Spatial analysis and modelling of land use distributions in Belgium

摘要: When statistical analyses of land use drivers are performed, they rarely deal explicitly with spatial autocorrelation. Most studies undertaken on autocorrelation-free data samples. By doing this, a great information that is present in the dataset lost. This paper presents spatially explicit, cross-sectional analysis Belgium. It shown purely regressive logistic models only identify trends or global relationships between socio-economic physico-climatic and precise location each type. However, when goal study to obtain best model fit distribution, autoregressive appropriate. this type deals appropriately autocorrelation as measured by lack deviance residuals model. More specifically, three types compared: (1) set binomial regression (one for modelled use) accounting proportion within neighbourhood cell; (2) multinomial autologistic accounts composition cell's neighbourhood; (3) stateof-the-art Bayesian Maximum Entropy (BME) based fully organization uses cell. The comparative shows BME approach has no advantages over other methods, our specific application, but essential obtaining an optimal fit. (C) 2006 Elsevier Ltd. All rights reserved.