Is Antarctica melting?
Abstract
Until recently, the mass balance of the Antarctic ice sheet was not well known. Here, I review recent progress in determining its magnitude and temporal evolution, the physical processes that control the observed changes in ice sheet mass balance, and the important knowledge gaps that remain. The results highlight that the linkage between climate change and the Antarctic ice sheet is more complex than anticipated and that major observational and numerical modeling advances will be needed before we can reliably predict its evolution in a warming climate. At present, the Antarctic ice sheet is losing mass at a rate almost comparable to that of the Greenland ice sheet, about 250 ± 31 Gt/year or 0.7 mm/year sea level rise, and the mass loss is increasing with time, at a rate slightly below that observed in Greenland, at 14 ± 2 Gt/yr2. The Antarctic ice sheet is therefore a major contributor to sea level rise and its contribution is slowly increasing with time. WIREs Clim Change 2011 2 324–331 DOI: 10.1002/wcc.110
This article is categorized under:
- Paleoclimates and Current Trends > Modern Climate Change
BACKGROUND
Determining whether Antarctica is melting or not is a topic of tremendous scientific and societal importance, even more so in the context of climate warming. Antarctica is 50% larger in area than the United States, or 12 million square kilometers. Precipitation on the continent is comparable to that of a desert, at 13 cm of water equivalent per year on average, compared to a global mean of 100 cm water equivalent. Yet this desert holds 70% of the world freshwater and if all ice in it were to melt to sea, it would raise global sea level by 56 m.
During the last inter-glacial, when global air temperatures were 3–5°C warmer than today, sea level was 7–9 m higher.1 The contribution from the Greenland ice sheet was likely in the 3–4 m range,2 which implies that a significant part of the Antarctic ice sheet must have melted to sea then. Climate warming of a similar order of magnitude over the next century, held there long enough, will almost certainly yield the same result.
Simplified numerical ice sheet models forced by global climate simulations predict that in a warmer climate, the Antarctic ice sheet would not melt unless global temperatures increase by at least 4–5°C; precipitation on the continent would increase as a result of enhanced evaporation from the oceans; and overall the Antarctic ice sheet could only grow and gain mass. This view has prevailed for the past several decades,3 but as early as in 1998 there has been mounting evidence that the Antarctic ice sheet is not in a state of mass equilibrium and almost certainly not growing.4, 5 In the last 3–4 years, major, independent surveys have confirmed this view using extensive measurements of ice sheet mass balance (i.e., whether the ice sheet is growing or shrinking with time) using a variety of independent techniques.6-13 These satellite observations and other data have unveiled an ice sheet on the move, losing mass rapidly to the ocean.
PHYSICAL CLIMATE
The question of whether Antarctica is warming or cooling at present has been the topic of intense debate and large uncertainty. In the dry valleys of Antarctica,14 reported a cooling of Antarctica between 1966 and 2000, which contrasts with the rapid regional warming of the Antarctic Peninsula.15 New insights were brought upon this issue by an examination of a much longer time record over a larger area. Steig et al.16 combined historical weather station data and satellite data over the entire continent to show that Antarctica as a whole has been warming over the last 50 years. According to their study, warming is most significant over the Antarctic Peninsula and northern West Antarctica, i.e., at low latitude, in low-lying regions (Figure 1). Warming is more subdued in East Antarctica. The observed trend is attributed to an increase in anthropogenic emissions of greenhouse gases in the global climate system combined with changes in stratospheric circulation and wind regime associated with the ozone depletion above the Antarctic.17
Long-term changes in yearly surface temperature in and around Antarctica. Places where it warmed over time are yellow to red, places where it cooled are blue, and places where there was no change are white.16
The present level of Antarctic warming is not sufficient to produce surface melt and runoff on continental ice, i.e., ice escaping the ice sheet in liquid form as in Greenland. Much of the East Antarctic plateau sits at high elevation and is unlikely to melt for decades to come. Sectors more clearly at risk of melting, however, include the vast floating expenses of glacier ice surrounding Antarctica, called ice shelves. Floating ice shelves play a fundamental role in ice sheet stability because they buttress the inland flow of ice. At present, many ice shelves experience some level of summer melt but most melt water refreezes in place. If warming were to take place in coastal regions, firn (old snow on top of ice) may become more permeable to water, water may accumulate at the surface of the ice sheet, surface cracks could get filled with water which after refreezing will propagate cracks downward until they ultimately sever the entire column of ice and lead to the collapse of the ice shelf. This sequence of events has been observed in the Antarctic Peninsula over the last few decades as a result of regional warming of air temperature and possibly ocean temperature.18-20
Ice shelves and glaciers interact not just with the atmosphere, but also with the ocean. Ice–ocean interactions are controlled by the temperature, salinity, and circulation of the ocean waters, which are affected by wind patterns and buoyancy forces.21 The rates of basal melting underneath ice shelves are orders of magnitude larger than mass changes at the surface.22 The rates of bottom melting are also highest near the glacier grounding line, which is where a glacier detaches from the bed and becomes afloat in ocean waters, because this is where the ice draft is the deepest, the pressure melting point of ice is the lowest, and basal ice is warm and close to the melting point. The grounding line region is also of critical importance to ice dynamics.23, 24 Migration of the grounding line induces large changes in basal resistance to flow and in turn has a major influence on the rates of glacier flow. Changes in ice–ocean interactions can therefore have a direct, profound impact on glacier flow and in turn on the mass balance of the glaciers and the entire ice sheet.
