IPCC Fourth Assessment Report: Climate Change 2007
Climate Change 2007: Working Group I: The Physical Science Basis Measured Balance of the Ice Sheets and Ice Shelves

Mass balance of the large ice sheets was summarised by Rignot and Thomas (2002) and Alley et al. (2005a). Greenland

Many recent studies have addressed Greenland mass balance. They yield a broad picture (Figure 4.17) of inland thickening (Thomas et al., 2001; Johannessen et al., 2005; Thomas et al., 2006; Zwally et al., 2006), faster near-coastal thinning primarily in the south along fast-moving outlet glaciers (Abdalati et al., 2001; Rignot and Kanagaratnam, 2006), and a recent acceleration in overall shrinkage.

Figure 4.17

Figure 4.17. Rates of surface elevation change (dS/dt) derived from laser altimeter measurements at more than 16,000 locations on the Greenland Ice Sheet where ICESat data from 2005 overlay aircraft surveys in 1998/1999 (using methods described by Thomas et al., 2006). Locations of rapidly thinning outlet glaciers at Jakobshavn (J), Kangerdlugssuaq (K), Helheim (H) and along the southeast coast (SE) are shown, together with an inset showing their estimated total mass balance (M∙ , Gt yr–1) between 1996 and 2005 (Rignot and Kanagaratnam, 2006).

Analysis of GRACE data showed total losses of 75 ± 26 Gt yr–1 between April 2002 and July 2004 (Velicogna and Wahr, 2005). Ramillien et al. (2006), also working from GRACE data, found a mass loss of 129 ± 15 Gt yr–1 for July 2002 to March 2005. Because of the low spatial resolution of GRACE, these include losses from isolated mountain glaciers and ice caps near the coast, whereas the results discussed next do not.

Mass loss from the ice sheet surface (net snow accumulation minus melt water runoff) has increased recently. Box et al. (2006) used calibrated atmospheric modelling and a single approximation for ice flow discharge to estimate average ice sheet mass loss of more than 100 Gt yr–1 during 1988 to 2004; they also found acceleration of surface mass loss during this interval of 43 Gt yr–1. A similar analysis by Hanna et al. (2005) for 1961 to 2003 found somewhat higher net accumulation but similar trends, with ice sheet growth of 22 ± 51 Gt yr–1 from 1961 to 1990, shifting to shrinkage of 14 ± 55 Gt yr–1 from 1993 to 1998 and shrinkage of 36 ± 59 Gt yr–1 from 1998 to 2003. Again, ice flow acceleration was not included in these estimates.

In a study especially using SRALT but incorporating laser elevation measurements from aircraft and a correction for the effect of changing temperature on near-surface density, Zwally et al. (2006) estimated slight growth of the ice sheet by 11 ± 3 Gt yr–1 from 1992 to 2002. However, they noted that mass loss of 18 ± 2 Gt yr–1 would be indicated if the thickness changes at higher elevations are largely low-density firn rather than high-density ice, as might apply if the effects of increasing accumulation rate were also taken into account (Hanna et al., 2005; Box et al., 2006). The more spatially limited results of Johannessen et al. (2005) from the same radar data indicated slightly less shrinkage or slightly more growth than found by Zwally et al. (2006) in regions of overlap. Krabill et al. (2000) also found thickening of central regions (~10 mm yr–1) from laser measurements covering 1993/1994 to 1998/1999.

Krabill et al. (2004) used repeat laser altimetry and modelled surface mass balance to estimate mass loss of about 45 Gt yr–1 from 1993/1994 to 1998/1999, with acceleration to a loss of 73 ± 11 Gt yr–1 during the overlapping interval 1997 to 2003. These values may underestimate total losses, because they do not take account of rapid thinning in sparsely surveyed regions such as the southeast, where mass budget studies show large losses (Rignot and Kanagaratnam, 2006). Thomas et al. (2006) extended these results to 2004 using ICESat data to include approximate corrections for density changes in the near surface. Results showed ice sheet mass loss of 27 ± 23 Gt yr–1 for 1993/1994 to 1998/1999, loss of 55 ± 25 Gt yr–1 for 1997 to 2003, with an updated loss of 81 ± 24 Gt yr–1 from 1998/1999 to 2004.

