The scientific community disagrees on the extent to which mélange directly impacts calving rates or vice versa . Some studies have pointed towards the mélange as a mechanical inhibitor of the calving, where calving happens after a removal or substantial weakening of ice mélange, thus placing it as a controlling factor . Multiple studies suggest mélange can prevent crevasse propagation and thereby hinder and limit calving. It has also been suggested that calving activity can influence the properties of the mélange through the generation of waves resulting in mélange dispersal and increased fracturing, thereby weakening the buttressing effect, potentially increasing further calving activities . suggest that ocean temperatures control the retreat and advance of marine-terminating outlet glaciers in Greenland, via changes in mélange rigidity that helps to modulate the velocity of the mélange as it moves down the fjord, pushed by the outlet glacier. However, other studies underline that the buttressing effect of the mélange depends on characteristics in the mélange such as presence of landfast sea ice . Studies in Antarctica for example, suggest that mélange is less significant than thick multi-year sea ice in exerting a buttressing effect on outlet glaciers and can also explain seasonal calving dynamics . The importance of mélange may also vary from glacier to glacier depending on the cohesion and density of the mélange . In this study we focus on the seasonal change in mélange velocities, driven by the formation and break up of landfast sea ice, which at many locations in Greenland acts as a seasonal glue, sticking together individual icebergs and smaller growlers and bergy bits within the mélange zone, and connecting the mélange to the wider fjord, ocean and sea ice system. Many studies of mélange in Greenland have focused on Jakobshavn Isbræ or Helheim glaciers. These are very large and constantly calving outlets of the ice sheet in long narrow fjords that may not be representative of other smaller and slower Greenland outlets where icebergs also calve to form a mélange zone that is seasonally frozen into landfast sea ice. Work by for example, also shows the importance of fjord geometry on mélange processes based on laboratory experiments. Similarly, analogies from Greenland may not be applicable at Antarctic ice fronts where, along with generally colder conditions that likely affect ice material properties, there is also a very broad range of geometries ranging from narrow fjords , to embayments and large open ice shelf fronts . It is thus important to consider local factors in studying ice mélange interactions with glaciers and landfast sea ice in order to make generalisable conclusions.

In this study we combine high time resolution field observations and analysis of satellite data from Northwest Greenland to assess the combined influence of ice mélange and landfast sea ice on typical calving outlet glaciers. Based on field observations in the winter period we suggest defining an inner mélange zone with very dense iceberg particles that can jam against each other as in e.g. and an outer mélange zone with a lower concentration of icebergs (see Fig. ). The inner zone is difficult to access and our field study is mostly limited to the landfast sea ice and between the large icebergs of the outer zone. The distinction of zones is clear from satellite imagery from the summer open water period (see Fig. ), but we note also that the mélange migrates considerably in the summer period after the break up of landfast sea ice and the distinction between inner and outer zones does not always hold (see also the Video Supplement).

Inglefield Bredning (Kangerlussuaq) is a wide gulf in Northwest Greenland with many glaciers discharging into the fjord that feeds it. Melville, Farquhar, and Tracy glaciers (Qeqertaarsuusarsuup Sermia, Toornaarsulissuup Sermii, and Tuttulipalluup Sermia in Greenlandic ) are located next to each other at around 77.6° N and 66.4° W, with their marine-terminating glacier fronts feeding into a common area at the head of Inglefield Fjord. The study area is visualised in Fig. .

