Australia: The Land Where Time Began

A biography of the Australian continent 

Thwaites Glacier, West Antarctica – Heterogeneous Retreat and Ice Melt

Over the last few decades glaciers that flow into the Amundsen Sea Embayment in West Antarctica have been increasing in speed and their grounding line has been retreating. Milillo et al. have used a constellation of satellites which have detected the evolution of the velocity of the ice, thinning of the ice, and retreat of the grounding lines of the Thwaites Glacier from 1992-2017. A complex pattern of retreat and ice melt, with sectors that have retreated at 0.8 km/year and floating ice that has melted at 200 m/year, while others have retreated at 0.3 km/year and ice melting 10 times slower. They interpreted the results in terms of buoyancy/slope-driven intrusion of seawater along preferred channels at tidal frequencies that have led to more efficient melting in cavities that were newly formed. Coupled ice sheet/ocean models do not yet have such complexities represented.

The Antarctic Ice Sheet is rapidly changing and contributing notably to global sea level rise (Church et al., 2011; Alley et al., 2015). With a potential sea level equivalent of 1.2 m. The Amundsen Sea Embayment (ASE) sector of west Antarctica is presently a dominant contributor to sea level rise, as well as for decades to come (Joughin, Smith & Medley, 2014; Rignot et al., 2014; DeConto & Pollard, 2016). A third of the mass loss (Sutterley et al., 2014) from the Amundsen Sea Embayment is accounted for by the Thwaites Glacier. The main trunk of the Thwaites Glacier, which is fast flowing, accelerated by 0.8 km/year, or 33%, between 1973 and 1996, and another 33% between 2006 and 2013 (Mouginot, Rignot & Scheuchl, 2014). The ice discharge increased by 2.2 Gt/year², and the rate quadrupled in 2003-2010 (9.5 Gt/year²). More recently, some parts of the glacier have been observed to thin by as much as 4 m/year (McMillan et al., 2014).

Beneath the Thwaites Glacier, as the bed topography is several hundred metres below sea level at the grounding line and is getting deeper inland (retrograde bed slope), this sector may be prone to rapid retreat (Hughes, 1981; Schoof, 2007). It has been suggested by several studies that the glacier is already in a stage of collapse and the retreat is not stoppable (Joughin, Smith & Medley, 2014; Rignot et al., 2014; Parizek et al., 2013). The rate of retreat of the grounding line is controlled by bed topography, dynamic ice thinning, and melt of the ice shelf by warm, salty circumpolar deep water (CDW), with a critical role being played by melting of the ice shelf (Seroussi et al., 2017). The grounding line retreated by 0.6 to 0.9 km/year between 1996 and 2011, along the sides of the glacier and the main trunk, respectively (Rignot et al., 2014). After 2011, there has been no adequate interferometric synthetic radar (InSAR) data to observe the grounding line retreat (Khazendar et al., 2016; Scheuchl et al., 2016).

In Fig. 1 of the article the rapid migration of ice was not expected because the bed is prograde at that location, i.e., the elevation of the bed rises in the inland direction, which should be conducive to a slower rate of retreat for a given rate of ice thinning (Alley et al., 2015; DeConto & Pollard, 2016; Schoof, 2007). At B, the migration is slower, with a lower rate of shelf thinning, though at that location the slope of the bed is almost flat or even retrograde, which would be expected to favour retreat that is more rapid for a given rate of thinning. According to Milillo et al. the newly formed cavity at B is thin, which does not favour intrusion of warm Circumpolar Deep Water from geostrophic flow and efficient vertical mixing (Joughin et al., 2016; Jenkins, 2011) and explains the low rates of ice shelf melt. Contrasting with this the prograde at A favours the opening of a new cavity in the ice shelf, stronger intrusion of CDW, and efficient mixing with melt rates that are 20 times higher than those at B. Melting of the ice shelf at A exceeds the values that were used in numerical ice sheet/ocean models by factors of 2 to 3 (Joughin, Smith & Medley, 2014; Khazendar et al., 2016). The melt rate at B is low versus numerical simulations. Milillo et al. also found that the intensity of the melt along the ice at A, that was newly ungrounded, is correlated linearly with the slope of the ice draft along the direction that is perpendicular to the gradient in melt, consistent with the plume theory (Jenkins, 2011). No such correlation exists at B. Geostrophic and buoyancy/slope-driven flows are inefficient in the thin cavity near B, which is likely to be dominated by tidal mixing (Holland, 2008).

