Past and future drivers of glacier mass changes in High Mountain Asia and their impacts on catchment hydrology

In High Mountain Asia (HMA), glaciers and seasonal snowpacks have seen a widespread decline in the last two decades. However, these changes have been highly heterogeneous: glaciers of the Pamirs and Karakoram have seen positive mass balances in the early 2000s but are now on a trajectory of decline, while moderate and rapid mass loss occurs in the Central Himalayas and southeastern Tibetan Plateau, respectively. To understand these contrasting behaviours, their future trajectory, and their impacts on streamflow, we combine several years of situ observations and climate reanalysis with highly resolved land-surface and ice-flow models. We simulate water fluxes from 1970 to 2100 across three catchments with contrasting climatic settings. The catchments are in the Northwestern Pamirs (Kyzylsu), Central Himalayas (Trambau-Trakarding), and Southeastern Tibetan Plateau (Parlung No.4), spanning elevations from 2100 to 6800 m a.s.l.. 

We simulate the largest mass loss acceleration at Parlung No.4 Glacier, from  -0.10 m w.e. yr-1 in 1970-1999 to -0.80 m w.e. yr-1 in 2000-2023. The limited mass loss at Kyzylsu in 1970-1999 accelerated after 2000, mostly driven by a rapid worsening of glacier health in 2018-2024 (-0.72 m w.e. yr-1). Mass loss remained moderate at Trambau-Trakarding, from -0.39 m w.e. yr-1 in 1970-1999 to -0.30 m w.e. yr-1 in 2000-2023. 

These heterogeneous glacier mass balance patterns were driven by different summer 0°C isotherm changes, and contrasting precipitation decadal variability. Our projections forced with downscaled CMIP6 show that continued warming will expand ablation areas and extend melting into former accumulation areas previously located above the freezing line. On-glacier summer snowfall is projected to decrease by up to 60% by 2100 due to changes in precipitation phase. However, this will largely be offset by increases in snowfall in other seasons, driven by the widespread rise in precipitation projected for HMA. Despite stable accumulation rates, enhanced melt will drive severe glacier volume loss, strongly reducing the likelihood of new prolonged periods of near-stable or positive mass balances seen in our 1970-2020 simulations. 

Glacier retreat, combined with snowline rising to 6000 m a.s.l., will shift meltwater generation to higher elevations, where mass turnover rates will intensify. High-elevation areas will maintain their role as water providers, but with diminished capacity for water storage. We find that the evolving runoff contributions can decouple catchment-scale peak water from glacier-scale peak water. Indeed, substantial increases in rainfall (from 49 ± 22 % at the catchment in the Pamirs to 290 ± 146 % in the southeastern Tibetan Plateau), caused by precipitation phase and amount changes, and relatively stable snowmelt compensate for ice-melt reduction and postpone the timing of maximum runoff. While we find that ice melt nears its peak at Kyzylsu, with considerable uncertainty due to the initial ice volume, we project catchment runoff to continue increasing at all sites until the end of the century. Our results highlight the importance of considering the glacier 'peak water' concept within a catchment or basin hydrological framework.