In this context, a central question for glaciology is to determine if the Southern Ocean has been warming up. Gille25 reports that the southern ocean has been warming up since the 1950s, faster than elsewhere in the world (Figure 2). Boning et al.26 confirm the hemispheric-scale warming of the southern ocean to depth of more than 1000 m using Argo data. A large share of the warming seems to be caused by changes in the water masses that make up the Antarctic Circumpolar Current, which is the dominant feature of ocean circulation in the southern hemisphere. Few data exist along the coast of Antarctica, however, it is not clear how far south these changes have affected the properties of ocean waters and how much extra heat may have been able to reach key Antarctic glaciers. This lack of knowledge in the southernmost portion of the Southern Ocean is a most fundamental gap in our current understanding of the Antarctic ice sheet evolution.
Temperature trends (°C/year) at 900 m depth using data collected from the 1930s to 2000s, including shipboard profile and Autonomous LAgrangian Current Explorer float data. The largest warming occurs in subantarctic regions, and a slight cooling occurs to the north. (Reprinted with permission from Ref 25. Copyright 2002 The American Association for the Advancement of Science)
ICE SHEET MASS BALANCE
There are three major techniques employed to measure ice sheet mass balance, i.e., how fast an ice sheet grows or shrinks with time. Historically, the first technique is the mass budget method, which compares the input and output of ice mass to the continent. The first attempts at closing the Antarctic mass budget in the 1960s until as recently as 1991 employed single measurements of ice velocity and thickness, and hand drawn maps of accumulation based on snow pits and shallow ice cores. These data were not sufficient to be conclusive even about the sign of the mass balance. The mass budget approach indeed compares two large numbers (total accumulation vs total loss) with large uncertainties, hence requires very precise measurements of both.
Major progress took place in the 1990s–2000s with the advent of satellite and airborne data and the maturation of regional climate models. Ice discharges or perimeter fluxes are now known along nearly the entire periphery of Antarctica from satellite measurements of ice velocity (Figure 3) and airborne measurements of ice thickness at the grounding line.8 Ice thickness, if not measured directly from airborne radio echo sounding, is deduced from surface elevation of floating ice measured by satellite altimeters.27 In addition, regional atmospheric climate models have provided corrections for firn depth,28 and grounding line positions have been mapped with unprecedented accuracy using satellite radar interferometry. Ice thickness is known within 10 m with radio echo sounding and 80 m from hydrostatic equilibrium. Ice velocity is measured within 1–10 m/year error to yield perimeter flux with 2–8% precision.
Ice velocity of the Antarctic ice sheet inferred from satellite radar interferometry collected by ERS-1/2, Radarsat-1, Radarsat-2, ALOS PALSAR, and Envisat ASAR.8, 29 Velocities are color coded on a log scale from brown (∼0) to green (a few 10 m/year) to blue (100 m/year), purple (several 100 m/year) to red (km/year) overlaid on a MODIS mosaic of Antarctica.
Snow accumulation maps are now derived using regional atmospheric climate models controlled by global climate reanalysis data. Prior maps based on surface data combined with proxy from surface reflectance data were largely in error along the wet, coastal sectors. The new numerical models use field data for evaluation purposes only. In Antarctica, the Regional Atmospheric Climate Model RACMO2/ANT forced by ERA-40 and ECMWF reanalysis data achieves a precision of 5–30% when moving from dry to wet areas, for an overall precision in accumulation of 71 Gt/year or 5%.30 The net accumulation of snow on the continent found by this model is not that much different from that derived in earlier assessments in total, but there are major differences in the spatial distribution of accumulation, in particular, much larger accumulation levels are found along the wet coastal sectors than in prior reconstructions.
