We manually outlined glacier front positions for the marine- and the lake-terminating glacier based on optical satellite images from Landsat 4, 5, 7 and 8 for the period 1992–2024. The images were downloaded using Google Earth Engine, and the margins were digitised using ArcGIS Pro. For Landsat 4 and 5, we used the near-infrared band (band 4) at a spatial resolution of 30 m. For Landsat 7 and 8, we used the panchromatic band (band 8), with a spatial resolution of 15 m. We identified additional front positions for 1987 based on orthorectified 2 m aerial images and for 1953 based on aerial images from The Danish Agency of Climate Data. The 1953 images were manually georeferenced based on tie points. We outlined a total of 389 and 308 front positions for MT and LT, respectively. While Landsat 4, 5 and 7 images only allowed for a relatively sparse temporal resolution with about 3–5 images per year, cloud-free Landsat 8 images were available on average every 15 (MT) and 17 d (LT), allowing for a seasonal analysis over the 11 years from 2014 to 2024. We quantified relative changes in the terminus positions using the variable box method as implemented in the Margin change Quantification Tool (MaQiT) . By considering the entire width of the glacier terminus, the method allows the detection of uneven changes along the glacier front. For the lake terminus, we identify major calving events by extracting glacier front position changes larger than 50 m between two images. Since the number of calving events is relatively small, we visually verify all instances based on satellite images.

We evaluated changes in ice surface elevations with two datasets. First, a 25 m resolution historical DEM . This DEM was constructed with stereo-photogrammetry based on ∼ 2 m aerial images covering the entire Greenland margin. The Danish Agency for Data Supply and Efficiency (SDFE) acquired the images in aerial campaigns between 1978 and 1987. In our region of interest, the underlying images of the DEM were taken in 1987. The DEM is co-registered to ICESat laser altimetry elevations, and the authors report an accuracy better than 10 m horizontally and 6 m vertically .

Second, we used the ArcticDEM strip collection, version 4.1 . The dataset is a compilation of 2 m resolution DEMs constructed with stereo photogrammetry based on submeter resolution Maxar satellite images . DEMs are available for single stereo image pairs between 2007 and 2023 (2012–2023 in our region). The DEMs vary in size and often do not cover both glacier termini at the same time. We accessed the DEMs via the FRIDGE portal of the Polar Geospatial Center (PGC, https://fridge.pgc.umn.edu/, last access: 10 July 2026). To reduce the influence of snow cover on the measured elevation changes we filtered the dataset to exclusively contain DEMs acquired in July, August, or September. Following that, we selected the first and the latest available DEM, which cover both glacier termini. The selected DEMs are from 14 August 2012 and 30 September 2023.

report an accuracy of the ArcticDEMs of about 4 m horizontally and vertically. To reduce the uncertainty in calculated elevation changes, we co-registered the DEMs with a method based on , as it is implemented in the Python package xDEM . For the co-registration, we masked out all glaciated areas and water bodies and only used assumed stable terrain (bedrock areas). Additionally, we applied the reliability mask, which is provided with the ArcticDEM, to remove photogrammetric artefacts. With the co-registration, we were able to reduce the median elevation change above stable terrain from -3.12 to 0.001 m for the ArcticDEM and from -3.17 to 0.002 m for the historical DEM. The normalised median absolute deviation (NMAD), which gives an indication of dispersion, was reduced from 0.92 to 0.51 m and from 3.89 to 3.55 m for the ArcticDEM and historical DEM, respectively. Thus, for the 2012–2023 period we can reliably detect changes greater than 0.51 m, and for the 1987–2012 period we can reliably detect changes greater than 3.55 m.

To avoid point sampling of local features such as crevasses, we resampled the co-registered ArcticDEMs to 25 m resolution using bilinear interpolation. Furthermore, we extracted width-averaged values of a 400 m wide band following the glacier centre lines (Fig. ).