Glaciers in the region have been well studied, particularly Tracy and Heilprin glaciers due to their contrasting recent retreat behaviour and to a lesser extent Melville and Farquhar glaciers have also been included in studies by for example, and . According to , Tracy glacier has recorded a consistent retreat since the 1980s and from 1980s–2014 it had the most rapid retreat of all glaciers in the area, with an average rate of 200 ± 1 m yr⁻¹, twice as fast as Farquhar (94 ± 3 m yr⁻¹). The onset of Farquhar's retreat took place after the separation of Farquhar and Tracy's shared glacier terminus in a floating ice tongue in the summer of 2002 . The geometry of the glacier fronts changed completely after the separation, as two separate tidewater glaciers were created, resulting in a notable change in the force balance of both Farquhar and Tracy. The new force balance combined with increasing meltwater enhancing basal sliding resulted in an ice velocity acceleration . Today, the ice mélanges of Farquhar and Tracy are also blended together in the winter season until the break-up of sea ice at the head of Inglefield Fjord in July. Melville is the glacier with the lowest retreat rate of the three. It is located west of Farquhar and experienced a general retreat of 48 ± 1 m yr⁻¹ from 1980s–2014 . As with virtually all glaciers in Western Greenland, the glaciers of Inglefield Bredning are characterised by vertical thinning and retreat over the last decades. According to , Tracy and Farquhar have the highest rates of elevation change in the fjord system at -3.91 ± 0.13 and -2.91 ± 0.18 m yr⁻¹ from 2001–2018, respectively. The elevation change of Melville during the same period was -1.73 ± 0.38 m yr⁻¹, similar to the mean rate of outlet glaciers in Inglefield Bredning but larger than the average thinning at glaciers across Baffin Bay. Ocean observations by suggest that the rapid retreat of Tracy glacier is in part related to ocean warming and plume dynamics at the glacier front. In contrast, they suggest the lack of retreat and acceleration at neighbouring Heilprin glacier is most likely due to an outlet glacier front grounded in shallower water, though we note bathymetric data at all outlet glaciers is not well constrained. The three study glaciers differ in size; Tracy is the largest glacier with an area of 222.93 km², while Melville and Farquhar have areas of 118.52 and 53.74 km², respectively . In our study period, the measured width of the calving fronts in 2023 from satellite imagery was approximately 1.5 km for Melville, 2 km for Farquhar and 4 km for Tracy. All three glaciers are found in their own narrow fjords that widen out into the head of the larger Inglefield Fjord. Farquhar and Melville have shorter internal fjords of about 2 km and 3 km each compared to Tracy Glacier's inner fjord length of 4.5 km (in 2022). Observations by confirm measurements reported in this study and in of a warmer Atlantic layer situated below 200 m depth throughout the fjord system, consistent with other fjords in Western Greenland, indicating that turbulent mixing caused by subglacial plumes likely play a key role in submarine melt at these glacier termini . Unfortunately, the ocean depths in front of all three retreating glaciers in the study region are poorly known and mostly inferred from model inversions and limited radar observations . Bathymetric mapping of Inglefield Bredning using single beam echo sounding has been carried out as part of the Ocean Melting Greenland project but survey lines do not reach the mélange zone or current frontal position of the Tracy glacier. The study glaciers represent a range of shallower and deeper grounded glaciers, all of which can be seen to transiently float, leading to the occasional calving of tabular icebergs, observed in satellite imagery (see Fig. ). These different types of outlet glaciers are representative for Greenland calving termini and the calving styles on display also match those classified by . Ice mélange is typically present throughout the year in this region and it's relative accessibility from the nearby town of Qaanaaq provides an ideal process study site.

We supplemented data on mélange deformation from in situ data with remote sensing observations. These included both pre-processed ice velocity and calving front data, as well as novel datasets produced in this study. The purpose of the data comparison was to observe calving events, to estimate the location of the calving fronts and the size of the icebergs, as well as to identify transient clearing of ice mélange and the seasonal break-up of sea ice. The occurrence of melt water plumes at the surface and the movements and characteristics of large icebergs within the mélange zones (see Table in the Appendix) were also recorded but not further examined in this study. We also used 's seasonal ice velocity product derived from feature tracking Sentinel-1 data to assess how representative the GNSS buoy velocities measured in situ are across the mélange zone of the study glaciers.

Previous work link break-up and calving dynamics with seasonal atmospheric drivers, in particular thinning of sea ice and glacier surface melt and run-off driven by warming summer temperatures and wind forcing of mélange movements. We use two climate datasets to evaluate the importance of the surface boundary layer on mélange movements. The Copernicus Arctic Regional Reanalysis (CARRA) is a very high resolution (2.5 km) climate reanalysis that is updated in near real-time and has been shown to be particularly effective at capturing near-surface winds and sea ice . In this study we use the 6 hourly output wind field to assess if mélange velocities can be explained by persistent wind forcing. As CARRA does not include complex glacier surface processes, we also use surface melt and SMB driven by output from the HIRHAM5 regional climate model (RCM) . In this study, HIRHAM5 is forced with the ERA-5 climate reanalysis data to identify the onset of melt and to quantify the amount of melt produced at the study glaciers. HIRHAM5 is run at a horizontal resolution of 0.05 × 0.05 ° (5.5 km) on a rotated polar grid and has been shown to resolve the local climate well in this region . The RCM outputs are further used to drive a detailed offline SMB model with 32 vertical levels , which, as analysis with in situ observations shows, reproduces observed firn processes like temperature and density well . Assessing uncertainties in climate, SMB and melt data is difficult in this region as we lack observations from low elevations, however a recent analysis of the simulation by shows that HIRHAM performs well in both timing of melt onset as well as total number of melt days at higher elevations when compared with ASCAT satellite data. In this study we use the modelled outputs largely to assess timing of the onset of melt and runoff and the relative magnitude of melt events.

