Evaluation of satellite-based rainfall estimates and their performance across Ethiopia

Precipitation plays a crucial role in Ethiopia’s rain-fed agriculture, ecosystems, and water resources; yet it exhibits, high spatial and temporal variability. However, the sparse gauge network makes accurate monitoring difficult. This paper evaluates widely used satellite precipitation products (SPPs), CHIRPS_v2, CHIRPS_v3, IMERG, GSMaP, PERSIANN, and AgERA5 (agrometeorological indicators derived from ERA5 reanalysis) against daily rain gauge observations from 95 Ethiopian stations between 2001 and 2024. Their performance was evaluated using statistical metrics ( \:R² , RMSE, bias), Probability of Detection (POD), False Alarm Ratio (FAR) and analyses of precipitation anomalies using PCA and empirical cumulative distributions (CDFs) across daily, monthly, seasonal, and annual scales, with stratification by elevation zones and land cover types. Results indicate that the performance of these products improve with temporal aggregation, showing low correlations at the daily level \:\:(R²≈\:0.1 ) but increasing significantly at monthly \:(R²\:=\:0.62-0.75) and annual scales \:(R²\:=\:0.28-0.60). At the daily scale, RMSE ranged from 12.0 to 16.3 mm, while relative bias ranged from − 36.2% for GSMaP to 32.8% for CHIRPS_v2. CHIRPS_v3 and IMERG consistently provide the most reliable estimates, combining high correlation, low bias, and strong rainfall detection skill across elevations and land covers. CHIRPS_v2 performs moderately, especially in semi-arid regions, while AgERA5 shows moderate skill but tends to overestimate precipitation. GSMaP systematically underestimates precipitation, and PERSIANN exhibits moderate positive bias with lower correlation. Seasonal evaluation indicates peak performance during the wet Kiremt season and lower skill in the dry Bega season. Stratified analyses reveal substantial differences in performance across elevation zones and land-cover types, with greater uncertainties in some lowland, densely vegetated, and shrubland environments. These findings guide the selection of suitable SPPs for hydrological modelling, agricultural planning, and climate risk assessment in data-scarce regions.