We extracted NASA ITS_LIVE ice surface velocities at both glacier fronts and about 8 km upstream of the marine terminus (locations indicated in Fig. 1). The ITS_LIVE dataset contains image pair velocities from Landsat 4, 5, 7, 8, 9, Sentinel 1 and 2. The velocities are constructed with the auto-RIFT feature tracking algorithm . With the launch of Sentinel 1 and 2 satellites between 2014 and 2017 the amount of available image pair velocities has increased substantially. Therefore, we constrained our analysis to the period 2016–2024 with several thousand available image pair velocities.

The difficulty in analysing ITS_LIVE data lies in the different temporal base lines of image pairs, making them not comparable to each other. While some image pairs might be separated by only a few days, others might be separated by more than a year. To address this issue, we make use of the Python package TICOI (Temporal Inversion using Combination of Observations and Interpolation) . The algorithm performs a temporal inversion to entangle the contribution of each image pair to the velocity at a given time. Furthermore, TICOI performs an interpolation in order to provide velocities at a regular, comparable interval.

The ITS_LIVE data is stored in cloud-optimised Zarr data cubes, from where we access the data directly through the TICOI Python package. Before running TICOI, we apply several filters to remove outliers. First, we filter out image pairs that have a flow direction differing by more than 45° from the median flow direction of all observations. Second, we filter out values below or above 80 % of the mean velocity. Finally, we filter out image pair velocities that have an error > 100 m in the error estimates provided by ITS_LIVE. We find that image pairs acquired by SAR sensors show a very large scatter for the slower flowing lake terminus. Therefore, we exclude Sentinel-1 image pairs for this glacier.

We measured three lake temperature profiles from the water surface to the bottom of the lake with an RBR concerto CTD measuring conductivity, temperature, and depth at 8 Hz. The manufacturer states an initial accuracy of ±0.002 °C for the temperature. Moreover, we measured a temperature and lake level time series for almost an entire year with a temperature and depth sensor (RBR Duet3). The sensor was placed inside a metal pipe and attached to a rock wall with Dyneema rope. The sensor was placed at approximately 5.4 m depth and measured temperature and pressure at a 30 min interval between 9 August 2024 and 1 August 2025. Measurements were compensated for atmospheric pressure with a barometric unit placed close to the lake during the measurement period. The locations of all measurements are indicated in Fig. .

We conducted a bathymetric survey of Lake Motzfeldt with a Kongsberg EA440 single-beam echo-sounder with a 38 and 200 kHz combi transducer. The instrument was mounted to a self-built aluminium frame attached to a 4.5 m long inflatable boat. An Emlid Reach RS3 GNSS receiver was used for location input at 5 Hz and was mounted on top of the transducer. During surveying, the boat was manoeuvred at approximately 9 km h⁻¹ aiming to follow a survey grid with a spacing of 200 m. However, large amounts of icebergs prevented us from doing so on several occasions, leading to areas with slightly more or slightly less spacing between survey lines. The measurements were corrected for sound velocity based on the CTD measurements. The data was manually cleaned of outliers in ArcGIS Pro. To produce a bathymetric map, we used the Python implementation of the Generic Mapping Tool (PyGMT) . We first calculated a block-median with a grid cell-size of 25 m before applying a spline interpolation function with a tension factor of 0.6. Finally, a grid was exported at 25 m resolution.

For the lake terminus we estimated the glaciers' floatation potential based on the ice thickness (assuming floatation) and the lake depth. Following Archimedes' principle of buoyancy we calculated the ice thickness below water (Hbw) from the freeboard (h), an ice density (ρi) of 917 kg m⁻³ and a water density (ρw) of 1000 kg m⁻³ as: 1Hbw=(ρw/(ρw-ρi)-1)h We derive the freeboard for the 1987 and 2012 DEMs by subtracting the respective lake levels from the glacier surface elevation, where an overlap with the bathymetry data was present. Consequently, we calculate the floatation index (I) as: 2I=Hbw-D where D is the lake depth. The glacier is floating where Hbw is smaller than D and grounded where Hbw exceeds D. The observed lake level changes introduce an error to D, which, however, is relatively small (< 2 m).

