

framework was developed at Wageningen University and first published in
1996 (Veldkamp and Fresco, 1996). The model is now one of the most
widely applied models with approximately 30 applications spread over the
different regions of the globe addressing a wide range of land-use change
trajectories including agricultural intensification, deforestation, land
abandonment and urbanisation. The model is a tool to better understand the

E. Koomen et al. (eds.), Modelling Land-Use Change, 321–335.

© 2007 Springer.
322 P.H. Verburg and K.P. Overmars

processes that determine changes in the spatial pattern of land use and
explore possible future changes in land use at the regional scale. The
methodology links the spatially explicit analysis of relations between land
use and its driving factors to a dynamic simulation technique to explore
changes in land use under scenario conditions.
Two different versions of the CLUE model have been developed to
account for differences in the structure of land-use data. Traditionally, land-
use data for study areas with a relatively small spatial extent are based on
land-use maps or remote sensing images that denote land-use types
respectively by homogeneous polygons or classified pixels. This results in a
land-use map in which only the dominant land-use type occupies each unit
of analysis. The validity of this data representation depends on the spatial
and categorical resolution. In a landscape with high spatial variability, e.g. a
small scale agricultural landscape with patches of natural vegetation, such a
representation is only valid if the data have a high spatial resolution (Moody
and Woodcock, 1994; Schmit et al., 2006). If high resolution data are not
available or a high spatial resolution is not feasible due to, e.g. a very large
extent of the study area, it is better to represent land use in a so-called ‘soft-
classified’ system (Pontius and Cheuk, 2006). In such a system, each spatial
unit, either a polygon or pixel, denotes the fraction of the different land-use
or land-cover types. The data can be generated through aggregation of fine
resolution data or based on a sample of high-resolution information such as
agricultural census data.
The CLUE model was originally developed to use soft-classified data and
the earlier applications were all based on census data for land use (Veldkamp
and Fresco, 1996; de Koning et al., 1999; Verburg et al., 1999) while the
extent was national (Ecuador, Costa-Rica, China) to supra-national (Central-
America) with spatial resolutions varying between 7x7 kilometres to 32x32
kilometres. Another land-use model that uses soft-classified data is the Land
Use Scanner, which was specifically developed for the Netherlands
(Hilferink and Rietveld, 1999). For the identification of critical areas of
land-use change, this data representation is sufficient. However, when the
impacts of land-use change on landscape pattern need to be assessed, a high-
resolution representation of land use with homogeneous spatial units is more
appropriate. This data representation also provides the possibility to
visualize the spatial pattern of various land-use types in a single map.
Furthermore, for impact assessment, it is often needed to exactly know the
characteristics of the location of change, e.g. the erosion potential after a
change in land use depends on the location of the land-use change.
Therefore, a modified version of the CLUE model, the CLUE-s model, was
developed that is based on high-resolution data in which each pixel only
contains one land-use type (Verburg et al., 2002; Verburg and Veldkamp,
18. Land-use change trajectories with the CLUE-s model 323

2004a; Overmars et al., 2007). This version has mainly been used in sub-
national case studies with a local to regional extent and a resolution ranging
from 20 to 1,000 metres. However, since computer capacity and data
availability have improved considerably during the last decade, it is now also
possible to simulate high-resolution land-use changes for larger areas.
Examples of applications of the model at the continental scale include the
studies for the European Union (Verburg et al., 2006) and Latin America
(Wassenaar et al., 2006).
This chapter describes the functioning of the CLUE-s model and provides
insights into the range of possible applications. The functioning of the model
is illustrated based on two newly developed applications of the model in
respectively the eastern part of the Netherlands and the Klang-Langat
watershed in Malaysia.

