<p>Assessing the performance of global precipitation products in semi-arid Mediterranean catchments is crucial for improving hydrological modeling and water resource management, particularly in regions where in-situ observations are scarce or nonexistent. This study evaluates five precipitation datasets: observed rainfall, CHIRPS (Climate Hazards Group InfraRed Precipitation with Station data), PERSIANN (Remotely Sensed Information using Artificial Neural Networks), GPM-IMERG (Global Precipitation Measurement Integrated Multi-Satellite Retrievals), and ERA5 over a 17-year period from 2001 to 2017, including one year of model warm-up for a semi-arid catchment in central Tunisia. The Soil and Water Assessment Tool (SWAT), a semi-distributed hydrological model, was calibrated using observed precipitation data and then applied with alternative datasets to quantify the influence of precipitation sources on model performance. Calibration with observed data yielded satisfactory results, with a Nash–Sutcliffe Efficiency (NSE) of 0.63 and a Kling–Gupta Efficiency (KGE) of 0.72. When alternative datasets were introduced, CHIRPS consistently provided acceptable results even without recalibration, achieving a KGE value of 0.32, which indicates moderate model performance although it remains below the commonly accepted satisfactory threshold. ERA5 and PERSIANN showed slight improvement after calibration but often yielded low KGE values. During calibration, CHIRPS and ERA5 also achieved satisfactory performance, with NSE values of 0.51 and 0.50 and KGE values of 0.58 and 0.61, respectively. In contrast, GPM-IMERG results remained largely unchanged, showing minimal sensitivity to calibration parameters. These findings confirm that precipitation is the principal source of uncertainty in hydrological modeling and that each dataset imposes specific ranges on calibrated parameters. For the Haffouz catchment, CHIRPS demonstrated the best spatial distribution and overall hydrological performance, highlighting the value of satellite-based precipitation products as reliable alternatives in data-scarce regions. This study provides critical insights, emphasizing the importance of carefully selecting precipitation datasets to reduce uncertainty and improve the reliability of hydrological simulations in semi-arid Mediterranean context.</p> Graphical Abstract <p></p> <p>Graphical abstract summarizes the assessment of the satellite-based precipitation products to hydrological modeling in data-sparse basins with SWAT model. Data are represented through the integration of multiple precipitation sources, including observed rainfall and satellite-based products (CHIRPS, ERA5, PERSIANN, and GPM), together with essential spatial and climatic inputs such as the digital elevation model (DEM), land use, soil characteristics, and temperature, which collectively define the study area and its hydrological context. Calibration and validation procedures are implemented to quantitatively assess model performance using statistical efficiency metrics (NSE and KGE), thereby enabling an objective comparison among the different precipitation datasets. Hydrological simulations are performed using the SWAT model, which combines all input data to reproduce watershed-scale processes. As a result, the model shows satisfactory performance when calibrated with observed data (NSE = 0.63, KGE = 0.72). When driven by satellite precipitation products, CHIRPS and ERA5 achieve satisfactory performance after calibration (NSE ~ 0.50, KGE ~ 0.60), while PERSIANN shows only slight improvement and GPM exhibits no significant change. CHIRPS provides acceptable results even without calibration and emerges as the best-performing dataset in terms of spatial consistency and hydrological response. Overall, the results highlight precipitation as the main source of uncertainty in hydrological modeling and confirm the relevance of satellite rainfall products, particularly CHIRPS, as reliable alternatives in data-scarce catchments such as Haffouz.</p>