These major improvements in our knowledge of perimeter losses and interior gains have allowed us for the first time to close the mass budget of Antarctica.8 The numbers are large in absolute terms but small compared to the annual turnover of mass in Antarctica (about 2200 Gt/year). This justifies a posteriori why high accuracy data were necessary to determine the mass budget of Antarctica. The mass budget method also provides insights about the exact partitioning of the mass balance between changes in snow accumulation and changes in glacier flow (see below). On the down side, improvements in regional atmospheric climate models are desirable in wet sectors to reduce uncertainties below 30%, and direct measurements of ice thickness are required on many glaciers to update calculations made based on hydrostatic equilibrium and reduce the corresponding errors.
A second technique of mass balance is to measure changes in surface elevation using satellite or airborne altimeters. Measurements started in the 1980s, but increased in quality in the 1990s with better instruments and orbit determinations. Laser altimeters improved upon the vertical precision of radar altimeters by one order of magnitude in 2002 (Figure 4). The altimetry technique provides a comprehensive coverage of the Antarctic, except the South Pole, and has accumulated nearly 20 years of observations. Several complications, however, make it difficult to convert the observed volume changes into mass changes. First, the density at which these volume changes are taking place is now well known. Fresh snow has a density of about 0.3; pure ice has a density of 0.9. If volume changes are caused by changes in precipitation, a density of close to 0.3 should be used. If volume changes are caused by an acceleration of ice flow along a glacier, a density of 0.9 should be used because it affects the entire column of ice. The uncertainty in converting volume changes into mass changes is therefore 50%. Secondly, changes in surface elevation are affected by a process called firn compaction, which is time-dependent and climate dependent and which must be corrected for to convert volume to mass.11 Third, the performance of radar altimetry degrades along the coast sectors of ice sheets, which are steep and rough compared to the interior. Fourth, radar altimeters penetrate the ice surface and the reflecting horizon is seasonally and climate dependent.31 Similarly, the reflecting horizon of laser altimeters has been shown to be influenced by the decaying power of the laser illumination, thereby introducing biases in long-term changes in ice sheet elevation that have not been fully evaluated yet. Finally, the altimetry technique is limited in spatial coverage, and the coastal sectors are generally not well surveyed which is a problem if most changes are taking along the coastal regions. A number of these issues will be resolved with the Cryosat mission launched in 2010. Despite the above limitations, it is also clear that satellite and airborne altimetry remain tremendously useful for detecting areas of elevation change around the continent and monitor changes in the rates of ice thinning and thickening. One of the greatest challenges of this technique, however, is to provide answers about the thickening rate of East Antarctica, where accumulation rates are low (10 cm/year), and hence decades of observations will be necessary before any conclusions may be reached about even the sign of the mass balance.
Satellite data shows fast ice thinning (red) along the coast of West Antarctica. The inset shows satellite tracks over Pine Island Glacier, from NASA's Ice, Cloud, and land Elevation Satellite, which reveal dynamic thinning concentrated on the fast-flowing areas. (Reprinted with permission from Ref 11. Copyright 2009 Nature Publishing Group)
A third technique emerged in early 2002 using time series of time variable gravity from the GRACE satellite. This approach measures mass changes directly, at a coarse spatial resolution (200–400 km), but independent of the density of volume changes, grounding line position (floating ice does not displace the gravity field) snowfall, sublimation, and evaporation. It integrates changes in ice dynamics and snow accumulation into a single number that represents the net change in mass. A drawback of the gravity-based approach is that it requires a correction for post-glacial rebound (GIA), i.e., the slow readjustment of the Earth crust in response to deglaciation during the past several thousand years. This signal is indistinguishable from the present-day mass balance and must therefore be estimated and removed. In Antarctica, the correction is larger than the signal itself, so its precision has a major impact on the mass balance results.6, 32, 33 In addition, the GRACE results published in the literature do not all agree, for various reasons, including: differences in processing schemes, differences or lack of proper calibrations of inversion schemes, analysis of different time periods, truncation of spherical harmonics series to a different order, and simply the natural temporal variability of the observed signal. The Antarctic record of precipitation reveals that a long time series of observations is needed to detect long-term trends in surface mass balance because of large inter-annual variations in snowfall, larger than the absolute mass loss we are trying to measure. It will therefore be a few more years before gravity techniques provide the most robust and reliable assessments of ice mass change in Antarctica.
