Multi-timescale evaluation of satellite precipitation products using 15-minute gauge observations in complex terrain, Ethiopia

This study evaluated the performance of six widely used satellite precipitation products, including GPM-IMERG V06 ER, IMERG V07 FR, CMORPH, PERSIANN-CCS, GSMaP-MVK, and MSWEP against observed sub-daily, and daily timescale rainfall. Thirty-six Automatic Weather Stations (AWSs), gauge observations were recorded at 15-minute intervals and aggregated to hourly, 3-hourly, and daily timescales for comparison with the satellite products across Ethiopia. A point-to-pixel validation approach was applied using statistical and categorical error metrics at hourly, 3-hourly, and daily temporal scales for the period 2016–2022. The findings revealed a pronounced west-to-east rainfall gradient and systematic seasonal biases among the products. Several products overestimate precipitation in lowland regions during the main rainy season (JJAS), likely due to sub-could evaporation of precipitation in an arid atmospheric profile. Satellite precipitation products such as IMERG V06 ER and V07 FR most closely matched the observed rainfall patterns, demonstrating moderate correlation and strong event detection capability (CC = 0.57 and 0.51; POD = 0.78 and 0.78; CSI = 0.55 and 0.51), respectively. These products are accompanied by comparatively lower RMSE, MAE, and FAR values. In addition, products such as CMORPH showed stable performance across all temporal scales, particularly at the daily resolution (CC = 0.5, POD = 0.61, CSI = 0.47). On the other hand, GSMaP-MVK exhibited a notable overestimation at 3-hourly scales but improved at daily scales, whereas PERSIANN-CCS consistently underestimated rainfall and showed weaker detection skill. Add the performance at the hourly timescale as well. This reflects the inherent limitations of infrared-based algorithms in capturing orographic warm rain over the Ethiopian highland. MSWEP showed weak performance at the daily scale, although it exhibited moderate correlation at the 3-hourly scale (CC = 0.33). Quantile-Quantile analysis, Cumulative Distribution Functions, and performance diagrams confirmed consistent underestimation tendencies in IMERG V07 FR, and PERSIANN-CCS, and overestimation in IMERG V06 ER and CMORPH. Overall, IMERG V06 ER, IMERG V07 FR, and CMORPH emerged as the most reliable SPE products for capturing rainfall variability across Ethiopia’s complex terrain. The findings highlight substantial inter-product variability, emphasize the need for algorithm refinement, and show the importance of product selection for hydrological applications and precipitation monitoring in data-scarce regions.