High spatial and temporal variability in Antarctic ice discharge linked to ice shelf buttressing and bed geometry | Scientific Reports – Nature.com

Ice discharge change in Antarctica

Several studies have recently reported ice discharge change across Antarctica over a variety of different timescales, using slightly different methodologies and velocity products. In this section, we briefly compare our results to previously published estimates of ice discharge change and explore the potential reasons behind any differences. In the WAIS we observe a 458 Gt year1 increase in ice discharge between 19992006 and 2018. Over a comparable time period, a recent study3 observed a larger (66 Gt year1) increase in ice discharge. Much of this discrepancy can be explained by the Pope, Smith and Kohler catchments (referred to as the Dotson and Crosson Ice Shelves in Ref.3) where Ref.3 collectively observed a 51% (21 Gt year1) increase in discharge, while we report a much smaller 15% (6 Gt year1) increase. The reason for this difference could be a consequence of the complex velocity patterns over the Crosson and Dotson ice shelves where sections of floating and grounded ice have been both accelerating and decelerating (Fig.3d). Ice discharge change in Ref.3 was calculated using a scaling factor over the fastest flowing sections of the ice to estimate changes in discharge, meaning the full spatial variability in ice flow across the catchment may not be fully captured.

A separate study5 reported a 308 Gt year1 increase in ice discharge in the WAIS between the MEaSUREs (~2008) velocity mosaic and a velocity mosaic based on image pairs from 2013 to 2016. We report a smaller increase of 14 Gt8 Gt year1 between 20072008 and 20132016. The largest differences in discharge change between our results and Ref.5 come from basin JJ that feeds the Ronne Ice Shelf and basin FG in Marie Byrd Land. In basin JJ we observe a decrease in discharge of 3 Gt year1, while Ref.5 observed a 2 Gt year1 increase. This small difference could be related to the underlying data in the MEaSUREs velocity mosaic used in Ref.5. While most of the velocity measurements are from around 2008, velocities from the late 1990s and 2009 are used for glaciers that feed the Ronne Ice Shelf13. At basin F-G we observe a 6 Gt year1 increase, while Ref.5 observed a 12 Gt year1 increase in discharge. The reasons for this discrepancy remain unclear but we note that, over the same timespan (2008 to 20132016), Ref.3 reported only a 5 Gt year1 increase in ice discharge at basin F-G, which is more consistent with our estimate. In the EAIS, our observations of limited change throughout the study period is consistent with the majority of previous studies3,5. The only exception to this is Ref.17 who reported a much larger increase in ice discharge in the EAIS (upwards of 50 Gt year1) for the period 2008 to 2015. The underlying reasons as to why the ice discharge change reported by Ref.17 differs markedly to our results and other studies3,5 remains unclear.

The majority of the regional trends we observe in ice discharge between 19992006 and 2018 are consistent with reported mass balance trends i.e. mass loss in WAIS and limited change in the EAIS1,3,5,6. This further reinforces the notion that changes in ice discharge have been driving the observed changes in the mass balance of WAIS, as oppose to surface mass balance, at least over decadal time periods1,3,5,6. In the EAIS our observation of limited change in ice discharge is consistent with reconciled estimates of a mass balance that is close to zero, albeit with large uncertainties1. Indeed, there is a considerable range in mass balance estimates of the EAIS, typically depending on the methods used, with some studies estimating a positive mass balance5,6 and others a more negative mass balance3. The only regional exception where trends in mass balance do not appear to be consistent with trends in ice discharge is in Wilkes Land where we observe a decrease in ice discharge between 19992006 and 2018. However, multiple studies have confirmed that this region has continued to lose mass over roughly the same time period1,3,5,6,18, suggesting that the decrease in ice discharge in Wilkes Land has not been enough to reverse the trend of mass loss. The decrease is in ice discharge that we detect is focused at the Totten catchment and much of the decrease in ice discharge has occurred in recent years (20152018; Fig.4s). We note that a reduction in ice discharge of a similar magnitude (around 11%) has been reported at Totten between 1996 and 200019, likely caused by intermittent contact between its ice shelf and bed obstacles20. This demonstrates that our observed reduction in discharge is not unprecedented over longer multi-decadal timescales and further highlights the considerable interannual variability in some of the outlet glaciers in Wilkes Land.

Observations and numerical modelling experiments have shown that changes in ice discharge from marine-terminating outlet glaciers are predominantly controlled by changes in ice shelf buttressing, which can manifest in response to changes in ice shelf thickness21,22, ice shelf extent23 and the structural integrity or damage of the ice shelf 24. Following a perturbation in buttressing, ice discharge responds instantaneously21,22, and the thickness changes induced, combined with the glaciers unique geometrical setting, then determine the overall magnitude of the ensuing change in ice discharge25,26,27,28. These associated feedbacks are transient and the response time for a glacier to reach a theoretical steady-state following a perturbation in ice shelf buttressing is dependent on its unique geometry26. In the following sections we highlight some examples as to how these constantly evolving processes are controlling the inter-annual and spatial variability in ice discharge that we have observed.

