Agricultural drought dynamics in East Gojjam: Insights on soil moisture, drought indices, and crop sustainability

East Gojjam Zone, Amhara Region, Ethiopia The research investigates the spatiotemporal dynamics of agricultural drought from 2013 to 2024 using an integrated approach that combines remote sensing indices (NDVI_Max, SWDI, SMA, and SPEI) and rainfall data with machine learning models. Random Forest and multiple linear regression were employed to assess the relationship between drought indicators and crop productivity. Sensitivity analysis and partial dependence plots were used to identify dominant predictors of vegetation health. The study revealed significant interannual variability in drought intensity, with 2022 identified as a critical drought year marked by widespread hydrological stress and reduced crop vigor. SWDI and NDVI_Max emerged as the most influential predictors of vegetation response, while rainfall showed a nonlinear impact. The findings highlight the spatial heterogeneity of drought impacts and underscore the need for woreda-specific mitigation strategies, including precision irrigation, drought-resilient crop varieties, and satellite-informed early warning systems. This research provides a replicable, data-driven framework for drought risk assessment and adaptive agricultural planning in moisture-stressed regions. This study investigates the spatiotemporal dynamics of agricultural drought across the East Gojjam Zone, Ethiopia, using integrated remote sensing indices and machine learning models. By analyzing NDVI_Max, SWDI, SMA, SPEI and rainfall data from 2013 to 2024, we identified significant interannual variability and localized drought intensities, particularly in Dejen, Debay Tilat, and Enarg Enawega. • SWDI validated with satellite SM to ensure realism and reduce uncertainty. • SMA, SWDI, SPEI, NDVI, rainfall used for holistic drought-yield modelling. • RF for feature selection; MLR improves interpretability in hybrid modelling. • Datasets harmonized across time/space for coherent multi-source analysis. • Landsat-8 NDVI used for fine-scale vegetation stress benchmarking.