Modeling Potential Glacial Lake Outburst Flood Process Chains and Effects From Artificial Lake‐Level Lowering at Gepang Gath Lake, Indian Himalaya

Glacial lake outburst floods (GLOFs) are a severe threat to communities in the Himalayas; however, GLOF mitigation strategies have been implemented for only a few lakes, and future changes in hazard are rarely considered. Here, we present a comprehensive assessment of current and future GLOF hazard for Gepang Gath Lake, Western Himalaya, considering rock and/or ice avalanches cascading into the lake. We consider ground surface temperature and topography to define avalanche source zones located in areas of potentially degrading permafrost. GLOF process chains in current and future scenarios, also considering engineered lake lowering of 10 and 30 m, were evaluated. Here, varied avalanche impact waves, erosion patterns, debris flow hydraulics, and GLOF impacts at Sissu village, under 18 different scenarios were assessed. Authors demonstrated that a larger future lake does not necessarily produce larger GLOF events in Sissu, depending, among other factors, on the location from where the triggering avalanche initiates and strikes the lake. For the largest scenarios, 10 m of lowering reduces the high‐intensity zone by 54% and 63% for the current and future scenarios, respectively, but has little effect on the medium‐intensity flood zone. Even with 30 m of lake lowering, the Sissu helipad falls in the high‐intensity zone under all moderate‐to‐large scenarios, with severe implications for evacuations and other emergency response actions. The approach can be extended to other glacial lakes to demonstrate the efficiency of lake lowering as an option for GLOF mitigation and enable a robust GLOF hazard and risk assessment.

Long-term stability conditions of icy peaks primarily depend on geology (lithology, structure, etc.), topography (especially slope inclination), and ice conditions (permafrost and glaciers). For more than a century, ice conditions have undergone rapid and deep-reaching changes, thereby reducing the stability of icy peaks . With this, slopes of these icy peaks, which were already in sub-critical conditions of failure considering their geology and topography, can now pass critical thresholds and fail. Advanced laboratory experiments document that the stability of frozen rocks with ice-filled cracks strongly decreases with rising sub-zero temperatures (Davies et al., 2001;Krautblatter et al., 2013). Particularly, critical failure conditions can occur in "warm permafrost" between about −1 and 0°C when rock, ice, and water coexist. Such critical conditions could have existed in the detachment zone of the Chamoli event 2021 in Uttarkhand, India , the Salkantay event 2020 in Peru (triggering a GLOF; Vilca et al., 2021), the Cengalo event 2017 in the Swiss Alps (Mergili, Jaboyedoff, et al., 2020;Walter et al., 2020), or the 2010 Hualcán GLOF event of Lake 513 in Peru (Carey et al., 2012). With continued warming, these critical conditions will further expand in space, climbing up icy peaks and destabilizing increasing parts of them in a "bottom-up" process . Due to the declining stability of high-mountain slopes resulting from permafrost degradation and glacial debuttressing Haeberli et al., 1997), the hazards associated with lakes located near potentially unstable slopes with a high possibility of having lake impacts are of great concern (Allen et al., 2019;. More importantly, as these lakes grow, they could approach unstable slopes, and further degradation of the moraine causing future hazard levels to change as the lake becomes more susceptible to impacts and Several Himalayan studies have identified potentially hazardous lakes (Dubey & Goyal, 2020;Mal et al., 2021;Rounce et al., 2016;Worni et al., 2013). In the western Himalaya, especially in the Indian state of Himachal Pradesh, studies on temporal lake changes show the Gepang Gath Lake to be one of the largest and fastest-growing glacial lakes (Kumar et al., 2021;Patel et al., 2017). GLOF risk assessments identified the lake to be highly susceptible to mass movement impact (S. K. Prakash & Nagarajan, 2017;Worni et al., 2013), while glacier bed mapping shows that the lake is expected to grow and double its size in the future (S. K. Pandit & Ramsankaran, 2020;present study). Therefore, evaluating the future hazard of the lake is of paramount importance. Previous GLOF modeling considering an ice avalanche impact volume (initiating from the same avalanche source area as considered in this study) of 4.7 M m 3 on the lake suggests moderate damage potential to Sissu village (Worni et al., 2013). However, there remains a gap in understanding the impacts of large hazardous events such as the recent Chamoli disaster (rock-ice avalanche of 25 M m 3 ), highlighting the impact of extreme events on glacial lakes in the Himalaya . What also remains is the gap in understanding the GLOF process chains (Emmer et al., 2022) and hazard reduction of larger lakes (similar to Gepang Gath) and how modeling approaches can be used to evaluate the effects of lake-level lowering.
Our study aims at understanding two critical questions on the GLOF hazard of the Gepang Gath Lake in the Himachal Pradesh, Western Himalaya (cf. Figure 1): 1. What is the current and future hazard potential of the lake? Here, we perform a detailed hazard assessment by identifying potentially unstable slope conditions around the lake based on mass movement susceptibility and permafrost/surface ice occurrence. This is followed by detailed process chain modeling of ice-rock avalanches of different magnitudes (low, moderate, and large), lake impact, erosion at the dam, and downstream propagation of the potential outburst floods (Figure 2). 2. What is the relative change in the potential downstream GLOF impact when the lake level is lowered? To address this, we consider two lake-lowering scenarios to evaluate the GLOF process chains and downstream impact: Scenario L10-the lake is lowered by 10 m compared to the current lake level. Scenario L30-the lake is lowered by 30 m compared to the current lake level (Figure 2). The change in the downstream GLOF impact and exposure is evaluated for different lake-lowering scenarios combined with different GLOF process chains.

Study Region and Its Importance
The Gepang Gath Lake is located at 32°31ʹ48″ N and 77°12ʹ31″ E, at an elevation of 4,080 m a.s.l in the Chandra Basin of Himachal Pradesh, Western Himalaya. The mean mass balance of the Chandra basin is estimated to be winter, pre-monsoon, monsoon, and post-monsoon respectively (Kaushik et al., 2020). An increasing trend in precipitation is observed in the monsoon season by 0.890 mm yr −1 . However, there is an overall fall in annual precipitation by 2.74 mm yr −1 (Kaushik et al., 2020). The nearest village to the lake, Sissu, is located ∼12 km downstream of the lake and has an elevation difference of ∼1,000 m from the lake (Figure 1). The village is located in the path of potential GLOFs from the lake (Kaushik et al., 2020;Worni et al., 2013). The Gepang Gath Lake has evolved from a small supraglacial lake to a large proglacial lake ( Figure S1 in Supporting Information S1). High negative mass balance (Tawde et al., 2017) led to downwasting and retreat of the glacier, causing rapid growth of the lake. Another dominating factor in lake expansion is ice loss due to calving at the glacier snout (Prakash & Nagarajan, 2017). The glacier's surface shows extensive transverse crevasses around the terminus, where it is in contact with lake water. These crevasses extend up to 0.5 km up the glacier from the terminus. The ablation zone is covered with debris that ranges from a very thin layer to a thickness of less than 1 m, containing a combination of cobbles, pebbles, and boulders (based on field investigation). The lowermost portion of the glacier shows a thin debris layer compared to the thick debris generally present at the terminus of many Himalayan glaciers. Ice cliffs and marginal crevasses are also present on the glacier, indicating complex glacier dynamics and the variable impact of supraglacial debris.

