The study, published in the journal Hydrology and Earth System Sciences, addresses a long-standing challenge in mountain hydrology. Snow and glacier meltwater sustain agriculture, drinking water supplies, groundwater recharge and hydropower across mountainous and high-latitude regions worldwide, and the cryosphere acts as a temporary water storage that buffers streamflow between wet, cold seasons and warm, dry ones. As temperatures rise, snow cover seasons are shortening and glaciers are losing volume across the Alps, shifting the timing and magnitude of melt contributions to rivers. Accurately separating how much streamflow comes from melting snow versus melting glacier ice is therefore essential for predicting how water availability will evolve, yet it remains one of the hardest quantities for models to pin down.

The research team built their framework around two established tools. The first is Raven, a flexible hydrological modeling framework in which they emulated the widely used conceptual HBV model. The second is the Global Glacier Evolution Model, or GloGEM, which simulates glacier mass balance and geometric change for individual glaciers worldwide using a temperature-index approach distributed over elevation bands. The coupling is one-way and mass-conserving: GloGEM produces daily glacier runoff, comprising ice melt, snowmelt and rainfall over the initially glacierized area, and this runoff replaces the precipitation input over glacier-covered parts of the catchment in Raven. The researchers implemented a new land-use category, called Masked Glacier, which adds the GloGEM runoff directly into the streamflow transfer function, bypassing intermediate soil and groundwater storages so that the timing and magnitude of glacier contributions are preserved.

On top of the coupling, the team integrated a snow redistribution scheme based on the SnowSlide concept, adapted for semi-distributed models. Conventional hydrological models typically ignore the fact that wind and avalanches continuously strip snow from exposed summits, cliffs and steep slopes. Without redistribution, models accumulate fictitious ever-growing snowpacks at high elevations, so-called snow towers, that store water which in reality would have moved downslope. The new scheme transports snow exceeding a slope-dependent holding depth from higher to connected lower-elevation units. Snow redistributed onto glacier surfaces in the uncoupled configuration delays ice melt, since no ice melt occurs while a glacier remains snow-covered, shifting melt from glacier ice toward snow. In the coupled configuration, such snow is removed instead, because GloGEM already accounts for snow processes on the glacier and double counting would otherwise occur.

The framework was applied across all gauged Swiss catchments with glacier cover but largely unaffected by hydropower infrastructure: 14 mesoscale headwater basins ranging up to 555 square kilometers, with mean elevations between 1811 and 2929 meters and glacier cover in the year 2000 ranging from 3.2 percent in the Rein da Sumvitg catchment to 60.1 percent in the Massa catchment. Six model configurations were calibrated for each catchment, varying in glacier representation, snow redistribution and the inclusion of a calibrated precipitation correction factor, a uniform multiplier that lumps together uncertainties such as gauge undercatch of solid precipitation, which can reach 50 percent, and poor constraints on high-elevation precipitation patterns. Calibration against daily streamflow used the Kling-Gupta Efficiency over 2000 to 2009, with validation from 2010 to 2020.

The validation produced a nuanced verdict. On streamflow metrics, the uncoupled HBV configurations came out slightly ahead, with median Kling-Gupta Efficiency values between 0.84 and 0.87 versus 0.81 to 0.87 for the coupled setups, a gap that largely vanished when the precipitation correction factor was included. On snow, however, the redistribution scheme proved decisive. Without it, catchment-average snow water equivalent diverged badly from the gridded SPASS reference dataset, with median root-mean-square errors of 243 and 184 millimeters for the two uncoupled redistribution-free configurations. Adding snow redistribution cut those errors to 142 and 134 millimeters, and per-elevation-band checks confirmed that the unrealistic high-elevation accumulation was eliminated.

The glacier melt comparison exposed the deeper problem. Evaluated against an independent Swiss dataset of annual glacier mass balances, GloGEM reproduced reference melt accurately, as expected given its calibration against geodetic mass balance derived from satellite-based elevation change measurements. The native HBV glacier module, by contrast, consistently overestimated melt in most catchments, in some cases exceeding both GloGEM estimates and observed glacier storage loss by factors of two to three. The researchers attribute this to compensatory error: the glacier routine offers extra degrees of freedom that the calibration exploits to offset underestimated precipitation, effectively using glacier ice as an adjustment variable to close the water balance. The uncoupled model wins the streamflow contest precisely by burying its forcing errors inside exaggerated ice melt.

Compounding the problem is severe parameter equifinality, the phenomenon in which many different parameter combinations produce similarly good streamflow simulations while implying radically different underlying processes. In the Rhone catchment, among the 20 best-performing parameter sets of the configuration with precipitation correction, simulated annual glacier melt contributions ranged from 11 to 38 percent of streamflow, and snowmelt contributions from 51 to 73 percent. Across catchments, glacier melt contributions varied by more than a factor of three among equally good parameter sets within a single configuration, and by up to an order of magnitude in weakly glacierized basins. This means a model can be judged excellent by conventional criteria while its water balance decomposition remains essentially unconstrained.

The coupled HBV-GloGEM configurations, despite their marginally lower streamflow scores, delivered glacier melt estimates consistent with the reference mass loss and substantially improved the identifiability of both glacier and snowmelt contributions. The authors are careful to note that the improved realism cannot be attributed to the coupling alone, since the coupled and uncoupled setups differ both in model structure and in the data constraining the glacier component, and GloGEM’s own calibration against geodetic mass balance is itself subject to some equifinality. Even so, the constraint works precisely because GloGEM resolves individual glaciers and can exploit per-glacier geodetic observations, something the native HBV routine, which treats glacier ice as an unlimited reservoir with a single degree-day parameter, cannot do.

For climate impact studies, the implications are stark. Models that perform well today through compensating errors misrepresent the present-day water balance and are unlikely to provide reliable projections as melt dynamics shift under continued warming, for example as catchments pass through peak water and glacier contributions decline. The authors suggest future refinements, including routing the ice-melt fraction of GloGEM runoff separately through a glacier storage, adding seasonal snowline and snow-cover observations as further constraints, and applying the framework to future climate scenarios. The core message, however, is already clear: in glaciated mountains, a good fit to historical streamflow is not enough, and independently constraining glacier evolution is what makes melt contribution estimates, and the water futures built upon them, trustworthy.

Subject of Research: Coupling a global glacier model with catchment-scale hydrological modeling to constrain snow and glacier melt contributions in Swiss Alpine headwaters

Article Title: Integrating a global glacier model into local hydrological modeling: Impacts on melt contributions

Article References: Berg, J., Horton, P., Kauzlaric, M., von der Esch, A., & Schaefli, B. (2026). Integrating a global glacier model into local hydrological modeling: Impacts on melt contributions. Hydrology and Earth System Sciences, 30(20), 6511-6530. https://doi.org/10.5194/hess-30-6511-2026

Image Credits: AI Generated

DOI: 10.5194/hess-30-6511-2026

Keywords: glacier melt, snowmelt, hydrological modeling, GloGEM, HBV, Raven, parameter equifinality, snow redistribution, Swiss Alps, geodetic mass balance, climate change impacts, water resources