We have presented new, detailed insights into the dynamic changes underway at five lake-terminating glaciers in south Vatnajökull for the period 2008/2010-2020. For all glaciers, our data illustrate an overall pattern of increasing velocities over the study period, as well as frontal retreat, proglacial lake growth, and surface thinning, but there is distinct variability in both the patterns and rate of change observed at each individual glacier in each year. In this section, we first compare our velocity, frontal position, and lake area change data to several previous datasets to assess the validity of our findings. We then investigate the dynamic response observed at each glacier and provide a detailed evaluation of the key forcing mechanisms before suggesting what the future response of these glaciers may be. Finally, we discuss the wider implications of our findings in regard to the other lake-terminating outlets of south Vatnajökull.
5.1 Comparison to Previous Data
5.1.1 Glacier Velocity
We compare our TerraSAR-X velocities with the ENVEO Icelandic velocity dataset, which was derived through offset tracking of Sentinel-1 SAR images (Wuite et al., 2022), and with the NASA MEaSUREs ITS_LIVE project, which provides continuous, near-global ice velocities generated using both optical (e.g., Landsat, Sentinel-2) and radar (e.g., Sentinel-1) imagery (Gardner et al., 2024) (Figure S8, S9). To both datasets, we apply the same method that was implemented in this study (detailed in Section 3.1.), but due to data availability we only compare velocities from 2016, 2018 and 2020. Overall, there is good agreement between the three datasets, particularly in terms of the spatial velocity patterns observed at each individual glacier, and how these evolve through time. However, there are some differences, for example over faster moving areas of ice (e.g., at the main ice falls), velocities are consistently higher in our data than in either of the other two datasets. In contrast, over slower moving areas of ice, there are a larger number of erroneous points and outliers in the other two datasets (but particularly ITS_LIVE) than observed in our data.
This is for two reasons: Firstly, where there are sharp velocity gradients present, such as over the ice falls of both Fjallsjökull and Kvíárjökull, or near the terminus of the eastern arm of Breiðamerkurjökull, the higher resolution of the TerraSAR-X imagery (~2 m) means such rapid changes in velocity can be more accurately tracked and reproduced (e.g., Figure S8A, S8C, S9A) (Nagler et al., 2015; Rohner et al., 2019). In contrast, the coarser resolution of the ENVEO (100 m) and ITS_LIVE (~250 m) data means the magnitude of these velocity gradients will have been smoothed over (Altena and Kääb, 2017; Joughin et al., 2018). Similarly, this higher resolution imagery is better able to track and reproduce velocities over slower moving areas of ice, such as accumulation areas, as well as on narrower parts of the main trunks of smaller glaciers (e.g., Kvíárjökull and Svínafellsjökull) than is possible in the other two datasets (Nagler et al., 2015; Millan et al., 2019). Indeed, these regions are less accurately reproduced in the coarser resolution ENVEO and ITS_LIVE data, resulting in a larger number of erroneous data points and outliers (e.g., over the accumulation areas of both Svinafellsjokull and Skaftafellsjökull, Figure S9B, S9C) (Friedl et al., 2021; Wuite et al., 2022). Secondly, our velocity data is generated from imagery acquired predominately in late summer (August-September) with a temporal separation of between 11-33 days (depending on image availability), whereas both the ENVEO and ITS_LIVE data are annually averaged velocity composites. As such, any particularly large velocity gradients will have been averaged out over the longer temporal baseline of both datasets (Nagler et al., 2015; Sugiyama et al., 2015). Yet despite this, our data still shows good agreement with both these datasets. Furthermore, our data also show good agreement with the few previous studies that have investigated the recent velocity change at several south Vatnajökull glaciers, including Brieðamerkurjökull (Baurley et al., 2020), Fjallsjökull (Dell et al., 2019) and Kvíárjökull (Kavan et al., 2024), providing further confidence in the validity of our findings.
5.1.2 Frontal Position and Lake Area Change
Our calculated rates of frontal position change show strong agreement with the values reported by Baurley et al. (2020) for Breiðamerkurjökull and Dell et al. (2019) for Fjallsjökull, as well as by Einarsson (2017) and Guðmundsson et al. (2019) for all south Vatnajökull glaciers (including Kvíárjökull, Svínafellsjökull and Skaftafellsjökull) over a similar period.