Field Investigation and Remotely Sensed Data
The field expedition to the Gepang Gath glacier was carried out on 25 June 2018. It was seen that the frontal area of the glacier contained debris cover ranging from heavy boulders to smaller debris, but no finer sediments were observed. Supraglacial debris thickness was measured manually at four locations in the frontal part of the glacier ( Figure S4.2 in Supporting Information S1). Crevasses as wide as 5-7 m were seen near the glacier terminus, depicting an active calving front. The lateral and frontal moraines comprised finer deposits compared to the supraglacial debris, and its hummocky surface suggests the presence of buried ice. The non-glaciated lateral sides were observed to be as steep as around 45°. Clean glacier ice was visible toward the higher reaches of the glacier on a gently sloping surface, followed by a steep glacier head. High streamflow is observed in supraglacial streams that reach the glacier terminus, contributing meltwater to the Gepang Gath Lake. Lake depths were measured manually in two locations at the frontal shallow end of the lake. The field-based information was further used to define the input parameters in the GLOF process chain modeling (see Section 3.4). A variety of remotely sensed data is used in the study, including Advanced Land Observing Satellite (ALOS)-Phased Array type L-band Synthetic Aperture Radar (PALSAR) digital elevation model (DEM) for topographic analysis and GLOF modeling. The DEM is a radiometrically terrain-corrected elevation product with a spatial resolution of 12.5 m, and was released globally in October 2014 by the Alaska Satellite Facility (https://asf.alaska.edu/datasets/derived-data-sets/alos-palsar-rtc/alos-palsar-radiometric-terrain-correction/). It has been previously used for dynamic GLOF routing in GLOF studies (Maskey et al., 2020;. For slope mapping and GLOF exposure mapping, imageries from Google Earth and PlanetScope, respectively, were utilized. Ice thickness data of the Gepang Gath glacier were obtained from Farinotti et al. (2019), and the ALOS PALSAR DEM has been used to map the frontal overdeepening in the bed of the glacier.

Bathymetric Reconstruction, Present, and Future Lake Extents
Due to the unavailability of the lake's in situ depth measurements, we rely on alternative approaches to derive the present and future lake volumes. Field-based lake depth measurements at point locations in the frontal part and analysis of high-resolution satellite imagery suggest that the lake has a shallow front, also evident from the partly exposed lake bottom sediments ( Figure 3f). Also, the accumulation of floating blocks of ice along an arc in the frontal part of the lake (approx. 400 m inwards from the frontal moraine) suggests shallow lake depths (Figures 3c and 3f). The spatially distributed bathymetry of the lake was derived from the glacier bed topography by subtracting the modeled distributed glacier thickness from the glacier surface elevation (DEM) for the year 2000 (consensus estimate from Farinotti et al. (2019)). Initially, the current lake extent (2020) is mapped using the latest Landsat-8 satellite imagery ( Figure S1 in Supporting Information S1), followed by the extraction of depth contours of the frontal overdeepening within the current lake boundary (Farinotti et al., 2019) (Figure 3a). Further, these depth contours are extrapolated and merged with the depth contours of the lake's frontal portion, and reconstructed based on field observations and visual interpretation of remotely sensed data to derive the 6 of 26 spatially distributed bathymetry of the current lake. We further employ nine different empirical approaches to evaluate the reconstructed lake volume (Table S1 in Supporting Information S1). Similarly, the future lake bathymetry (maximum possible lake extent) is derived by merging the spatially distributed depths of the entire frontal overdeepening of the Gepang Gath glacier and the reconstructed present lake bathymetry (as in 2020). A similar approach to estimating lake volume using ice-thickness and surface DEM has been employed in previous glacial lake studies (e.g., Sattar, Goswami, et al., 2021).

Mapping Mass Movement Source Zones: Topographic Potential and Permafrost Characteristics
In the present study, the potential avalanche source zones were identified based on combined criteria of topographic potential and permafrost occurrence on the surrounding slopes of the Gepang Gath Lake. Here, the present and the future lake extents (see Section 3.2) were used to calculate the topographic potential of the surrounding slopes based on the angle of reach of mass movements to the lake. We consider a slope threshold of 30° to identify the surrounding slopes, susceptible to mass movements. The concept of topographic potential assumes an impact on a lake is possible from any slope >30°, from which the overall slope trajectory to the lake is >14° (tanα = 0.25) (Allen et al., 2019;Sattar, Goswami, et al., 2021). These conservative values are based on typical ice and/or rock avalanches reported globally, although mass movements from gentler slopes or with longer runout distances are possible in exceptional cases (cf. Kääb et al., 2021).
For permafrost mapping, we use outputs from S. K. Allen, Fiddes, et al. (2016) to derive spatially distributed permafrost zones, where state-of-the-art numerical models were driven by downscaled atmospheric data sets (ERA-Interim), enabling large area simulations of the ground surface temperatures at high spatial resolution. The permafrost distribution outputs from S. K. Allen, Fiddes, et al. (2016) extend to the Gepang Gath glacier valley. Permafrost conditions are inferred to prevail where modeled Mean Annual Ground Surface Temperature (MAGST) is calculated below 0°C. We further compare the inferred permafrost distribution with large-scale modeled permafrost data sets (Obu et al., 2019;Ran et al., 2022) and with local occurrences of viscous creep features (rock glaciers) in perennially frozen talus/debris. Here, we identify potential threats to the lake as mass-movement impacts originating from steep slopes, based on the calculated topographic potential of the surrounding slopes and modeled the permafrost distribution of the slopes. The slopes with the highest topographic potential and coexisting permafrost were identified as a source for low-and moderate-magnitude avalanches. For a large magnitude avalanche (Chamoli type, cf. Shugar et al. (2021)), we identify the nearest threat to the lake based on both the highest topographic potential and permafrost occurrence. Further, we employ high-resolution Google Earth imagery (dated 12 September 2019) for visual evaluation of these source zones and to map speculatively those slopes that exhibit permafrost creep as a sign of previous slope failures as well as to identify areas with a high density of crevasses, cracks, and scarps depicting previous slope failures.

