We used the global positioning system (GPS) dataset of Tsutaki et al. (2019), who conducted a kinematic survey with a differential GPS (DGPS; GEM-1, GNSS Technologies) across the on- and off-glacier terrains during the 19–22 September 2011 field campaign. The base station for this survey was installed to the west of Thorthormi Glacial Lake (Fig. 1a). These GPS data points were used to validate the satellite and photogrammetry digital elevation models (DEMs) and compute the surface elevation changes (Sect. 3.2 and 3.3).

We conducted a helicopter photogrammetry survey on 24 March 2018. Four action cameras (GoPro HERO5 Black) were attached to the skids of a helicopter and acquired 4000×3000 pixel images in 1 s shooting mode. The shutter speed, focal length, and ISO were fixed at 1/1250 s, 28.3 mm, and 100, respectively. We obtained ∼7500 photos in total and cropped each photograph by preserving the central 2500×2500 pixel area to eliminate the “fisheye effect” of the GoPro camera lens (Girod et al., 2017). We finally employed 3560 images based on the image quality and glacier coverage. These images were processed in Agisoft Metashape Professional Edition 1.7.1 (Agisoft LLC), and the sky view was masked for the terrain data processing.

We extracted ground control points (GCPs) from a Pléiades panchromatic orthoimage (0.5 m resolution), which was acquired on 7 November 2017 (Berthier et al., 2014), and its DEM (2.0 m resolution) for the photogrammetry terrain data processing. We first generated a GPS-derived DEM (GPS-DEM) to assess the vertical accuracy of the Pléiades-derived DEM (PL-DEM). The 2011 GPS data points (UTM Zone 46N, WGS84) were interpolated using the inverse distance weighted method and then exported to the same grid size as the PL-DEM in ArcGIS (Tshering and Fujita, 2016; Sato et al., 2021). We employed the standard deviation (SD; σ) of the elevation difference between the PL- and GPS-DEMs on the off-glacier stable terrain as the vertical accuracy of the PL-DEM. We did not use the grid cells with steep slopes (>30∘; Fujita et al., 2008; Nuimura et al., 2012). We then eliminated the validation points that were greater than ±3σ from the mean elevation difference as extreme outliers (Mertes et al., 2017). Berthier et al. (2014) reported that the vertical accuracy of the PL-DEM was improved by shifting the DEM horizontally. We therefore shifted the PL-DEM by ±2 pixels (±4 m) in the northing and easting directions, computed the elevation difference against the GPS-DEM, and confirmed that there was no improvement in the vertical accuracy. Finally, the PL-DEM vertical bias (mean elevation difference: MED) was assessed for 12 009 grid cells, yielding a mean value of 0.26±3.86 m (MED ± SD). We extracted the GCP coordinates from the orthoimage and bias-corrected PL-DEM. Specific topographic features (e.g. boulders, river bending points, and dense vegetation spots) on the stable ground were used as GCPs for the photogrammetry terrain data processing.

We used the structure from motion (SfM) software in Agisoft Metashape to generate orthoimages and a DEM from the helicopter photogrammetry data (hereafter HP-ortho and HP-DEM, respectively). We overlaid the 77 GCPs that were extracted from the Pléiades products onto the helicopter photogrammetry images (Fig. 1b) and generated both the HP-ortho and HP-DEM at a 0.5 m resolution (Fig. 1a and b). We employed the same approach used in the PL-DEM evaluation to evaluate the vertical bias and accuracy of the HP-DEM by re-generating a 0.5 m resolution GPS-DEM. The vertical accuracy of the HP-DEM was 0.25±3.70 m (N=25474 GPS-DEM grid cells; Fig. S1a in the Supplement); we also applied an elevation change correction (Sect. 3.3) to correct for the vertical bias of the HP-DEM.

We calculated the surface elevation change rates (dh/dt) by comparing the 2011 GPS-DEM and 2018 HP-DEM (both at 0.5 m resolution). We used 9491 and 15 604 grid cells to calculate dh/dt for Thorthormi and Lugge glaciers, respectively (red tracks in Fig. 1b). We then compared our results with previous elevation change studies. Tsutaki et al. (2019) calculated dh/dt from the overlapping 2004 and 2011 DGPS data; they also computed the spatial distribution of dh/dt for the same period using Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER)-derived DEMs. We also employed the long-term dh/dt data from Brun et al. (2017) and Maurer et al. (2019) to assess the thinning trends of these two glaciers. Brun et al. (2017) computed dh/dt for the 2000–2016 period over the Hindu Kush Himalaya region using ASTER-based DEMs, and Maurer et al. (2019) calculated dh/dt for the 1975–2000 and 2000–2015 periods using satellite-based DEMs. These datasets (hereafter RS-based dh/dt) are provided as 30 m resolution raster data. We extracted the dh/dt data from our DEMs at the same positions for a comparative analysis.

