Abstract
Rapid urban growth in tropical megacities is putting serious pressure on critical ecosystem services, thereby complicating the implementation of sustainable urban planning frameworks. To better understand and address these challenges, we used artificial intelligence (AI) and satellite data to map and predict land-use changes in Chittagong City Corporation (CCC), Bangladesh, from 2000 to 2048. We combined Random Forest (RF) classification with a Cellular Automata–Artificial Neural Network (CA–ANN) model. The Random Forest method classified four land-cover types: vegetation, built-up areas, barren land, and open water bodies, with over 97% accuracy. These accurate maps formed the base for future simulation. Our results show that vegetation dropped sharply from 48% in 2000 to 31% in 2024. During the same period, built-up areas grew from 35 to 62%. Barren land decreased steadily, while water bodies remained mostly stable. Predictions suggest that by 2048, built-up land will reach 65%, causing further loss of vegetation and green space. These changes point to rising habitat fragmentation and increased ecological stress. The trends we found match global patterns of urban land conversion and show how cities are moving closer to critical ecological tipping points. These shifts threaten biodiversity, reduce climate resilience, and break the connection between green spaces. Our model offers a practical tool for city planners and decision-makers. It helps predict land cover changes and supports actions that balance development with conservation. To manage urban growth sustainably, planning must shift from reactive steps to forward-looking strategies. Our study supports the use of AI-powered geospatial tools to guide land-use decisions, protect ecosystems, and promote resilient urban futures, both in Bangladesh and in other rapidly growing cities around the world.