GLOF Process Chain Modeling for the Gepang Gath Lake
Here, we employ the open-source mass flow simulation tool r.avaflow  for detailed modeling of complex chains of processes. Therefore, we use the Pudasaini and Mergili (2019) multi-phase model for an integrated simulation of (a) propagation of the initial avalanche; (b) avalanche-lake interaction, that is, formation and propagation of an impact and displacement wave; (c) overtopping and retrogressive erosion of the moraine dam; and (d) debris flow or hyperconcentrated flow downstream. Here, debris flow or hyperconcentrated flow events were considered for the Gepang Gath Lake considering that (a) the valley is devoid of dense vegetation, making the valley floor susceptible to erosion, and (b) the moraine is hummocky in nature with available glacial debris in front of the moraine that can be subjected to erosion during a GLOF event.
Depending on the GLOF process chain scenario (see below), we assume the mixture of one or two solid phases (rock component of the initial avalanche and entrainable material as phase 1; ice component of the initial avalanche as phase 2) and one fluid phase (lake water as phase 3). A simplified entrainment model is employed, a product of the flow momentum and the empirical entrainment coefficient . The parameterization of the simulations builds on the back-calculation of various well-documented GLOF events (Mergili, Jaboyedoff, et al., 2020;Mergili et al., 2018;Vilca et al., 2021;Zheng, Mergili, et al., 2021).
As inputs to the mass-movement model, we consider avalanche scenarios (referred to as acronym "SC" here) of varying magnitudes with release thickness/volume of 25 m/6.2 M m 3 (low-magnitude; SC-1), 50 m/12.4 M m 3 (moderate-magnitude; SC-2), and 100 m/25 M m 3 (large-magnitude; SC-3) originating from the surrounding susceptible slopes (as identified in Section 3.3) ( Table 1). The above-mentioned avalanche scenarios with different magnitudes were considered to evaluate GLOF process chains for the present and future conditions of the Gepang Gath Lake. As these potential avalanche source zones can have overlying glacier ice, their ice-rock-volume ratios are calculated based on the total glacier-volume present within the potential failure zones (Farinotti et al., 2019).
Here, the large-magnitude avalanche scenario (100 m/25 M m 3 ) is chosen to represent the events experienced in the near past in the Himalaya .
Further inputs to the model include lake dimensions (bathymetry) for the present and the future. The reconstructed bathymetry (see Section 3.2) are used here as inputs, combined with the scenarios of lake evolution: we do not only consider the current lake (SC-1-L0, SC-2-L0, and SC-3-L0) but also the future lake (SC-4-L0, SC-5-L0, SC-6-L0) without lake lowering to evaluate the GLOF process chains ( Figure 2, Table 1). In the present study, we consider scenarios of GLOF process chains with not only varied (a) volume and material composition of the initial   avalanche; but also (b) lake level lowering modeling (see next section) that accounts for the range of possible events/uncertainties in the initial avalanche and possible downstream impact.
The erosion parameters in the model are defined based on the available entrainable material of the moraine and glacier. The frontal moraine of the Gepang Gath Lake exhibits a relatively flat geometry with a crest width of ∼150 m. Further, Worni et al. (2013) reported that a significant overtopping flow could initiate the erosion process at the southern end of the lake's frontal moraine and eventually erode the dam. In many of the reported GLOFs globally, it is seen that the damming moraine is subjected to erosion upon the impact of the displacement wave that overtops the moraine. In that context, assuming that an avalanche-induced displacement wave would initiate erosion at the dam crest followed by retrogressive erosion of the moraine, we define the section of the moraine above lake level as entrainable material having a grain density of 2,700 kg m −3 (see Figure S4.1 in Supporting Information S1). However, the final erosion depths would depend on the momentum of the particular process and are controlled by the entrainment coefficient (see below). Here, the lake is assumed to contain clear water with a density of 1,000 kg m − ³. We do not account for the presence of sediments at the lake's bottom surface as such assumptions are complex and not supported by the available data and have minimum influence on the model results considering the overall GLOF process.
Furthermore, our field investigation shows that the frontal part of the Gepang Gath glacier along the trajectory of the potential avalanche modeled in this study is covered with loose supraglacial debris surrounding the Gepang Gath Lake. These deposits mostly comprise glacial till, rock fragments, and lateral morainic material ( Figure  S4.2 in Supporting Information S1). Assuming that an avalanche can entrain these loosely available materials on the glacier surface, we define the available debris as potential entrainable material in the model based on Rounce et al. (2021). We assume that the supraglacial debris is potentially erodible. However, erosion only occurs when the avalanche interacts with the supraglacial debris (i.e., over the flow surface of the avalanche) and the potentially eroded debris varies depending on the momentum/energy of the avalanche (scenarios). Here, the supraglacial debris is mostly exposed to avalanches originating from the headwall of the Gepang Gath glacier (SC-1 and SC-2). Whereas in SC-3, SC-4, SC-5, and SC-6 there is very minimal interaction of the avalanche with the supraglacial debris ( Figure S4.2 in Supporting Information S1) as it initiates from the sidewall directly impacting the lake. Also, the spatially distributed debris thickness is less than a meter (ranging between 0.01 and 0.75 m) in the major portion of the debris-covered region that interacts with the avalanches (SC-1 and SC-2) ( Figure  S4.2 in Supporting Information S1). Higher debris thickness is seen in a small portion along the periphery of the lake (northern shore of the lake), which is directly exposed to high energy rock avalanche considered in this study (SC-3 and SC-6). It is likely that this supraglacial debris layer will be subjected to erosion considering the magnitude of the avalanche considered in the study. Considering the nature of debris as investigated in the field, the grain density of the entrainable material is assumed to be 2,700 kg m −3 . The final entrainment of this debris, however, depends on the entrainment coefficient and the momentum (see below). Thus, we construct a detailed process-chain model to evaluate the potential impact of avalanche material on the lake and the influence of eroded material on the avalanche-lake interaction and the downstream flood or hyperconcentrated flow.
Other input parameters include basal friction angle (φ) and internal friction angle (δ) that govern the rheology of the flow. Here, we set φ = 25°, δ = 10° for the initial stage of the process chain dominated mostly by the solid phase, that is, avalanche, avalanche impact on the lake, and moraine erosion. For the downstream process from the moraine to the village of Sissu and further, we set φ = 25°, δ = 1° to model the flow as a water-saturated debris flow. The entrainment coefficient at the moraine dam and the supraglacial debris was set to 10 −6.5 . The flow channel from the lake to Sissu village is devoid of dense vegetation, as interpreted from satellite imagery and also witnessed in the field. Therefore, uniform channel friction (similar to Manning's n) of 0.05 is considered along the channel down to the Sissu village. All the simulations are executed at a spatial resolution of 20 m for efficient processing, and the total duration is set to 45 min (2,700 s) from the avalanche initiation. The domain of the model is constructed such that it completely includes the avalanche sources and the Sissu village. Finally, to evaluate the plausibility of the model results obtained in terms of flow depth and discharge, we define three cross-sections (the cross-sections are referred as XS in the manuscript) along the flow channel located 1 km (immediately downstream of the moraine; XS-1), 9 km (before Sissu village; XS-2), and 11 km (after Sissu village; XS-3) downstream of the lake. To evaluate the erosion dynamics at the moraine site, we define one reference point (R1) located at the crest of the moraine (the highest point of the moraine). We further calculated the total eroded volume, including the entrained supraglacial debris and the eroded material from the moraine. The modeled flow heights are converted to GLOF intensity levels.

Modeling Lake Level Lowering
Apart from the process chains modeled with no lake lowering (SC-1-L0 to SC-6-L0; Section 3.4), we evaluate the GLOF process considering scenario-based lake lowering of Gepang Gath Lake where the lake level is lowered. Aiming to understand the changes in the flow hydraulics and potential downstream impacts, we repeat the process chain simulations described in Section 3.4 for GLOF scenarios where the lake is lowered to defined depths both for the present and the future. The lowered lake bathymetry ( Figure S3 and Table S2 in Supporting Information S1) are further used as model inputs for process chain modeling to evaluate the associated change/ reduction in the downstream GLOF impact (Section 4.3). We consider two lake lowering scenarios, L10-the lake is lowered by 10 m, and L30-the lake is lowered by 30 m, and apply these scenarios to each process chain model (SC-1 to SC-6) (see Table 1 for lake lowering scenario definitions). The inputs and model parameters given in Section 3.4 were applied, except for the lake bathymetry, which were obtained by lowering of the original lake surface. Thus, the freeboard increases by 10 and 30 m in the L10 and L30 scenarios in the lake lowering modeling, respectively. Further, intercomparison of the flow hydraulics and erosion dynamics was performed for the no-lowering (L0) and lake-lowering scenarios (L10 and L30). To illustrate the effect of lake lowering on downstream GLOF impact and exposure, modeled flow heights were converted to GLOF intensity levels. Following the approach outlined in GAPHAZ (2017), the zone of high GLOF intensity is defined where the modeled flow height (H) > 1 m and the moderate intensity zone where H < 1 m.