We computed the surface velocity field using the ImGRAFT (Image GeoRectification and Feature Tracking) open-source feature tracking toolbox in MATLAB (Messerli and Grinsted, 2015). The normalized cross-correlation algorithm (NCC; Heid and Kääb, 2012) in the feature tracking toolbox (Templatematch) identifies the displacement patterns of the glacier surface features and computes their magnitude from a pair of images. We selected a Sentinel-2 image pair that was acquired on 16 November 2016 and 11 November 2017 (post-monsoon seasons). After visual trial and error, we chose a 20×20 template size (200×200 m) and 75×75 search window size (750×750 m) to compute the surface feature displacements and calculate the annual flow velocity. We set the correlation value for image matching (r>0.6) and signal to noise ratio (SNR <0.7) and eliminated the low-quality pixels, all of which served as a confidence level threshold for successful image matching. We then estimated the uncertainty of the glacier surface velocity and corrected systematic error by checking the stable-ground (off-glacier) displacement (e.g. Liu et al., 2020). We calculated the stable-ground surface displacement (slopes <20∘; Quincey et al., 2009) and set the corrected median values of Vx (east–west component) and Vy (north–south component) to zero. The flow speed V (m a-1) is calculated as follows: 1V=Vx2+Vy2. The mean and median stable-ground V values were 2.2 and 1.6 m a-1, respectively, after the displacement correction. The velocity profiles were extracted along the glacier central flowlines every 10 m, and the pixel values where the flow directions differed by >90∘ from the flowlines were eliminated. We also eliminated the velocity data along the upper section of Lugge Glacier (>5100 m elevation) because of its heavy snow cover, which can cause incorrect image matching via feature tracking (e.g. Nuimura et al., 2017). We extracted the surface velocity from the regional velocity product derived from the ITS_LIVE (Inter-Mission Time Series of Land Ice Velocity and Elevation) project (Gardner et al., 2019), which covered the entire HMA region. The ITS_LIVE velocity product is generated from the Landsat series with the auto-RIFT feature tracking processing chain yielding a 240 m spatial resolution (Gardner et al., 2018). We extracted the annual velocity data along the central glacier flowlines from the ITS_LIVE product (2010–2018) to trace the temporal changes in flow velocity; the ITS_LIVE velocities possess mean uncertainties of 1.0 m a-1 (maximum of 3.1 m a-1) and 0.6 m a-1 (maximum of 6.0 m a-1) along Thorthormi and Lugge glaciers, respectively. We also employed the velocity data produced by Tsutaki et al. (2019), which were calculated annually from the ASTER-derived optical satellite images at 15 m resolution during the 2002–2011 period.

We delineated the glacial lake area from Landsat 7 and 8 (ETM+: Enhanced Thematic Mapper Plus; OLI: Operational Land Imager) images with false-colour image composites that were acquired between November 2012 and November 2018 (30 m resolution). We then combined the proglacial lake polygons before 2012 (Tsutaki et al., 2019) and traced the annual lake area changes for the entire 18-year study period. The total lake area uncertainties were estimated to be ±0.14 and ±0.08 km2 for Thorthormi and Lugge glaciers, respectively, depending on the user-induced error and satellite image resolution (Paul et al., 2013). We removed the DEM and velocity data where the glacier surface turned into the lake surface in successive images. We compared the recent lake expansion rates with a glacial lake inventory (High Mountain Asia Glacier-lake inventory: Hi-MAG; Chen et al., 2021), which was generated for the entire Himalayan region using data from the 2008–2017 period. We finally chose the 2011 and 2017 proglacial-type lakes (n=832) and calculated the expansion rates between 2011 and 2017 in the eastern Himalaya region (including the Lunana region).

