Land Use/Land Cover Change Modeling and Prediction Using Artificial Neural Networks: A Case Study of the Birjand Watershed

چکیده مقاله

Human development and land transformation have turned land use and land cover (LULC) change into one of the major global environmental challenges, affecting essential biogeochemical cycles such as water, carbon, and energy. This study aims to model and predict LULC changes in the Birjand watershed, located in eastern Iran, using an Artificial Neural Network (ANN) within the Land Change Modeler (LCM) framework. Landsat satellite imagery (TM and ETM+) was used to detect historical changes, while eight explanatory variables digital elevation model, slope, soil texture, and distance from roads, rivers, faults, residential areas, and agricultural lands were applied to simulate transition potentials for the 2000–2010 calibration period. A Markov chain analysis was then employed to estimate the probability of transitions between land use classes, and predicted maps were generated for 2020. The results revealed that barren lands were predominantly converted into residential areas, reflecting rapid urban expansion. Validation of the simulated 2020 map against observed data indicated strong model performance, supported by accuracy metrics such as the Kappa coefficient and Overall Accuracy. Based on the validated model, LULC projections for 2040 show a continued decline in agricultural lands alongside the expansion of urban and barren areas. The projected areas amount to approximately 73.82 km² of agricultural land, 68.81 km² of urban land, and 3,282.39 km² of barren land. These projections underscore the urgent need for sustainable land-use planning and management strategies to minimize environmental degradation in arid regions.