After initial tests to automatically detect lake ice based on SAR or optical satellite data, we find it unsuitable for Lake Motzfeldt because of the immense amount of icebergs permanently covering the lake. Therefore, we fall back to a manual detection based on a combination of optical satellite images and air temperature. We use GEEDiT to manually label Sentinel-2, Landsat-7, 8 and 9 images with “full ice cover”, “partial ice cover” and “no ice cover”. Based on long-term climatic means from the Narsarsuaq weather station, we solely inspect images after mid-April (lake ice breakup) and after the first of October (lake freeze up). Once the lake ice cover is established in the fall, we assume that it persists until breakup in May or June. We verified this by checking all instances where daily mean temperatures rose above zero for more than 5 consecutive days. Despite hindering automatic detection, the movement of the icebergs is an excellent indicator of the presence of lake ice in the manual classification. Without lake ice, icebergs move due to wind and water currents. However, as soon as the lake ice cover is established, the icebergs are locked in place. To identify lake ice, we also used the visual colour difference between ice and open water, cracks in the lake ice, and snow cover. Since a partial lake ice cover likely only has a minor buttressing effect, we report the lake ice duration as the time between the first and last image labelled with “full ice cover” for every winter season.

We acquire mean daily air temperature data for Narsarsuaq from the Danish Meteorological Institute (DMI) . The station with the ID 04270 is located at Narsarsuaq Airport at an elevation of approximately 34 m above sea level and approximately 12 km west of the marine terminus and 18 km west of the lake terminus (Fig. ). The available air temperature data spans from 1961 to present.

We obtained freshwater runoff for Qooqqup Sermia from the Regional Atmospheric Climate Model (RACMO2.3p2), statistically downscaled to 1 km . We spatially average daily runoff data for the entire Qooqqup Sermia based on the catchment from .

In the period analysed between 1953 and 2024, the two glacier termini show clear differences in their overall evolution of the terminus position, as well as in their retreat and advance patterns (Fig. ). At first, both glacier fronts show a net advance between the first aerial observation in 1953 and the first satellite observations in 1992. Subsequently, however, they develop differently. The marine terminus reached its most advanced position in 1994, from where it gradually retreated about 1.3 km until 2004. Since 2004, the terminus has remained in a consistent position, with a seasonal advance and retreat pattern (Fig. a, blue shading). The lake terminus, on the other hand, held a stable position until 2012, when it abruptly retreated about 3.2 km within one year. Since this event, the glacier front has transitioned into a mode of long phases of gradual front advance followed by abrupt, large calving events (Fig. a, red shading), producing tabular icebergs.

Based on a comparison of digital elevation models, we observe pronounced thinning at both glacier termini (Fig. ). At the marine-terminating glacier, the highest thinning rates are observed close to the glacier front, with a rate of 2.2 m yr⁻¹, or a total of 80 m, between 1987 and 2023. With increasing elevation, the thinning rates decrease continuously to approximately 1.7 m yr⁻¹ at a distance of 8 km from the glacier front. At the lake terminus, the lowest thinning rates are observed across the floating glacier tongue, only 0.3 m yr⁻¹ or 7.7 m in total before its disintegration in 2012. These thinning rates amount to approximately 10 % of the thinning above grounded areas, due to to buoyant adjustment. At LT, the highest thinning rates are observed upstream of what became the new glacier front after the disintegration in 2012, with a rate of 2.3 m yr⁻¹ or a total of 82 m (1987–2023).

By comparing thinning rates for both glaciers at the same elevation (> 200 m above sea level), we find that between 1987 and 2012, the thinning rates are similar, with 2.1 m yr⁻¹ (LT) and 2.0 m yr⁻¹ (MT). However, between 2012 and 2023, the thinning rate of the lake-terminating glacier exceeded that of the marine-terminating glacier by 0.4 m yr⁻¹, with rates of 1.7 and 1.3 m yr⁻¹, respectively. Both glaciers clearly show lower thinning rates after 2012.

The bathymetric survey of Lake Motzfeldt reveals an exceptionally deep lake with a maximum lake depth of 368 m (Fig. a, red dot) and an average of 131 m. The lake is generally shallower south of the lake bend and gradually deepens toward the north. The lake morphology is characterised by steep side walls and a flat, deep middle part. At the position where the glacier was stagnant until 2012, a clearly visible ridge extends from the western lake shore to the middle of the lake (Fig. a). However, the composition of this ridge is unknown.

