Using extreme gradient boosting for predictive urban expansion analysis in Rustenburg, South Africa from 2000 to 2030

Since the end of apartheid in 1994, South Africa has experienced significant urban population growth, with major cities undergoing particularly rapid expansion. Understanding spatial urban expansion in developing cities, such as Rustenburg, is essential for sustainable infrastructure planning, environmental management, and service provision. This study modeled the spatial extent of urban growth in Rustenburg from 1994 to 2022 using Extreme Gradient Boosting (XGB) and predicted future urban expansion from 2022 to 2030 through the Cellular Automata Simulation in the MOLUSCE plugin. The results indicate that in 1994, water bodies covered 88,303 hectares, developed areas covered 125,763 hectares, and vegetation occupied 2,513,336 hectares. By 2022, vegetated regions covered 5,104,145 hectares, developed areas increased to 586,017 hectares, and mining zones expanded to 113,224 hectares. Furthermore, the landscape featured disjointed urban areas, varied natural ecosystems, and significant fragmentation in undeveloped zones. In 2022, the number of water patches increased to 27,845, built-up areas to 60,690, vegetated areas to 102,119, bare land to 80,921, and mining areas to 23,296 patches. The Kappa statistics ranged from 0.93 (93%) to 0.99 (99%), demonstrating high reliability in simulating urban growth patterns. An overall accuracy of 95% and a Kappa statistic of 0.93 are indeed higher than the mean scientifically accepted accuracy level of 85% for urban growth analysis. These findings highlight significant changes in land use and landscape fragmentation over time and provide critical insights into urban growth patterns and their implications for water security, sustainable development, and community livelihoods in Rustenburg, South Africa. By examining these patterns, we can better plan for sustainable infrastructure, manage environmental impacts, and improve service provision in rapidly expanding urban areas.