We then compare our digitised proglacial lake areas (and calculated rates of proglacial lake growth) to those of Guðmundsson et al. (2019) for the year 2018 (Table S6). Again, we find very good agreement between the two sets of data, particularly when comparing the digitised lake areas from both studies. There is slightly more variation in the calculated rates of proglacial lake growth recorded by both studies (our growth rates are higher in general), but this is because our growth rates are calculated over a shorter period (8-10 years) than those of Guðmundsson et al. (2019) (16-27 years). Yet despite these differences in time period, both sets of growth rates still show good agreement. Furthermore, (and as above) our data also show very good agreement with those studies which have also investigated proglacial lake change at individual south Vatnajökull glaciers, including Brieðamerkurjökull (Baurley et al., 2020), Fjallsjökull (Dell et al., 2019) and Kvíárjökull (Kavan et al., 2024), again providing further confidence in our findings.
5.2 Glacier Response 2008/2010-2020 and Future Outlook
5.2.1 Brieðamerkurjökull
It was suggested by Baurley et al. (2020) that the recent retreat and subsequent increase in velocity observed at the eastern arm of Brieðamerkurjökull is directly related to the rapid growth of Jökulsárlón (particularly in depth) as the glacier retreated down a reverse bed-slope into the 100-300 m deep bedrock trough it formed during the LIA. This led the authors to propose that velocities may have reached their maximum towards the middle of the last decade in response to this significant deepening of the lake. However, the data from this study importantly indicate that such a dynamic response is ongoing, with velocities and terminus retreat both continuing to increase over recent years (Figure 2A, 4A, S5C). For example, mean near-terminus velocities in 2015 were 1.64 m d-1 (Baurley et al. (2020), whereas in 2020 they were ~2.45 m d-1 (this study, ~50% increase). Similarly, the eastern arm retreated by ~400 m between 2014-2018 (Baurley et al., 2020), yet between 2018-2020 it receded by ~250 m (this study), which is over half the retreat observed in 2014-2018.
Based on these observations, it is likely that initial retreat into deeper water resulted in an increase in buoyant forces acting on the terminus, reducing the effective pressure (and consequently the basal drag), leading to an increase in velocity (Sugiyama et al., 2011; Trüssel et al., 2013; Tsutaki et al., 2013; 2019). This, in turn, will have caused the glacier to extend and thin, steepening the ice surface and causing a further increase in velocity by increasing the driving stress (Benn et al., 2007; King et al, 2018; Minowa et al., 2023). This will have resulted in increased fracture propagation at the terminus, leading to an increase in calving activity and subsequently, the rate of retreat (Carrivick and Tweed, 2013; Dell et al., 2019; Liu et al., 2020).
Continued retreat into deeper water will have triggered a further increase in buoyant forces, causing an additional increase in velocity, thinning, calving, and retreat (as observed in our data) and resulting in the implementation of a positive feedback mechanism termed “dynamic thinning” (e.g., Trüssel et al., 2013; Tsutaki et al., 2019; Pronk et al., 2021) that is driving the current unstable dynamic behaviour of the glacier (Baurley et al., 2020). This may explain why the near-terminus region of the eastern arm thinned by ~100-150 m between 2010-2020 (Figure 5C) which is significantly greater than the rate of change observed at the calving front of the other glaciers in this study. This response may be exacerbated by the fact that ice flow from the interior cannot balance the substantial losses occurring at the terminus, further increasing the ice velocity and rate of retreat (Nick et al., 2007; Baurley et al., 2020) and providing clear evidence that the dynamic behaviour of the eastern arm has become decoupled from local climate.
In addition, such a dynamic response is likely to continue in future as the eastern arm continues to retreat through its deep bedrock trough (Storrar et al., 2017), which radio-echo sounding surveys undertaken in the 1990s revealed was ~20 km long (Björnsson 1996). As of 2020, this bedrock trough still extends some ~12 km back from the terminus, with the first ~2 km characterised by a depth of ~200 m and the remaining ~10 km characterised by a depth of ~100-150 m (Figure 6). Therefore, and assuming a similar retreat rate as observed in this study, over the next 10 years, the eastern arm will still be retreating through one of the deepest parts of its trough (Figure 6) and as such the dynamic processes observed here will continue to drive the dynamics and retreat patterns of this arm of Brieðamerkurjökull. This unstable dynamic behaviour will continue until it begins to retreat out of the deep bedrock trough into shallower water, which modelling studies suggest will not be until at least ~2100 (e.g., Flowers et al., 2005; Schmidt et al., 2019), at which point it may then begin to stabilise.