2. MODEL DESCRIPTION

The modelling framework is based on the dynamic simulation of spatial

patterns of land-use change in reaction to pre-defined changes in demand for
land by different sectors (e.g. agriculture, wood harvesting). Figure 18-1
provides a schematic representation of the functioning of the modelling
framework. During each time step, the model determines for each location
(grid cell) the most preferred land use based on a combination of the
suitability of the location itself and the competitive advantage of the
different land-use types, which is a function of the land requirements (or
demand). If the most preferred land-use type requires a land-use conversion
which is not realistic or not allowed due to spatial policies and restrictions,
the next most preferred land-use type is selected. Spatial policies, restrictions
and a matrix listing which conversions are possible need to be specified by
the user in advance (Verburg and Veldkamp, 2004a). After allocating the
preferred land use to all locations in the study area, the aggregate demand for
land use is compared with the allocated areas. If the demand is not correctly
allocated, the competitive advantage of the different land-use types is
modified in such a way that the land uses for which the demand was not met
obtain a higher preference. Overrepresented land uses get a lower
preference. This procedure is repeated iteratively until the demand equals the
allocated area. The demand for the different land-use types needs to be
specified by the user in advance. The calculation of the changes in demand is
exogenous to the CLUE-s modelling framework and can be based on
different techniques ranging from simple trend extrapolation to advanced
multi-sectoral modelling.
324 P.H. Verburg and K.P. Overmars

The algorithm of the CLUE-s model is applicable to different scales and

capable of simulating different types of land-use change trajectories. The
actual configuration of the model is dependent on the case study, the
dominant land-use change processes and the available information and data.
Especially the specification of the location specific preference is different for
each case study. The location specific preference is determined by a
selection or combination of four different methods (Figure 18-1).
(1) In case information on the (proximate) determinants of land-use
change is lacking or if the scale of application is much coarser than the scale
of the understanding of the land-use change processes, empirical methods
are used. The empirical analysis is used to estimate the contribution of
different location characteristics (explanatory variables), such as soil
conditions and accessibility, to the suitability of a location for a specific
land-use type. The possible determinants are selected based on either theory
or knowledge of the study area and the relations are quantified with logit
models (Verburg et al., 2004d). Upon a change in the value of one of the
determinants (e.g. improved accessibility), the new location preferences can
be calculated. The empirical specification of location preference does not
necessarily lead to causal relations and assumes that the relations based on
the current land-use pattern remain valid during the simulation period.
(2) In cases with more information on the determinants of the location
preference of land-use types, the empirical analysis can be replaced by
decision rules that reflect the knowledge of the processes. These decision
rules should specify the relative preference of the different land-use types for
each location as a function of the land-use history, biophysical and/or socio-
economic conditions of the location (Overmars et al., 2005).
(3) Land-use data often show spatial autocorrelation (Anselin, 2002;
Munroe et al., 2002). This can be a result of trends in the underlying factors
e.g. the land tenure structure or spatial processes like agglomeration effects
in residential land use and imitation among farmers. Spatial interactions can
be represented in the model by the incorporation of a neighbourhood
function as a determinant of the location preference. The relation between
the location preference and the neighbourhood composition is determined
either by (calibrated) decision rules, as in most cellular automata models, or
based on statistical analysis (Verburg et al., 2004c). At each time step of the
analysis, the location preferences need to be updated to account for changes
in land-use structure.
(4) Finally, current land use is in many cases an important determinant of
the location preference. Land conversions are often costly, e.g. after
establishing a plantation of fruit trees, the owner will not consider to ‘move’
his plantation to a nearby location due to the high investment costs involved.
Other conversions are almost irreversible, e.g. residential area is not likely to
18. Land-use change trajectories with the CLUE-s model 325

be converted back into agricultural area. Therefore, conversion elasticity is

assigned to each land-use type, which increases the preference for locations
where this land-use type is found in the current situation. During the
simulation, these preferences are updated to account for the simulated
changes in land use.

Figure 18-1. Overview of the CLUE-s modelling procedure.

The final location specific preference used in the simulations is

determined by a mix of these four components based on the preferences of
the model user and the available knowledge and data. This mix can be
different for each land-use type and each application. This makes the
326 P.H. Verburg and K.P. Overmars

modelling framework very flexible and enables configurations that classify

the model as either a (constrained) cellular automata model, an empirical-
statistical model or a dynamic simulation model (Lambin et al., 2000).

