Spatiotemporal analysis of rural–urban land conversion

Abstract

Understanding the complexity of urban expansion requires an analysis of the factors influencing the spatial and temporal processes of rural–urban land conversion. This study aims at building a statistical land conversion model to assist in understanding land use change patterns. Specifically, GIS coupled with a logistic regression model and exponential smoothing techniques is used for exploring the effects of various factors on land use change. These factors include population density, slope, proximity to roads, and surrounding land use, and their influence on land use change is studied for generating a predictive model. Methods to reduce spatial autocorrelation in a logistic regression framework are also discussed. Primarily, an optimal sampling scheme that can eliminate spatial autocorrelation while maintaining adequate samples to allow the model to achieve the comparable accuracy as the spatial autoregressive model is developed. Since many of the previous studies on modeling the spatial complexity of urban growth ignored temporal complexity, a modified exponential smoothing technique is employed to produce a smoothed model from a series of bi‐temporal models obtained from different time periods. The proposed model is validated using the multi‐temporal land use data in New Castle County, DE, USA. It is demonstrated that our approach provides an effective option for multi‐temporal land use change modeling and the modeling results help interpret the land use change patterns.