The GNSS position observations show a consistent pattern of movement spatially and temporally in both 2022 and 2023 in Figs.  and , with mélange moving slowly but continuously away from the glacier fronts, with occasional abrupt jumps in position recorded. The mélange velocities when bounded by landfast sea ice have a uniform and steady flow, but are slightly faster, particularly during abrupt movements, closer to the glacier fronts, and slower further away, with the direction of movement suggesting that the larger Tracy glacier has a greater influence on ice mélange velocity than the other two termini. This would also indicate that Tracy is less sensitive to any buttressing effects from mélange and fast ice. The large abrupt extensional movements that intersperse the steady ice mélange drift are shown in Figs.  and . The buoys in the mélange have slightly different median velocities in the range of 7.6–11.3 m d ⁻¹ in 2022 and 1.4–9.7 m d⁻¹ in 2023. Note though that between the two field seasons the buoys were placed in different relative positions across the fjord, due to access differences between the seasons. Our results indicate however that proximity to the glacier front as well as location relative to centre line of fast-flowing ice can explain differences in velocity, but a longer term and more spatially dispersed dataset is necessary to fully determine this. Future deployments will allow us to increase the area observed in both time and space. The large number of large icebergs close to Josephine Peary Island shown in both Figs.  and are likely grounded, and form a stabilising location for the mélange where extensional movements gradually fall off. The location coincides with the former location of the ice tongue that broke up in 2002 and is probably a submarine moraine ridge. The abrupt movement events that we have identified in the record happen within the same 10–30 min time intervals of all the GNSS buoys. It is therefore unfortunately not possible given the precision of the instruments to assess how quickly the signal travels between the buoys. We observe though, that in addition to happening concurrently, the measured speeds decrease with distance from the glacier. Our data indicates that the largest of the measured abrupt movements represented up to 804 m of horizontal movement near Tracy Glacier during the half hour measurement period. These large abrupt movements correlate with large calving events at the glacier front (see Table ). The observations in Figs.  and also do not show much change in average velocity of the mélange through the season until the landfast sea ice break up approaches in mid July.

We summarise the observed calving events and distances moved by the specific buoys as a result in Table . It is difficult to give full uncertainty estimates on the position data as we do not have a differential GNSS system in operation at this site. We reduce noise introduced into the system by individual measurements by averaging all calculated velocities, derived from position information over each day. This smooths out noise introduced by GNSS signal variability. We note also that the abrupt jumps as well as the steady state drift of all buoys is consistent across all positions, suggesting that the stated position accuracy and precision of the buoys themselves is likely underestimated in Table . We have opted to omit the uncertainty bars on the figures in this paper for clarity, but full error estimates are available in the raw datafiles (see “Code availability” and “Data availability” sections). Buoys in front of Farquhar follow vectors that show an influence from Tracy as in Figs.  and but are also being partly steered by the closer glacier. The buoys in front of Melville in 2023, show no influence from Tracy, only their local Melville glacier, but given the local topography and their position inside the local Melville glacier fjord arm, this is not surprising. We attribute the erratic back and forth movements of Ismaage (Fig. ) in 2023 to its location on very thin sea ice over a bedrock sill in relatively shallow (150 m depth, measured in the field) water, where tidal mixing in the fjord below the ice causes regular melt and fracture of the ice on the corner of the fjord between Farquhar and Tracy glaciers (see satellite timelapse video in the Video Supplement). The in situ observed velocities are helpful to compare with and evaluate remote sensing derived glacier and mélange velocities (Fig. ). The Sentinel-1 data has a 12 d repeat orbit, which means it cannot be used to directly assess the abrupt events identified in the GNSS buoy data, though the table in the Appendix (Table ) and calving events visualised in Fig.  do allow some relationships to be defined. Our analysis in Fig.  shows the GNSS velocities are similar to and centred around the satellite derived velocities of the lower part of Tracy glacier and the mélange in front of Tracy, with a tendency towards generally higher velocities in the GNSS observations of mélange. While the GNSS data have very similar average velocities to the satellite derived velocities, they have much higher daily variability, which it is not possible to capture given the image repeat time. Interestingly, our results confirm that the sea ice in the mélange zone and the outer fast ice zone is moving away from the glacier fronts slightly faster than the outlet glaciers are advancing, even when the landfast ice is at its seasonal maximum thickness. This extensional flow has also been observed by and in remote sensing observations. We interpret this to indicate that the buttressing effect of mélange is limited in winter though it may still exert some resistive force on the glacier termini that suppresses, though does not entirely eliminate calving activity (see also Figs.  and ).

