Hydrological model intercomparison in cold regions: challenges, insights, and future paths – a case from Nelson-MIP II

Intercomparison studies have evaluated both utility and constraints of existing hydrological and climate models worldwide. While some of these hydrological studies focused on cold Northern latitude regions, the initial phase of the Nelson Churchill River Basin (NCRB) Model Intercomparison Project (Nelson-MIP) was the sole effort to examine unregulated watersheds in this transboundary basin in Southwestern Canada and Northwestern United States. Thus far, a lack of clarity remains regarding the performance of various models in regulated watersheds around the globe and specifically within this complex region. Therefore, the primary aim of this study, constituting the second phase of the Nelson-MIP project, is to assess the performance of hydrological models in both regulated and unregulated watersheds within the NCRB, an area characterized by 95 reservoirs and managed lakes, numerous diversions and irrigation withdrawals, and significant areas of the North American Prairie region with millions of unmanaged lake and wetland systems. Four hydrological models (HYPE, BLENDED-LUMPED, RAVEN-HYPR, and RAVEN-UWFS) contribute to this phase and are set up using standardized input datasets to facilitate accurate comparisons. The performance of the models is evaluated against streamflow, actual evapotranspiration (AET), and snow water equivalent (SWE) simulations. Results reveal that HYPE and BLENDED-LUMPED models exhibit diminished streamflow performance across the Prairie region in comparison to the broader NCRB. While AET and SWE simulations were good across the entire NCRB (average Kling Gupta efficiency; KGE > 0.7), BLENDED-LUMPED encounters challenges in replicating AET patterns across the Prairies, a factor not reflected in the streamflow evaluation. This investigation highlights structural limitations within the participating models, particularly their insufficient representation of flow regulation and contributing area dynamics, which underscores the need for future enhancements. Moreover, this study underscores the complexities involved in conducting large-scale process-based intercomparison studies and offers strategies for addressing these challenges in future research endeavors. The models struggle with Prairie streamflow simulation despite calibration.Models lack accurate representation of (un)regulated water bodies.Models often compromise AET/SWE simulation to achieve good streamflow simulation.Future MIPs should adopt automated workflows to promote FAIR principles.Better benchmarking of WSC regulation data is needed. The models struggle with Prairie streamflow simulation despite calibration. Models lack accurate representation of (un)regulated water bodies. Models often compromise AET/SWE simulation to achieve good streamflow simulation. Future MIPs should adopt automated workflows to promote FAIR principles. Better benchmarking of WSC regulation data is needed.