Modeling land use change using an eigenvector spatial filtering model specification for discrete responses

Abstract

Spatial discrete choice models best suit unordered categorical response variables like land use change, but little literature is available about their application to large datasets. This chapter focuses on addressing the computational issues of a spatial discrete choice model and large datasets. An eigenvector spatial filter specification with coarser resolution has been used that accounts for spatial autocorrelation between neighboring land pixels. Variables of the built environment and socioeconomic and demographic characteristics are used as covariates in the spatial multinomial logistic regression.