Quantification of Different Uncertainties in Streamflow Simulations and Their Propagation Across Multiple Catchments

ABSTRACT Quantifying various sources of uncertainty in simulations of hydrological models is crucial for effective risk management and water resource planning. This study makes the assumption that the relative contributions of different sources of uncertainty in streamflow simulations, including input data, model structure and output observations, differ depending on the hydrological conditions of the catchment. To investigate this, we develop and apply a novel approach that incorporates time‐dependent random errors into observational data, in order to separate and estimate input and output uncertainties in streamflow predictions. Two distinct methods, namely the multiple model approach (MMA) and the multiple simulation approach (MSA), are used to quantify model structural errors. Cross correlation function (CCF) analysis is utilised to investigate the relationship between input uncertainties and streamflow simulations. The findings demonstrate that for four watersheds in arid regions, estimated uncertainty in precipitation contributed more to streamflow uncertainty than numerical differences across six conceptually similar lumped hydrologic models. During periods of low precipitation, model structural differences were the dominant source of streamflow uncertainty. The research also highlights significant limitations in using single models for simulating flows, revealing their inability to fully capture complex flood event dynamics. We also observe that, for some hydrological models, the forcing data and the corresponding response exhibit a delayed response, and we utilised this property to estimate uncertainty in streamflow simulations. The developed approach can prove beneficial in quantifying predictive uncertainties and supporting more informed decision‐making in water resource planning and risk management.