Methodology

Abstract

Remote sensing, GIS, and land change models (LCMs) are critical for mapping urban land use/cover and simulating “what if” urban growth scenarios, particularly in developing countries experiencing rapid urbanization. The purpose of this chapter is to describe briefly the methodology used to produce land use/cover maps, and simulate land use/cover changes for selected metropolitan areas in Asia and Africa. Land use/cover maps were classified from Landsat imagery for 1990, 2000, 2010, and 2014 using the random forest (RF) classifier. Quantitative accuracy assessment was not conducted for the 1990 land use/cover maps due to lack of reference data. However, qualitative and quantitative accuracy assessment was performed for the 2000, 2010, and 2014 land use/cover maps based on Google Earth imagery. Overall land use/cover classification accuracy for all land use/cover maps ranged from 70 to 90%. Land use/cover changes were simulated based on the boosted regression trees-cellular automata (BRT-CA) and RF-CA LCMs. We evaluated the goodness-of-fit of transition potential maps, and validated the simulated land use/cover changes based on robust statistical measures. Generally, the BRT-CA and RF-CA LCMs for all metropolitan areas in Asia and Africa performed relatively well. In particular, the BRT-CA and RF-CA LCMs for metropolitan areas in Africa had the best performance. The modeling and simulation results presented in this chapter provide an initial exploration of BRT-CA and RF-CA LCMs in Asia and Africa. This chapter demonstrates the significance of robust calibration, validation, and simulation of spatial LCMs for all metropolitan areas in Asia and Africa.