We calculated the emergence velocities along Thorthormi Glacier to evaluate the change in glacier dynamics since its detachment from the terminal moraine. We estimated the emergence velocity of a given section (Ve, m a-1) from the ice fluxes along the upper and lower boundaries of the section as follows (e.g. Nuimura et al., 2011; Vincent et al., 2016; Brun et al., 2018): 2Ve=qin-qoutW‾⋅dx, where qin and qout are the ice fluxes (m3 a-1) along the upper and lower boundaries, respectively, and W‾ and dx are an averaged glacier width (m) and length (200 m in this study) for the analysed Ve section, respectively. The ice flux q (qin or qout, m3 a-1) is calculated as follows: 3q=W⋅h⋅Vc, where W, h, and Vc are the glacier width (m), ice thickness (m), and depth-averaged ice velocity (m a-1), respectively. We then applied a simplified assumption that the glacier width is constant such that Eq. (2) can be rewritten as follows: 4Ve=hup⋅Vc,up-hlow⋅Vc,lowdx, where hup/low are the ice thicknesses (m) and Vc,up/low the depth-averaged ice velocities (m a-1) along the upper/lower boundaries. We assumed that both the glacier thickness and width were constant in the transverse and longitudinal directions, respectively, to calculate the emergence velocities along the central flowline. The ice thickness h along the central flowline was calculated from the HP-DEM-derived glacier surface elevation and estimated bedrock elevation in Tsutaki et al. (2019). Tsutaki et al. (2019) estimated the glacier-bed topography using the Farinotti et al. (2009) ice thickness model and tuning a model parameter based on the observed lake depth (ice thickness). Tsutaki et al. (2019) simulated that the basal velocity reaches 90 % of the surface velocity of Thorthormi Glacier. We therefore calculated the emergence velocity using two assumptions regarding the surface velocity: the depth-averaged velocity is 90 % of the surface velocity based on Tsutaki et al. (2019), and 100 % is assumed for a floating condition after terminus detachment. We then used the surface velocity component of the same vector in the central flowline direction. The sections without flow velocities (2520–3020 m from the 2002 terminus) were linearly interpolated using the surface velocities of the surrounding upglacier and downglacier sections. We calculated the emergence velocity for a 200 m section by shifting the section in 50 m increments and obtained a mean emergence velocity around the current terminus (2400–3500 m from the 2002 terminus). We also calculated Ve in 2011, when Thorthormi Glacier was still a land-terminating glacier, to compare the land- and lake-terminating conditions. We estimated the ice thickness from the glacier surface elevation of the ASTER-derived DEM acquired on 6 April 2011 and the glacier-bed elevation along the central flowline. The depth-averaged ice velocities were calculated from the surface velocities in Tsutaki et al. (2019) (Sect. 3.3).

We evaluated the ice flotation potential of the Thorthormi Glacier terminus based on the ice flotation thickness (hf, m), which was calculated as follows (Boyce et al., 2007; Watson et al., 2020): 5hf=ρwρid, where ρw is the density of water (1000 kg m3), ρi is the density of ice (917 kg m3) (e.g. Boyce et al., 2007; Carrivick et al., 2017), and d is lake depth (m). We then defined an index of potential ice flotation as follows: 6Pf=hfh×100, where Pf is the potential ice flotation (%). The glacier can attain flotation when the glacier ice thickness reaches hf such that Pf is ≥100%. We extracted the lake surface elevation (4415 m a.s.l.) from the HP-DEM-derived lake perimeter and estimated the lake depth from the glacier-bed elevation (Tsutaki et al., 2019). We then calculated Pf in 100 m intervals in 2011 and 2018 along the glacier central flowline in the terminus region (up to 3500 m from the 2002 terminus).

We traced the lake expansion for the 2000–2018 period (Fig. 2a and b). The proglacial lake areas at the termini of Thorthormi and Lugge glaciers were 3.05 and 1.58 km2 in 2018, an increase of 2.01 (193 %) and 0.48 km2 (44 %) from the 2000 lake areas, respectively (Fig. 2c). Both lakes have expanded throughout the study period, and the lake expansion rates (dA/dt) were calculated via a linear regression of the cumulative areas during the 2000–2011 and 2011–2018 periods (Fig. 2c). Lugge Glacial Lake steadily expanded during the 2000–2018 period, with 0.03 and 0.02 km2 a-1 observed before and after 2011, respectively. However, there has been accelerated expansion of Thorthormi Glacial Lake since 2011, with 0.07 km2 a-1 observed before 2011 and 0.13 km2 a-1 observed after 2011. A comparison of these observations with the Hi-MAG data (Chen et al., 2021) indicates that the expansion rates are in the upper 2.5 % (Thorthormi) and 10 % (Lugge) of the observed proglacial lakes across the eastern Himalayas.