Bathymetric Reconstruction and Lake Lowering
The modeling of the frontal overdeepening of the Gepang Gath glacier and further extraction of the region lying within the lake extent of 2000 and 2020 indicates a maximum depth of ∼114 m where a deep depression on the glacier bed is seen. The total lake volume calculated for this section is 28 × 10 6 m 3 (Figure 3b). The shallow nature of the lake front can be inferred from the exposed lake-bed sediments and the existence of stranded ice chunks along a transverse arc in the frontal portion of the lake (satellite photo in Figure 3f). These stranded ice chunks can be due to either early formation of lake ice ( Figure 3c) on a shallower part of the lake or accumulation of calving ice originating at the terminus ( Figure 3f). The depth within this part of the lake varies from ∼0 to ∼1 m. The uneven lake bottom topography in the frontal part resulting from the hummocky nature of the lake bed leads to varying water depths at places. GIS-based spatial interpolation performed to derive the spatially distributed lake bathymetry for the present lake resulted in a total volume of 28.4 × 10 6 m 3 (Figure 3d). This volume, when compared to the ensemble mean of the nine empirical estimates (see Table S1 in Supporting Information S1), has a difference of ∼17.8%. Our lake volume estimates are substantially higher than the volume estimated for the early 2000s by Worni et al. (2013) because the lake has grown 65%, from 0.58 to 0.96 km 2 in 2020.
Future lake bathymetry is derived by merging the present lake bathymetry and the frontal overdeepening of the Gepang Gath glacier (extracted for the year 2020) (Figures 3a and 3e). The present lake is 2.31 km long along its major axis and its anticipated future growth of another 2.28 km is indicated (Figures 3d and 3e). The volume of the frontal overdeepening calculated for the year 2020 is 28.2 × 10 6 m 3 . The total volume of the future lake (volume of the frontal overdeepening + present lake volume as calculated above) is calculated to be 56.6 × 10 6 m 3 . The reconstructed present and the future bathymetry are further used for the process chain modeling. It is to be noted that lake volumes are based on the lake bed and not the lake bottom sediment surface.
Lake lowering by 10 m (L10) and 30 m (L30) of the present lake results in a total reduction in the lake's volume by 15.5 × 10 6 and 22.4 × 10 6 m 3 , respectively. Similarly, lake lowering of the future lake resulted in a 41 × 10 6 m 3 (L10) and 50.7 × 10 6 m 3 (L30) reduction in the future volume.

Mapping Mass Movement Source Zones-Topographic Potential and Permafrost Characteristics
Based on the calculation of topographic potential (see Section 3.3) of the surrounding slopes, we identify slopes with a high angle of reach to the present lake (tan α = 0.25-0.80) located at the glacier headwall ( Figure 4a). We identified this site as a source of low (SC-1) and moderate magnitude (SC-2) avalanches that can impact the current lake ( Figure 4d). The identified site shows evidence of previous mass movements, indicated by freshly exposed scarps and lineaments (Figure 4e). Since this zone lies within the glacier boundary, the avalanches are assumed to be a mixture of ice and bedrock. The ice-rock ratio depends on the total ice present (Farinotti et al., 2019) within the avalanche release zone (see Table 1). Further, the slope with the highest topographic potential (tan α = 2.0 to 2.41) lies slightly north-east of the lake at a distance of 1.8 km from the lake (Figure 4a). This slope lies outside the Gepang Gath glacier boundary and is characterized by permafrost with higher MAGST (0°C to −2°C) ( Figure 4c). Permafrost mapping indicates that permafrost may extend down to ∼4,000 m a.s.l in favorable instances, which compares well with the observed lower elevation limit from the mapped rock glaciers around the glacier basin (Figures S2.1-S2.3 in Supporting Information S1). The high-resolution Google Earth imagery investigation indicates that this permafrost slope has developed huge lateral fractures, and a transverse crack is observed from the top (Figure 4g). This is similar to that reported in the Chamoli event, where crack development at the periphery of the failure zone was observed and reported as precursory signs of slope failure . Here, we identify this slope as a potential source of large magnitude (SC-3) avalanche consisting of only bedrock and no ice that can potentially impact the current lake.
Similarly, for future scenarios, we evaluate the topographic potential of the surrounding slopes calculated considering the future lake extent. The identified slope for low magnitude (SC-4) and moderate magnitude (SC-5) avalanches is located in a region with a high angle of reach to the future lake (tan α = 0.77 to 1.20) (Figures 4b  and 4d). The total ice within the release areas is used to calculate the ice-rock ratio (Table 2). For a large magnitude scenario of a potential future avalanche, we selected the same slope as that of SC-3 characterized by permafrost and the highest angle of reach to the future lake (tan α = 0.77 to 1.80) (Figures 4b and 4c).

Current Scenarios
Evaluation of the present GLOF hazard of the Gepang Gath Lake reveals that, without lake lowering, potential impact from a low magnitude avalanche with a volume of 6.2 Mm 3 (SC-1-L0) produces a GLOF peak of 23,475 m 3 s −1 immediately downstream of the lake (at cross section XS-1, located 1 km downstream of the moraine; see Figure 6). Here, the maximum modeled flow depth is ∼5.5 m. The GLOF outflow peak reduces by 63% when the lake is lowered by 10 m (SC-1-L10) and 90% when lowered by 30 m (SC-1-L30) ( Figure 6, Table 2). Similarly, the flow depths immediately below the moraine reduce by 69% and 70% when the lake is lowered by 10 and 30 m respectively. The GLOFs terminate before reaching Sissu (at cross section XS-2, located 8.5 km downstream of the moraine; see Figure 6) in lowering scenarios (L10 and L30) ( Figure 5). In a low-magnitude GLOF process chain without lake lowering (SC-1-L0), the maximum erosion depth of the frontal moraine is 1.9 m. The erosion depth decreases by 62% and 87% as the lake levels are lowered by 10 and 30 m respectively (Figure 9a).
In moderate magnitude GLOF process chain scenarios (SC-2) where an avalanche with a volume of 12.4 M m 3 potentially impacts the lake without lake lowering, the outflow peak immediately downstream of the lake reaches up to 60,119 m 3 s −1 , with a maximum flow depth similar to SC-1-L0 (∼5.5 m). The GLOF peak reduces by 64% and 81% when the lake is lowered by 10 m (SC-2-L10) and 30 m (SC-2-L30), respectively ( Figure 6, Table 2). At Sissu the peak discharge reduces by 89% in 10 m lowering scenario (SC-2-L10). The GLOF terminates before reaching Sissu only in the 30 m lowering scenario ( Figure 5). Also, in the moderate-magnitude GLOF process chain without lake lowering (SC-2-L0), the maximum erosion depth of the frontal moraine is 3.7 m. The erosion depth reduces by 50% and 83% when the lake is lowered by 10 and 30 m respectively (Figure 9a).
In a large magnitude GLOF process chain, where an avalanche with a volume of 25 Mm 3 potentially impacts the lake without lake lowering (SC-3-L0), it produces a GLOF peak of 190,190 m 3 s −1 immediately downstream of the lake. Here, the maximum modeled flow depth reaches up to ∼12.9 m. The GLOF peak reduces by 27% when the lake is lowered by 10 m (SC-3-L10) and 88% when lowered by 30 m (SC-3-L30). At Sissu, the peak discharge reduces by 81% and 98% when the lake is lowered by 10 m (SC-3-L10) and 30 m (SC-3-L30) respectively. Similarly, the maximum flow depth at Sissu reduced from 14.3 to 11.2 m (with 10 m lowering) and 5.7 m (with 30 m lowering) (Figures 5 and 6, Table 2). The erosion of the frontal moraine is maximum (13.5 m) in the large magnitude GLOFs (SC-3) when compared to the moderate and low magnitude scenarios. The maximum erosion depth reduces by 55% and 78% when the lake is lowered by 10 and 30 m, respectively. It is to be noted that the distance from the impact site (where the avalanche first interacts with the lake; see Figure 6b) to the frontal moraine for all present GLOF scenarios (SC-1 to SC-3) is 2.28 km. The displacement waves are unidirectional in all present GLOF scenarios, that is, from the impact site toward the frontal moraine (Figure 6b).