In contrast, the recent changes at Western A are more muted, with significantly slower velocities, less retreat and little change in proglacial lake area observed over our study period (Figure 2B, 4A), which agree with the finding of Baurley et al. (2020). This likely reflects the specific bed topography of Breiðárlón, which is much shallower than Jökulsárlón (<40 m deep in 2018), as well as its ongoing sedimentation, which has caused it to remain stable over recent years (Guðmundsson et al., 2019). However, velocities over this arm of the glacier did increase steadily over our study period, particularly in the near-terminus region (333% increase, Figure 2B), suggesting that over recent years the influence of Breiðárlón on the near-terminus dynamics of Western A may have also increased (e.g., through calving).
Importantly, this influence may become more pronounced in future as this arm of the glacier continues its retreat through its ~9 km long, max. ~30 m deep bedrock trough (Björnsson 1996). Indeed, while the 2020 terminus is located in relatively shallow water (~9 m deep), ~1 km back from this location the bed slope continues to reverse into deeper water (~30 m deep), resulting in a long and narrow bedrock trough that extends ~4 km into the interior of the glacier (Figure 6). Therefore, although in 10 years’ time the glacier terminus will still likely be grounded in relatively shallow water (assuming a similar rate of retreat as observed in this study), there is the possibility that in future the influence of Breiðárlón may develop to such an extent that it can begin to impact the dynamics and retreat patterns of this arm of the glacier, similar to what is currently underway at the eastern arm, but at a much smaller scale and magnitude.
5.2.2 Fjallsjökull
A similar dynamic response to that observed by Baurley et al. (2020) at the eastern arm of Breiðamerkurjökull has also been observed at the neighbouring glacier Fjallsjökull by Dell et al. (2019). Indeed, the authors suggest that the increased velocities and heightened retreat rate observed since the early 2000s directly corresponds to the rapid expansion of Fjallsárlón and subsequent retreat of the glacier back into its ~200 m deep bedrock trough, resulting in an increase in buoyant forces and the implementation of the same positive feedback mechanism described previously. Furthermore, the data from this study also suggest that such a dynamic response is ongoing, just like at the eastern arm of Breiðamerkurjökull.
Indeed, we observe an almost 500% increase in mean near-terminus velocities since 2008, which peak at ~0.82 ±0.09 m d-1 in 2020 (Figure 2C). This is a similar magnitude to, but noticeably larger than, the ~0.5 m d-1 observed by Dell et al. (2019) over the same region in 2017/2018. Our data also seem to indicate that from 2018 onwards, velocities in the near-terminus region (within ~2 km of the front) begin to increase up to the calving front, rather than decrease (Figure 2C). This is similar to the pattern observed at the eastern arm of Breiðamerkurjökull over the study period and may reflect the ongoing dynamic evolution of the glacier as it continues its retreat through its deep bedrock trough. We also observe a further increase in the both the size of Fjallsárlón and the cumulative retreat of the glacier since 2016 (Figure 4B, the last year that Dell et al. (2019) obtained this data), providing further support to the assertion that the dynamic response originally observed by the authors is ongoing.
In addition, the data from this study also seem to indicate that there is a non-uniform pattern of negative surface elevation changes near the terminus of Fjallsjökull (Figure 5B). Indeed, the most negative elevation changes are found over both the central, and in particular, the southern part of the terminus (~5.5 ±0.24 m a-1), encompassing a ~2 km x ~2 km region, whereas significantly smaller elevation changes (in both magnitude and extent) are observed over the northern part of the terminus (~2.8 ±0.24 m a-1). This contrasts with the more homogenous pattern of negative surface elevation changes observed near the terminus of the eastern arm of Breiðamerkurjökull. Such a pattern is likely the result of a deeply incised bedrock channel that sits within the main ~3 km x 4 km bedrock trough found under the glacier (Magnússon et al., 2012). Importantly, this deep bedrock channel, which is ~2 km by ~2 km and ~120 m deep at its maximum, directly underlies the southern part of the present-day terminus (Dell et al., 2019; Baurley, 2022). As such, the terminus is currently retreating through the deepest part of the channel in this region, meaning velocities, and consequently the rate of surface thinning, will be elevated (Figure 7) (e.g., Pronk et al., 2021; Minowa et al., 2023). Indeed, the location of this channel coincides with the region of the terminus where velocities are at their highest, suggesting that dynamic thinning may have recently been initiated in this region of Fjallsjökull (Dell et al., 2019; Baurley, 2020), and that the dynamic behaviour of this region of the glacier may have also become decoupled from the local climate as a result.