3. CASE STUDY 1: THE ‘ACHTERHOEK’ REGION

IN THE NETHERLANDS

This case is a study in the framework of a research project “Sustainable

multiple functions in the rural countryside”, which is jointly undertaken by
WUR (Wageningen University and Research Centre, Netherlands) and
INRA (l’Institute National de la Recherche Agronomique, France). The
’
study area comprises four municipalities in de Achterhoek’ in the
Netherlands and covers approximately 420 km2. The main land-use type in
this rural region is dairy farming. Currently, the Government is purchasing
areas for nature conservation and subsidizing farmers to use farming regimes
favourable for nature conservation.
In this case study, the CLUE-s model is used to compare two scenarios.
Both scenarios are based on the same quantity of land-use change, but differ
with respect to the assumed spatial policies that influence land allocation.
The timeframe for the simulations is from 2000 to 2018 with yearly time
steps and a spatial resolution of 50x50 metres. Initial land use is presented in
Figure 18-2. Several land-use types in the area are considered to have high
values for biodiversity and landscape conservation. The areas classified as
nature consist of forests and other natural vegetations; management areas are
agricultural lands that are managed to increase the natural value; reserves are
newly purchased areas which are actively managed to create conditions for
the development of natural vegetations.
The land requirements (i.e. demands) are determined based on the
objectives of the provincial authorities regarding the improvement of the
ecological quality of the area and include an overall decrease in agricultural
area and increases in the land-use types nature, management, reserves and
residential.
For the land-use types management, reserve, nature and residential, the
location characteristics are determined by calculating the location specific
preferences for each land-use type based upon a logistic regression with the
current land use as dependent and location factors as independent variables.
The factors included in the analysis to explain the location preference are the
distance to highway, distance to provincial road, distance to cities, distance
to creeks and streams, slope, elevation, soil characteristics and ground water
level. For the land-use type agriculture, no logistic regression analysis was
made, instead the preference of each location for agriculture (mainly
18. Land-use change trajectories with the CLUE-s model 327

grassland in this region) was assumed to be dependent on the production

potential for grass derived from a quantitative land evaluation system.
In addition to the derived location preferences, the following
neighbourhood functions were included: land-use changes into management
and reserve are considered to be more preferred in the neighbourhood of
other management, reserve and nature areas. New nature areas are
considered to be preferred close to other nature areas while new residential
areas will most likely be positively influenced by the quantity of other
residential areas in the neighbourhood. The neighbourhood values make up
10% of the total location preference used in the simulations.
The difference between the two scenarios in this study is the
incorporation of spatial policies. In scenario 1, no spatial policies were
included. The spatial policy used in scenario 2 is based on the EHS
(Ecological Main Structure) that aims to develop nature preferably at
locations that enable connections between existing natural areas and thus
reduce fragmentation of nature. This spatial policy cannot be implemented
as a strict rule since it is often implemented as higher efforts to purchase
land or higher subsidies for changes in farm management. Therefore, in
scenario 2, the location preference for the land-use types management,
reserve and nature are increased in the designated EHS areas.
The results of this analysis are presented in Figure 18-2 by maps that
only show the areas that changed between 2000 and 2018. The designated
EHS areas of scenario 2 are indicated as well. The changes to reserve in
scenario 2 are all located within the EHS areas. In area A, conversions to
reserve area are dominant in scenario 1, whilst in scenario 2, there is hardly
any change in this area. On the other hand, in area B, changes into nature are
found in both scenarios. Area C does not show changes in both scenarios.
This may indicate that it is relatively difficult and costly to achieve changes
in this region. Area D shows changes to more natural land use types in both
scenarios. Establishing a corridor in this region is probably much easier as
compared to area C.
The results visualize an important issue for policy making concerning the
selection of locations to target subsidies and spatial policies. Given the
trends in land use, some areas within the designated EHS area will change
into reserve or management even without explicit spatial policies or support.
Other areas only change when the policies are implemented and some areas
will not change anyway. For these areas higher subsidies can be considered.
Another option is to target resources to other locations where change is
achieved more easily. Herewith, the model can be used to assess, discuss and
fine-tune policies.
328 P.H. Verburg and K.P. Overmars

Figure 18-2. Observed land use in the study area in 2000 and land-use change as result of
scenario 1 and scenario 2.