We use a high-resolution ice-shelf thickness time-series16 to extract anomalies in ice shelf thickness and then compare these to ice discharge anomalies (see Methods). We focus on nine examples: the Thwaites and Pope Glaciers in the Amundsen Sea; Land Glacier in Marie Byrd Land; Moscow University, Totten and Denman Glaciers in Wilkes Land; and Rennick and Cook Glaciers in George V Land. We justify our selection of these example catchments because their ice shelves are large enough to be captured in the ice shelf thickness dataset16 and they represent a selection of the both warm-water ice shelves (Amundsen Sea, Marie Byrd Land and Wilkes Land) and cold-water ice shelves (George V Land). In general, we observe a clear pattern whereby periods of anomalous ice shelf thinning coincide with increases in ice discharge, while periods of anomalous ice shelf thickening coincide with decreases in ice discharge (Fig.5). The only exceptions to this are at the Rennick (Fig.5h) and Cook East Glaciers (Fig.5i), where there is little relationship between ice shelf thickness anomalies and ice discharge. At Rennick, this might be explained by the comparatively small magnitude in ice shelf thickness variations (<1m). In contrast, at the Cook East Ice Shelf there are no clear inter-annual anomalies in ice shelf thickness, meaning any corresponding anomalies in ice discharge are not expected.

Linearly de-trended anomalies in ice-shelf thickness (m) versus linearly de-trended anomalies in ice discharge (%) for the (a) Thwaites, (b) Pope, (c) Land, (d) Moscow University, (e) Totten, (f) Denman, (g) David, (h) Rennick and (i) Cook East Glaciers. Note different scales on the y-axes and that the Cook catchment has been seperated between the East tributary and West tributray. This is because the West tributary does not have an ice shelf. The original ice-shelf thickness data has been taken from dataset produced in16.

We interpret the periods of anomalous ice shelf thickening and negative ice discharge anomalies (e.g. in Fig.5ag) as a direct response to relatively cooler oceanic conditions and lower basal melt rates, whereas periods of anomalous thinning and increased ice discharge are linked to relatively warmer oceanic conditions and higher basal melt rates. At warm ice shelves, wind-driven variations in the transport of warm modified Circumpolar Deep Water (mCDW) onto the continental shelf have the potential to cause variations in basal melt rates underneath warm-water ice shelves29,30,31,32. These wind-driven variations in the transport of mCDW onto the continental shelf are linked to large-scale atmospheric patterns33,34,35 and, as a result, this mechanism can operate over a large spatial scale. This explains why multiple nearby neighbouring catchments can undergo similar patterns of interannual variability in ice discharge in the Amundsen Sea (Fig.4dg). The similar coherent response of outlet glaciers in Wilkes Land (Fig.4rt), where the continental shelf can also be flooded with mCDW is also indicative of a common large-scale atmospheric driver.

For cold-water ice shelves, inter-annual variations in basal melt rates are driven by high salinity shelf water (HSSW) and seasonal warming of the upper layers of the ocean near the ice front16. These variations in cold-water ice shelves can be linked to highly localised sea-ice and polynya processes36,37,38. These highly localised processes have the potential to drive a more localised response in ice discharge. For example, the variability in ice discharge at David Glacier is not observed in any of its neighbouring catchments. These processes have the potential to be driven by both external forcing e.g. katabatic winds37,38, but also internal ice sheet processes e.g. iceberg calving. For example, the calving of the Mertz Ice Tongue resulted in a large change in polynya persistency and resulting oceanic conditions39,40.

Changes in ice shelf extent can directly influence ice discharge rates if dynamically important sections of floating ice are lost or enlarged23. An example of this process is seen at the Rayner catchment in Oates Land, where modelling experiments have shown that its entire floating ice shelf is dynamically important23 and where the observed inter-annual variation in ice discharge can be explained by its calving cycle (Fig.6a). Between 2005 and 2014, the glacier advanced continuously while ice discharge decreased. However, between 2014 and 2016 ice discharge increased as the Rayner ice-front began to rift and break-up, before a final calving event in 2016 resulted in its ice front retreating~10km (Fig.6ac). After this event the ice front re-advanced and ice discharge started to decrease (Fig.6ac). It is important to note, however, that not all calving events result in an increase in ice discharge. If the calved portion is passive and offers limited buttressing, only a limited velocity response from ice inland would be expected23,41. For example, we detect no change in ice discharge rates following the calving of the Mertz ice tongue in 201042.

(a) Time-series of ice discharge change (%) and ice-front position change between 2000 and 2018 for Rayner Glacier. (b) and (c) are Landsat-8 images showing the progression of a calving event at Rayner between 2014 and 2017. (d) Time-Series of ice discharge and ice-front position change for Commandant Glacier, (e) and (f) are Landsat-7 (2009) and Landsat-8 (2018) images showing the rapid growth of the Commandant ice tongue in response to persistent landfast sea-ice. The red line in (a) and (d) are cubic spline trends of ice discharge.