Furthermore, such a dynamic response will continue in future as this part of the glacier continues its retreat through the southern bedrock channel. As of 2020, the channel still extends ~1 km back from the terminus, with much of this characterised by a depth of ~80-100 m (Figure 7). The main bedrock trough under Fjallsjökull then extends for another ~1.5 km after this, but at a much shallower depth (<50 m) (Figure 7). Therefore, and assuming a similar retreat rate as observed in this study, over the next 10 years this region of Fjallsjökull will still be retreating through one of the deepest parts of the southern bedrock channel (Figure 7), meaning the dynamic processes observed here will continue to drive the dynamics and retreat patterns of this region of glacier until it retreats into shallower water.
A similar response may also be observed at the main flowline (centreline) of the glacier, which is currently retreating through the main bedrock trough. As this trough still extends for another ~2.5 km back from the terminus, with the majority of this characterised by a depth of ~50-80 m, it is likely that in 10 years’ time this region of the glacier will also still be retreating through one of the deepest parts of its main bedrock trough (Figure 7), with similar consequences for the dynamics of this region of the glacier. Finally, while another, deeper, bedrock channel is found under the northern part of the terminus (~200 m deep), the glacier front in this region is presently grounded in very shallow water (Figure 7). This may explain why relatively slow velocities (Figure S3) and less-negative elevation changes (Figure 5B) are currently observed in this region. Although this is unlikely to change significantly in the near future (i.e., 10 years’ time, Figure 7), once the terminus does begin to retreat into deeper water then it is likely that the dynamic behaviour will evolve in a similar way to what is presently observed over the southern bedrock channel, with elevated velocities, frontal retreat, and thinning rates occurring as a result.
5.2.3 Kvíárjökull
In contrast to both Fjallsjökull and the eastern arm of Breiðamerkurjökull, the dynamics of Kvíárjökull are unlikely to have been driven by proglacial lake expansion and the retreat of the glacier into deeper water. Indeed, the area of Kvíárjökulslón only increased by ~0.25 km2 over the study period, whilst the terminus receded by only ~76 m (Figure 4B), which represents both the smallest increase in proglacial lake area and change in front position observed in this study. Furthermore, the terminus is presently situated at the outer edge of its proglacial lake, which when combined with the small variations in lake growth and frontal position, suggest that other factors are driving the dynamic behaviour of this glacier.
Previous research has demonstrated that the dynamics of Kvíárjökull are primarily controlled by a narrow flow corridor located along its central axis, which is surrounded by slower moving or stationary lateral and latero-terminal regions (Bennet and Evans, 2012; Phillips et al., 2017). This active flow corridor does not move as one complete unit, rather it comprises several individual lobes that move independently (or ‘pulse’) in surge-like movements down-glacier, with flow directed towards the northeastern part of the margin (Phillips et al., 2017; Kavan et al., 2024). Therefore, the glacier is characterised by periods (or ‘pulses’) of increased ice flow separated by periods of ‘quiescence’, with the pulse-like activity occurring over decadal timescales (Phillips et al., 2017).
Based on the data from this study, the most recent period of increased ice flow likely occurred at some point between 2012 and 2016. This is evidenced by mean-terminus velocities peaking at ~0.42 ±0.09 m d-1 in 2016 (increasing from ~0.19 ±0.09 m d-1 in 2012, Figure 3A, S3), and by a clear re-advance of the northeastern part of the margin over the same period which narrowed the connection between the northern and southern parts of Kvíárjökulslón (Figure S6C). Mean-terminus velocities then begin to decrease, marking the end of the speed-up event, however, the northeastern part of the margin remains relatively stable up until 2020, despite the presence of Kvíárjökulslón at its northern and southern boundary. This is likely a result of the continual movement of mass down-glacier by the active flow corridor (e.g., Bennet and Evans, 2012; Phillips et al., 2017), as well as the insulating effect of the thick layer of supraglacial debris cover in this region (e.g., Reznichenko et al., 2010; Nicholson and Benn, 2013), which may also explain why the most negative elevation changes are found below the main ice fall, not at the terminus as has been observed at other glaciers in this study (Figure 5B).