Glacier dynamics and calving activity at each of our three glaciers is similar to, but slightly varying from the characteristic seasonal pattern found by other studies for example, at around 80 % of Greenlandic glaciers. identify 3 typical calving styles at Greenland outlet glaciers and using manual inspection of optical satellite imagery summarised in Table we observe all styles occurring at all three glaciers. We also identify predominant styles at each glacier. Tabular rifting is a characteristic of both Melville and Tracy glacier, serac collapse is a common characteristic of calving at all glaciers but especially at Farquhar and slab capsize appears to be more common at Tracy glacier than the other termini. We see contrasting behaviour in the glacier front position at Tracy, Melville and Farquhar glaciers and these differences probably reflect the balance between long and short-term variability as well as contrasting calving styles. At Tracy (Fig. ) and Melville glaciers (Fig. ), there is a typical small winter re-advance of the terminus, though often also associated with calving activity, with a larger early summer advance followed by late summer and autumn retreat, often by a few large individual calving events. The full length of the terminus is shown in (Fig. ) and the pattern of slow readvance and retreat is summarised in Fig. . Farquhar glacier shows a similar but slightly less clear pattern in Fig.  with an obvious faster retreat on the western side of the calving front than on the east. This may be related to a series of small islands that can be seen appearing on the eastern side of the calving front and suggests the glacier is grounded in shallower water on that side of the front, which will influence the calving style . The crevasse patterns in Fig.  also suggest slower moving ice in this part of the glacier. Interestingly, the analysis of the automated calving front in Fig. b shows a general but rather fuzzier retreat pattern, possibly suggesting the deep learning algorithm found it harder to map the calving front at Farquhar glacier.

With the combination of the automated-extracted calving fronts , buoy data showing abrupt velocity changes, and Sentinel-2 satellite imagery, we manually identified different calving events happening in total on 40 d from March–September in 2022 and 2023. Some events may have occurred over multiple days but are noted as one event happening on the same date due to the available temporal and spatial resolution of satellite imagery. These calving events and changes in velocity are summarised in Fig. . We analysed correlating factors for 21 of the 40 identified events and these are summarised in Table  in the Appendix. The measured sizes of the icebergs produced in these calving events are also indicated. This effort allows us to closely connect changes in mélange and landfast ice with calving events.

There are five large calving events on Tracy in 2022 and 2023. The events detected on Tracy happen from June to September and are larger than the events detected at Melville and Farquhar. Events at Melville Glacier were detected from March to the end of September. Melville is the glacier with most events both in 2022 and 2023 (10 in 2022 and 14 in 2023). Farquhar produces the smallest calving events on average, and we detected four in 2022 and seven in 2023. The majority of the events happen in the summer when there are positive daily air temperatures, but all three glaciers show calving events well outside of the melt season, including during the period with thickest sea ice and mélange. This conclusion is supported by both the satellite imagery and in situ buoy data. Tracy's front for example retreated almost a kilometre on the northern side during the winter of 2022–2023 indicating winter calving (Fig. ). A similar 1 km retreat is detectable at Melville glacier between 2022 and 2023 (Fig. ). 11 out of 15 events detected in 2022 happen after sea ice break-up, while the calving is more equally distributed in 2023, with 14 events after sea ice break-up and 10 before, of which 3 occurred around the identified break-up date. We note that smaller calving events, which in fact likely make up the majority of all calving events, are undetectable by the manual methods chosen here but are integrated into the calving front changes.

Glacier mélange properties are likely influenced by ocean and atmospheric processes. We therefore include observations of ocean processes and state as well as information from atmospheric reanalysis in our study.