The dh/dt values of both glaciers were calculated from the 2011 GPS-DEM and 2018 HP-DEM, with the 2002 terminus position used as the base position for the comparison. We also extracted the calculated dh/dt values from previous studies that had focused on different time periods (Fig. 3 and Table 1). The thinning rate of Lugge Glacier was more than 3 times greater than that of Thorthormi Glacier for the 2004–2011 period, when Thorthormi Glacier was a land-terminating glacier, and was then comparable (-2.78 m a-1) to that of Lugge Glacier (-2.87 m a-1) for the recent 2011–2018 period, when Thorthormi Glacier had evolved into a lake-terminating glacier (Fig. S1). There was a 2-fold increase and 0.61-fold decrease in the thinning rates of Thorthormi and Lugge glaciers between the 2004–2011 analysis by Tsutaki et al. (2019) and our presented 2011–2018 analysis. The RS-based dh/dt values for the 2000–2016 period (-3.81 to -3.50 m a-1; Brun et al., 2017; Maurer et al., 2019) are similar to the Tsutaki et al. (2019) values (Table 1). The RS-based dh/dt values of Thorthormi and Lugge glaciers are -0.16 and -1.20 m a-1 for the 1975–2000 period, respectively, which suggests that the lower section of Thorthormi Glacier experiences minimal thinning before 2000 (Maurer et al., 2019). The spatial distribution of dh/dt along Thorthormi Glacier exhibited a decreasing trend in the upglacier direction during the 2011–2018 period, whereas the dh/dt values during the 2004–2011 period were almost constant across the same region. The thinning rate was >4 m a-1 (dh/dt of less than -4 m a-1) in the upglacier area (>2500 m from the 2002 terminus) during this later period. The dh/dt profiles obtained in previous studies do not reveal such a remarkable trend; however, similar dh/dt plots are independent of the distance from the terminus (Fig. 3a). The results of this study reveal a large spatial variability compared with the RS-based dh/dt distributions of previous studies (Brun et al., 2017; Maurer et al., 2019), which is likely due to the differences in the spatial resolution of the data. Tsutaki et al. (2019) reported that Lugge Glacier has a heavily crevassed, bumpy surface. We consider that the high-resolution photogrammetry data (0.5 m) identified the displacements due to these steep surface features.

We calculated the surface velocity field between November 2016 and November 2017 and extracted the velocities along the central flowline (Figs. 4 and S2). We also plotted the ITS_LIVE product (2010–2018) and Tsutaki et al. (2019) velocity profiles. We found fast flow velocities (>200 m a-1) from the 2017 terminus to the middle part (∼4200 m from 2002 terminus) of Thorthormi Glacier. The ITS_LIVE velocity also exhibited a similar flow-velocity magnitude near the 2017 terminus; however, the ITS_LIVE velocities decreased more rapidly in the upglacier direction than our calculated results. We are able to confirm the large displacement of a surface feature (∼200 m a-1) from the Sentinel-2 satellite images at ∼3500–4000 m from the 2002 terminus (Fig. S3), which suggests that the flow-velocity profile for Thorthormi Glacier that is calculated in this study should be more reliable than the automatically derived ITS_LIVE flow-velocity profile. A comparison of our velocity profile (2016–2017) with the 2002–2010 velocity profile (Tsutaki et al., 2019) reveals a substantial 2–4-fold increase at ∼2400–4000 m from the 2002 terminus. The ITS_LIVE flow-velocity profiles indicate that until 2016 the Thorthormi Glacier shows similar flow velocity profiles to the 2002–2010 average, and remarkable increases are observed in 2017 and 2018 (Fig. 4a). Tsutaki et al. (2019) projected the flow velocities of Thorthormi Glacier under the assumption of a “lake-terminating condition” (dashed line in Fig. 4a). Although the magnitude is ∼100 m a-1 less than that in this study, the increasing flow velocities toward the calving front are similar to the trend in the recent velocity profile.

