Australia: The Land Where Time Began

A biography of the Australian continent 

 Greenland – Accelerating Changes in Ice Mass and Sensitivity of Ice Sheet to Atmospheric Forcing

The total mass of ice in Greenland declined at a progressively increasing rate between early 2003 and mid-2013 then followed an abrupt reversal, when there was very little loss of ice over the next 12 to 18 months. Observations using the Gravity Recovery and Climate Experiment (GRACE) and global positioning system (GPS) revealed a similarity between the special patterns of the sustained acceleration in the mass loss and the deceleration of mass loss. The phase of the North Atlantic Oscillation (NAO) tracked the strongest accelerations. Summer warming and insolation are enhanced by the negative phase of the NAO while reducing snowfall, especially in west Greenland, which drives surface mass balance (SMB) more negative, as illustrated by use of the regional climate model MAR. The geography of NAO-driven shifts in atmospheric forcing and the sensitivity of the ice sheet to the forcing are reflected in the spatial pattern of accelerating mass changes. Bevis et al. infer that southwest Greenland will become a major contributor in the future to sea level rise.

The GRACE mission has been used to monitor loss of ice in Greenland by inferring near surface changes of mass from temporal variations in gravity measured in space (Velicogna & Wahr, 2006; Khan et al., 2010; Harig & Simons, 2012; Wouters et al., 2014; Van den Broeke, 2016). These measurements were remarkably consistent prior to the middle of 2013, with a mass trajectory model (Bevis & Brown, 2014) that consisted of an annual cycle, which was represented by a 4-term Fourier series, superimposed on a quadratic or “constant acceleration” trend with an acceleration rate of -27.7 ± 4.4 Gt/y². Mass was being lost at a rate of about -102 Gt/y from the Greenland Ice Sheet and the outlying ice caps in early 2003, and this rate had increased nearly 4-fold to about -393 Gt/y by 10.5 years later, which accounted for the acceleration of sea level rise that had been observed (Chen et al., 2017).

The Greenland GPS Network (GNET), which senses changes in mass by measuring the response of the solid Earth to changing surface loads, also observed the abrupt slowdown in deglaciation (Bevis et al., 2012). A combination of:

i)                   Glacial isostatic adjustment (GIA), i.e., the delayed viscoelastic response to changes in ice loads in the past, and

ii)                Instantaneous, elastic adjustment to contemporary ice mass changes.

Over decadal and shorter timescales GIA rates are nearly constant, with the possible exception of Kangerdlugssuaq Glacier where there are extremely low mantle viscosities (Khan et al., 2016). As a result, the vertical accelerations observed in GNET displacement time series (Bevis & Brown, 2014; Bevis et al., 2012; Khan et al., 2014) represent to a very large extent elastic adjustments to accelerating changes in the mass of ice.

Between 2008.4 and 2013.4, which excludes the summer of 2013, the estimates of Bevis et al. for the mean acceleration in uplift were positive at about 75% of GNET stations, and the largest positive accelerations were almost 3 times larger in magnitude than the most negative accelerations. Contrasting with this, for the 5 year period from 2010.4 to 2015.4, which includes the summer of 2013, negative accelerations were sensed by more than 90% of GNET stations, and the most negative accelerations had almost 3 times the magnitude of the most positive accelerations. Comparison of the cumulative distributions for each time period can be used to assess the ubiquity of the shift in mean vertical acceleration rates. Sign reversal is not strongly sensitive to the limits of these intervals of time.

It is suggested by the GRACE time series that the ~10 year episode of mass loss ceased, and the 2013-2014 Pause in the recent deglaciation of Greenland began near the middle of 2013. It is hard to be more precise, given the level of scatter in the GRACE residuals. An independent means to estimate the time of onset of the Pause is provided by the GNET data. Bevis et al. defined the station uplift anomalies by the use of a reference period that begins in or after 2017.0, ending at 2013.4 – the final epoch was determined a posteriori, following a series of experiments, in order to establish a self-consistent result. They fitted the vertical displacement (up) time series for each GNET station during the reference period with the same trajectory model that had been used to model the GRACE data. Bevis et al. then projected this model forward in time. The uplift anomaly is defined as the difference between the observed and model displacements. They used a travelling window of width 0.1 years to combine the daily displacement anomalies for 46 GNET stations, after which they computed the 25^th, 50^th, and 75^th percentiles of this point cloud. They found that the 50^th percentile curve (i.e. the median anomaly) deflects below the zero line near epoch 2013.4 and remains negative thereafter.