Future Scenarios
Evaluation of the GLOF future hazard of the Gepang Gath Lake reveals that, without lake lowering, a low magnitude avalanche impact on the future lake (SC-4-L0) produces a GLOF peak of 8,274 m 3 s −1 immediately downstream of the lake (at cross section XS-1, located 1 km downstream of the moraine; see Figure 8). Here, the maximum flow depth reaches up to 3.2 m. The GLOF terminates before reaching Sissu (Figure 7). Lake-lowering scenarios reveal that low magnitude (SC-4) scenarios do not result in overtopping of the moraine when the lake is lowered by 10 m (SC-4-L10) and 30 m (SC-4-L30) (Figure 7). In low magnitude GLOF process chain, the erosion of the frontal moraine is very minimal in the future (less than 0.5 m) even without lake lowering (SC-4-L0). There is no erosion at the moraine when the lake is lowered by 10 and 30 m (Figure 9b) as the GLOF wave does not overtop the damming moraine (also see discussion Section 5.2).
The moderate magnitude future GLOF process chain scenarios (SC-5) show that without lake lowering the GLOF peak immediately downstream of the moraine (SC-5-L0), reaches up to 58,091 m 3 s −1 with a maximum flow depth of 6 m (Figures 7 and 8, Table 2). At Sissu (at cross section XS-2, located 8.5 km downstream of the moraine; see Figure 8), the GLOF peak is 19,856 m 3 s −1 with a maximum flow depth of 7.5 m. There is no overtopping when the lake is lowered by 10 m (SC-5-L10) and 30 m (SC-5-L30) resulting in no erosion at the moraine (Figure 9b). However, without lake lowering (SC-5-L0), the maximum erosion depth of the frontal moraine reaches up to 3 m.
In large magnitude future GLOF process chain scenario without lake lowering (SC-6-L0) produces a GLOF peak of 198,703 m 3 s −1 immediately downstream of the lake (Table 2, Figure 8). Here the maximum modeled flow depth is ∼13 m. The GLOF peak reduces by 29% when the lake is lowered by 10 m (SC-6-L10) and 94% when lowered by 30 m (SC-6-L30). At Sissu, the peak discharge reduces by 88% and 99% when the lake is lowered by 10 m (SC-6-L10) and 30 m (SC-6-L30) respectively. Similarly, the maximum flow depth at Sissu is reduced by 50% and 70% when the lake is lowered by 10 and 30 m respectively ( Table 2). The erosion of the frontal moraine without lake lowering (SC-6-L0) is 10.7 m. It is reduced to 4 and 1.6 m when the lake is lowered by 10 and 30 m respectively (Figure 9b). It is to be noted that the distance from the impact site to the frontal moraine for the future low magnitude (SC-4) and the moderate magnitude (SC-5) GLOF scenarios is 4.6 km. Here, the impulse wave is unidirectional, that is, from the impact site toward the frontal moraine ( Figure 8b). However, the impulse wave is bidirectional in case of large magnitude GLOF scenarios (SC-6), as the GLOF impact occurs in the middle of the future lake. Here, the impulse waves propagate in opposite directions from the site of impact that is, toward the moraine (2.28 km) and also toward the glacier terminus (2.31 km) (also discussed in Section 5.2). The GLOF arrival times for all simulations at three cross-sections along the channel are given in Table S3 in Supporting Information S1.

Permafrost, Thermal Conditions, and Stability Aspects of Surrounding Slopes
In the Chandra basin, there are only a few recognizable landforms (rock glaciers; cf. Cicoira et al., 2021;Haeberli et al., 2006), indicating active creep of perennially frozen talus/debris with morphological expressions of viscous flow and bright over-steepened fronts exhibiting fresh debris. For example, (a) on a south-exposed slope near the lake, two small but active features ending slightly above 4,600 m a.s.l., and a less active feature (partial vegetation cover) at about 4,500 m a.s.l are visible ( Figure S2.1 in Supporting Information S1); (b) further north, large exposed rock glaciers with active fronts at about 4,000-4,200 m a.s.l are observed ( Figure S2.2 in Supporting Information S1); and (c) a couple of east-oriented active rock glaciers with their fronts at about 4,010 m a.s.l and 4,025 m a.s.l are identified on the south side ( Figure S2.3 in Supporting Information S1). These findings agree with the regional permafrost modeling by S. K. Allen, Fiddes, et al. (2016). Also, they correspond to experiences from the European Alps and comparable estimates carried out at Dzhimarai-Khokh/Karmadon (Haeberli et al., 2004), Chamoli/Uttarakhand , and Chamlang/Nepal (personal analysis by WH). For further considerations in the case of Gepang Gath, we assume the lower limit of permafrost to approximately correspond to a mean annual subsurface temperature of 0°C and to occur at about 4,100 m a.s.l. on especially cold slopes with heavy shadow, at about 4,600 m a.s.l. on especially warm slopes fully exposed to the sun, and at about 4,300 m a.s.l on slopes with intermediate (west or east) orientations (see Section 3.3 for the spatial distribution of MAGST) (S. K. Allen, Fiddes, et al., 2016). These numbers relate to cold microclimates on slopes covered with ventilated coarse debris or wind-blown ridges/steep faces with thin to absent winter snow cover. Sites with fine-grained material and topographic depressions with thick winter snow can be warmer (cf. Boeckli et al., 2012). The altitudinal difference of roughly 500 m between the warmest and the coldest side of the mountains is estimated to reflect a difference in mean subsurface temperature at an identical altitude of about 3°C. Under the influence of this temperature difference, heat flow in sharply east-west-oriented ridges tends to primarily be sub-horizontal from the warm/sunny to the cold/shady side of the mountain tops (Noetzli & Gruber, 2009). Further the regional large-scale permafrost data sets are compared with S. K. Allen, Fiddes, et al. (2016) (used in the present study), which provides a high-resolution simulation of ground surface temperature with a grid cell resolution of 30 m ( Figure S2.4 in Supporting Information S1). Such high-resolution permafrost distribution is required for our study in order to be able to make a reasonable interpretation of permafrost conditions on a given mountain face or slope. In contrast, the large-scale data sets produced for the entire northern Hemisphere by Obu et al. (2019) or Ran et al. (2022) provide simulations at a grid cell of approximately 1 km, which is unable to represent the complex variation in permafrost conditions that occur across different elevations. In view of the strong topographic smoothing induced using these coarser large-scale approaches with elimination of steep/high peaks, their general indication provided agrees well with the general characteristics produced by the more detailed analysis applied in the present case study.
Although it must be emphasized that in addition to the presence of permafrost, the formation and evolution of rock glaciers require additional topographic, microclimatic, and temporal criteria. Because of the slow dynamical downslope motion of the rock glaciers, the lower limit does not necessarily correspond directly to the lower limit of permafrost (Haeberli et al., 2006); therefore, the rock glacier zone can extend below the currently expected permafrost limits. Due to the generally slow creep rates involved, this effect of downslope movement is minimal. The effect of global warming is much stronger because rock glacier permafrost still relates to cold Holocene temperature, especially during the Little Ice Age. With the global temperature rising over time, the isotherms today are at comparatively higher elevations. Furthermore, the existence of cold to polythermal hanging glaciers can cause local anomalies with complex thermal and hydraulic conditions (cf. Carey et al., 2012;Haeberli et al., 2004;Shugar et al., 2021).
The topographic potential for ice-and/or rock-avalanching changes in the future as the lake approaches the steep slopes due to its continued growth. There can be other locations from where avalanches can initiate, making it challenging to identify locations that can result in mass movement unless there are obvious signs. For instance, we realize that in the future, the area in the middle of the glacier tributary is topographically highly susceptible to failure. However, the selected avalanche source located on the glacier headwall (considered in the present study) has an avalanche drop height of ∼1,450 m whereas, for the source in the middle of the glacier, the drop height is 576 m. For a conservative approach, we selected the source with higher drop height considering potentially higher energy events. The consideration of mass movement source zone is speculative considering that the present study is largely scenario-based and understanding how a certain magnitude of mass will interact with the lake. An approach has recently been developed to roughly estimate probabilities of occurrence concerning impacts from large rock-ice avalanches into specific lakes . As slope failures are non-repetitive events, resulting from cumulative developments, reference must be made to frequencies and recurrence times over susceptible areas as obtained in well-documented regions (Coe et al., 2018;Fischer et al., 2012). Per km 2 of susceptible area with icy peaks, steep slopes, glaciers, and permafrost, correspondingly calculated recurrence times of large rock-ice avalanches are 10 3 -10 4 years. With a susceptible area of around 30 km 2 around the Gepang Gath and trend to less stability with continued warming, this results in a probability of occurrence around 0.003 to 0.03 per year or even higher, which corresponds to a centennial to decadal time scale. Such time scales are clearly policy relevant.
Further, the geology of the surrounding area shows numerous dissected slopes (visible in the high-resolution Google Earth imageries). Also, the potential avalanche source zones (identified in the present study) show heavily crevassed surfaces with numerous vertical and transverse cracks. Further, evidence of exposed scarps shows previous instabilities in these slopes (Figures 4e and 4g). Penetration of liquid water into these cracks followed by freezing and thawing cycles can accelerate weathering processes leading to slope failures. Another phenomenon is where the frost cracking window moves upwards in warming atmosphere, and therefore frost cracking occurs more frequently at higher elevations but also less at lower elevations. This affects daily and seasonal freeze/thaw cycles and, hence, near-surface effects within the uppermost decameters to meters below the surface and resulting in rock fall at stone to block size. The effect is probably strongest where surfaces become exposed by retreating ice for the first time and freeze-thaw cycles start increasing in number from zero. Large-volume rock-ice avalanches as discussed in our paper relate to processes at greater depths (tens of meters) and, hence, to warming permafrost and enhanced water penetration deep into the mountain slopes. Therefore, considering these factors, we consider the ice and/or rock avalanche of higher magnitude (SC-3 and SC-6) presented here to have a low to very low likelihood of occurrence, but with an increasing tendency under a warming climate .
The fact that such large-magnitude events have occurred in the past, is a sufficient basis to suggest that these events cannot be neglected in comprehensive hazard assessments. It is seen that slope destabilization can cause failure leading to large-magnitude mass movement events where failure depths can go up to 100 m or more. For instance, the Chamoli disaster total avalanche volume of ∼25 M m 3 containing 20% ice and 80% bedrock (failure depth up to 180 m) was released from a steep slope that caused huge devastation downstream . Other events with even higher magnitudes have been reported in the recent past, for example, 10 7 M m 3 rock avalanche on Siachen glacier, in September 2010 , the ∼50 M m 3 ice-rock avalanche in the Sedongpu valley, in March 2021, Tibetan Plateau (Zhao et al., 2021), and the Kolka-Karmadon rock/ice slide  ; the direction of the impulse wave in SC-4 and SC-5 is unidirectional, that is, propagating only toward the frontal moraine of the lake in all the scenarios, and bidirectional in SC-6, that is, both toward the frontal moraine (2.28 km) and the glacier terminus (2.31 km).
of September 2002 with a mass flow volume >100 M m 3 (Haeberli et al., 2004). Therefore, considering a range of avalanche scenarios (with varying volumes and ice-rock ratios, see Table 1) enables us to comprehensively evaluate the potential hazard associated with the Gepang Gath Lake.

