Incorporating Freeze–Thaw Adaptations Into the <scp>HBV</scp> Model for a High Arctic Permafrost Catchment: Benefits and Limitations

ABSTRACT The widely‐used HBV hydrological model faces limitations in cold regions where permafrost and freeze–thaw dynamics significantly influence runoff processes. This study addresses these challenges by adapting the HBV model for the Fuglebekken catchment in Svalbard, a High Arctic basin undergoing rapid transition due to climate change. We introduced two simple, process‐based adaptations: (1) a seasonal ice reservoir to represent the freeze–thaw cycle of subsurface water, and (2) a routine to block infiltration and simulate the impermeability of frozen ground. To test these adaptations and investigate system non‐stationarity, four model configurations (the standard HBV and three adapted versions) were evaluated using a split‐sample test (SST) across two distinct periods, 2014–2018 and 2019–2024. Results show that while statistical improvements were modest, the adapted models tended to produce more physically consistent runoff simulations. The adaptations were especially effective at capturing the high‐flow dynamics of spring melt and autumn rain events, outperforming the standard model, particularly in years with otherwise poor evaluation results. The SST also provides compelling evidence of non‐stationarity. Model parameters were not transferable between the two periods, revealing a fundamental shift in the catchment's rainfall‐runoff relationship over the last decade. This study underlines the necessity of including process‐based adaptations in hydrological models for Arctic catchments in transition. Even though these simple additions are valuable, the observed non‐stationarity highlights the limitations of static calibration and the critical need for adaptive modelling strategies to ensure more reliable predictions in rapidly warming environments.