SB: Importance of melting water in the Third Pole transboundary fluvial floods

River floods (fluvial floods) are a global concern, inflicting substantial harm on human safety and societal progress [1]. Unfortunately, river floods have been amplified by the increase in extreme precipitation events induced by global climate warming [2,3], including the Third Pole (TP) region. Furthermore, TP is home to the most extensive glaciers and snow cover outside the Arctic and Antarctic, supplying abundant meltwater to several major Asian transboundary rivers (e.g., Indus, Ganges-Brahmaputra, Salween, and Mekong) [4]. In addition to the increase in extreme precipitation, warming-induced accelerated melting of snow and ice has led to a significant increase in the risk of transboundary compound river floods (defined as river floods contributed to by multiple processes, Text S1 online) in TP [5,6]. These river floods are pertinent to multiple riparian densely-populated [7] and heavily-irrigated Asian countries [8], where there is considerable hydro-political tension [9]. Consequently, the increased risk of transboundary river floods in TP threatens the safety of nearly one billion people in downstream countries, as well as downstream agriculture [[4], [5], [6],8].

Many studies have analyzed the mechanisms of pluvial or nival floods in different basins of the world [1,[10], [11], [12], [13], [14]]. Pluvial floods are prevalent across most regions globally, primarily triggered by intense rainfall. However, studies on flood mechanisms in high-mountain basins reveal that snowmelt plays a non-negligible role in flood generation [[11], [12], [13], [14]]. Rising temperatures have accelerated snow cover ablation and increased the frequency of rain-on-snow events, exacerbating the risks of nival or mixed pluvial-nival floods in alpine basins [11]. Furthermore, snowmelt significantly contributes to shaping antecedent soil moisture preconditions, which modulate the occurrences and magnitudes of floods [[12], [13], [14]]. Despite these complexities, contemporary global flood modeling efforts often overlook glacier melt impacts. This oversight is particularly critical in the Third Pole region, where the interplay of rainfall, snowmelt, and glacier melt introduces exceptional challenges in understanding flood dynamics within transboundary river basins [6].

Previous modeling studies have consistently underscored the substantial contributions of glacier and snow melt to the annual total runoff in most TP basins and their great importance to downstream agricultural and social development [5,8,15]. However, most of the cryosphere-hydrological models used in these studies have a coarse temporal resolution (daily or coarser) and a simplified parameterization scheme for cryospheric processes (particularly the temperature-index glacier module), which cannot meet the requirements for the quantitative attribution analyses of compound river floods in TP (that request hourly or even finer-scale process-based studies) [6,15]. Consequently, the mechanisms (especially the role of glaciers and snow melt) of compound river floods in TP remain unclear.

To address this issue, we combine models and observations to conduct a comprehensive analysis of compound river floods during the period 1981–2020 in TP basins (Text S2, Table S1, and Fig. S1 online), using a high-resolution (hourly, 5 km) physically-based distributed cryosphere-hydrological model (WEB-DHM) (Text S3 and Fig. S2 online). This model, which incorporates a three-layer energy-balance snow module and an energy-balance glacier module for both clean and debris-covered glaciers, underwent rigorous evaluation and validation against gauge discharge observations (daily and hourly) and remote-sensing-derived glacier-area and snow-cover data, as well as reanalysis soil-moisture data (Texts S4, S5 and Figs. S3–S7 online). We carefully examined the roles of different runoff components (particularly glacier/snow melt, which are defined as the meltwater from glacier/snow induced by energy exchange, Text S3 online) in shaping the flood volume and mechanisms of compound river floods with this observation-constrained cryosphere-hydrology model. This analysis encompassed both annual-maximum and peak-over-threshold river floods (Text S6 and Fig. S8 online).