Conversely, the calculated surface velocities of Lugge Glacier are ∼50 m a-1 up to ∼2000 m from the 2002 terminus in this study (2016–2017). There also appears to be a gradual decrease up to ∼2700 m from the 2002 terminus, and it possesses a similar velocity magnitude/trend to the 2002–2010 mean velocity calculated by Tsutaki et al. (2019; Fig. 4b). The ITS_LIVE velocity profiles (2010–2018) show flow velocities of <5 m a-1 for the entire glacier, which is probably due to the coarser resolution (240 m) of the velocity field compared with that in this study (10 m) and Tsutaki et al. (2019; 15 m). Although the terminus position of Lugge Glacier has retreated almost 1 km since 2002, the mean velocity profile appears to have remained persistent between the 2000–2010 (Tsutaki et al., 2019) and 2016–2017 observation periods.

We calculated the emergence velocity of the Thorthormi Glacier terminus under the assumption that the depth-averaged velocity equals either 90 % or 100 % of the surface velocity. The resultant Ve values are -0.69±11.65 and -0.77±12.94 m a-1, respectively (2400–3500 m from the 2002 terminus; Fig. S4 and Table 2), although there are large variations depending on the computational area (Fig. S4). The land-terminating condition yielded Ve values of 5.20±3.78 and 5.78±4.20 m a-1, respectively, for the above-mentioned depth-averaged velocity assumption (Table 2). These results suggest that the mean Ve has decreased and become negative after transitioning to a lake-terminating glacier. Tsutaki et al. (2019) also estimated Ve via numerical modelling of the lake- and land-terminating conditions, yielding 2.2±1.9 and 6.3±2.2 m a-1, respectively.

We also estimated the potential ice flotation index in the terminus area of Thorthormi Glacier (up to 3500 m from the 2002 terminus). The mean Pf values for 2011 (land-terminating) and 2018 (lake-terminating) are 86 % and 97 %, respectively, with this increase attributed to the surface lowering of the terminus area during the 2011–2018 period. As a result of surface lowering, some parts of the ice in the terminus area reached ice flotation thickness (Pf>100%) by 2018.

Thorthormi and Lugge glaciers possess contrasting dh/dt trends and flow velocities even though they are adjacent to each other. The dh/dt trend and the flow-velocity magnitude and spatial distribution of Lugge Glacier are approximately constant between the 2004–2011 and 2011–2018 periods and 2002–2010 and 2016–2017 periods, respectively (Figs. 3b and 4b). Conversely, remarkable increases in the thinning rate and flow velocity are observed across Thorthormi Glacier over the same study periods (Figs. 3a and 4a). Such a drastic velocity increase within a decade has not been reported in the Himalayas, although the multi-decadal acceleration of glacier thinning and deceleration of glacier flow have been reported (Dehecq et al., 2019; Maurer et al., 2019).

Tsutaki et al. (2019) performed finite-element simulations of present (land-terminating) and future (lake-terminating) Thorthormi Glacier dynamics. Their simulations reproduced the flow velocities for the land-terminating condition with a small root-mean-square error (<10 m a-1) using satellite-based flow velocities. Their future prediction for a lake-terminating condition, which suggested an increase in flow velocity, is inconsistent with our 2017 velocity analysis (Fig. 4a). However, Tsutaki et al. (2019) highlighted that changes to the sliding coefficient and ice thickness parameters could alter the flow velocity significantly as their sensitivity tests demonstrated that the simulated flow velocity increased (decreased) by 33 % (51 %) if the sliding coefficient and ice thickness changed by +30 % (-30 %) for the land-terminating condition of Thorthormi Glacier. Therefore, the difference between the observed and simulated velocities is likely due to the uncertainties in the sliding coefficient, ice thickness, and state of the terminus position. Despite this underestimation, Tsutaki et al. (2019) have reasonably demonstrated the change in sliding conditions associated with the transition from land- to lake-terminating conditions.

