Modeling Spatial Land Use Pattern Using Autologistic Regression

The significance of land cover as an environmental variable has made land use change an important subject in global environmental change and sustainable development. Modeling land use change has attracted considerable attention. Currently, empirical estimation models using statistical techniques are one of mostly used spatial models to simulate land use pattern and its changes. Empirical estimation methods can model the relationships between land use changes and the drivers. However, existing logistical regression models often ignore the spatial autocorrelation among land use data, which affect the goodness of fitting and accuracy of fitting of land use modeling. In this study we incorporate components describing the spatial autocorrelation into existing logistical regression and form an autologstic regression. Taking the Yongding County, Hunan province, China as study area, we simulate spatial pattern of different land use types using autologistic regression and compare with the existing logistical regression method. The results indicate that autologistic regression has better goodness of fitting and higher accuracy of fitting than the existing logistical regression method. Autologistic regression can improve the modeling result in some degree reasonably.