One of the chief strength of gravity, however, is its unique ability to examine whether the interior of Antarctica is gaining or losing mass. The mass budget method relies on regional atmospheric climate models to determine mass gain or loss; the altimeter technique requires decades of data to detect a signal likely to be in the range of a few millimeter to centimeter thickening or thinning of an entire continent. GRACE data examined so far suggests no change in mass of interior East Antarctica. This is an important result, independent of the GIA correction, i.e., a constant signal. It suggests that global climate model predictions of Antarctic growth may be called into question. The GRACE result is also consistent with long-term trends in surface mass balance, or lack thereof.
TEMPORAL EVOLUTION
The surface mass balance of Antarctica exhibits no long-term trend for the 1980–2009 and the 1958–2009 periods.30, 34 Although this does not preclude changes in precipitation in the future, it suggests that global climate models do not predict Antarctic climate well.
Satellite radar altimetry studies have suggested growth of East Antarctica over 1995–2002, but this is largely caused by decadal variability in snowfall. Surface mass balance records show that we cannot conclude positively on the mass balance of the ice sheet with only 7 years of data (Figure 5). More recent results with satellite laser altimetry suggesting growth of East Antarctica are called into question for the same reason, and in addition because of possible biases introduced by instrument degradation.
Estimates of the mass balance of the Antarctic ice sheet in Gt/year (trillions of tons per year) from GRACE (red points), and InSAR/RACMO2 (black points) from 1992 to 2010. The data clearly show an increase mass loss with time, amidst considerable inter-annual to decadal variations in mass balance driven by the large inter-annual variability in Antarctic snowfall. Notice that over the period 1995–2002, the mass balance was positive but that this trend is not reflective of the longer term trend.13
Coastal discharge is increasing in several key sectors because the glaciers are accelerating. Acceleration took place in the Antarctic Peninsula following the collapse of the Larsen A and B ice shelves.35 In the Amundsen Sea sector of West Antarctica, glacier acceleration has been attributed to the presence of warm circumpolar deep water that drives large rates of bottom melt near the grounding lines.5, 36, 37 We do not know what processes control the intrusion of warm water on the continental ice shelf, and whether and when they have changed in the past. In other parts of Antarctica, glaciers are slowing down,38 but this phenomenon is an aclimatic response that does not carry much long-term significance because of its apparent cyclic nature. In the Antarctic Peninsula, the glaciers that lost their ice shelves speed up 300–800%39 and have maintained high velocities since. In the Amundsen Sea, the acceleration is 80% in 35 years for Pine Island Glacier, but the speed up has increased every year since 1992. Simple models predict that the velocity of this large ice stream will triple in the next 5–10 years as the glacier ungrounds from an ice-plain area only a 30–50 m above sea level.36
The numbers from the mass budget method have recently been reconciled with the GRACE data by comparing them over a common time period to validate a longer time record.31 From 1992 to 2008, the mass loss of Antarctica increased at a rate of 14 ± 2 Gt/yr2 to reach 250 ± 31 Gt/year in 2010 (Figure 5). The mass loss from Antarctica is now comparable in magnitude to the mass loss for Greenland. It is only 15% of the annual input of mass to Antarctica, i.e., a small leakage of ice to the ocean, but this is sufficient to contribute 0.7 mm/year sea level rise or one fourth of the total observed sea level rise.
PHYSICAL CONTROLS
Over the last several decades, snowfall and ice melting have not been major participants in the long-term trend in Antarctic mass balance. The mass budget of the Antarctic ice sheet has been largely controlled by the evolution of its outlet glaciers. Glacier velocities are in turn controlled by internal dynamics and ocean temperature, but we do not fully understand these interactions.
Ice–ocean interactions certainly play a central role. More heat delivered to the glacier grounding lines melts ice from underneath, ungrounds the glaciers from their bed, reduces buttressing of the inland ice, and allows faster rates of ice discharge to the sea. This chain of events is the lead explanation for the ongoing glacier acceleration in the Peninsula and northern West Antarctica.
The southern ocean, near the ice sheet edge, is largely devoid of observation, because it is hard to get to, this part of the ocean is covered by sea ice more than half of the year, and few satellites extend their coverage of the oceans that far south. Although the intrusion of warm water in Pine Island Bay is known,40 we do not know how and when it changed in the past. Recent surveys of the sea floor have revealed the presence of previously unknown ridges that may have played an important role in initiating the ongoing retreat of the glacier.41 These new data illustrate how little we know about the coastal sectors of Antarctica and yet how important it is to map them better, in particular, the sea floor bathymetry in order to understand ice–ocean interactions.