The floatation index shows that in 1987, extended parts of the glacier front were grounded on and upstream of the ridge, as well as on both lateral margins (Fig. b). On the ridge, the floatation index reaches values of up to 200 m, which means that the theoretical ice thickness, assuming floatation, exceeds the lake depth at that position by 200 m. Due to substantial thinning (see the previous section), most of the pixels classified as grounded in 1987 are classified as floating in 2012. The grounded area on the ridge was largely diminished with a maximum floatation index value of 80 in 2012. The DEM from 2012 shows that, at that time, the lateral margins of the glacier had crumbled apart, leading to the loss of grounded areas on the sides.

The three CTD measurements show a cold lake with a depth-averaged temperature of about 0.65, 0.68 and 0.71 °C for CTD 1, 2 and 3 respectively (Fig. in the Appendix). All CTD casts are coldest at the top and gradually become warmer toward the bottom. The coldest temperatures are found closest to the glacier front (CTD 1), and the warmest are farthest away (CTD 3). CTD 1 and 2 remain below 0.75 °C at all depths, while CTD 3 shows slightly warmer water at the bottom, with a maximum temperature of 0.86 °C. The time series of the temperature sensor that was placed from 2024 to 2025, shows that temperatures decreased from about 0.7 °C at deployment in August to 0.2 °C in November (Fig. ). Between November and mid May, the lake water stays below 0.2 °C, with very little daily variation. Following that, temperatures rise again to a maximum of 0.8 °C in the end of July before starting to decrease again in September. The lake level, inferred from the sensor depth, remains approximately constant between the beginning of October and the beginning of July. In July, the lake level rises by 1.6 m compared to the winter low. In August and September, the lake level is generally high (1 m above winter low); however, it shortly lowers by 0.6 m at the end of August.

We tracked changes in glacier front position with high temporal resolution between 2014 and 2024 and found a characteristic ice-shelf calving style at the lake-terminating glacier (Fig. ). Following the disintegration of the ice tongue in 2012–2013, the LT glacier front has experienced prolonged phases of gradual front advance, interrupted by abrupt, large calving events. In some years, such as 2020, the advance phase persisted through the summer, with the glacier front advancing approximately 500 m over the entire year without any noticeable calving events. As shown in Fig. , the glacier advances primarily in the centre of the lake, with no lateral contact with the lake shore.

Between 2014 and 2024, we identified 15 major calving events, each resulting in a front retreat exceeding 50 m (Fig. ). The largest single-event retreat measured 923 m, while the most substantial cumulative retreat measured about 1.4 km in the summer of 2017 during three consecutive events. From satellite images, we observed that these events produced large tabular icebergs that do not capsize. All calving events coincided with the runoff season and occurred during lake ice-free conditions on Lake Motzfeldt (Figs. , ).

The duration of the full lake ice cover in Lake Motzfeldt averaged 212 d, ranging from 187 d in 2015 to 230 d in 2021 (Fig. ). Lake ice cover generally developed in October, with one exception in 2015 where it developed in early November. The earliest onset of the full ice cover was observed on 8 October 2022. Lake ice break up usually occurs in May or June, the earliest recorded on 2 May 2021 and the latest on 21 June 2015.

In the observed period 2016–2024, the ice flow velocities at the lake terminus ranged between 380 and 610 m yr⁻¹, with a gradual decline in the average annual velocity from ∼ 544 m yr⁻¹ in 2016 to ∼ 472 m yr⁻¹ in 2024. A clear seasonal signal can be identified with a 15 % increase from winter (DJF) to summer (JJA) velocities. Ice flow velocities usually increase with the beginning of the runoff season and drop again at the end of the runoff season. In some cases, calving events coincide with small flow velocity increases (Fig. ).

