Integrating Sentinel-2 Data And CA–Markov For Spatial Simulation Of Urban Expansion And Environmental Change: A Case Study Of Sanandaj

Abstract

This study investigates and forecasts land use and land cover (LULC) dynamics in Sanandaj, a medium-sized mountainous city in western Iran, over the period 2020–2040. Multi-temporal Sentinel-2A satellite imagery, combined with Cellular Automata (CA) and Markov chain models, was used within the TerrSet and ArcGIS environments to simulate spatiotemporal patterns of land transformation. LULC was classified into seven major categories using the Maximum Likelihood Classification (MLC) algorithm. Model accuracy was evaluated using the Kappa coefficient and Receiver Operating Characteristic (ROC) curve. Between 2020 and 2024, approximately 5,180 hectares of LULC change occurred. The most significant increases were observed in built-up areas (612 hectares, 11.8%) and barren lands (3,945 hectares, 76.2%), while dense vegetation (1,082 hectares, -20.9%) and water bodies (436 hectares, -8.4%) declined notably. The CA–Markov model projects that by 2040, an additional 3,960 hectares of land conversion will take place. Positive transitions are expected in built-up (14.2%) and barren (6.1%) categories, while negative changes are projected in vegetative covers and hydrological areas. The validation metrics yielded a Kappa coefficient of 0.61 and ROC AUC of 0.65, confirming moderate-to-good model reliability. These findings underscore the critical need for proactive land management and urban planning strategies to mitigate ecological degradation and ensure sustainable urban growth in semi-arid and mountainous contexts.