Thermal Conditions in the Slopes Above Gepang Gath Lake
Close inspection of local terrain characteristics together with the available local information from rock glacier mapping and spatial permafrost modeling as described above can be combined with the existing knowledge and understanding about microclimatic aspects in complex/rugged topography (Gärtner-Roer et al., 2022;Haeberli et al., 2004), borehole temperatures available in mountain regions elsewhere (Etzelmüller et al., 2020;IPCC, 2019) and spatial permafrost modeling at various scales (Boeckli et al., 2012;Magnin et al., 2017;Noetzli et al., 2007) in order to infer differentiated information for the investigated site. The highest rock ridges and peaks around the Gepang Gath at altitudes between 5,500 and 6,000 m a.s.l., in places covered with thin ice, are likely to be characterized by temperatures of approximately −5°C to −12°C. They are frozen down to several hundred meters into the mountain with predominantly horizontal heat flow from warm to cold sides. Warm and relatively shallow periglacial permafrost with temperatures close to 0°C and frozen depths of a few tens of meters probably characterizes the zone near the upper limit of alpine meadows. On the sunny slopes north of Gepang Gath, this probable belt of warm permafrost seems to correspond to the starting zones of smaller active debris flows. The lake with its moraines and the valley bottom leading to Sissu is probably a permafrost-free terrain.
The glaciers around Gepang Gath are most likely polythermal, with cold parts on the steepest slopes. Permeable firn zones, where latent heat from percolating and refreezing meltwater has strong impacts, can be much warmer Figure 9. (a) Erosion of the moraine measured at reference point R1 located at the crest of the frontal moraine (see Figure 6 for the location of R1) for the current scenarios (SC-1 to SC-3) and their corresponding-lowering scenarios (L10 and L30); (b) erosion of the moraine measured at reference point R1 located at the crest of the frontal moraine (see Figure 8 for the location of R1) for the future scenarios (SC-4 to SC-6) and their corresponding-lowering scenarios (L10 and L30); the total eroded volume given in bar blots is the sum of the eroded supraglacial debris (before lake impact) and the eroded volume of the moraine. and, in places, even temperate. Overall, the glacier cover with its flow, crevasse patterns, and firn patches tends to strongly influence the permafrost of the steep rock walls. This can lead to a spatially complex thermal pattern within cold deep-frozen sites adjacent to warm to even unfrozen subglacial rocks (cf. Haeberli et al. (2004) concerning the detachment zone of the Kolka-Karmadon event or Shugar et al. (2021) concerning the Chamoli event).
Regional warming during the past century was around +2°C, similar to that in the European Alps. Detailed borehole information in European mountains (Etzelmüller et al., 2020;IPCC, 2019), as well as numerical modeling (Noetzli & Gruber, 2009), document that this atmospheric temperature rise caused subsurface warming to a depth of roughly 50-100 m, producing over this depth range a marked thermal anomaly (reduction to even inversion of temperature gradients and heat flow). The retreat and disappearance of surface ice and glaciers can have contrasting effects. One effect may be marked bedrock warming where thin and cold ice aprons melt away, inducing so far impossible summer warming above 0°C with resulting freeze/thaw cycles. Another effect may be cooling and permafrost formation, where permeable warm firn areas turn into impermeable/cold ice or vanish entirely. With continued warming, the already existing deep thermal anomaly will continue to strengthen and further progress at depth. The belt of warm permafrost with temperatures at, or close to, thawing conditions will affect slopes at increasing elevation.

Permafrost Degradation and Slope Stability
Thermal conditions of warm permafrost with ground temperatures between about −1°C and 0°C and with the coexistence of rock, ice, and water, are similar to conditions estimated for the detachment zone of the recent Chamoli event . Therefore, the possibility of a future large and high-energy Chamoli-like event reaching the Gepang Gath cannot be excluded. The probability of such a potentially disastrous event can at present roughly be estimated to be 0.003 to 0.03 per year or even higher, which corresponds to a centennial to decadal time scale which is increasing with continued warming and ice loss. Small-scale rock falls with volumes up to tens or hundreds of cubic meters can already be observed at several places in the area today (Figures 4e  and 4g).
With changes in frequency and volume, such minor events could in the future become precursory events of larger-scale slope instabilities (Walter et al., 2020) and hence need careful observation. With the surface lowering (Shean et al., 2020) and vanishing of thick glaciers in the valley bottoms, glacial de-buttressing can increasingly destabilize over-steepened adjacent slopes (Deline et al., 2021). Larger events could develop due to ice thickness losses up to about 200 m at the main glacier, which is presently still directly connected to the Gepang Gath.

