Antarctica – Thwaites Glacier Basin, West Antarctica, Marine Ice Sheet Collapse Potentially Underway

Australia: The Land Where Time Began

The West Antarctic Ice Sheet covering a deep marine basin has for a long time been considered to be prone to instability. Joughin et al. have used a numerical model to investigate the sensitivity to ocean melt and if unstable retreat has already begun. Observed losses were reproduced by the model when forced with ocean melt comparable to estimates. Simulated losses were moderate, less than 0.25 mm/year sea level over the 21^st century, though they generally increased after that. The simulations indicated that early stage collapse has already begun, with the possible exception of the lowest-melt scenario. The timescale is less certain, with onset of rapid, more than 1 mm/year of sea level, and collapse for the different simulations with a range of 2-9 centuries.

Along the Amundsen Coast, Antarctic, glaciers are thinning (Pritchard et al., 2009; Shepherd, 2002), which produced the majority of the contribution of Antarctica to sea level rise (Rignot, 208; Shepherd et al., 2012). According to Joughin et al. it is likely that much of this thinning is a response to the increased presence of warm modified circumpolar deep water (CDW) on the adjacent continental shelf (Jacobs et al., 1011; Thomas et al., 2008), which is melting and thinning the floating ice shelves that buttress the ice sheet (Rignot et al., 2013; Pritchard et al., 2012). Ice shelves that are thinner have a reduced ability to restrain flow of ice from the interior, which contributes to feedbacks that increase the discharge of ice to the ocean (Shepherd et al., 2014; Payne et al., 2014; Joughin et al., 2010; Joughin et al., 2014; Gagliardini et al., 2010). The Thwaites and Haynes glaciers are referred to collectively as the Thwaites Glacier in this paper, produce just under half (52 Gt/year in 2007) of the total losses from the Amundsen Coast (105 Gt/year in 2007) (Rignot, 2008; Shepherd et al., 2012; Medley et al. 2014; Mouginot et al., 2014), which makes it one of the largest contributors to sea level change. Along with the glacier that is immediately adjacent to it, the Pine Island Glacier (Shepherd et al., 2002; Rignot, 2008), were identified as being potentially unstable several decades ago (Hughes, 1981).

The present grounding line of the Thwaites Glacier – the location at which the ice reaches the ocean and begins to float – rests on a coastal sill ~60 m below sea level (bsl) (Holt et al., 2006). This sill gives way to a deep marine basin, >1,200 m bsl, at~ 60-80 km further inland, which yields the potential for marine ice sheet instability (Joughin & Alley, 2012; Hughes, 1981; Schoof, 2007; Weertman, 2004; Mercer, 1978). The discharge of ice is non-linearly proportional to the thickness of the grounding line. Therefore there is a potential for instability where the bed of the ice sheet lies below sea level and steepens towards the interior in such a way that the retreat into deeper water forms a feedback which leads to more thinning and retreat. Therefore, as there are only 10s of kilometres separating the grounding line from the deepest regions of the marine basin, and ongoing thinning, it is possible that collapse of the Thwaites Glacier may have been already initiated, though at a relatively moderate Rate for now. Joughin et al. used a basin-scale ice-flow model in order to explore this possibility and to evaluate whether collapse is already underway or, alternatively, stabilising factors may limit the retreat of the glacier.

The response of the Thwaites Glacier to subshelf melt by the use of a prognostic, finite-element, depth-averaged, shallow-shelf model (Joughin et al., 2010; MacAyeal, 1989; Model details are provided in the supporting online material). Joughin et al. determined the basal shear stress and the ice shelf rheological parameters that best matched the about 1995 (1994-1996) observed velocity (Joughin et al., 2010; Joughin et al., 2009) and grounding line (Rignot et al., 2011). They used a simple-depth parameterised melt function that was scaled by a coefficient, m. This function produced steady state behaviour for the neighbouring Pine Island Glacier (Joughin et al., 2010). Maximum melt rates with m=1 for the Thwaites Ice shelf are just over 200 m/year in the deepest regions and total melt at 32 Gt/year at the commencement of the simulation, which makes it comparable to a 1992-1996 steady state estimate of 31 Gt/year for the highest–melt area (Rignot & Jacobs, 2002) of the shelf.

Since it appears that the thinning at the Amundsen Coast is driven by increased melting of the ice shelf (Shepherd et al., 2004; Payne et al., 2004), many of the experiments that were carried out were designed to examine this sensitivity to melt. Joughin et al. first examined the direct effect melting had on the position of the grounding line with no feedback or response from the glacier by setting the velocity at its initial value throughout the simulation. It was revealed by these experiments that the position of the grounding line is relatively insensitive to the direct effect of melting (Gagliardini et al., 2010; Schoof, 2007), producing nearly the same pattern of retreat for m = 1 and 4. Any retreat that does occur is largely driven by the non-steady state fixed velocity that was imposed at the start of the simulation.

The next experiment involved evaluating the response of the model to melt (m=0.5-4) with coupling (i.e. velocity that is evolving freely, to the ice sheet. There is a much greater sensitivity for these cases, in which the grounding line approaches the deepest parts of the trough for the higher melt simulations. Ice loss at rates of <0.25 mm/yr sea level equivalent (sle) for the first century, beyond which there is a period observed in each strong-melt simulation when there is an abrupt retreat of the grounding line, which produces greater ice loss (0.25 -0.5 mm/yr sle). With the exception of a few decades in the m=1 simulation, Ice loss for the lower melt simulations (m=0.5 and 1) was less than observed in 2010,

Accumulation rates in the Antarctic are projected to increase over the 21^st century (Genthon et al., 2009). Joughin et al. simulated a 20% linear increase in the rate of accumulation over the first 100 years, with a fixed rate thereafter, in order to evaluate any stabilisation effects such a change might have. The higher accumulation moves the low (m=0.5-1) melt simulations closer to balance. In the case of the higher melt (m=4) the transition to large losses is delayed by it.