Between 2014 and 2024, the marine-terminating glacier showed a distinct seasonal cycle, with front advance occurring on average between October and May and retreat from June through September (Figs. and ). Notable exceptions are the years 2019, 2020 and 2021, where retreat already started in February, April, and March, respectively. The average magnitude of front position change between winter and summer positions is 280 m. The largest differences occur in the years 2014–2016 with 350, 400 and 410 m difference between minimum and maximum positions.

Ice flow velocities at the glacier front ranged between 1.4 and 2.1 km yr⁻¹, with the lowest annual average velocities in 2024 (1.5 km yr⁻¹) and the highest during 2019–2021 (∼ 1.9 km yr⁻¹) (Fig. ). During these peak years, velocities continued to accelerate even after the runoff season, reaching a maximum in December. As a result, the highest monthly velocities over the full period were on average during October through December. However, no clear seasonal signal can be identified.

Eight kilometres upstream, the ice flow velocities are notably lower, ranging from 0.7 to 0.9 km yr⁻¹. Similarly to the terminus, the highest annual velocities occurred in 2019–2022 and the lowest in 2024. Although velocities are on average highest in May, June, and July, seasonal variability is low, less than 4 %. In the upstream velocities, we see a strong acceleration each year coinciding with the beginning of the runoff period.

To assess potential drivers of terminus position variation, we compared monthly retreat rates with air temperature and runoff. We found a positive correlation of front position and runoff (r=0.6), and air temperature (r=0.5). As shown in Fig. , the runoff period aligns closely with the retreat period.

The flat morphology of the glacier surface, the production of large tabular icebergs, as well as the buoyancy compensated thinning rates of LT suggest that prior to 2012, a floating ice tongue was present at the lake terminus (Figs. and ). Although we are unaware of other descriptions of floating ice tongues in Greenlandic lakes, they are a common phenomenon in other regions like Patagonia , Alaska , and Iceland . In Greenland, fully floating ice tongues are found at marine-terminating glaciers, but only in the very north following the disintegration of several ice tongues further south .

We believe that the formation and sustained presence (at least 1987–2012) of the ice tongue in Lake Motzfeldt in the very south of Greenland was possible for two main reasons. First, lake Motzfeldt is exceptionally deep and therefore provides the necessary geometry. Second, subaqueous melt rates in Lake Motzfeldt are low, therefore limiting thinning rates and promoting stability. This falls in line with who suggest that floating ice tongues in lakes can occur in more temperate regions further south compared to marine-terminating glaciers due to colder water temperatures and the lack of salinity-driven circulation, which strongly limit subaqueous melt as well as calving as a result of undercutting of the glacier front. And indeed, our CTD and temperature measurements in Lake Motzfeldt show a very cold lake, with a depth averaged temperature at the glacier front of < 0.7 °C in August 2025. The lake can sustain such a cold temperature throughout the year due to meltwater input and the immense amount of icebergs present. Any additional energy input through solar radiation or heat advection is likely transferred into the icebergs, efficiently cooling the lake.

Nonetheless, in 2012 and 2013, several large calving events led to the disintegration and massive loss of more than 3 km of the glacier terminus. Our bathymetry data and the historical DEMs indicate that the disintegration of the floating ice tongue was due to the separation from a pinning point at the glacier front. These pinning points exert important backstress on the glacier, and their loss is known to have destabilising effects . In 1987, the glacier was well grounded at the pinning point. However, gradual thinning led to the almost complete separation from the the pinning point by 2012 (Figs. and ). Since the disintegration of the floating ice tongue, the glacier has changed to a mode of long advance phases followed by abrupt, large calving events. These calving events still produce large tabular icebergs, which indicates that a floating tongue develops each time during the advance phase, again pointing to low subaqueous melt rates.

We observe a distinct calving pattern at the lake terminus that can be described as follows: The glacier experiences long advances phases that can persist throughout the melt seasons and can last longer than a full year. These advance phases are interrupted by the calving of large tabular icebergs (most larger than 200 m). Some calving events follow the formation of clearly visible rifts, while others don't show clear rift formation. Almost all icebergs detach from the glacier front and drift away without capsizing. This calving style is typical for fully floating ice tongues in Greenland and resembles that of large fully floating ice tongues in North Greenland, like Peterman e.g. and Ryder Glacier e.g..