4. CASE STUDY 2: THE KLANG-LANGAT

WATERSHED IN MALAYSIA

The Klang-Langat Watershed is located in the mid-western part of

Peninsular Malaysia and covers eight administrative districts of Selangor,
Negeri Sembilan and the Federal Territory of Kuala Lumpur. The Klang-
Langat Watershed represents the most highly urbanized region of the
18. Land-use change trajectories with the CLUE-s model 329

country. Kuala Lumpur, the capital city of about 1.5 million people is
located at the confluence of Klang and Gombak rivers. The suburbs of Kuala
Lumpur extend southward to the coast of Selangor and westward into the
Langat river basin. The spillover development of Kuala Lumpur has spread
into the surrounding districts of Selangor which have become highly
urbanised. The whole region has about 4.2 million inhabitants. Some of the
urban developments have been located in abandoned mining areas. The area
was once a centre for tin mining activities. The study area has several
patches of upland forest, lowland forest and mangrove forest. Besides the
built-up areas and forests, the other major land use type in the area is
agriculture. The main agricultural crops are oil palm, natural rubber, coffee,
cocoa and coconut. Oil palm and rubber estates have been established
through forest conversion and currently, many of these estates are in their
turn converted into urban, residential, recreational and industrial areas. The
Klang-Langat Watershed is also vital in terms of domestic water supply to
the most densely populated area in the country.
The CLUE-s model application for this region was made to test different
frequently used land-use allocation mechanisms in a region with opposing
agricultural and urban developments. Eight different land-use types were
considered and two different allocation mechanisms for urban land were
tested. Both applications related to the period 1990-1999, a period during
which the area of urban land expanded very rapidly (Figure 18-3).
Simulation 1 determines the location of change as a function of the
neighbouring land uses. Based on observed land-use patterns, a set of
neighbourhood rules was derived that define the location specific
preferences for urban land. These preferences are updated each year, which
makes the model function similar to a constrained cellular automata model.
In simulation 2, another approach is chosen to determine the location
specific preferences for urban area. Here a logistic regression analysis based
on the location of urban land in 1990 is used to quantify relations between
the location of urban land and location characteristics such as accessibility
and slope.
The results were quantitatively compared with the observed land-use
change during the 1990-1999 period using the multi-resolution method for
map comparison (Costanza, 1989; Pontius et al., 2004). The results were
also compared with two different reference models: the first assumes a
complete persistence of land use at the 1990 situation and the second
assumes random allocation of the observed changes. Both simulations with
CLUE-s performed better than both reference models. Simulation 2 was
much better able to predict the spatial pattern of urban development as
compared to simulation 1. Obviously, the neighbourhood relations only
330 P.H. Verburg and K.P. Overmars

capture part of the processes that govern urban expansion, while the
empirical approach based on location conditions was better able to capture
the full complexity of the involved processes. The estimated preferences for
urban areas, related to the land-use pattern of 1990, apparently offered a
reasonable prediction for the processes over the dynamic period 1990-1999.

Figure 18-3. Observed land use in the Klang-Langat watershed in 1990 and 1999 and
simulated land use for 1999 using two different algorithms.

The case study is an illustration of the use of the model to test different
representations of the land-use change processes in a modelling environment
and evaluate the outcomes. Such analysis may inform further research on the
processes of land-use change and on the techniques needed to represent
these processes in models.
18. Land-use change trajectories with the CLUE-s model 331