Rignot and Kanagaratnam (2006) combined several data sets, with special focus on the acceleration in velocity of outlet glaciers measured by Synthetic Aperture Radar (SAR) interferometry. Starting from an estimated excess ice flow discharge of 51 ± 28 Gt yr–1 in 1996, these authors estimated that the ice flow loss increased to 83 ± 27 Gt yr–1 in 2000 and 150 ± 36 Gt yr–1 in 2005. Adding surface mass balance deviations from the long-term average as calculated by Hanna et al. (2005) yielded mass losses of 82 ± 28 Gt yr–1 in 1996, 124 ± 28 Gt yr–1 in 2000 and 202 ± 37 Gt yr–1 in 2005. The more pronounced ice flow accelerations were restricted to regions south of 66°N before 2000 but extended to 70°N by 2005. These estimates of rapid mass loss would be reduced somewhat if ice surface velocities are higher than depth-averaged velocities, which may apply in some places.

Greenland Ice Sheet mass balance estimates are summarised in Figure 4.18 (top). Most results indicate accelerating mass loss from Greenland during the 1990s up to 2005. The different estimates are not fully independent (there is, for example, some commonality in the isostatic corrections used for GRACE and altimetry estimates, and other overlaps can be found), but sufficient independence remains to increase confidence in the result. Different techniques have not fully converged quantitatively, with mismatches larger than formal error estimates suggesting structural uncertainties in the analyses, some of which were discussed above. The SRALT results showing overall near-balance or slight thickening, in contrast to other estimates, may result from the SRALT limitations over narrow glaciers discussed earlier.

Assessment of the data and techniques suggests a mass balance for the Greenland Ice Sheet ranging between growth of 25 Gt yr–1 and shrinkage of 60 Gt yr–1 for 1961 to 2003, shrinkage of 50 to 100 Gt yr–1 for 1993 to 2003 and shrinkage at even higher rates between 2003 and 2005. Lack of agreement between techniques and the small number of estimates preclude assignment of statistically rigorous error bounds. Interannual variability is very large, driven mainly by variability in summer melting, but also by sudden glacier accelerations (Rignot and Kanagaratnam, 2006). Consequently, the short time interval covered by instrumental data is of concern in separating fluctuations from trends.

Figure 4.18

Figure 4.18. (Top) Mass balance estimates for Greenland. The coloured rectangles, following Thomas et al. (2006), indicate the time span over which the measurements apply and the estimated range, given as (mean + uncertainty) and (mean – uncertainty) as reported in the original papers. Code: B (orange; Box et al, 2006), surface mass balance, using stated trend in accumulation, ice flow discharge (assumed constant), and standard error on regression of accumulation trend, with added arrow indicating additional loss from ice flow acceleration; H (brown; Hanna et al., 2005), surface mass balance, with arrow as for B; T (dark green; Thomas et al., 2006), laser altimetry, showing new results and revision of Krabill et al. (2004) to include firn densification changes; Z (violet; Zwally et al., 2006), primarily radar altimetry, with uncertainty reflecting the difference between a thickness change due to ice everywhere and that due to low-density firn in the accumulation zone; R (red; Rignot and Kanagaratnam, 2006), ice discharge combined with surface mass balance; V (blue; Velicogna and Wahr, 2005) GRACE gravity; RL (blue; Ramillen et al., 2006) GRACE gravity; J (magenta dashed; Johannessen et al., 2005), radar altimetry without firn densification correction and applying only to central regions that are thickening but omitting thinning of coastal regions. (Bottom) Mass balance estimates for grounded ice of Antarctica. Coloured rectangles show age span and error range as in the top panel. Code: Z (violet; Zwally et al., 2006), radar altimetry, with uncertainty reflecting the difference between a thickness change due to ice everywhere and that due to low-density firn everywhere; RT (dark green; Rignot and Thomas, 2002), ice discharge and surface mass balance, with dashed end line because some of the accumulation rate data extend beyond the time limits shown; RT2 (dark green; Rignot and Thomas, 2002), updated to include additional mass losses indicated by Thomas et al. (2004) and Rignot et al. (2005), dashed because the original authors did not produce this as an estimate for the whole ice sheet nor are accumulation rates updated; V (blue; Velicogna and Wahr, 2006), GRACE gravity; RL (blue; Ramillen et al., 2006), GRACE gravity. Antarctica