Remarkably, all documented calving events occur between mid May and mid October, coinciding with the runoff period. This could indicate that also buoyant forces play a role in triggering calving events, as it was observed at several other lake-terminating glaciers . During the runoff period, buoyant forces are increased through thinning of the glacier from the surface as well as through a rise in lake level due to increased meltwater input (Fig. ). Furthermore, we observe that all calving events occur outside of the lake ice covered period. Although lake ice could provide some backstress to the glacier, it is unclear as to what magnitude and what effect it would have on calving. However, we believe that lake ice possibly functions as a binder, preventing icebergs from drifting away, helping to keep the glacier front together during winter.

Ice flow velocities are generally slower at LT compared to the center of MT, which already is apparent at the beginning of LT where it branches off from MT. These velocity differences may stem from the bedrock geometry. The overall elevation drop is likely higher for MT than for LT, as the latter terminates in Lake Motzfeldt about 160 m above sea level, whereas MT extends to sea level. Bedrock data from BedMachine (BM) version 5 indicates a small bump at the start of LT which could slow down the flow from MT into the valley of LT (Fig. ). Unfortunately, bedrock measurements in South Greenland and on Qooqqup Sermia in particular are notoriously difficult to obtain, and BM relies on poorly constrained indirect estimates . However, recent measurements by with AIRETH, a redesigned helicopter-borne ground-penetrating radar, show that the bedrock elevation at LT is even higher than reported in BM and that the bedrock slope toward the glacier front is likely less steep (Fig. ). Measurements with the same system along the marine terminus of Qooqqup Sermia could not recover reliable bed returns, indicating a greater ice thickness. As a result, the bedrock geometry beneath LT is now better constrained, while BM remains unreliable at MT, making a comparison of bed geometry between the two termini difficult. In addition to the bedrock geometry, it appears as LT is fed by the slower lateral ice from the margins of MT rather than by its fast central trunk.

Despite their large magnitude, calving events do not appear to have an effect on the glaciers flow velocities, as flow velocities only seem to follow a pattern of a swift increase at the onset of the runoff season and a progressive decrease at the end of the season (Fig. ).

The historical evolution, the seasonal front position, as well as the ice flow velocity largely resemble those of other marine-terminating glaciers in Greenland, with differences discussed below. Dissimilar is the net advance during the pre-satellite era (here 1953–1992) of about 640 m. Notably, describe a retreat of the glacier front between 1942, 1953 and 1981, indicating that the advance must have occurred only between 1981 and 1992, surpassing the initial retreat. Although mean annual air temperatures were considerably lower than average for a few years in the early 1980s, it is difficult to pinpoint the cause of the advance. For the neighbouring Eqalorutsit Kangilliit Sermiat, also describes an anomalous advance (during a more extended period 1942–2000), noting a complicated precipitation pattern with large spatial variations over a small area as a possible cause.

After reaching its most advanced position in 1994 the glacier recedes about 1.3 km until 2004, where it remains stable until today. Although the cause of the retreat of other marine-terminating glaciers in Greenland during that period is often induced by warming ocean temperatures , we lack oceanographic data in the inner fjord to confirm this at Qooqqup Sermia. Furthermore, it appears that the position before 2004 was an overextension of the glacier to an unstable position without any visible pinning points. An indication of this is that the glacier front did not remain at a specific position before 2004, but rather changed every year.

Since 2004, the glacier remained at approximately the same position with seasonal ice-front advance during winter (October–May) and retreat during summer (June–September). The position around which the glacier front oscillates is characterised by a visible narrowing of the fjord through a bedrock protrusion on either side. Seasonal advance and retreat of glacier fronts are common phenomena , observed at more than 80 % of marine-terminating glaciers in Greenland . This seasonality of glacier front positions has been suggested to be mainly driven by meltwater runoff and the presence or absence of an ice mélange , both of which affect the calving rate. In agreement with these findings, we observe that the majority of Qooqqup Sermia’s frontal retreat takes place during the runoff season, whereas most of its advance occurs outside of the runoff season (Figs. and ), indicating that the ice-front position is influenced by submarine melting due to meltwater runoff. This is supported by observations during both field campaigns and on satellite images, where we observed meltwater plumes for extended periods during the melt season (see Fig. A3). These plumes, formed by runoff entering the fjord subglacially, enhance submarine melting and can trigger calving by undercutting the ice-front . Figure A3b also shows strong undercutting of the glacier front at the location of the plume. The calving that we observe at MT can be classified as serac collapse type with frequent calving of small icebergs.

