Evaluating Multi‐Model and Multi‐Metric Approaches to Low‐Flow Simulation in Irish Catchments

ABSTRACT Climate change has increased the stress on water resources, making reliable predictions, particularly of low flows vital for sustainable management, challenging. This study evaluates 47 hydrological models in 125 catchments in the Republic of Ireland with the MARRMoT framework and uses two low‐flow objective functions (inverse Kling–Gupta efficiency and log‐transformed Nash–Sutcliffe efficiency). The results show that the moderate complexity models (e.g., GR4J, HYMOD, Collie3) performed best in both the calibration and validation periods. By comparing objective functions, the log‐transformed Nash–Sutcliffe efficiency demonstrated greater stability between the validation and calibration periods (mean difference = 0.007, SD = 0.093) compared to the inverse Kling–Gupta efficiency (mean difference = −0.204, SD = 0.233). The performance of the models was also evaluated against several catchment attributes that represent the climate, topography, land use and hydrogeology of the catchment using the Spearman correlation coefficient. Higher model performance was generally associated with lower mean annual precipitation and higher potential evapotranspiration. Statistically significant positive correlations were also observed between model performance and catchment steepness, whereas negative correlations were found with base‐flow index and reservoir storage. In general, our findings advocate for multi‐objective calibration, ensemble modelling and improved representations of the groundwater, wetlands and urban hydrology process to improve the prediction of low flows.