Improving SLEUTH Urban Growth Model Using Logistic Regression and Land Density Function

Abstract

Multi-criteria evaluation (MCE) probability maps based on various driving forces of urban growth were combined with the exclusion layer of the SLEUTH urban growth model. The extended model was further tested by integration of the land density function (LDF) to explore the possibility of modeling accuracy. Four different versions of SLEUTH were compiled from two different approaches as within the exclusion layer comprising areas where urban growth is restricted such as water bodies and natural reserves, and within the slope layer applicable to cities with a relatively flat topography as in the chosen test site. Inclusion of the MCE layer upgraded the SLEUTH predictions since more suitable pixels for urbanization were selected and changes in other land-use/land-cover categories based on constraints incorporated into the MCE excluded layer were prevented. Five metrics were selected  including class area (CA), number of patches (NP), largest patch index (LPI), edge density (ED), and contagion index (CONTAG) to measure and describe the level, pattern, heterogeneity, and complexity of the urban area. Cohen’s kappa (k) was used as an evaluation coefficient to measure accuracy assessment of model prediction. Results showed that in cases of exclusion input, the standard model with MCE coupled with LDF provided the highest kappa hat coefficient at 0.7888. In cases of slope input, the MCE model provided highest overall accuracy at 82.60%, with mean overall accuracy of 82.00% for the four extended models.