It is revealed by the TDX data at D that there is formation of a subglacial channel prior to ungrounding, followed by rapid melting along the sides near C and E. The D channel is initially 1.2 ± 0.2 km wide. There is no change in speed along the channel, therefore no dynamic thinning; melt by the ocean reflects ice thinning (Millgate et al., 2013). Ice shelf melt is high along prograde slopes along the sides – as for A – and low along retrograde slopes – as for B – where there less efficient cavity formation. A process of ice melt via channel intrusion that differs from the diffusive process that takes place along the grounding zone near A is revealed by these observations.

The melt rates of the ice shelf that have been discussed were calculated by use of a Eulerian framework, i.e., at a fixed location in space, in order to capture melt rates of the ice shelf and ungrounds of the ice shelf. Milillo et al. also calculated the melt rates in a Lagrangian framework, where ice blocks are tracked with time and corrected for flow divergence. The Lagrangian calculation does not apply on land, on areas that are partially grounded, or where ice is depressed flow flotation then rebounds during retreat. Milillo et al. confirmed ice shelf melt rates of up to 50 m/year on the butterfly and up to 200 m/year near the main trunk, with large spatial variations, away from these zones.

The observations of Milillo et al. contrast with the traditional view on the interaction between ice and ocean at grounding lines.

1.     Melt channels 1 to 2 km wide and cavities that were newly formed less than 100 m high would require ocean models to operate at the subkilometre horizontal scale and sub-100 vertical scale in order to replicate the melt processes responsible for forming the cavities, which is a challenge.

2.     As the melt rates in the main trunk are 2 to 3 times higher than those in models limits the ability of models to reproduce the ice retreat at those locations.

3.     Ocean-induced ice melt in the main trunk of the Thwaites Glacier occurs over a 2.5 km wide grounding zone, whereas fixed grounding lines are used by numerical ice sheet models, i.e., are not affected by tidal mixing (Joughin, Smith & Medley, 2014; DeConto & Pollard, 2016; Schoof, 2007; Parizek et al., 2013; Seroussi et al., 2017), with zero melt applied to the grounding line.

The existence of wide grounding zones, with a distinctive melt regime, narrow cavities, and non–zero ice melt over the entire grounding zone is revealed by the results of Milillo et al. The ice shelf rates may be lower along retrograde slopes than along prograde slopes, which is another observation to be explored in detail with ice-ocean models. Milillo et al. concluded that the shape of the cavity, including the slope of the bed, bumps, and hollows in the bed, influences the ocean heat access to the glacier as well as ocean-induced melt rates.

This study detected the highest rates of retreat at the heads of major channels of transport of CDW to the main trunk and TEIS (Schmeltz, Rignot & MacAyeal, 2001), with slow retreat in between (E to F), where ice is grounded on a ridge. Recent numerical models (Seroussi et al., 2017) have replicated the more rapid retreat of TEIS verses the remainder of the glacier and the mean retreat rate of 0.8 km/year since 1992; therefore it demonstrates that model skills and boundary conditions have improved considerably. However, the recent models do not replicate the fast retreat rate along the main trunk of the Thwaites Glacier, partly because in that region the ice shelf buttressing is limited. Milillo et al. report that in that region heterogeneous melt up to 200 m/year, with large tidal migration of the grounding line as well as significant melting of ice over the entire zone of tidal migration. Milillo et al. suggest that this configuration calls into question the concept of a fixed grounding line with zero melt because models that use melt at the grounding line predict retreat that is more rapid. Therefore, detailed studies of the grounding zone and its specific regime of ice melt will be critical to explore in more detail using numerical models, data from remote sensing, and in situ observations to improve the characterisation of retreat of the Thwaites Glacier near its grounding line, its rate of mass loss, and, in turn, its contribution to global sea level rise in decades to come.

Sources & Further reading

Milillo, P., et al. (2019). "Heterogeneous retreat and ice melt of Thwaites Glacier, West Antarctica." Science Advances 5(1): eaau3433.