In comparison, the southeastern part of the margin is far more sensitive to the presence of Kvíárjökulslón, which is primarily due to the relatively flat and thin nature of the ice surface (Phillips et al., 2017; Kavan et al., 2024). This means that the lake can often inundate the glacier front, causing it to destabilise and possibly disintegrate due to the processes of frontal ablation and related thermo-mechanical processes (e.g., Carrivick et al., 2020). This may explain why the area of Kvíárjökulslón grew so rapidly between 2010-2012 (from 0.41 to 0.60 km2, Figure 4B, S7C) and why the southeastern part of the margin also retreated significantly over the same period (Figure S6C). In contrast, the readvance of the northeastern part of the margin between 2012-2016 may have caused the southeastern part to temporarily stabilise, and, as was observed in 2018, even undergo a slight readvance (Figure S6C), which was primarily driven by an overriding flow unit (or ‘lobe’) immediately up-glacier (Phillips et al., 2017). Such a response may explain why the area of Kvíárjökulslón, as well as the position of the terminus in this region, remained relatively stable during this time (Figure 4B, S6C, S7C). However, between 2018-2020 the terminus in this region retreated relatively rapidly (~100-150 m), resulting in a relatively large increase in the area of Kvíárjökulslón (Figure 4B), and indicating that this region is again being impacted by the processes of frontal ablation and the related thermo-mechanic properties of the lake (e.g., Kavan et al., 2024).
Like many of the other southern outlets of Vatnajökull, Kvíárjökull is also underlain by a relatively large bedrock trough, which extends ~4 km back from the present-day terminus and has a maximum depth of ~100 m (Magnússon et al., 2012). Yet despite the 2020 terminus being located above the deep reverse-sloping part of the trough (Figure 8), it is unlikely that the glacier will undergo a similar dynamic response to what is currently underway at Fjallsjökull and the eastern arm of Breiðamerkurjökull. This is because the southeastern part of the terminus (where the deepest parts of the trough are located) is likely to be floating, meaning it is not physically grounded in, or retreating into, deeper water. Indeed, field observations indicate that the terminus is relatively flat and thin in this region (and therefore stagnant), which allows lake water to propagate under and up into the glacier, inundating the ice surface and resulting in the calving of large tabular blocks (e.g., Kavan et al., 2024).
Although it is unclear when the switch from a grounded to a floating ice front occurred, it was likely driven by a change in ice thickness relative to water depth (i.e., thinning), and a subsequent increase in buoyant forces, resulting in uplift of the glacier terminus (e.g., Boyce et al., 2007; Tsutaki et al., 2013). As such, over the next 10 years the southeastern part of the margin will continue to retreat rapidly in response to the inundation and destabalisation by Kvíárjökulslón (Figure 8), as has recently been observed by Kavan et al. (2024). This may have significant implications for the overall stability of the floating portion of the glacier margin, which may undergo complete terminus break-up and disintegration in future. In contrast, the northeastern part of the margin will likely remain stable for the foreseeable future due to the continual inflow of mass into the region, as well as its thick layer of supraglacial debris cover (e.g., Nicholson and Benn, 2013; Phillips et al., 2017).
5.2.4 Svínafellsjökull
Of the five glaciers investigated in this study, it is the dynamics of Svínafellsjökull that have undergone the least change over our study period, and in fact have remained relatively stable. Indeed, our data indicate a gradual increase in mean near-terminus velocities, a relatively modest rate of surface thinning, and only ~97 m of terminal recession in this time. Although proglacial lake expansion (and related dynamic effects) can explain some of the observed dynamic variations, it is likely that several other factors may also be exerting a key control.
With a mean slope of ~9.0°, Svínafellsjökull is the one of the steepest outlets of Öræfajökull (Hannesdóttir et al., 2015), and the steepest outlet investigated in this study. Yet despite this, the lower ~6 km of the glacier is characterised by a relatively gentle surface slope (~3°) (Figure 8), which would result in a down-glacier reduction in driving stress. This, combined with the narrow valley in which the glacier is situated (which increases the lateral resistive stresses), may explain why we observe relatively low velocities in the lower part of the glacier (Benn et al., 2007; Adhikari and Marshall, 2012; Dehecq et al., 2019). However, this reduction in driving stress does not explain why we observe a gradual increase in near-terminus velocities over our study period, from ~0.12 to ~0.19 ±0.09 m d-1 (58% increase, Figure 3B). Instead, this gradual acceleration is likely a result of the growth and expansion of the two proglacial lakes on the northern and southern side of the terminus (Figure S7D). This would have increased the area of the terminus that is in contact with the lakes, and consequently, the extent to which these lakes could influence near-terminus dynamics, leading to a further reduction in the effective pressure, increased basal sliding and resulting in the observed gradual increase in velocity (Sugiyama et al., 2011; Carrivick and Tweed, 2013).