Present and Future GLOF Exposure at Sissu: Lake-Level Lowering and Potential GLOF Risk Reduction
Sissu village is exposed to high-intensity GLOF (flow height > 1 m) under all current scenarios (SC1-SC3) and the moderate and large future scenarios (SC5-SC6) (Figures 10 and 11, Figure S5.1-S5.3 in Supporting Information S1). Considering 10 m of lake lowering, the hazard to Sissu is greatly reduced for all except the largest scenarios (SC-3 and SC-6), although an area of 1 km 2 remains exposed to medium GLOF intensity (H < 1 m) under SC-2 ( Figure S5.4 in Supporting Information S1). For the largest scenarios, 10 m of lowering has little effect on the medium-intensity flood zone but reduces the high-intensity zone by 54% and 63% for the current (SC-3) and future (SC-6) scenarios, respectively. As seen for SC-6 in Figure 10, lowering of 10 m would reduce the risk to a number of houses and cropland adjacent to the Sissu River, and around 600 m of the main highway passing through the village. Under 30 m of lowering, both the high and medium intensity zones reduce further, with an overall reduction in the affected area of 78% for SC-3 and 93% for SC-6. The inundation of the buildings and agricultural land is significantly reduced (Figure 10). The remaining affected land area is primarily within the uninhabited area of the active river channel, although there is infrastructure, riverbank camping sites, and a water reservoir (Sissu lake) located on the outwash fan of the Sissu river. Notably, in the absence of lake lowering, the helipad is within a high-intensity zone under all moderate-to-large scenarios, and even with up to 30 m of lowering, it remains threatened under SC-3 ( Figure 10). This exposure of the helipad could have severe implications for heli-evacuations and other emergency response actions following a GLOF event, and possible relocation should be considered. It is to be noted that in the future, even if the lake grows substantially, the impact and exposure at Sissu in the large magnitude event (SC-6) remain similar to the large magnitude event under current conditions . This is because of the bidirectional propagation of the impact wave, where the rear part of the future lake can actually take up and attenuate some of the impact energy ( Figure 8b). Also, in the case of future moderate (SC-5) and small magnitude (SC-4) scenarios, the downstream impact is comparatively less when compared to the respective present scenarios (SC-2 and SC-1). This is because the impact occurs at the rear end of the future lake, which is shallow. Also, the impact wave travels 2.28 km more until it reaches the frontal moraine when compared to the present lake. This can significantly reduce the energy of the impulse wave propagation toward the damming moraine leading to reduced peak discharge, eventually leading to reduced or no overtopping. This, in turn, reduces the erosion potential of the frontal moraine, and therefore in future scenarios, there is comparatively less erosion of the frontal moraine compared to the present scenarios ( Figure 9). In the case of Gepang Gath Lake, it has a very different geometry as compared to a typical glacial lake as frontal portion of the Figure 10. Zones of high (H > 1 m) and medium (H < 1 m) glacial lake outburst flood (GLOF) intensity classified according to GAPHAZ (2017), based on modeled GLOF events from Gepang Gath, under the large-magnitude scenarios SC-3 and SC-6, and considering the effects of lake lowering; also shown are the potentially exposed elements at Sissu; background: Planet Scope imagery (Planet Scope; https://www.planet.com); for small-and medium-magnitude events, GLOF intensity zones are given in Figures S5.1 and S5.2 in Supporting Information S1. lake (∼500 m) is very shallow and dammed by a flat and wide moraine (see Figure 3f). It is highly unlikely that internal failure of the moraine would cause a GLOF event. In the case of moraine collapse that may occur due to melting of the ice core, the outflow volume would be much lower when compared to impulse waves caused by avalanche impacts. Most of the lake water is stored near the glacier terminus (deeper portion) and erosion of the frontal shallow end is very unlikely. Therefore, the GLOF hazard due to internal failure of the moraine remains low. Hence, in the present study we consider only GLOF process chain scenarios that involve potential avalanche impacts on the lake.

Previous Glacial Lake Lowering in the Himalaya-Challenges and Alternative GLOF Mitigation Strategies
Previous efforts at glacial lake lowering in the Himalaya were implemented on several lakes including the Imja Tsho in 2016 (Lala et al., 2018) and Tsho Rolpa in the year 2000 (https://www.ctc-n.org/products/artificial-lowering-glacial-lake) in Nepal Himalaya where the lakes were lowered by 3.4 and 3 m respectively. Both these lakes were identified as potentially dangerous (Khadka et al., 2019). An open channel was constructed at the moraine   (Shrestha et al., 2013). However, there has not been a significant change in the lake area even after the construction of the open channel at the moraine of the lake. Also, modeling of lake lowering scenarios of the Imja Tsho shows that lowering the lake by 3 m does not have a significant reduction in the downstream hazard. The reduction in the potential downstream impact is seen in the 10 m lake lowering scenario, and 20 m lake lowering is required to further minimize the downstream hazard (Somos-Valenzuela et al., 2015).
In Bhutan Himalaya, lake lowering has also been implemented on glacial lakes including the two glacial lakes in the Lunana complex namely the Thorthomi Lake and the Raphstreng Tsho where the lakes were lowered by 3.63 m (UNDP, 2011) and 4 m (Ghimire, 2004) respectively. The Raphstreng Tsho was initially lowered by 0.95 m in 1996 and by 1998, the lake was lowered by 4 m (Singh & Karki, 2004). The targeted lowering for the lake was 20 m. However, this couldn't be achieved in a feasible time frame and would have taken 7-8 years to lower the lake by 20 m. The lake was drained manually by widening the lake outlet channel. Initially, a pumping method was employed to drain out the lake's water, however, due to its ineffectiveness and high cost, the method was later dropped. For Thorthomi Lake, the targeted lowering was 5 m, of which 3.63 m was achieved. The major challenges faced during the operation were related to the harsh weather and terrain conditions. The distance from the nearest road to the lake was a trek of 9 days. Further, lowering of the lake was carried out manually using simple tools. Also, the transportation of equipment to such high altitudes was challenging as they were carried using horses and yaks. Health risk is another challenge of working at high altitudes. There have been 3 casualties reported in 2010 in the Thorthomi Lake lowering expedition of which 2 were because of altitude sickness (UNDP, 2011).
The Poiqu basin (a transboundary Himalayan basin) hosts a number of potentially dangerous lakes and also has a history of at least six major GLOF events in the past . The Jailongco Lake is one potentially hazardous lake in the basin (Allen et al., 2019) from which two GLOFs occurred in 2002 (Chen et al., 2013). Starting in 2018, the lake was lowered by 16 m, the frontal moraine was largely removed, and the outlet channel was stabilized with concrete. Recent modeling efforts show that there is a minimal effect on the GLOF magnitude and arrival time in a worst-case GLOF triggered by a large ice avalanche with a failure volume of 18 M m 3 .
Given these examples above, it is to be noted that the effectiveness of lake lowering as a mitigation strategy can vary from lake to lake depending on the geomorphological setting of the lake and its surroundings. In the present study, we consider lake-lowering scenarios of 10 and 30 m, which would be beyond what has been achieved in other cases. Even with extreme lowering of up to 30 m, our results also show that lake lowering can reduce downstream impact on the Gepang Gath valley but not completely eliminate downstream risk when large magnitude events are considered. Hence, lake level lowering should not be relied on as a single mitigation strategy, and in line with Indian national guidelines on the management of GLOFs (NDMA, 2020), a comprehensive package of structural and non-structural risk reduction strategies is required. This includes Early Warning Systems (EWS), spatial planning to prevent critical infrastructure from being built in exposed areas, and programs to strengthen local capacities and raise awareness. Importantly, experiences and lessons learned from disaster risk reduction implementation show that the local population must be closely involved from the beginning in the design, planning, and implementation of risk reduction strategies, and their risk perceptions and knowledge must be respected. The implementation of structural risk reduction and mitigation measures, for example, the construction of artificial dams or siphoning of the lakes can involve huge costs, and therefore it is important to analyze cost-effectiveness before ground implementation. Also, high costs can be involved in the maintenance of the EWS, and institutions may be unwilling to absorb these costs once external project funding ends. Cost-effective options are therefore likely to involve both structural and non-structural measures (such as, involvement of local society and NGOs for raising awareness and capacity building) to mitigate and reduce risks under a range of future scenarios.