5. DISCUSSION

This chapter described one particular tool for land-use change modelling.
However, this land-use model is illustrative for a much larger group of land-
use change models. Approximately one decade ago, the choice of land-use
models was fairly small and most models were specifically developed for a
particular case study. Currently, a much larger selection of land-use models
is available (Briassoulis, 2000; Agarwal et al., 2001; Verburg et al., 2004b).
In spite of conceptual differences, there is a tendency to develop models as
frameworks in which the user can specify the representation of the land-use
processes considered. The recent version of the CLUE-s model presented in
this chapter is representative of this trend: the user can decide on the
relevance of either empirically derived relations between location factors
and land use, neighbourhood rules or the use of location preference maps
derived from theoretical or observed processes of land-use change. There
has always been a relatively strict subdivision between urban and rural
models of land-use change. The modelling procedures used in urban and
rural models were distinctly different: urban models were dominantly based
on neighbourhood functions such as cellular automata and infrastructure
access (White et al., 1997; Clarke and Gaydos, 1998; Agarwal et al., 2001;
Dietzel et al., 2005) while rural land-use models primarily used land-quality
assessments as the basis of simulation (Veldkamp and Fresco, 1997; Alcamo
et al., 1998; Pontius et al., 2001). The possibility to combine different
procedures for land allocation within one modelling framework allows for a
more integrated assessment of land-use dynamics that can explicitly address
the competition between land uses in the urban fringe (Dietzel and Clarke,
2006). The development of such integrated approaches also facilitates the
collaboration of urban and rural modellers with different expertise and
complementary knowledge on the land-use system.
In many spatially explicit models, such as the CLUE-s model, the unit of
analysis is an area of land, either a polygon representing a field, plot or
census track, or a pixel as part of a raster-based representation. The
disadvantage of this ‘land-based’ approach is the poor match with the agents
of land-use change. Individual farmers or plot owners are usually not
represented explicitly and the simulations usually do not match with the
units of decision making. A rapidly expanding group of models use
individual agents as units of simulation (Parker et al., 2003; Bousquet and
Le Page, 2004). These so-called multi-agent systems emphasize the
decision-making process of the agents and the social organization and
landscape in which these individuals are embedded. An agent can represent
any level of organization (a herd, a village, an institution), and does not
332 P.H. Verburg and K.P. Overmars

necessarily have to be an individual. A disadvantage of the agent as the basic

unit of simulation is the difficulty to adequately represent agent behaviour
and to link it to the actual land areas. Well established multi-agent models to
predict changes for real landscapes have become available only recently
(Huigen, 2004; Castella et al., 2005). Such models are very data demanding
and the specification of realistic agent behaviour and diversity is very
challenging, especially when it comes to the behaviour of agents at higher
levels of organizations such as institutions. The choice of either a pixel-
based or an agent-based model is therefore dependent on the objectives of
the study and the available information and resources. In many cases, it may
even be most appropriate to use different model approaches to study the
same region. Comparing the outcomes of such models may lead to a better
and more complete understanding of the system dynamics and inform each
other (Castella et al., 2007). Agent-based models can explore mechanisms
that can, later on, be included in spatial simulation models.
It is important to acknowledge that no single technique or approach can
sufficiently describe the different processes at all spatial and temporal scales.
Recent validation studies have indicated that most spatial models still
contain a high level of uncertainty while multi-agent models sometimes have
not been validated at all (Pontius et al., 2004; Pontius et al., 2007). The wide
selection of models and modelling approaches that has become available
provides the researcher with the opportunity to select the modelling
approach that best fits the research questions and characteristics of the study
area. The CLUE-s model enables the user to test different mechanisms of
simulation in one single modelling framework. The model can easily be
implemented in new case study areas and is used in policy support studies
(Klijn et al., 2005). However, the model is first and foremost a tool to
support the analysis of land-use processes and interactions and evaluate
different mechanisms of land allocation. Therefore, it is not possible to
classify this new generation of ‘hybrid’ models into a single model category
as proposed by Lambin (1997; 2000), Agarwal et al. (2001) or Briassoulis
(2000). It is certain that CLUE and other modelling approaches will see more
development and the implementation of new modelling concepts in the
future. However, it is important to keep in mind that applications of such
modelling frameworks will only be successful if a good balance between
transparency of the model and representation of the complexity of the land-
use system is achieved.
18. Land-use change trajectories with the CLUE-s model 333

ACKNOWLEDGEMENTS

The authors would like to thank all that have contributed to the
development of the CLUE-s model and specifically Alias Mohd Sood for his
contribution to the Kuala Lumpur case study.