In addition, the valley of Svínafellsjökull widens ~2 km up-glacier from the terminus, allowing the glacier to spread and resulting in an extensional flow regime (e.g., Sato et al., 2022) (Figure S6D). Importantly, this extensional regime may influence near-terminus dynamics through increased surface thinning, which when combined with surface melt may have caused the southern part of the terminus to be subjected to increased buoyant forces, which continued to evolve until partial floatation occurred (e.g., Boyce et al., 2007; Trüssel et al., 2013). Surface thinning will also result in a reduction in the effective pressure, meaning it is possible that this thinning may have also contributed to the observed gradual increase in near-terminus velocities (e.g., Sugiyama et al., 2011; Tsutaki et al., 2019). Although terminus floatation can result in rapid ice marginal retreat and terminus disintegration via calving (e.g., Warren et al., 2001; Motyka et al., 2002; Boyce et al., 2007), the reason this has yet to occur at the southern margin of Svínafellsjökull is because parts of the terminus are grounded on bedrock at the outer extent of the lake, despite the terminus itself being relatively flat and thin (Figure 8). This, combined with the continual inflow of mass to the region (due to the gradual increase in velocities), means the high stresses present at the margin can be accommodated, allowing the terminus to remain relatively stable (e.g., Boyce et al., 2007). This continued inflow of mass to the terminus may also explain the relatively low rates of thinning observed in this region (Figure 5A).
Furthermore, parts of the both the central and northern terminus are also grounded on bedrock at the outer extent of the lake (Figure S6D) and as such these regions have also remained stable, which may help to explain why the margin of Svínafellsjökull only retreated by ~97 m over our study period. Whilst the rate of retreat was consistent over the study period (Figure 4B), it was not homogenous across the entire terminus, with much of the recession focused over the lateral margins of the northern and southern parts of the terminus where it terminates in a lake but is not grounded on bedrock (Figure S6D). In these regions the influence of the thermo-mechanical properties of the lake are greatest, and as such it is likely that calving is actively occurring, either through thermal melt and notch formation, or through buoyant forces acting on the terminus and the propagation of basal crevasses (e.g., Warren et al., 2001; Röhl, 2006; Baurley, 2022; Minowa et al., 2023). Therefore, whilst the overall pattern is one of terminus retreat, the relatively stable nature of large parts of the terminus means the retreat rate is low.
An additional factor which may have influenced the observed dynamic variations is the occurrence of a large landslide in 2013, which caused a ~1.7 km2 area of the ice surface to be covered in a thick layer of debris (Figure 2 in Ben-Yehoshua et al., 2022). While the ice underneath the debris has been efficiently insulated and protected from surface melt, the ice immediately surrounding it has seen enhanced melt due to the fine layer of dust that settled on the surface post-landslide (e.g., Reznichenko et al., 2010; Nicholson and Benn, 2013; Fyffe et al., 2020). This resulted in a 35 m difference in surface elevation between the two regions by 2020 (Ben-Yehoshua et al., 2022). Although the impact of the landslide is not present in our surface elevation data (which cover the period 2010-2019), it’s probable that its occurrence will have contributed to several of the other dynamic variations observed in this study. For example, the difference in elevation between the debris covered- and clean-ice areas will have increased the surface slope, causing an increase in the driving stress and thus velocities, which in turn will have provided additional resistive stresses to the partially floating southern terminus, helping it to remain stable. In addition, this landslide material is continually being advected down-glacier (it was advected ~1 km between 2013-2020), and as such these processes are likely to continue in future as it is transported towards the southern margin, where it will likely lead to further stabilisation, incremental stagnation, and the potential formation of a dead-ice environment (Ben-Yehoshua et al., 2022).
Based on the data presented in this study, it is unlikely that the dynamics of Svínafellsjökull will significantly change or evolve over the coming decade, despite the presence of a 300 m deep bedrock trough under the glacier, which extends ~6 km back from the present-day terminus (Magnússon et al., 2012; Guðmundsson et al., 2019). This is because rapid retreat and terminus disintegration is unlikely whilst much of the margin remains grounded on stable bedrock at the lake edge, a factor that will be further influenced by the continued down-glacier advection of the landslide material, which as mentioned previously may further stabilise the southern margin through the formation of a dead-ice environment. As such, by 2030 this part of the margin is still likely to be grounded at the lake edge (Figure 8).