Uncertainties and Future Direction
We note the uncertainty in the present study from various sources, including (a) initial mass movement scenario definition; (b) lake bottom topography and volume; and (c) model parameters. Among these, one major source of uncertainty in predictive modeling, like in the case of Gepang Gath, is the initial scenario definition of GLOF triggers (avalanche source zone, composition, and magnitude). One major challenge is defining realistic GLOF process chain scenarios in predictive modeling (Emmer et al., 2022). Therefore, we here consider various combinations of these factors to define the initial release of the avalanches. We identified avalanche source zones using two criteria (highest topographic potential and permafrost occurrence). To further cover the uncertainty in potential GLOF triggers of the Gepang Gath Lake, we considered events with varied magnitude (release volume) for both present and future scenarios. Also, consideration of the varied composition of the avalanches (ice-rock ratio) enabled the consideration of a wide range of potential events. Also, depending on the composition of the avalanche that is, ratios of the rock and ice component, the processes can vary. For instance, for higher-density avalanches, impact can be higher, which in turn will affect the impulse waves created at the initial impact site of the lake. Further, the presence of ice in the avalanche can alter the GLOF process. As the density of ice is lower than the lake's water, the initial impact would be comparatively of smaller magnitude when compared to rock avalanches or rock-ice avalanches where GLOF discharge can peak higher. Also, water-ice interaction can lead to ice melting and can contribute an additional volume to the downstream flood. A classic example can be the Chamoli event where a rock-ice avalanche transformed into a hyperconcentrated flow due to additional water from the melted ice of the avalanche . Here we rely on assumptions of rock-ice composition of the avalanche based on the overlying ice (based on ice thickness; Farinotti et al., 2019) and the assumed thickness of underlying rock. The present study shows that GLOF processes can be different depending on the GLOF impulse wave which varies based on the magnitude, nature, and site of initial mass movement impact. What remains uncertain in predictive modeling is the flow transitions due to phase changes. Also, GLOF processes are lake specific and the GLOF hazard potential of a particular lake can change over time depending on the changes in the lake and its surrounding conditions (e.g., permafrost degradation or geological instabilities). These changes in the surrounding slopes in turn can affect the trigger sources and their failure mechanisms. However, it is very difficult to determine the tipping point of potential hazard for a particular lake (i.e., the time point at which a lake has the maximum hazard potential) as determining future changes in the surrounding slopes is challenging. Also, the triggering source, magnitude, and mechanisms remain highly uncertain. For characterizing avalanche sources in the bedrock, no in situ temperature measurements are available and for obvious logistic reasons, no such measurements are easily possible within potential detachment zones. Estimates, therefore, have a semi-quantitative character as described in the text. Advanced knowledge can come from (a) borehole sites elsewhere with longterm measurements (which is rare in the Himalaya), (b) local field evidence (visually recognizable phenomena of permafrost creep, cold ice aprons), and (c) results from large-scale spatial modeling of permafrost occurrence in rugged terrain. The general fact that estimates of air temperature and related subsurface temperatures are quite robust due to efficient atmospheric mixing, the estimates concern quite general conditions in cold mountains and can be considered realistic, at least as a first-order assessment. Another important source of uncertainty is the lake-bed topography and the lake volume. As field-measured lake bathymetry was unavailable, we relied on glacier ice thickness to calculate the lake's present and future volume. It is to be noted that the bathymetry was derived from the glacier bed and it does not consider the debris at the base of the lake. Further uncertainties in the glacier lake-bed can also arise due to ice-thickness and DEM using which the glacier bed of the lake was derived. We also note that uncertainties in an individual GLOF scenario can be due to the model parameters.
The results of the individual process chain simulations are sensitive to variations of key model input: initial conditions such as release volumes and solid content of the release, flow parameters including basal friction angle, and entrainment coefficient. Here, the attribution of rock-ice avalanche magnitudes to individual sites is highly tentative based on assumptions including avalanche release volumes, and avalanche composition. Also, we define the model inputs based on previous studies on GLOF process chains (Mergili et al., 2018;Zheng, Mergili, et al., 2021). Still, the simulation results have to be interpreted with caution: in contrast to high-frequency processes, where the back-calculation of large numbers of events of different magnitudes may yield fairly robust guiding parameter values, the simulation of complex high-mountain process chains is not a standard task, and parameter sets derived from a still limited number of GLOF simulations with a steadily evolving simulation framework such as r.avaflow (Mergili, Jaboyedoff, et al., 2020;Mergili et al., 2018;Vilca et al., 2021;Zheng, Mergili, et al., 2021) are connected to significant uncertainties which are often hard to quantify. Concerted simulations of well-documented GLOF process chains and comprehensive parameter sensitivity analyses are needed to further improve the basis for predictive simulations.
For future work, we recommend regular monitoring of the lake and its surroundings using remotely sensed data. InSAR techniques can be helpful to monitor slope movements or moraine instability (Bertone et al., 2022;Scapozza et al., 2019;Wangchuk et al., 2022). Real-time water-level sensors can be used to monitor the lake-level fluctuations and downstream water levels. A precise real-time Differential Global Positioning System can be used to monitor movements in the damming moraine. The Cirenmaco lake in the Central Himalaya is an example where lake-level sensors and moraine monitoring devices were installed as an integral part of the monitoring and EWS (Wang et al., 2022). Also, a field-based detailed evaluation of the lake's frontal moraine and bathymetry is necessary for further assessments. For further precision in modeling, we recommend the acquisition of high-resolution DEMs to determine the lake hazard potential. The presented study is a first-order detailed modeling assessment of the Gepang Gath Lake along with lake-lowering effects that will help policy/decision-makers understand the hazard complexity in the valley. The study serves as a basis for further detailed assessments to establish a EWS network in the region to avoid and reduce the impact of future catastrophes from the lake.
It is evident that not only lakes are expanding, but also the infrastructure is expanding higher into alpine valleys in the GLOF-exposed regions of the Himalaya. This is also true in the case of the Gepang Gath valley, where infrastructure and agricultural land have developed at Sissu in a span of 12 years ( Figure 11). Therefore, there is a clear need for assessment to identify the exposed infrastructure and community to glacial-related hazards such as GLOFs, as a basis for the design of risk reduction strategies. Also, important is prior identification of the safe or low-risk zones for future development in the Himalayan valleys. Several studies have identified high-risk glacial lakes in the Himalaya that have the potential to impact the downstream community, for example,  and Mal et al. (2021). The presented approach can go further for these lakes, by considering very large outburst scenarios, and by demonstrating the possible effectiveness of lake lowering as a robust GLOF risk reduction strategy.

Conclusion
The study presented a holistic assessment of the GLOF hazard of the Gepang Gath Lake in Himachal Pradesh, Western Himalaya, India. Results reveal that the Gepang Gath Lake is a critical glacial lake and potential GLOFs from the lake can impact buildings and communities along the banks of the flow channel at Sissu. It is to be noted that despite significant lake growth expected in the future, GLOF intensities in Sissu remain similar to the current situation. However, the GLOF probability of occurrence is likely to increase (rather than the magnitude) with continued lake growth and deep warming of the steep frozen rock walls. In the case of lake lowering by 10 and 30 m significantly reduces the exposed elements at Sissu. The main concern is the exposed elements such as the Sissu helipad, Sissu Lake, and the riverbank camping sites located in the fan area where the channel meets the Chenab River. These infrastructures are exposed to large-magnitude events even after lake lowering. Also, these infrastructures remain exposed in 10 m lowering in the moderate-magnitude event. It can be concluded that lowering the Gepang Gath Lake considerably reduces the downstream exposure but does not completely eliminate the GLOF risk when large-magnitude events are considered. It is to be noted that the study was based on assumptions of mass movements into the lake of different magnitudes; predicting an anticipated future event is highly uncertain. However, the present study can provide a basis for applying disaster risk reduction strategies in the valley and cost-benefit evaluations of lake-lowering efforts, further recommending a more detailed investigation of the area.