The 2013.4 epoch falls 18 days after the peak of the purely cyclical component of the model mass curve, and 21-25 days following the annual onset of negative mass balance (for Greenland as a whole) that was inferred from GRACE in 2004-2012. Bevis et al. suggested it took about 4 days for the deviation between predicted mass change and the actual mass change (in 2013) to be resolved clearly by GNET, i.e., for the trends in the percentile curves to emerge from the oscillatory noise that is seen in these curves prior to 2013.4, as only a small fraction of the net mass loss accumulated during the “mass loss season” accumulated in the first 21-25 days of that season.

It is implied by GRACE and GNET that the 2013-2014 Pause arose as a result of the expected season of negative mass balance that is associated with summer in the decade before 2013 did not develop, or hardly developed, during the summer of 2013 that was (recently) “anomalous”. If the mass balance curve produced by GRACE is examined, the magnitude of this deviation can be assessed by averaging the residuals in the interval 2013.79-2014.45. 

It was found by Bevis et al. that the mass loss that accumulated (in Greenland as a whole) in the summer of 2013 was smaller by 284 ± 43 Gt than expected based on the accelerating trend that had been observed in the previous decade. During the Pause the total ice mass fell by no more than ~75 Gt. According to Bevis et al. during the Pause there was little or no net change in ice mass, which does not imply that there was no mass loss anywhere within Greenland, but rather that local changes in ice mass tended to cancel out. By early 2015 the Pause had ended, though it is difficult to determine the end time of the Pause with any great precision, given the emergent onset of renewed ice loss, as well as the temporally correlated noise in the GRACE residuals.

It was noted by Van Angelen et al. (Van Angelen et al., 2014) that the accelerating ice loss that was observed by GRACE through the year 2012 correlated with summertime North Atlantic Oscillation (NAO) index that was increasingly negative during the 6 successive summers. The prevalence of clear sky, high pressure conditions increases due to the negative phase of the summertime NAO (sNAO) index, which enhances surface absorption of solar radiation and decreasing snowfall, and it causes the advection of warm air from southern latitudes into west Greenland. Higher temperatures are promoted by these changes, a longer ablation season and enhanced melt and runoff (14). It was concluded by Van Angelen et al. (Van Angelen et al., 3014) that if the summer sNAO switched back to positive values after 2012, then there might be a partial recovery of the surface mass balance (SMB). Indeed, not only did the June to August (JJA) and June to September (JJAS) NAO induce a turn to positive in 2013, but each of these sNAO indices from 2012 to 2014 was the single largest interannual change that had been recorded since 1950. Also, in 2015 when the sNAO index again turned strongly negative, there was a reestablishment of serious ice loss.

Topography modulates the impact of atmospheric warming

Over much of Greenland the negative phase of the NAO in summer enhances melting, especially in west Greenland (Van Angelen et al., 2014; Fettweis et al., 2013). The progressive warming of summers in west Greenland pre-2013 was less spatially focused than the strongest negative mass accelerations. The spatial distribution of ablation is controlled to a large extent by the spatial distribution of air temperature and solar radiation. Surface elevation has a strong influence on the sensitivity of the ice sheet surface warming. E.g., if the surface warms from -1 to 3^oC, then the impact of 4^oC of warming is vastly greater than if the surface warms from -5^oC to -1^oC. This is the reason simple models of melting are often expressed in terms of seasonal sums of positive degree-day (Brathwaite, 1995; Lacavalier et al., 2014). The amount of melting that is induced by an increase in temperature depends strongly on initial surface temperature, and therefore on latitude and elevation, as well as the time of year. A powerful positive feedback enhances the influence that surface elevation has on melting and runoff. The ice that is exposed in the ablation zone has a lower albedo than snow surfaces, which results in a greater absorption of solar radiation. In the ablation zone the largest source of melt energy is the solar radiation that is absorbed, and not on the sensible heat from the air (Van den Broeke, Smeets & van de Wal, 2011). The primary control on the geometry of the ablation zone is, nevertheless, air temperature, and, near surface temperature is controlled to a large extent by latitude and elevation, at a given time of year. In a given latitude zone, lower temperature gradients near the ice sheet margins lead to a wider ablation zone, which therefore act as primary controls on the spatial extent of the albedo feedback.

