Uncertainty Analysis of Internal Hydrological Processes in Two Agricultural Catchments in Ireland via Stochastic Calibration of <scp>SWAT</scp> +

ABSTRACT Hydrological models are a valuable tool for providing insights into hydrological processes, particularly at different scales for which observations are not available. Calibration and uncertainty analysis of hydrological models typically focus on streamflow at the catchment outlet. However, extension of uncertainty analysis to internal model surface and subsurface hydrological processes remains less commonly practiced. The primary objective of this paper is to propagate parameter uncertainty from stochastic model calibration for river discharge simulation to quantify, analyse and compare the modelled uncertainty of hydrological fluxes via different pathways and provide more robust insights into hydrological processes. This paper focuses on two agricultural catchments in Ireland representing different scales: the River Owenabue (143 km 2 ) and River Bandon (598 km 2 ) catchments. A sensitivity analysis is conducted to identify important SWAT+ parameters for hydrologic simulation. To identify behavioural parameter sets for uncertainty analysis of hydrological processes, stochastic calibration using the SUFI‐2 procedure in R‐SWAT and evaluation of SWAT+ model performance for daily river discharge simulation are carried out. Overall, the SWAT+ model achieved satisfactory performance for prediction of daily mean river discharge in the study catchments. Uncertainty analysis of hydrological fluxes identified lateral flow as an important hydrological pathway in the study catchments, particularly in the Bandon Catchment, while agricultural tile drainage was identified as an important hydrological pathway in the Owenabue Catchment. This paper demonstrates how quantification and comparison of lower and upper bound hydrological pathway contributions provide a more comprehensive understanding of specific hydrological processes driving streamflow generation in river catchments as well as the degree of uncertainty associated with individual processes across multiple behavioural parameter sets for streamflow simulation. The approach is transferable to other hydrological models, optimisation methods and catchments in different hydroclimatic regions.