Quantifying Precipitation-Driven Uncertainty in Streamflow Simulations: Application to the Reno River Basin (Italy)

Several factors inherently influence the accuracy of hydrological model simulations –including uncertainties in input data, model parameterization, and the (unavoidably simplified) representation of physical processes. Among these, precipitation data used as input play a crucial role as they directly influence the magnitude and timing of streamflows. This study aims to unravel the propagation of uncertainty in hydrological modelling from precipitation data to streamflow simulations. 

To this end, we built a semi-distributed and process-based hydrological model using the Hydrological Predictions for the Environment (HYPE) code of the Reno river basin in Italy. Moreover, we used four gridded precipitation datasets —ERG5 (5 km), CHIRPS (5 km), E-OBS (0.1°), and ERA5 (0.1°)—to calculate mean annual and seasonal precipitation at the sub-basin scale for the period 2001–2022. Despite similarities in seasonal patterns, notable differences emerge during the wet season (especially in winter) and in annual averages, particularly in the small and mountain sub-basin. ERA5 and CHIRPS generally underestimate precipitation during the wet season, while E-OBS exhibits strong correlation with the observed ERG5 dataset.

Observed daily streamflow data were used to calibrate (2001–2010) and validate (2011–2022) the hydrological model. While the Kling-Gupta Efficiency (KGE) values were overall acceptable, we found larger uncertainties across all sub-basins. In the small and mountainous sub-basin, simulated streamflow shows greater variability and peak flows are often overestimated during the winter. This might be attributed to limitations in gridded datasets, such as the density of gauge stations and the capturing of snow precipitation. These uncertainties also propagate into the dry season, where the variability in simulated streamflow is relatively larger compared to the entire basin for the same season. These findings underscore the significant influence of uncertainty in precipitation data on hydrological simulations, especially in areas with complex orography. We also discuss the importance of addressing such uncertainties in hydrological modeling across different scales.