The wide availability and near real-time updating of remote sensing data provided us with an excellent opportunity to monitor outlet glaciers and derive insights into important cryosphere processes and to evaluate earth observation data products. Machine learning offers an automated means to analyse this data and the application of deep learning to track calving front position is one obvious benefit. We note a few limitations of both manually determined and deep learning extracted calving fronts. While useful for tracking calving fronts on seasonal to annual timescales, the automated calving fronts did not have a precision high enough to identify calving front dynamics on a daily time scale. This is likely due to the fact that the three glaciers analysed are not part of the training data used for the deep learning model, which reduces the accuracy and consistency of the delineation. In addition, the strong ice mélange and calving activity make these glacier fronts difficult to delineate, even manually. We note however, that the technique shows great promise in reducing labour intensive tasks associated with manually digitising calving fronts and improvement and application of the technique will certainly be beneficial in the future for monitoring and research efforts. Our study also demonstrates the great promise in the deployment of cheap, robust sensors in collaboration with local communities in Greenland. Our analysis though also highlighted the importance of careful calibration of such sensors to avoid over-interpreting what turned out to be artefacts in the GNSS data from these sensors.

The conceptual framework and analysis of is relevant to our interpretation of the results. Field observations confirm the existence of fractures consistent with ploughing through the mélange with a crumple zone of ice rubble and cracks spreading out more widely from the mélange area itself as shown in the photo in Fig.  and in the animations of satellite images in the Video Supplement.

suggests that thinning sea ice triggers the onset of the calving activity at the end of the season. Our results do not completely confirm this framework, as we detect calving activity all through the winter, even when sea ice is at its maximum extent and thickness (see e.g. Fig.  and the Video Supplement too). Our analysis of velocities shown in Figs.  and   does not show a steady increase or decrease in the measured velocities with time. Instead, we observe two distinct regimes, one where the mélange velocity is largely stable and there are a few abrupt movements, indicating sparse but definite calving events, noted in Fig.  in the winter period, and a second summer regime where the abrupt shifts in position become more common and of a higher magnitude, correlating also with more frequent calving events. There is a break point where the mélange shifts between these two regimes in mid-June in both years, suggesting that the onset of enhanced calving is related to the break up of landfast sea ice in the mélange zone, rather than the properties of the mélange itself. This shift coincides with the start of the melt and runoff seasons at the study glaciers, suggesting potentially an atmospheric influence on both sea ice thickness and glacier velocity that has a dynamic effect, as well as the possibility of submarine melt influences. The ocean measurements shown in Fig.  show a thick cold upper layer. A 30 m high (above sea surface) iceberg (likely more than the maximum observed in the field) may reach depths of about 300 m below the surface and therefore could potentially reach into the warm Atlantic water layer observed by CTD measurements. This warmer layer could potentially accelerate melt, but the vast majority of the mélange is sitting in the cold and fresh upper layers of the fjord where there is little potential for mixing until the sea ice cap has broken up sufficiently. We therefore interpret the temperature and salinity in the deep water as being unimportant for the seasonal changes in mélange, certainly in the outer zone and especially for the sea ice which had an observed thickness of around a metre during deployment of the GNSS buoys. Thinning and melt of the winter ice in summer, and well as formation and growth in the late autumn and winter is likely largely driven by atmospheric processes in this local context. We suggest on the basis of our observations however that the presence of the landfast ice is more important than the presence of glacier mélange in the winter time in modulating activity, though largely in preventing the removal of calved icebergs. While the landfast sea ice and mélange may exert some back pressure on the calving front, the extensional flow of the mélange away from the glacier fronts, interrupted by abrupt jumps associated with calving, even during the winter period, suggests that at least in these fjords, the effect of mélange is minimal in controlling ice dynamics to any meaningful extent. This conclusion is supported by analysis of the velocity field in the mélange zone and the principal strain rates across glacier front and mélange zone derived from satellite data and shown in Fig. . The average velocity of the mélange zone is close to the velocity of the outlet glaciers, particularly Tracy Glacier, through much of the winter period, as measured by the PROMICE velocity data. This suggests a tightly coupled system between glaciers and landfast sea ice. Satellite data confirms the extensional flow in the mélange zone but cannot measure the abrupt movements measured by the GNSS buoys and our analysis shows the importance of very high time resolution and the application of in situ data to accurately measure these events. Results from our GNSS buoys show an almost instantaneous response to the same forcing, in spite of their wide (km scale) spacing. The abrupt movements, corresponding to the release of stress related to fracture of the mélange, occur within the same half hour period. While it is difficult to be absolutely certain if the abrupt movements reflect stress relief in the mélange that then leads to calving, or if the iceberg calving leads to the abrupt jump in the position of the mélange, the abrupt jumps do correlate with calving activity at the glacier fronts. Even with a higher time resolution, the OMBs do not resolve sufficiently the propagation of sudden fracture events through the mélange, but both OMBs register the same shift as the lower resolution TRUSTED buoys, which gives us additional confidence in the interpretation of the movements. The magnitude of the total distance moved decreases very steadily with distance from the glacier fronts as Table in the Appendix also makes clear. This observed pattern of movement, suggests that it is calving that is driving the abrupt shifts in mélange position, and not vice versa. The rigidity and coherence of the mélange due to the presence of fast ice is thus likely a more important factor to consider for glacier calving dynamics than the buttressing properties of unfrozen mélange, as also for example suggested by . After the sea ice break up dates, the movement of the mélange buoys is very clearly different and once afloat, they typically left the fjord within a few days and were collected by partners in the local community of Qaanaaq. With the Sentinel-2 satellite imagery (see Video Supplement) after the landfast ice breaks up, we observe both extensional, compressional and uniform flows of mélange, sometimes simultaneously but usually spatially distributed across the fjord. The difference in the flow regimes before and after sea ice break-up have also been proposed by and our observations support this analysis.