Lateral lakes had formed on both sides of the Thorthormi Glacier terminus several years before 2011 (Fig. 2a and c). The ice thickness was near flotation (Pf>85 %) such that the flow velocities could accelerate near the terminus. However, the flow velocities decreased toward the terminus, producing a longitudinal compression field and subsequent surface lowering that might have been less than that of lake-terminating Lugge Glacier (Tsutaki et al., 2019). The longitudinal stress field regime changed from compressional to extensional after the terminus detached from the terminal moraine and transitioned to a lake-terminating condition, and its flow increased owing to efficient basal sliding. Although the satellite imagery shows that the glacier terminus detached from the terminal moraine in 2011 (Fig. 2a), the rapid increase in flow velocity since 2017 suggests that the glacier terminus was in contact with the terminal moraine underwater until 2016 and detached between 2016 and 2017 (Fig. 4a). Furthermore, the lakes that formed on both sides of the glacier terminus may also have reduced the lateral resistive stresses that prevented glacier flow (e.g. Adhikari and Marshall, 2012). These factors might have led to the observed dramatic increase in flow velocities (Fig. 4a). Such an increase in flow velocities due to proglacial lake formation has been observed in other regions (e.g. Boyce et al., 2007; Tsutaki et al., 2011; Sakakibara and Sugiyama, 2014); however, this is the first observation of such a phenomenon in the Himalayan region.

The rapid increase in flow velocities may have enhanced the ice flux towards the glacier terminus due to the longitudinal strain. Positive emergence velocities are distributed up to 2400–3500 m from the 2002 terminus for the 2011 land-terminating condition (Sect. 4.4). However, Ve decreased in 2017 and became negative due to the increase in flow velocities toward the terminus. The unchanged thinning rate and velocity regime of Lugge Glacier (Figs. 2b and 3b and Table 1) suggest that any recent climatic changes in the Lunana region could not have yielded a significant increase in surface ablation. Therefore, the increased thinning rate of Thorthormi Glacier can be largely attributed to the decrease in Ve. However, the decrease in Ve (approximately -6 m a-1; Table 2) seems to be too large to account for the increased thinning rate from 2004–2011 to 2011–2018 (-1.38 m a-1; Table 1). We quantified the change in emergence velocity by hypothesizing that this change occurred in the last 2 years (2017 and 2018), during which time the surface velocity accelerated (Fig. 4a). The weighted average of the emergence velocity for the 2011–2018 period (Ve,2011–2018, m a-1) is described as follows: 7Ve,2011–2018=ΔtlandVe,land+ΔtlakeVe,lakeΔtland+Δtlake, where Ve and Δt are the emergence velocity and duration of the land- or lake-terminating conditions, respectively. We obtained a time-weighted mean emergence velocity of 3.52 m a-1 for the 2011–2018 period based on emergence velocities of Ve,land=5.20 and Ve,lake=-0.69 m a-1 (assuming a depth-averaged velocity that is 90 % of the surface velocity; Table 2) and periods of Δtland=5 and Δtlake=2. This means that Ve decreased by -1.68 m a-1 around 2011, which is consistent with the 2004–2011 to 2011–2018 change in dh/dt of -1.38 m a-1. The highly variable Ve profiles suggest that there are large uncertainties in the estimates (Fig. S4 and Table 2); however, our first-order evaluation can explain the cause of the drastic change in the thinning rate of Thorthormi Glacier.

We simply assumed a constant glacier width to calculate Ve along the central flowline. However, this glacier terrain tends to widen in the downglacier direction, yielding an extensional velocity regime. The lateral proglacial lakes on both sides of the terminus before it transitioned to a lake-terminating condition may have further contributed to a more negative Ve than that estimated along the central flowline. Despite these favourable conditions to enhance dynamic thinning, the surface lowering of Thorthormi Glacier has likely been suppressed by the compressive flow regime of the land-terminating condition. The transition to a lake-terminating condition should have caused a 2-fold increase in the thinning rate during such a short period (Fig. 2a and Table 1).

These above-mentioned mechanisms might cause a positive feedback between glacier thinning and the increase in flow velocity by enhancing each other. Therefore, increased glacier thinning and surface velocity speed-up will continue along Thorthormi Glacier in the future. The dynamic thinning of lake-terminating glaciers has been discussed in other HMA regions (e.g. Nuimura et al., 2012; King et al., 2018; Liu et al., 2020). However, our study is the first reported observation of the dynamic changes during the transition from land- to lake-terminating conditions, which have led to the enhanced thinning of a Himalayan glacier.

This study employed a modelled ice thickness (lake depth) that was tuned using point measurement data (Tsutaki et al., 2019) to estimate the dynamics of Thorthormi Glacier. Previous studies have suggested that the surface flow velocity of lake-terminating glaciers is sensitive to the terminus ice thickness and lake water depth (Benn et al., 2007; Pronk et al., 2021). Therefore, constraints on the lake bathymetry may allow us to better understand past and current terminus conditions and quantify the dynamic thinning process.