Incorporating the multi-cross-sectional temporal effect in Geographically Weighted Logit Regression

Abstract

Sustainability concerns have aroused great interest among policy makers in the interrelated land use change systems. Land use change analysis becomes a hot topic. In recent years, considerable attention has been devoted to consider the spatial effect. Less attention has been paid to analyze the temporal effect. Learning how the temporal effect affect the accuracy of land use change model is thus of our great interesting. This paper incorporates the temporal effect in Geographically Weighted Logit Regression (GWLR), which can consider spatial nonstationarity in land use change modeling. Specifically, there will be two intervals to consider the temporal impacts in GWLR on land cover change modeling. Firstly, different time intervals in different land cover change process on one point will cause different results. Secondly, the other kind of time intervals is the time interval between the land use changes. In this study, for the first kind of temporal interval, we try to add an 'age' variable to the regression model. For the second, we proposed a new weighting function that combines the "geographical space" and "temporal space" between one observation and its neighbors, such that (1) neighbors with greater geographical distances from the subject point are assigned smaller weights, and (2) at a given geographical distance, neighboring points with less temporal interval to that of the subject point are assigned larger weights. The GWLR with weight function which consider both the two kind temporal effects will be implemented to be compared with the others. Case studies of land use change patterns in Calgary, Canada will be implemented. The results indicate that the GWLR model considering the temporal effect performs better than the other models. © 2013 Taylor & Francis Group, London, ISBN 978-1-138-00115-2.