The calving front positions of all glaciers, but especially Melville and Tracy show a similar seasonal timing in both 2022 and 2023 (Figs.  and ), with a net advance from mid-September/start October to mid/late July and a net retreat from mid/late July to mid-September/start October. We note that the pattern in frontal position shown in Fig.  is similar to the terminus change time series shown in Fig. 1 of for the three different styles of calving they identify as prevalent in Greenland. Our analysis also therefore supports theirs that although local factors are important for each individual terminus, controls on calving processes can be inferred from general first principles. As all glaciers have similar atmospheric and ocean properties, these appear to be the dominant control on seasonal variability in the front position. The seasonal cycle confirms earlier studies on other glaciers e.g., although Melville, Farquhar, and especially Tracy show a later onset of retreat than other termini in Greenland, possibly related to their relatively northerly position. For example, showed that the onset of summer retreat patterns occurs later as you move north on the west side of Greenland. also concluded that the onset of retreat generally corresponds with break-up of sea ice and onset of higher surface melt, which makes it difficult to distinguish the underlying importance of both factors (see also Fig. ). Other studies have suggested an increase in glacier velocity could lead to enhanced calving and given the importance of melt and glacier sliding on the velocities in this region it is likely that these processes are also moderating the calving activity, particularly at Tracy and Melville Glaciers. We examined the relation between sea ice break-up, onset of melt and other possible driving factors in the overview of calving events displayed in Table and visualised in Figs.  and . These clearly show that the drivers of calving front position and calving activity are likely multi-factorial. As another example, our manual inspection of satellite images clearly showed that glacier front instability through changes to the geometry was an important factor in driving subsequent calving events after a single large event causes the glacier front geometry to be in an unstable configuration, eventually leading to another calving event. An individual calving event may have multiple causal factors, many of which are coincident in time and space. These factors include ice velocity, terminus topography, bathymetry and basal hydrology as well as local climate and have been identified in other studies (see for example, ). We found no diurnal or tidal influence on mélange velocities or calving events during the fast ice season and cannot find a strong diurnal velocity signal in the PROMICE velocities due to the multi day smoothing. However, we do detect a small increase in the glacier velocity in 4 out 10 events detected in 2022. We note also the results of , whose three dimensional full stokes model of Sermeq Kujalleq (Jakobshavn Isbræ) showed the relative unimportance of mélange in controlling calving dynamics and terminus position compared to local geometry.