However, in some years, the coupling of ice-front position and runoff is slightly less evident. In the years 2019–2021 the front position starts to retreat before the beginning of the melt season. Similar observations at other glaciers showed that this can be caused by an early clearing of the ice mélange and the accompanied reduction of resistive stresses at the glacier front . However, we have no time series documenting the presence of an ice mélange at Qooqqup Sermia to confirm this. Therefore, we conclude that the seasonality of glacier front positions is influenced by subglacial runoff, however, with other processes likely also playing role.

As is typical for calving glaciers, Qooqqup Sermia's flow velocity also increases towards the terminus, thereby more than doubling its speed from 8 km upstream to the glacier front. The upstream ice-flow velocities remain relatively constant throughout the observation period, with a notable spike at the beginning of the runoff season each year. This short-term increase in flow velocities is also observed in other regions of Greenland and can be linked to the efficiency of subglacial drainage when meltwater or rainwater enters the system . At the glacier front, ice flow velocities begin to accelerate in late spring and continue to accelerate even after peak runoff is reached, indicating a decoupling from runoff after initial acceleration (Fig. ). This is observed for many glaciers in Greenland as shown by large scale and case specific studies e.g.. show that for this category of glaciers (called “type-1”), ice flow velocity is closely coupled to changes in ice-front position. Also, at Qooqqup Sermia, we observe that the highest flow velocities coincide with the most retreated glacier front positions. This can be explained by reduced resistive stresses when the ice front retreats behind the pinning point .

In previous sections, we illustrated that the lake- and marine-terminating glacier exhibit contrasting dynamics, despite being subject to the same climatic conditions. The key differences can be summarised as the following: We observe the marine terminus to be a fast flowing, grounded glacier with frequent small-size calving events, mainly in the form of serac failure type. The lake-terminating glacier, on the other hand, has a floating extension and experiences rare, but very large calving events, producing large tabular icebergs. Furthermore, MT shows a clear seasonality with glacier front retreat in summer and advance in winter as well as a coupling of ice flow velocities to the glacier front position. In contrast, LT does not have seasonal ice front fluctuations, and ice flow velocities do not seem to be coupled to the front position.

Differences in bedrock geometry and ice thickness certainly partly determine the dynamics at the two termini. However, in addition, we identified the following physical differences between the two environments, which potentially influence the morphology and dynamics at the glacier termini. The following considerations apply to this study and, in general, to the comparison of marine- and lake-terminating glaciers.

Freezing point depression: In the marine environment, the freezing point of ocean water is lowered due to the salt content. With a salinity of 32 PSU the freezing point drops to about -1.8 °C. A lower freezing point increases the thermal driving (or thermal excess), consequently increasing submarine melt rates. Even in a cold ocean environment this freezing point depression leads to higher melt rates compared to a freshwater environment. For example, assuming an ocean temperature of 0.6 °C (average temperature of the top 100 m measured by in Tunulliarfik Fjord, close to Qooroq Fjord) and a lake temperature of 0.7 °C (measured in lake Motzfeldt) results in a thermal driving that is 3 to 4 times higher in the ocean compared to the lake (2.4 °C vs. 0.7 °C above the freezing point).

The differences outlined above clearly demonstrate that subaqueous melt rates must be higher at MT than at LT, even with cold ocean conditions. The presence of a meltwater plume, the undercutting of the ice front, and the correlation between surface runoff and glacier front position all suggest that submarine melting affects the dynamics of the marine-terminating glacier. At LT on the other hand, low melt rates create favourable conditions for a floating extension of the glacier through limited thinning and calving rates. This comparison indicates that the difference in subaqueous melt rate contributes to the contrasting dynamics that we observe.