Hydrological evaluation of top-down and bottom-up rainfall products in West Africa: Model performance, parameter range and uncertainty propagation

The West African Sahel is characterized by high rainfall variability and sparse observational networks, challenging hydrological modelling and water resources management. This study focuses on three major river basins in Burkina Faso (the Mouhoun, Nakanbé, and Sirba) which are tributaries of the Volta and Niger systems and are critical for regional agriculture, livelihoods, and water security. We assessed the hydrological value of four top-down satellite precipitation products (ARC2, RFE2, CHIRPS, TAMSAT) and three bottom-up products derived from SM2RAIN (ASCAT, CCI, GPM) using the semi-distributed SWAT model. Baseline simulations were forced with synoptic gauge observations (SIM_OBS). The results show that model skill varied across basins, with KGE = 0.83/0.78 (Mouhoun), 0.70/0.52 (Nakanbé), 0.78/0.83 (Sirba) in calibration/validation. Several gridded products outperformed SIM_OBS: CHIRPS (0.90/0.80) and RFE2 (0.81/0.85) in the Mouhoun, and CCI (0.83/0.72) and CHIRPS (0.80/0.75) in the Sirba. Several gridded products outperformed gauge data in reproducing discharge and flow signatures (Q5, Q95, FDC slope), while others remain unreliable, particularly in the Nakanbé. Accurate products maintained narrow posterior ranges for runoff and routing parameters, whereas uncertain inputs induced compensatory broadening of groundwater controls. Error propagation analysis revealed that rainfall uncertainty could be either attenuated or amplified, but parameter uncertainty accounted for most discharge spread. These findings demonstrate that carefully selected rainfall products can enhance hydrological modelling and water resource planning in Sahelian basins. • We evaluated the hydrological performance of top-down and bottom-up rainfall products. • CHIRPS, RFE2, and CCI achieved comparable performance to gauge-forced baseline. • Rainfall forcing shaped calibration behaviour and optimal parameter range. • Error propagation analysis showed basin-dependent amplification of input errors.