Landfast sea-ice conditions can also play an important role in determining changes in ice shelf extent for some glaciers with heavily damaged ice shelves or ice tongues43,44,45,46. One of the most rapid changes in ice discharge was at the Commandant Glacier, a small glacier within the Adelie Coast catchment, where we observe a 152% decrease between 2009/2010 and 2018 (Fig.6df). This coincided with the formation of an ice tongue, of which growth continued unabated until the end of our observational period in 2018 (Fig.6d). This is indicative of the growth of the ice tongue providing additional buttressing and driving a dynamic slow-down. The growth of the ice tongue appears to be anomalous. In all available satellite imagery since 1973, no comparable ice tongue is present, with the exception of a much smaller tongue in 1973 (Fig. S4). The anomalous ice tongue growth is, however, likely to be linked to an abrupt change in sea ice conditions. Prior to 2009 sea-ice cleared away each austral summer and sometimes resulted in small calving events, but post-2009 a band of multi-year landfast sea ice has formed which has remained consistently fastened to the ice-front and inhibited calving (Figs. S4, 6f). It is unclear what has triggered this abrupt change in sea-ice conditions, but we hypothesize a positive feedback whereby an initial cooler period enabled sea-ice to survive the summer and in doing so trapped detached icebergs in the embayment. These trapped icebergs may have then helped further strengthen the sea-ice by negating the impact of damaging oceanic swell47,48,49,50. Therefore, changes in ice shelf extent linked to changes in landfast sea-ice can provide a direct and rapid link between external forcing, ice shelf buttressing and ice discharge.

The link between landfast sea-ice and ice shelf extent may also be important in determining ice discharge variability for glaciers whose ice shelves are both damaged and float in confined embayments that are favourable for persistent landfast sea-ice formation. A 52% reduction in ice discharge between 20112012 and 2015 at the western section of the Cook Glacier, for example, can be explained by a multi-year landfast sea-ice promoting ice-front advance51 and reducing ice discharge (Fig. S5). However, at the neighbouring Frost and Holmes catchments, which also underwent large calving events in response to landfast sea-ice break-up44, we do not observe any obvious relationship with ice discharge variability, indicating that the glacial ice lost was passive. We suggest that the interaction between landfast sea-ice and ice shelves may become a more important driver of variability in ice discharge in the future if ice shelves weaken and retreat into confined embayments and/or if landfast sea-ice were to decrease52.

The response of a glacier to an initial velocity perturbation associated with a change in ice shelf buttressing is strongly modulated by the geometry of the glacier in its topographic setting. One potentially important aspect is the local slope of the topography at the grounding line. This is particularly the case for outlet glaciers with unconfined or weak ice shelves, which may be susceptible to rapid grounding line retreat along retrograde slopes53,54. However, for outlet glaciers with ice shelves that are able to provide sufficient buttressing, the local bed slope becomes less important in determining grounding line stability27,55. Therefore, if ice shelves are weakened sufficiently, local bed slope can be an important factor in determining grounding line migration. Retreat of the grounding line can cause further feedbacks because the associated loss in basal traction can lead to further acceleration and thinning56,57, thus creating a positive feedback. The acceleration associated with a loss in basal traction, in addition to reduction in ice shelf buttressing, has been shown to be an important factor in explaining the longer-term observed accelerations of both the Denman58 and Pine Island Glaciers28. Therefore, whilst changes in ice-shelf buttressing are likely to be the primary driver of the high spatial and temporal variability in ice discharge that we observe, the local bedrock slope at the grounding line is an important secondary factor in explaining the precise rate of response of individual glaciers.

On a localised scale, variable bed topography may explain why some neighbouring catchments can simultaneously undergo opposing trends in ice discharge, when variations in external forcing might be expected to be similar. For example, the mostly consistent acceleration of the Matusevich (Fig.4l) catchment between 2003/2006 and 2015 is anomalous amongst its neighbouring catchments (e.g. Cook, Slava, Rennick). A similar anomalous acceleration is also observed at the Hull catchment in Marie Byrd Land, where we observe a consistent increase in ice discharge throughout our observational period (Fig.4h), but a much more varied discharge of the neighbouring Land Glacier (Fig.4i). In both of these examples, the grounding lines of the glaciers displaying a spatiality anomalous acceleration, Matusevich (Fig.7a) and Hull (Fig.7c), have been retreating rapidly along retrograde slopes59,60. In contrast, there has been comparatively little grounding line migration in their respective neighbouring catchments, Rennick (Fig.7b) and Land (Fig.7d), which rest on a flat or prograde bedrock slopes. This would suggest that the spatially anomalous acceleration of Matusevich and Hull catchments has been strongly influenced by their underlying bed topography, which is conducive for rapid grounding line retreat.

Bed topography profiles from BedMachine14 extracted along the central flow line of the (a) Matusevich, (b) Rennick, (c) Hull and (d) Land Glaciers. The grey vertical lines are InSAR derived grounding line positions59,60. On the x-axis, zero represents the earliest measured position of the grounding line, positive values are the bedrock elevations advanced of the grounding line and negative values are the bedrock elevation of grounded ice.

See the rest here:

High spatial and temporal variability in Antarctic ice discharge linked to ice shelf buttressing and bed geometry | Scientific Reports - Nature.com