Evaluation of accuracy and changing rainfall trends of gridded satellite precipitation products: a comparative analysis with ground observations over the Modjo catchment, central Ethiopia

Analyzing long-term rainfall variability and trends under global climate change is critical, particularly when integrating satellite-based precipitation products. The present study evaluates the performance and detects trends in annual and seasonal rainfall across Modjo catchment central Ethiopia by comparing ground observations with two satellite datasets (CHIRPS v2.0 and PERSIANN-CDR) from 1985 to 2020. The study employed different statistical metrics to investigate the performance of satellite precipitation products. The Sen Slope Estimator, Mann–Kendall test, and Innovative Trend Analysis (ITA) were used to assess the trends within the data series. Results indicate that CHIRPS v2.0 outperformed PERSIANN-CDR, exhibiting stronger correlations (R = 0.885–0.915), lower errors (RMSE: 40.16–47.72 mm; MAE: 24.68–28.01 mm), higher model efficiency (NSE: 0.770–0.816), and minor underestimation (PBIAS: 6.22–7.65%). In contrast, PERSIANN-CDR showed weaker correlations (R = 0.849–0.876), higher errors (RMSE: 45.24–56.32 mm; MAE: 27.94–31.95 mm), and systematic overestimation (PBIAS: − 1.49% to − 12.77%). Trend analysis using the Mann–Kendall test, Sen’s slope estimator, and ITA revealed divergent patterns: observed data indicated decreasing annual trends at Ejere (Z = − 0.503) and Modjo (Z = -0.053) but a weak increase at Chefe Donsa (Z = 0.395). The Belg (short rainy) season showed consistent declines, notably at Chefe Donsa (Z = − 0.694), while Kiremt (long rainy) rainfall increased only at Chefe Donsa (Z = 0.912). Satellite data, particularly CHIRPS v2.0, suggested strong annual and Kiremt increases at Ejere (Z = 1.743 and 2.628) and Modjo (Z = 2.547 and 2.683), whereas PERSIANN-CDR uniformly indicated decreasing trends. The Bega season (dry season) exhibits mixed trends, with observed declines at Chefe Donsa (Z = –0.449) and Ejere but a slight increase at Modjo (Z = 0.169), while satellite data diverge, highlighting challenges in dry-season estimation. The study concludes that satellite precipitation products show strong potential for hydro-meteorological applications, though further refinement is needed. For climate trend analysis, the present study recommend rigorous conjunction with ground observation datasets to ensure more reliable results.