Simultaneous Evaluation of Streamflow Data Assimilation for Addressing Precipitation Error Propagation and Hydrological Model Equifinality

Quantitative precipitation estimates (QPE), the main input driving hydrological model simulations, are known to have different levels of uncertainty across spatial and temporal scales. These uncertainties propagate through model simulations and thus lead to erroneous estimations of hydrological variables and extreme events. The role of equifinality—where different model structures or parameter sets produce similarly acceptable hydrological results—needs further research in the context of precipitation error propagation. Additionally, while data assimilation (DA) is a well-established method to improve model predictive performance by addressing various sources of uncertainty, its application to precipitation error propagation under the influence of model equifinality has received limited attention. This study investigates these gaps by leveraging the Raven hydrological modelling framework in combination with the dynamically dimensioned search (DDS) algorithm to calibrate streamflow at the outlets of multiple catchments across Southern Manitoba. Hence, different sets of optimized parameters are identified for each catchment, reflecting equifinality in the model structure and calibration. Subsequently, the calibrated model is driven by precipitation estimates from various satellite-based and reanalysis precipitation products to examine the propagation of precipitation errors through hydrological simulations. Finally, the study evaluates the effectiveness of streamflow data assimilation in correcting precipitation-induced errors in streamflow and improving the accuracy and robustness of the hydrological model. By systematically addressing the interplay between precipitation uncertainty, model equifinality, and data assimilation, this work provides novel insights into improving hydrological simulations.