The contributions of snow and glacier melt to annual-maximum river floods vary substantially across different TP basins, with their joint proportions ranging from 1.1% to 52.3% (Fig. 1d and Table S2 online). This variation is primarily determined by the differences in precipitation characteristics and glacier storage within each basin. Compared with the upper Indus, the Mahakali and Upper-Ganges basins have greater glacier coverage (Fig. 1b), yet the contributions from glacier melt to annual-maximum river floods in these two basins are smaller, at 10.8% and 12.3%, respectively (Fig. 1d and Table S2 online). This is because heavy summer (June–August) rainfall in these two basins dilutes the contribution from glacier melt (Fig. 1c and Fig. S9 online). Similarly, although the long-term mean amount of glacier melt during annual-maximum river floods in the Gandaki basin (4.5 mm) is similar to that in the Upper Ganges basin (4.9 mm) (Fig. 1e), the contribution percentage of glacier meltwater in the Gandaki basin (5.5%) is only approximately half that of the Upper-Ganges basin (10.8%) (Fig. 1d) owing to much higher rainfall in the Gandaki basin (Fig. 1c and Fig. S9 online). Apart from the upper Indus basin, where snowmelt makes the largest contribution (32.3%), snowmelt also makes a substantial contribution of 18.4% to annual-maximum river floods in the upper Salween basin. The Upper-Ganges and Mahakali basins, which receive substantial winter (December–February) snowfall (Fig. S9 online), have snowmelt contributions of 9.5% and 12.5%, respectively, to annual-maximum river floods (Fig. 1d and Table S2 online). In the other six TP basins (Yamuna, Karnali, Gandaki, Koshi, upper Brahmaputra, and Mekong), annual-maximum river floods are predominantly driven by rainfall, accounting for over 90% of the contributions (Fig. 1d and Table S2 online).

We further quantified the specific roles of snow and glacier melt during both flood processes and pre-flood periods in each historical compound peak-over-threshold river floods in TP basins (Fig. 2a, b online). We found that the meltwater contributions displayed a notable degree of variability during the period 1981–2020 (Fig. 2a). Snow and glacier melt are substantial in some compound flood events, even in basins (Gandaki, Koshi, and upper Mekong) where annual-maximum river floods are predominantly driven by rainfall. In particular, in the upper Mekong basin, the maximum contribution of snowmelt to compound peak-over-threshold river floods approaches 30% (Fig. 2a). For the upper Indus basin, the contributions of glacier and snow melt to five historical compound peak-over-threshold river floods are 4.0%–42.6% and 2.8%–38.6%, respectively (Fig. 2a). In the upper Salween basin, the maximum contribution of snowmelt even exceeds that in the upper Indus basin, contributing 40.2% to the compound peak-over-threshold river floods of June 2001 (Fig. 2a).

In addition, we found that the role of snowmelt in shaping the preconditions of compound peak-over-threshold river floods is more crucial than its role during the flood processes in most TP basins, contributing more than 20% of water during pre-flood periods for nearly half of the 47 historical compound peak-over-threshold river floods (Fig. 2b), even in basins where rainfall dominates the contributions to flood processes during the compound peak-over-threshold river floods. For example, in the Yamuna and Karnali basins, although the direct contribution of snowmelt to flood processes is often negligible, snowmelt accounts for the large contributions (up to 22.8% and 30.7%, respectively) to the formation of preconditions suitable for floods during 1981–2020 (Fig. 2b), which further highlights the substantial contribution of meltwater to compound peak-over-threshold river floods in TP basins.

In summary, combining a high-resolution state-of-the-art cryosphere-hydrology model with observations, we reveal that glacier and snow melt exacerbate TP’s river floods through not only directly supplying floodwater but also shaping the catchment preconditions (enhancing relative soil moisture saturation). During the period 1981–2020, meltwater contributed over 20% to annual-maximum river floods in 4 of 10 studied basins. In particular, in the upper Indus basin, meltwater from glaciers and snow contributed 52.3% to annual-maximum river floods, surpassing the contribution of rainfall. Meanwhile, snowmelt accounted for over 20% during pre-flood periods in 23 of 47 historical peak-over-threshold river floods. Due to the complex terrain and harsh climatic conditions of the Third Pole basins, there are some uncertainties in quantifying meltwater contributions to flood events. These uncertainties primarily stem from errors in input meteorological data and simplifications in our meltwater contribution quantification framework (Text S6 online). Specifically for the physical model employed, beyond the well-documented uncertainties in precipitation and air temperature forcing, the parameterization of energy-related radiative forcing data (solar and downward longwave radiation) introduces additional variability in simulating snow/glacier ablation processes, thereby propagating uncertainties into meltwater contribution estimates. Overall, despite the uncertainty in the contribution values, we conclude that the role of snow and glacier melt in flooding is not negligible. These findings underscore the consideration of snow and glacier dynamics besides rainstorms in mountain flood predictions, which are critically important for achieving Sustainable Development Goals in the riparian countries (e.g., Pakistan, India, Nepal, and China).