The trend of ice loss would be more pronounced at the southwest margin even if the southwest and the southeast margins of the Greenland Ice Sheet were exposed to similar positive temperature trends, because it has a much greater area of low elevation ice surface per unit length of margin than does the southeast margin. The low elevation and surface slopes that prevail at the northeast margin similarly ensure that it incorporated a much greater of low elevation ice surface than is the case for a segment of the northernmost margin of the ice sheet that is of similar size, or a similarly sized segment of the east margin where elevations are greater than 2 km loom over the nearby edges of the ice sheet. This helps to explain the localised centre of sustained negative mass acceleration in the northeast. The localised sensitivity of the northeast margin of the ice sheet to atmospheric forcing, relative to areas that are immediately adjacent, was also apparent in the correlated 2010 melting day and uplift anomalies, that were reported by Bevis et al. (Bevis et al., 2012).

Transient regional warming has less impact on the higher portions of the surface of the Greenland Ice Sheet than on lower portions. In central east Greenland the high mountains damming the ice sheet in central east Greenland ensure that there is very little ice per unit distance along the general trend of this ice margin, compared to the adjacent margins to the north and south. To a large extent, this explains the near zero mean mass acceleration rates that were inferred from east Greenland.

In summary, it is suggested by Brevis et al. that the geographical distribution of the progressive warming prior to 2013, which was focused mostly in the west of Greenland, as well as the spatial structure of the sensitivity to atmospheric warming, which is dominated by the topography near its margins, jointly explain most of the spatial pattern of the Surface Mass Balance trend and the mass acceleration field that was sensed by GRACE prior to 2013. The recent history of the runoff within the Taseriaq Basin of southwest Greenland (Ahlstrøm et al., 2017) supports this interpretation.

Atmospheric forcing, DMB and SMB

Changes in SMB and DMB drive accelerations in total ice mass change. (Note that DMA = -D, where D is the discharge, so total ice mass balance = SMB + DMB = SMB – D.) Changes in DMB are commonly driven by:

i)                   Changes in the circulation of the ocean and temperature, and

ii)                Changes in the floating portion of the ice sheet and the mélange of ice bergs and sea ice, which modulates their buttressing effect.

Calving rates and the velocity of outlet glaciers are affected by both changes, also causing inland changes in the thickness of the ice.

In northeast Greenland the secondary mass acceleration peak has already been associated with dynamic thinning in and near the outlet glaciers of the Northeast Greenland Ice Stream (Khan et al., 2014), though this does not rule out a role for atmospheric forcing. The observation that the mass anomaly field associated with the Pause has its 3^rd largest centre of mass gain in northeast Greenland, close to a centre of accelerating mass loss in the previous decade, suggests that this area was also affected by the shifting phase of NAO (Tedesco et al., 2016). In northeast Greenland the 3 GNET stations close to the margin of the Greenland Ice Sheet recorded accelerated uplift from their installation date through 2012 (Khan et al., 2014), and negative uplift anomalies by all of them after mid-2013. According to Bevis et al. this reversal occurred rather later than 2013.4-2013.5, presumably as a result of the arrival of summer later in this region than it does in central or southern Greenland, and therefore the nondevelopment of a negative SMB season that was previously typical that would not be evident until later in the year. The fact that it is evident that a sustained acceleration followed by an abrupt deceleration for northeast Greenland in both the GRACE and GNET time series suggests a connection to the charges driven by the NAO that were identified in southwest Greenland. It is indicated by the MAR SMB trend field that there was greater mass loss acceleration in the northeast sector than in either adjacent sector of the ice margin, though this is less pronounced as could be expected based on the GRACE results.

Bevis et al. suggested that sustained warming prior to 2013 drove a shift in DMB, as well as SMB in northeast Greenland. There are at least 2 possible mechanisms:

i)                   Regional warming drove a reduction in the extent of the floating ice sheet before the 2013 summer, and this diminished the buttressing effect on the outlet glaciers, which prompted increased rates of discharge which thinned the ice, as observed in the Antarctic Peninsula (22,23), and

ii)                Increases in the production of meltwater can modulate changes in ice mass.