To predict the evolution of these glaciers, we also need a detailed characterization of their geometry, where they are grounded below sea level and how far inland they remain grounded below sea level, where bumps and hollows in bed topography upstream may slow or hasten their retreat. Basic observations of ice thickness in Antarctica's critical dynamic sectors are still vastly lacking today. Major efforts are underway to fill in this data gap, but it will be sometime before these data are incorporated into numerical models to make them more realistic.
Large-scale models of Antarctica have been operating at spatial resolutions that are too coarse to resolve glaciers and ice streams, and they used simplified ice physics to enable long-term simulations. Recent observations and progress in numerical modeling suggest that simplified ice physics is not sufficient to explain the evolution of present-day glaciers.42-44 A new generation of higher resolution, higher order numerical ice sheet models is needed, forced by satellite observations that can help constrain unknown model parameters, and coupled with ocean numerical models capable of prescribing the spatial and temporal distribution of ice-shelf melting.45 Until the coupling of ice and ocean is realistically implemented and numerical models employ realistic geometries for the outlet glaciers, predicting the evolution of Antarctic glaciers will remain very difficult.
CHALLENGES AHEAD
Major advances have been made in mapping glacier velocities, grounding line positions, accumulation of snow, mass and surface elevation changes at the continental scale over the past two decades in Antarctica. These observations have shown three largely unanticipated results: (1) snowfall has not been increasing; (2) outlet glaciers are the dominant control on ice sheet mass balance; and (3) a main driver of glacier changes is the Southern Ocean, but we do not understand it well.
Although much attention has been paid to the Antarctic Peninsula and the Amundsen Sea sector of West Antarctica, other sectors of Antarctica are possibly equally important to the future. One of them is the Wilkes Land sector of East Antarctica. Wilkes Land is a marine sector, like West Antarctica, prone to rapid changes. It drains more ice than the entire West Antarctic ice sheet. Some of its most prominent outlet glaciers (Totten, Cook, Denman) are thinning and losing mass at present, possibly more now than 10 years ago.11, 12 We do not know what processes are driving these changes and over what time scale.
Significant data collection efforts resulted from the International Polar Year 2007–2009, but critical observational capabilities are being lost. The GRACE mission has only 3 more years to go, with no follow-on mission for the next several years. InSAR monitoring of ice sheets is at an all time low since 1992 because several missions shifted to the commercial sector or are terminated. And there is no mission capable of mapping ice thickness from space.
The European Space agency is flying Cryosat-2, a new altimetry mission, which will bring new data about Antarctica's ice elevation. ESA will launch Sentinel-1, an InSAR mission, in late 2012. NASA will fly its own InSAR mission in 2017–2019. In the meantime, NASA is flying IceBridge, an airborne program dedicated to bridging the gap between ICESAT-1 and ICESAT-2. The Canadian Space Agency is flying RADARSAT-1, -2, and soon RADARSAT-3. Satellite programs are essential to the study of the Antarctic continent given its shear size and the complexity of its climate and ice regime. In addition, major efforts are still underway to improve our characterization of Antarctic climate and changes in the southern ocean.
CONCLUSIONS
The Antarctic ice sheet is melting, but not in the way we anticipated. Ice is being discharged too rapidly to the ocean in some sectors; the interior regions experience large year-to-year fluctuations but exhibit no long-term trend. Air temperature is on the rise, but we do not know if the Southern Ocean has been delivering more heat to the glaciers in the recent past.
If we use the current rate of increase in mass loss of Antarctica to project future losses in the coming decades, we obtain a contribution to sea level of 30–40 cm by 2100, and a third of that by 2050. This calculation is highly uncertain because ice sheet evolution is largely nonlinear, the observational record is only two decades long, and only a new generation of advanced numerical models will lead us to produce more realistic predictions. We do not know how the surface mass balance of Antarctica may be changing in the future. A small increase in interior snowfall could offset the losses from enhanced glacier flow. Similarly though, we do not know how fast some of these glaciers may flow as they retreat, e.g., in Pine Island Bay. We do know that their Greenland counterparts flow ten times faster.
This leaves us with considerable uncertainty about the future of Antarctica, how much it will melt, how fast it will melt. Recent progress has probably raised more new issues than solved old ones. In major contrast with assessment conducted as recently as the last IPCC, however, we now have a much better idea that the Antarctic continent is a strong contributor to changes in sea level; it is not growing as suggested by climate models, but melting away, very slowly.
