Linking models in land use simulation - Application of the Land Use Scanner to changes in agricultural area

When we model land use change, we utilize – consciously or unconsciously – other models as well. The variables we regard as exogenous are often generated endogenously by a different model. We are not always fully aware of the implications of this for our modelling exercises. The model which generated the demographic growth that we use in forecasting the need for residential space may have used assumptions that are at variance with ours. The model resulting in claims for agricultural land may have already taken competing claims into account – whereas our land use model may simulate this competition all over again. The data used for different models may not be compatible. Conversely, our land use simulation exercises can also be used by others as input. A model for the agricultural sector, for instance, must consider the constraint of available land – especially whether the land required is available in a particular area which is regarded as optimal for a particular production line. Land use models can provide that input. The Agricultural Economics Research Institute in The Hague, uses a number of models at various spatial levels – from the individual farm to the global economy – and for different purposes. Recently, the linkages between these models have received more attention, which also lays bare the compatibility problems between them. In order to examine both the possibilities and the problems inherent in these linkages, a research project on this 'model train' has been undertaken. Based on two opposing scenarios prepared by the Dutch Central Planning Bureau, the study calculates the long-term consequences of these scenarios: beginning with a general equilibrium model at global level (GTAP) through a sectoral model at national and regional scale - the Dutch Regionalized Agricultural Model (DRAM) – to a model assessing ecological effects in a local area (SOMMA). The Land Use Scanner, a land use information system and simulation model for the Netherlands, has been used to predict changes in the agricultural area for the regions used in DRAM. The land claims, which are an exogenous variable in the Land Use Scanner, were generated from projections of future population and GDP, on the basis of their historical correlation with land use. This project has led to interesting insights into the problems of linking models. It is hoped that these insights will help to improve the models we use – including land use models. The paper highlights the importance of making modelling assumptions explicit, such that the outcome of one model can indeed be a useful input into another one. The integrated modelling approach yields more consistent projections of land use.