In contrast, it is more likely that the northern part of the terminus will undergo rapid retreat in future, as although the terminus is grounded (i.e., not floating), it is beginning to detach from the surrounding bedrock, while at the same time its proglacial lake is growing relatively rapidly (Figure S6D, S7D). This indicates an increasing influence of the lake and calving on terminus stability, and as such in the near future this part of the glacier may begin retreating down its reverse bed slope into deeper water, leading to increased velocities, thinning, and thus further calving and retreat (e.g., Sakakibara et al., 2013; King et al., 2018; Baurley et al., 2020). Indeed, it is likely that calving will play an important role at both proglacial lakes in future as they continue to grow, resulting in continued terminus retreat and the potential detachment of large parts of the terminus from its surrounding bedrock, with implications for the stability of the lower part of the glacier.
5.2.5 Skaftafellsjökull
In contrast to Svínafellsjökull, the dynamic behaviour of Skaftafellsjökull has evolved considerably over the study period, with the second largest change in front position (~380 m), second highest rate of terminus thinning (~7.6 ±0.17 m a-1), and third highest growth in proglacial lake area (~1.1 km2) observed at this glacier. Surface velocities also increase significantly over the study period, particularly near the terminus (Figure S5A), suggesting that the dynamics of the glacier are likely being influenced by the growth of the proglacial lake and retreat of the terminus into deeper water.
In 2010, the terminus of Skaftafellsjökull was grounded in shallow water (~12 m deep) on a relatively flat region of bedrock near the edge of the lake (Figure 8). As such the influence of the lake on the dynamics of the glacier were limited, which may explain why near-terminus velocities were low in 2010 and remained as such in 2012 (~0.05 ±0.09 m d-1). Instead, it is likely that during this time the dynamics of the glacier were primarily controlled by air temperatures (e.g., Hannesdóttir et al., 2015). Yet while this would have resulted in the observed terminus recession of ~95 m between 2010-2012 (Figure 4B), as well as some of the observed surface thinning (via surface melt), overall, the glacier was relatively stable during this time (Figure S6E). However, at some point between 2012 and 2016, the continued retreat of the glacier caused the grounded terminus to recede past the flat region of bedrock and begin retreating down a much steeper, reverse bed slope into deeper water (Figure 8).
This would have increased the buoyant forces acting on the terminus, reducing the effective pressure and causing velocities to increase (Sugiyama et al., 2011; Trüssel et al., 2013). Increased velocities will, in turn, have caused the ice surface to extend and thin, leading to increased calving, terminus retreat, and a further increase in velocities (Benn et al., 2007; King et al., 2018; Minowa et al., 2023). Such a response can be observed in our data, with mean near-terminus velocities increasing by ~160% from ~0.05 to 0.13 ±0.09 m d-1 (Figure 3C), and a steady increase in both the rate of terminus retreat and proglacial lake growth over the same period (Figure 4B). Importantly, our data indicate that this dynamic response has continued, at least until the end of the study period, with a further increase in velocities, terminus retreat and lake growth observed between 2016-2020 (Figure 3C, 4B, S6E, S7E), strongly suggesting that the same positive feedback mechanism already underway at both Fjallsjökull and the eastern arm of Breiðamerkurjökull may have recently been initiated at Skaftafellsjökull.
This increase in velocity may also explain why we observe such high rates of surface thinning at the terminus of Skaftafellsjökull over the study period (Figure 5A). Indeed, while some of the observed thinning can be attributed to surface melt (e.g., Hannesdóttir et al., 2015; Aðalgeirsdóttir et al., 2020) the magnitude of this thinning (~7.6 ±0.17 m a-1) can only have occurred in response to ice dynamics, i.e., through dynamic thinning (e.g., Tsutaki et al., 2019; Liu et al., 2020). Furthermore, thinning itself also reduces the effective pressure, and therefore it is likely that the high thinning rates will have also contributed to the observed rapid increase in near-terminus velocities (e.g., Sugiyama et al., 2011; Tsutaki et al., 2019), providing further evidence that the growth of the proglacial lake and retreat of the glacier into deeper water are now driving the dynamic behaviour of the glacier.
