Merits and limits of SWAT-GL: application in contrasting glaciated catchments

Abstract. The recently released SWAT-GL aims to overcome multiple limitations of the traditional hydrological model SWAT (Soil Water Assessment Tool) in glaciated mountainous catchments. SWAT-GL intends to increase the applicability of SWAT in these catchments and to reduce misapplication when glaciers have a significant role in the catchment hydrology. It thereby relies on a mass balance module, based on a degree-day approach similar to SWAT's snow melt module, extended by a glacier evolution component that is based on the delta-h (Δh) parameterization. The latter is a mass-conserving approach that enables the spatial distribution of ice thickness changes and thus dynamic glacier retreat. However, the extended SWAT version has not yet been comprehensively benchmarked. Hence, our paper aims to evaluate SWAT-GL with four different benchmark glaciers, which are part of the United States Geological Survey Benchmark Glacier Project. The benchmarking considers a comprehensive evaluation procedure, where SWAT-GL is optimized on glacier mass balance and hypsometry as well as snow cover. Snow cover is included to consider snow–glacier feedbacks. In addition, a sensitivity analysis using elementary effects (or the Morris method) is performed to give a detailed picture of the importance of the introduced glacier processes, as well as the relevance of the interactions with the already-existing snow routine. We intentionally did not include discharge in the optimization procedure to fully demonstrate the capabilities of SWAT-GL in terms of glacier and snow processes. Results demonstrate that SWAT-GL is able to represent the characteristics of contrasting glaciated catchments, which underlines SWAT-GL's applicability and transferability. We could further show its strong (non-linear) interactions with the existing snow routine, suggesting a simultaneous calibration of the snow components. While snow and glacier processes were adequately represented in the catchments, discharge was not necessarily represented sufficiently when excluded from the optimization procedure. However, SWAT-GL has been shown to be easily capable of reproducing discharge when used in a stand-alone optimization, although this may come at the expense of model consistency. Lastly, SWAT-GL significantly outperformed a standard SWAT model used for benchmarking purposes in high mountain environments.