REFERENCES
Agarwal, C., Green, G.M., Grove, J.M., Evans, T.P. and Schweik, C.M. (2001) A Review and
Assessment of Land Use Change Models. Dynamics of Space, Time, and Human Choice.
Center for the Study of Institutions, Population, and Environmental Change, Indiana
University and USDA Forest Service, Bloomington and South Burlington.
Alcamo, J., Leemans, R. and Kreileman, E. (1998) Global Change Scenarios of the 21st
Century, Results from the IMAGE 2.1 Model, Elsevier, London.
Anselin, L. (2002) Under the hood: issues in the specification and interpretation of spatial
regression models, Agricultural Economics, 27: 247–267.
Bousquet, F. and Le Page, C. (2004) Multi-agent simulations and ecosystem management: a
review, Ecological Modelling, 176: 313–332.
Briassoulis, H. (2000) Analysis of land use change: theoretical and modelling approaches, in
Loveridge, S. (ed) The Web Book of Regional Science, at http://www.rri.wvu.edu/
regscweb.htm. West Virginia University, Morgantown.
Castella, J.-C., Boissau, S., Trung, T.N. and Quang, D.D. (2005) Agrarian transition and
lowland-upland interactions in mountain areas in northern Vietnam: application of a multi-
agent simulation model, Agricultural Systems, 86: 312–332.
Castella, J.-C., Pheng Kam, S., Dinh Quang, D., Verburg, P.H. and Thai Hoanh, C. (2007)
Combining top-down and bottom-up modelling approaches of land use/cover change to
support public policies: Application to sustainable management of natural resources in
northern Vietnam, Land Use Policy, (in press). doi:10.1016/j.landusepol.2005.09.009
Clarke, K.C. and Gaydos, L.J. (1998) Loose-coupling a cellular automaton model and GIS:
long-term urban growth prediction for San Francisco and Washington/Baltimore,
International Journal of Geographical Information Science, 12: 699–714.
Costanza, R. (1989) Model goodness of fit: a multiple resolution procedure, Ecological
Modelling, 47: 199–215.
de Koning, G.H.J., Verburg, P.H., Veldkamp, A. and Fresco, L.O. (1999) Multi-scale
modelling of land use change dynamics for Ecuador, Agricultural Systems, 61: 77–93.
Dietzel, C., Herold, M., Hemphill, J.J. and Clarke, K.C. (2005) Spatio-temporal dynamics in
California's Central Valley: empirical links to urban theory, International Journal of
Geographical Information Science, 19: 175–195.
Dietzel, C. and Clarke, K. (2006) The effect of disaggregating land use categories in cellular
automata during model calibration and forecasting, Computers, Environment and Urban
Systems, 30: 78–101.
Hilferink, M. and Rietveld, P. (1999) LAND USE SCANNER: An integrated GIS based
model for long term projections of land use in urban and rural areas, Journal of
Geographical Systems, 1: 155–177.
334 P.H. Verburg and K.P. Overmars