In addition, this dynamic response may have been enhanced between 2018-2020 due to an increase in the gradient of the reverse bed, which led to a more rapid increase in water depth between the two years (Figure 8). Indeed, in 2018 the water depth at the terminus was ~24 m, whereas in 2020 it was ~30 m. Consequently, while terminus water depth increased by ~150% between 2010-2020 (12 m to 30 m), one third of this increase occurred between 2018-2020 (i.e., within two years). This would have triggered a rapid increase in buoyant forces, further reducing the effective pressure and resulting in an additional increase in velocities, calving, and retreat (Benn et al., 2007; Minowa et al., 2023). Importantly, the impact of this rapid increase in water depth is clearly observed in our data. For example, near-terminus velocities increased from ~0.17 to ~0.30 ±0.09 m d-1 over the two years (~76% increase, Figure 3C), whilst the terminus itself retreated by ~120 m over the same period, which means that one third of the total retreat that occurred between 2010-2020 did so between 2018-2020 (Figure 4B, S6E). Such a dynamic response illustrates how small, but rapid, changes in water depth can have a significant impact on the dynamics of lake-terminating glaciers.
However, in spite of these recent variations, there’s the possibility that the dynamics and retreat of Skaftafellsjökull may begin to stabilise towards the end of the decade, despite the presence of the ~200 m deep, ~6 km long bedrock trough under the glacier (Figure 8) (Magnússon et al., 2012, Guðmundsson et al., 2019). This is because immediately up-glacier of the 2020 terminus the gradient of the bedrock slope reduces significantly due to the presence of a ~400 m long region of relatively flat bedrock (Figure 8). Consequently, rapid increases in water depth, as observed between 2018-2020, will not be able to occur. This means that the likelihood of the glacier undergoing rapid changes in ice dynamics may also be reduced. Indeed, based on the retreat rate calculated in this study, by 2030 the terminus of Skaftafellsjökull will most likely be grounded on this region of flat bedrock (Figure 8) and, therefore, its dynamics may have begun to stabilise.
On the other hand, there is also the possibility that the observed recent rapid retreat may have caused the dynamics of the glacier to become partly decoupled from the local climate, meaning such a dynamic response will be maintained regardless of any future change in the gradient of the bedrock slope. This is supported by our observations from Fjallsjökull and the eastern arm of Breiðamerkurjökull, both of which see a continuous increase in their velocities and terminus retreat over the study period, despite both glaciers having retreated over the deepest parts of their respective bedrock troughs (i.e., the water depth has decreased) (Figure 6, 7). As such, there is the strong possibility that the dynamics of Skaftafellsjökull will continue to evolve in future as it continues its retreat through its deep bedrock trough, resulting in a heightened dynamic response that is decoupled from climate, similar to that already observed at Fjallsjökull and the eastern arm of Breiðamerkurjökull over recent years.
5.3 Implications for the other outlets of South Vatnajökull
The findings of this study highlight the importance of proglacial lake growth in driving the dynamics and retreat patterns of glaciers in Iceland, with such a pattern likely to continue in future as they further grow and develop. In addition, there is the strong possibility that the other southern outlets of Vatnajökull will also undergo a similar dynamic response in future, particularly those to the east of Breiðamerkurjökull. Indeed, many of these outlets also have reverse-sloping beds that sit some 200-300 m below the current elevation of their termini, including Skálafellsjökull (~3 km long, ~200 m deep), Heinabergsjökull (~11 km long, ~200-300 m deep), Fláajökull (~5 km long, >200 m deep) and Hoffellsjökull (~7 km long, ~250 m deep), and as such they have also seen the rapid growth and expansion of proglacial lakes at their margins in recent years (Hannesdóttir and others, 2015; Guðmundsson et al., 2019). This has also resulted in accelerated terminus retreat and mass loss via calving, although as was observed in this study, different glaciers often display contrasting dynamic behaviour, highlighting the need for further work in this region.
As a result, in order to better understand the influence of proglacial lake growth on the dynamics of the southern lake-terminating glaciers of Vatnajökull, and to determine whether their contribution to the overall mass loss of the ice cap may increase in future, additional, multi-method and multi-temporal analyses are required, such as those by Dell et al. (2019), Baurley et al. (2020), and the work presented here. In addition, detailed in-situ field measurements of lake depth, above-waterline ice thickness and ice surface slope in the vicinity of the calving front, as well as observations of specific calving style, are needed to better understand the factors controlling the dynamics of induvial glaciers. Such data could then be used to help further constrain calving processes in glacier and ice sheet models, allowing the future patterns of retreat and mass loss, and subsequently the SLR contribution, of these rapidly changing lake-terminating glaciers to be more accurately quantified.