Although our results focus on a single large fjord with multiple glaciers in Northwest Greenland, the geographic situation is much more typical of most Greenland outlet glaciers than the very large outlets typically studied . We therefore use this analysis to draw conclusions about outlet glaciers in Greenland more widely and, with some reservations to outlet glaciers in Antarctica. Firstly, our results confirm the importance of sea ice in seasonally limiting to some extent, calving from active glaciers. The recent acceleration of Crane glacier on the Antarctic peninsula as well as the seasonal cycle in peninsula glaciers have all alternatively been attributed to the loss of rigid multi-year sea ice. propose, however, that the direct buttressing effect of landfast sea ice on glaciers in the Larsen B embayment (Antarctica) is minor and that the observed acceleration of glaciers after the disintegration of landfast sea ice is due to secondary processes like increased swell activity in open water, which in turn destabilizes the ice shelves. Alternatively, postulate that landfast sea ice has a stabilizing effect on the ice mélange in front of the glaciers (for example, by inhibiting export of icebergs) and that stable ice mélange in turn slows down the glaciers. In Greenland, several studies at other glaciers suggest an important role for mélange in modulating calving activity but these have been overwhelmingly carried out at very large and dynamic outlet glaciers such as Sermeq Kujalleq (Jakobshavn Isbræ) and Helheim Glacier and Kangerlussuaq glacier in Eastern Greenland . These glaciers have very large particle sizes in their mélange zones, which may promote particle jamming more than in the mélange zones of our study glaciers. The effect of inhibiting if not completely preventing calving might be a factor also in the ice mélange zone we observed in front of Tracy, Farquhar and Melville glaciers, though, perhaps because the seasonal landfast sea ice is thinner than mulæti year ice in Antarctica, we see little evidence of mélange slowing ice velocity. This aligns with who also assess little back stress from landfast ice thinner than 2 m. While the mélange at our study glaciers includes many icebergs and some very large ones, the landfast ice was consistently measured to be around a metre thick at all GNSS buoy deployment sites. Similarly, our results do not support the idea that glacier mélange can act as a significant buttress to outlet glaciers in the absence of landfast sea ice during marine ice sheet retreat and/or following the loss of ice shelves as for example proposed by , though models of mélange buttressing typically postulate very thick layers of mélange with a high packing density . We note the importance of the topographic configuration of our study area. The three glaciers are relatively fast moving, but found in short but narrow and straight fjords with a single outlet to the ocean. They are all in relatively deep water (as measured in the field), where the terminus likely reaches the observed core of warmer Atlantic water at around 200–300 m depth. Larger termini in more open embayments may respond differently to the presence of mélange , and for example the Totten Glacier in Antarctica with a large ocean facing ice shelf does display a sensitivity to the presence of landfast sea ice in seasonal velocity patterns. The period of this study is too short to show any trends related to climate or otherwise in the fjord system, however, anecdotal evidence from interviews with local hunters suggests that the fast sea ice season in this fjord system has changed over the last 50 years. Observed sea ice thickness was around 1 m throughout the fjord system in both years. The sea ice is therefore typically much thinner compared to photographs showing almost 2 m thick ice in the 1960s. The fast ice has also formed later and later over the last decade when observations in the fjord monitoring programme run by DMI have been made in this region. In addition, winter storms that have broken up early season thin sea ice have been noted in the last ten years, while being unknown in the period since settlement in Qaanaaq in the 1950s. These anecdotal results suggest that thinning sea ice over a shorter season may affect the retreat of the outlet glaciers as it allows the mélange to be evacuated from the fjords more quickly.

While several models incorporating iceberg mélange exist , few of them have been tested against a range of in situ observations. Of the studies of Greenland mélange that exist, there is a strong summer bias in field studies of glacier mélange for obvious reasons of accessibility as well as a bias towards the very large outlet glaciers. Where other more typical Greenlandic or Alaskan tidewater glaciers have been studied, and particularly in the winter to spring period, their results are more consistent with our findings . This is important for the development of models that can make projections of ice sheet evolution under climate change. Such models are currently reliant on quantities such as mélange thickness, which is hard to define in practice given the heterogeneity of the granular material. Future work will address the definition of mélange thickness in more detail with photogrammetry based on UAV mapping that was carried out concurrently with the buoy deployments, from which we can derive a size distribution of particles that can be applied in a modelling framework. Satellite data has proved immensely useful for testing these models but our results also show that for detailed process understanding, the time resolution of Earth Observation data is insufficient for resolving the rapid movements of the mélange, though demonstrate how useful it can be for generalising across termini. In situ observations of ice mélange are rare and difficult and expensive to obtain and we are therefore committed to open data access to datasets such as this one. We also urge the mélange community to use such in situ observations where available to develop and test numerical models.