Intercomparison and evaluation of gridded precipitation products and enhanced drought analysis in the Shabelle Basin, Somalia-Ethiopia

This study compiled gauge-based precipitation observations (1998–2016) for the Shabelle River Basin in eastern Ethiopia and compared them with multiple satellite and reanalysis gridded precipitation products. Each dataset's performance was evaluated against ground measurements using diverse statistical metrics (correlations, error indices, efficiency scores) across daily, monthly, seasonal, and annual scales, as well as categorical indices for rainfall event detection. Results indicate that a satellite–rain gauge blended product (CHIRPS) and a modern reanalysis (ERA5) achieved the highest agreement with station observations, accurately capturing spatial rainfall gradients and bimodal seasonal cycles, whereas a gauge-only analysis (NOAA CPC Unified) consistently underestimated precipitation. Using the best-performing datasets, the Standardized Precipitation Index (SPI) and Standardized Precipitation–Evapotranspiration Index (SPEI) were calculated to conduct an enhanced drought analysis. These indices derived from CHIRPS and ERA5 closely mirrored ground-station drought histories, successfully capturing the timing and severity of major drought events from 1999 to 2016. Furthermore, multiple machine learning models were applied for station-level drought prediction; notably, a Naïve Bayes classifier outperformed more complex methods (random forest, k-nearest neighbors, linear regression) in predicting short-term SPI variations. The findings provide a comprehensive evaluation of precipitation data uncertainties in the Shabelle Basin and offer practical guidance on selecting optimal datasets for drought monitoring in data-sparse regions.