Land use and land cover changes in Notwane watershed, Botswana, using extreme gradient boost (XGBoost) machine learning algorithm

Abstract

Botswana’s Notwane watershed is crucial for supplying water to Gaborone and surrounding areas. However, urbanization and climate variability threaten its water resources. Land Use and Land Cover (LULC) changes contribute to climate variability, impacting food security, water supply, and weakening rural economies and socio-cultural structures. This study employs geospatial techniques to model LULC changes and inform sustainable environmental management. This study utilized Extreme Gradient Boost (XGBoost), Support Vector Machine (SVM), and Random Forest (RF) algorithms to classify multi-temporal Landsat images from 1984 to 2022. The XGBoost model achieved the highest overall accuracy of 0.81%. Results indicated a 21,239.2% increase in built-up areas between the year, while agricultural land and natural vegetation decreased significantly by 9.38%. These shifts are driven by urbanization, which heightens climate change through increased greenhouse gas emissions and reduced carbon sinks. Variations in water-covered areas were, influenced by dam construction, droughts, and cyclones. A strong correlation between built-up areas and population data highlights the impact of urban expansion. To ensure sustainable urban growth and mitigate negative effects on biodiversity, urban planners must integrate sustainable land use strategies. These findings highlight the necessity for informed decision-making to balance development with environmental sustainability in the Notwane watershed.