Coupling advanced earth observation, Machine learning, GeoAI, and CA-ANN model for assessing Spatio-temporal LULC change in Gilgel Gibe I Watershed, Upper Omo-Gibe Basin, Ethiopia

Abstract Land use and land cover (LULC) change represents one of the most significant manifestations of human-environment interactions and a primary driver of global environmental change. Rapid LULC transformation in the Gibe I Watershed has created substantial challenges for ecosystem service provisioning, biodiversity conservation, and sustainable development. Although LULC change prediction using geospatial artificial intelligence (GeoAI) and multi-factor transition potential modelling has received increasing attention, studies integrating these approaches in the Omo-Gibe basin of Ethiopia remain limited. This study integrates cloud-based remote sensing classification with a Cellular Automata-Artificial Neural Network (CA-ANN) geospatial model to comprehensively assess historical, contemporary, and projected LULC dynamics in the Gibe I Watershed. Multi-temporal Landsat and Sentinel-2 imagery were processed through Google Earth Engine using Random Forest supervised classification to generate high-accuracy LULC maps for 2004, 2014, and 2024, achieving overall classification accuracies of 88.45%, 90.62%, and 93.8% respectively, and with substantial Kappa coefficients exceeding 0.89 in these years. Transition potential modeling revealed pronounced Spatio-temporal transformations, with built-up and agricultural areas expanding substantially at the expense of forest, grassland, and bare land. Forest cover declined by 49,851.53 hectares (58.15%) between 2004 and 2024, while agricultural land increased by 83,219.68 hectares (32.25%) during the same period. The CA-ANN model incorporated six driving factors including elevation, slope, and proximity variables—and was validated against 2024 LULC classifications before projecting future scenarios for 2034 and 2044. Projections indicate continued land transformation, with built-up areas expanding further while forest and grassland areas decline progressively. These LULC changes have profound implications for ecohydrological regimes and threaten ecosystem services, livelihoods, and environmental sustainability. The findings provide critical decision-support tools for land use planning, water resource management, and environmental conservation in rapidly transforming tropical watersheds.