Huigen, M.G.A. (2004) First principles of the MameLuke multi-actor modelling framework
for land use change, illustrated with a Philippine case study, Journal of Environmental
Management, 72: 5–21.
Klijn, J.A., Vullings, L.A.E., van de Berg, M., van Meijl, H., van Lammeren, R., van
Rheenen, T., Eickhout, B., Veldkamp, A., Verburg, P.H. and Westhoek, H. (2005)
EURURALIS 1.0: A scenario study on Europe's Rural Areas to support policy discussion.
Background document. Alterra report 1196, Wageningen University and Research
Centre/Environmental Assessment Agency (RIVM).
Lambin, E.F. (1997) Modelling and monitoring land-cover change processes in tropical
regions, Progress in Physical Geography, 21: 375–393.
Lambin, E.F., Rounsevell, M.D.A. and Geist, H.J. (2000) Are agricultural land-use models
able to predict changes in land-use intensity? Agriculture, Ecosystems and Environment,
82: 321–331.
Moody, A. and Woodcock, C.E. (1994) Scale-dependent errors in the estimation of land-
cover proportions: implications for global land-cover data sets, Photogrammetric
Engineering & Remote Sensing, 60: 585–594.
Munroe, D.K., Southworth, J. and Tucker, C.M. (2002) The dynamics of land-cover change
in western Honduras: exploring spatial and temporal complexity, Agricultural Economics,
27: 355–369.
Overmars, K.P., Verburg, P.H. and Veldkamp, A. (2007) Comparison of a deductive and an
inductive approach to specify land suitability in a spatially explicit land use model, Land
Use Policy, (in press). doi:10.1016/j.landusepol.2005.09.008
Parker, D.C., Manson, S.M., Janssen, M.A., Hoffman, M. and Deadman, P. (2003) Multi-
agent systems for the simulation of land-use and land-cover change: a review, Annals of
the Association of American Geographers, 93: 314–337.
Pontius, R.G., Cornell, J.D. and Hall, C.A.S. (2001) Modelling the spatial pattern of land-use
change with GEOMOD2: application and validation for Costa Rica, Agriculture,
Ecosystems and Environment, 85: 191–203.
Pontius, R.G., Huffaker, D. and Denman, K. (2004) Useful techniques of validation for
spatially explicit land-change models, Ecological Modelling, 179: 445–461.
Pontius, R.G., Boersma, W., Castella, J.-C., Clarke, K., de Nijs, T., Dietzel, C., Duan, Z.,
Fotsing, E., Goldstein, N., Kok, K., Koomen, E., Lippitt, C.D., McConnell, W.,
Pijanowski, B., Pithadia, S., Sood, A.M., Sweeney, S., Trung, T.N. and Verburg, P.H.
(2007) Comparing the input, output, and validation maps for several models of land
change, Annals of Regional Science, (in press).
Pontius, R.G. and Cheuk, M.L. (2006) A generalized cross-tabulation matrix to compare soft-
classified maps at multiple resolutions, International Journal of Geographical Information
Science, 20: 1–30.
Schmit, C., Rounsevell, M.D.A. and La Jeunesse, I., (2006). The limitations of spatial land
use data in environmental analysis, Environmental Science and Policy, 9(2): 174–188.
Veldkamp, A. and Fresco, L.O. (1996) CLUE-CR: an integrated multi-scale model to
simulate land use change scenarios in Costa Rica, Ecological Modelling, 91: 231–248.
Veldkamp, A. and Fresco, L.O. (1997) Exploring land use scenarios, an alternative approach
based on actual land use, Agricultural Systems, 55: 1–17.
Verburg, P.H., Veldkamp, A. and Fresco, L.O. (1999) Simulation of changes in the spatial
pattern of land use in China, Applied Geography, 19: 211–233.
Verburg, P.H., Soepboer, W., Limpiada, R., Espaldon, M.V.O., Sharifa, M. and Veldkamp, A.
(2002) Land use change modelling at the regional scale: the CLUE-S model,
Environmental Management, 30: 391–405.
18. Land-use change trajectories with the CLUE-s model 335

Verburg, P.H. and Veldkamp, A. (2004a) Projecting land use transitions at forest fringes in
the Philippines at two spatial scales, Landscape Ecology, 19: 77–98.
Verburg, P.H., Schot, P., Dijst, M. and Veldkamp, A. (2004b) Land use change modelling:
current practice and research priorities, Geojournal, 61: 309–324.
Verburg, P.H., de Nijs, T.C.M., Ritsema van Eck, J., Visser, H. and de Jong, K. (2004c) A
method to analyse neighbourhood characteristics of land use patterns, Computers,
Environment and Urban Systems, 28: 667–690.
Verburg, P.H., Ritsema van Eck, J., de Nijs, T., Dijst, M.J. and Schot, P. (2004d)
Determinants of land use change patterns in the Netherlands, Environment and Planning
B, 31: 125–150.
Verburg, P.H., Schulp, C.J.E., Witte, N. and Veldkamp, A. (2006) Downscaling of land use
change scenarios to assess the dynamics of European landscapes, Agriculture, Ecosystems
& Environment, 114(1): 39–56.
Wassenaar, T., Gerber, P., Verburg., P.H., Ibrahim, M. and Steinfeld, H. (2007) Projecting
land-use changes in the neotropics: the geography of pasture expansion into forest,
Global Environmental Change, (in press). doi:10.1016/j.gloenvcha.2006.03.007
White, R., Engelen, G. and Uijee, I. (1997) The use of constrained cellular automata for high-
resolution modelling of urban land-use dynamics, Environment and Planning B, 24:
323–343.