Modelling land Use Change : Improving the prediction of future land use patterns

Modelling land Use Change: Improving the prediction of future land use patterns. Man has been altering his living environment since prehistoric times and will continue to do so. It is predicted that by 2030 about 90,000 ha will be needed for residential developments in the Netherlands and 55,000 ha for industry, offices and commerce. Moreover, nearly 250,000 ha of agricultural land will be required for the development of the Ecological Main Structure. Where will these developments take place? What effect will these developments have on our environment, natural landscapes, ecosystems and human health? Land use simulation models are used to depict future spatial developments, ideally revealing conflicting interests between urbanisation, nature conservation, industrialisation, water management, agriculture and a healthy living environment. Decision makers relying on the results of these models should have some concept of the validity and accuracy of these models. The overall objective of the thesis is: to improve the quality of land use simulation models and the prediction of future land use patterns. The land use model selected in this thesis to simulate the land use developments in the Netherlands is the Environment Explorer. Statistical analysis of land use developments over the last years revealed the main determinants of land use change in the Netherlands: current or initial land use, the land use in the neighbourhood and spatial policy. This thesis shows that the influence of the neighbourhood on the probability of new residential developments depends on cluster-size. Locations near large residential areas are more likely to change than locations near small residential areas, while locations near large airports are less likely to change to residential area than those near small airports. The suitability of the soil seems to be less important for urban developments in the Netherlands except for the peat areas where the soil subsides due the oxidation of the peat. The probability of urban developments is highest on building sites and on the undeveloped sites in towns and on allotments, sporting fields and agricultural land on the outskirts of the towns. Calibration of land use models is a necessary, complex and time-consuming business in which Single Parameter Impact Estimation (SPIE) should be used to estimate the influence of the neighbourhood in the land use model. The simulated urban land use developments can and should be validated using Zipf’s Law describing the relation between the occurrence and the size of the urban land use clusters. The uncertainty in the simulated urban land use developments is large. The uncertainty is low for airports and water as these land uses are unlikely to change as well as for building sites because these sites will change most likely. The uncertainty is highest for the undeveloped sites at the outskirts of the towns and cities. Future research should focus on empirical spatial analysis of land use developments for various land use categories at different locations, spatial and temporal resolutions. Furthermore, calibration and validation of land use models may improve most from the direct application of successive remote sensing images, avoiding the effect of misclassification in the translation of the images into land use maps