Increased transport of warm circumpolar deep water onto the continental shelf and not by direct warming of the CDW (Thoma et al., 2008) largely drives the currently elevated melt rates on the Amundsen Coast. Melting should be reduced if the conditions that are responsible for this transport abate. Therefore Joughin et al. simulated 100 years of high melt (m=3 and 4) after which melt is reduced (m=1) for the remainder of the simulation. While the rate of loss was slowed by the reduction of melting, at the end of the simulations of 250 years, losses were substantially greater relative to the sustained m=1simulation.

Substantial weakening that is due to either rheological softening (e.g., fabric or strain heating) (Larour et al., 2005) or mechanical damage (e.g., crevassing or rifting) (Borstad et al., 2013). Weaker margins of the initial ice shelf is included in the model of Joughin et al., though as the shelf expands into the ice that was originally grounded, the margins of the ice shelf that are newly formed remain strong. An ad hoc weakening scheme implemented (Model details are provided in the supporting online material) in order to evaluate the sensitivity to weakening of the margin, and repeated their standard set of experiments. Grounding line retreat that was more extensive was produced by weakening of the margins, for the m=3 simulation. For about 212 years into the simulation, for the highest melting case (m=4), there is rapid grounding line retreat to the deepest regions of the basin, which yielded a contribution to sea level of more than 1 mm/year.

When losses in the simulation are greater than 1 mm/year sle, whiting a few years there are generally much greater losses. However, it is difficult to model such rapid collapse when using the basin-scale model of Joughin et al. Therefore 1 mm/year sle was taken to be the threshold, that once crossed, marks the onset of rapid decades long collapse as the grounding reaches the deepest region of the marine basin. Only the simulation with the highest melting, weak margin simulation reaches the critical threshold in the 250 year simulations. Therefore the remaining simulations were extended to determine when the threshold is reached. The onset of rapid collapse begins within 1,000 years for all but the lowest melt simulations (m=0.5).

The losses that were observed between 1995 and 2013 fall between the results from the highest (m=3 and 4) melt simulations by Joughin et al. The average simulated melt of 84 Gt/year for m=4, over this period, is in agreement with recent melt estimates of 69-97 Gt/year (Rignot et al., 2013; Depoorter et al., 2013), indicating that the early stages of the higher melt simulations reasonably approximate present conditions. Therefore the close argument that losses are driven by melt (Shepherd et al., 2004) is strengthened by the agreement between models and observation. In turn, the grounding line is caused to retreat by this initial speedup, which results in a loss of traction, and much greater speedup and retreat. It was noted by Joughin et al. that prior to 1995 the ice stream was already out of balance, which may have resulted from thinning that caused the ice to unground several decades or more ago from a ridge seawards of the grounding line of the present (Rignot, 2008; Tinto & Bell, 2011).

The simulations of Joughin et al. are not coupled to a global climate model to provide forcing, and they don’t include an ice shelf cavity circulation model to derive melt rates. There are at present few if any of such fully coupled models (Joughin et al., 2012). As such, the simulations by Joughin et al. do not constitute a projection of the sea level in the future in response to climate forcing that has been projected. However, it is indicated by the results the type of behaviour that is likely to occur. In particular, it is shown by the stepping back of the grounding line in stages with concurrent increases in discharge, consistent with other models and observations of the retreat of palaeo ice streams (Jamieson et al., 2012; Parizek et al., 2013). In these simulations the intensity of melt regulates the timescale over which this retreat pattern occurs. Therefore, while cavity circulation models that are driven by regional ocean circulation that are coupled to global climate models might yield different spatio-temporal variation in melt, they should produce retreat patterns that are similar to those that have been simulated by Joughin et al., though with tighter constraints on the timing.

In the numerical simulations of Joughin et al. an important feature is that they reveal a strong sensitivity to mechanical and/or rheological weakening of the margins, which can lead to an accelerated rate of collapse by decades to centuries. Therefore, in order to project accurately the rise of sea level models in the future will require careful treatment of shear margins. It is also assume d by their simulations that there is no retreat of the iceshelf front. More rapid retreat than have been simulated Joughin et al. should lead to full or partial collapse of the iceshelf. Also, they have not modelled ocean-driven melt extending immediately upstream of the grounding line, which could also accelerate retreat (Parizek et al., 2013).

Strong evidence that the process of destabilisation of marine ice shelves is already underway on the Thwaites Glacier, in large part in response to high subshelf melt rates. Rapid collapse (> 1 mm/year sle) will ensue after the grounding line reaches the deepest regions of the basin, which could occur within centuries, though over the next century losses are likely to be relatively modest (<0.25 mm/year sle). It is likely that such rapid collapse would spill over to adjacent catchments undermining much of West Antarctica (Holt et al., 2006). It is possible that similar behaviour may be underway on the neighbouring Pine Island Glacier (Joughin et al., 2010; Favier et al., 2014). It is difficult to foresee a stabilisation of the Thwaites system, even if there are plausible increases in surface accumulation, unless melt is reduced well below present levels by sufficient receding of the CDW. While it is suggested by the simple parameterisation of Joughin et al. that a full collapse of this sector may be inevitable, large uncertainty in the timing remains. Therefore, ice sheet models that are coupled fully to ocean/climate models are required to reduce uncertainty in the chronology of a collapse. It is suggested by the similarity between the highest melt rates of this study and the present observations that collapse may be closer to a few centuries than to a millennium.

Sources & Further reading