There is a much greater area of low elevation surface of the northeast margin of the Greenland Ice Sheet than the sectors of margin on either side, and this would expand the area of enhanced production of meltwater. The viscosity of the ice sheet is lowered by the increased surface melting via the advection of latent heat to its interior (Phillips et al., 2013), and in thinner portions of the ice sheet associated with low surface elevation this mechanism will be volumetrically concentrated. Meltwater can also accelerate the flow of ice by modifying the mechanical conditions at the base of the ice sheet (Parizek & Alley, 2014; Schoof, 2010; Bartholomew et al., 2012). The development of subglacial lakes can, in extreme cases lift portions of an ice sheet or an ice cap from its bed (Schoof, 2010; Willis, Herried, Bevis & Bell, 2015). In Prudhoe Land in northwest Greenland, the hypothesis that increases in discharge, dynamic thinning, and glacial retreat has been invoked (30).

Discussion

As we regress to the mid-1900s the coverage and quality of our meteorological, glaciological, and geodetic datasets decline, as does the ability to track the relative importance of SMB and DMB as drivers of deglaciation. In spite of this it is clear that the sustained acceleration in mass loss that was recorded by GRACE prior to mid-2013 was completely unprecedented (Khan et al., 2015), as was the collapse of seasonally adjusted mass rate from its peak value to almost zero in the following 12-18 months. Mass rate scales with SMB and DMB, so mass acceleration scales with the trend or rate of change of SMB and DMB. The air-sea-ice system of Greenland crossed one or more of the thresholds or tipping points near the beginning of this millennium, which triggered deglaciation that was more rapid. Increased runoff in summer dominated the pronounced negative shift in spatially integrated SMB. Over most of the flanks of the Greenland Ice Sheet runoff increased, most noticeably in southwest Greenland, where in 2003 the margin was gaining mass though by 2012 it was strongly losing mass. The total glacial discharge integrated over southwest Greenland is not only very low (9.5 ± 1.5 Gt/year) compared with other areas (King et al., 2018), it has also been unusually stable. South of JI from 2000 onwards mass acceleration was dominated by falling SMB. A bit further north, discharge rates that had been adjusted seasonally at JI increased by ~44% from early 2000 to early 2006, though barely changes from 2006 to 2012 (King et al., 2018). In this 2^nd 6 year time interval it was SMB that was falling strongly, not DMB (Nielsen et al., 2013). Similar considerations apply in southwest Greenland (King et al., 2018).

In southwest Greenland the decadal acceleration in mass loss arose as a result of sustained global warming and positive fluctuations in temperature and insolation driven by the NAO. Bevis et al. developed an analogy with the global coral bleaching events that were triggered by every El Niño since that of 1997-1998, though not by any earlier El Niño event. The NAO has worked in concert with global warming to trigger major increases in the runoff in summer. The air was too cool for NAO to do the same prior to 2000. Global warming will be able to drive 2012 levels of runoff in a decade or 2, with little or no assistance from NAO. It can be inferred in the short term that the next time NAO turns strongly negative, over west and especially southwest Greenland, SMB will trend strongly negative, just as warming of the shallow ocean is expected to have its largest impact, via DMB (Holland et al., 2008; Straneo & Heimbach, 2013), in southwest and northwest Greenland. Brevis et al. inferred that within 2 decades this part of the Greenland Ice Sheet will become a major contributor to sea level rise, because ice sheet topography equips southwest Greenland with greater sensitivity to atmospheric forcing. It has also been suggested that enhanced melting in summer may induce more sustained increases in discharge rates.

Significance

The recent deglaciation of Greenland is a response to both oceanic and atmospheric forcings. Ice loss was concentrated in the southeast and northwest margins of the ice sheet from 2000 to 2010, as a result, in large part, of the increasing discharge of marine-terminating outlet glaciers, which emphasises the impedance of oceanic forcing. The largest sustained (~10 years) acceleration that was detected by Gravity Recovery and Climate Experiment (GRACE) was found to be in southwest Greenland, an area that is largely devoid of such glaciers. The sustained acceleration and subsequent, abrupt, and deceleration that was even stronger were driven mostly by air temperature changes and solar radiation. Bevis et al. suggest continued atmospheric warming will lead to southwest Greenland becoming a major contributor to sea level rise.

Sources & Further reading

Bevis, M., et al. (2019). "Accelerating changes in ice mass within Greenland, and the ice sheet’s sensitivity to atmospheric forcing." Proceedings of the National Academy of Sciences 116(6): 1934-1939.