Recent estimates of Antarctic Ice Sheet mass balance are summarised in Figure 4.18 (bottom). Rignot and Thomas (2002) combined several data sets including improved estimates of glacier velocities from InSAR to obtain antarctic mass budget estimates. For East Antarctica, growth of 20 ± 21 Gt yr–1 was indicated, with estimated losses of 44 ± 13 Gt yr–1 from West Antarctica. The balance of the Antarctic Peninsula was not assessed. Combining the East and West Antarctic numbers yielded a loss of 24 ± 25 Gt yr–1 for the region monitored. The time interval covered by these estimates is not tightly constrained, because ice input was estimated from data sets of varying length; output data were determined primarily in the few years before 2002.

Zwally et al. (2006) obtained SRALT coverage for about 80% of the ice sheet, including some portions of the Antarctic Peninsula, and interpolated to the rest of the ice sheet. The resulting balance included West Antarctic loss of 47 ± 4 Gt yr–1, East Antarctic gain of 17 ± 11 Gt yr–1 and overall loss of 30 ± 12 Gt yr–1. If all the ice thickness changes were low-density firn rather than ice, the loss would be smaller (14 ± 5 Gt yr–1). Davis et al. (2005) compiled SRALT data for about 70% of the ice sheet, and did not interpolate to the rest. The same pattern of East Antarctic thickening and West Antarctic thinning was observed (Figure 4.19). Davis et al. (2005) suggested that the East Antarctic change was primarily from increased snowfall. Assigning all thickness change to low-density firn produces growth of the monitored portions of the ice sheet of 45 ± 8 Gt yr–1; if all thickness changes were ice, this growth would be 105 ± 20 Gt yr–1. Following the suggestion that the East Antarctic changes are from increased snow accumulation and the West Antarctic changes are more likely to be ice-dynamical would yield growth of monitored regions of 33 ± 9 Gt yr–1. Note, however, that Monaghan et al. (2006) did not find the strong increase in snow accumulation suggested by Davis et al. (2005) in arguing for use of low-density firn in East Antarctic changes.

Figure 4.19

Figure 4.19. Rates of surface elevation change (dS/dt) derived from ERS radar-altimeter measurements between 1992 and 2003 over the Antarctic Ice Sheet (Davis et al., 2005). Locations of ice shelves estimated to be thickening or thinning by more than 30 cm yr–1 (Zwally et al., 2006) are shown by red triangles (thickening) and purple triangles (thinning).

Rignot et al. (2005) documented discharges 84 ± 30% larger than accumulation rates for the glaciers that feed the Wordie Ice Shelf on the west coast of the northern Antarctic Peninsula (which shrank greatly between 1966 and 1989), a region largely absent from the SRALT studies. Consideration of strong imbalances in glaciers feeding the former Larsen B Ice Shelf across the Peninsula, and extrapolation of the results to undocumented basins, suggested mass loss from the ice in the northern part of the Antarctic Peninsula of 42 ± 7 Gt yr–1. Observation of widespread glacier front retreat in the region (Cook et al., 2005) motivates the extrapolation, although mass loss would be overestimated if snow accumulation has been systematically underestimated (van de Berg et al., 2006).

Taking the Rignot and Thomas (2002), Zwally et al. (2006) and Rignot et al. (2005) results as providing the most complete antarctic coverage suggests ice sheet thinning of about 60 Gt yr–1, with uncertainty of similar magnitude to the signal. Consideration of acceleration of some near-coastal glaciers, discussed below, and the difficulty of SRALT sampling of such regions, might allow slightly faster mass loss. The time interval considered is not uniform; the Rignot et al. (2005) results include changes after the collapse of the Larsen B Ice Shelf in 2002, more recent than data in the other studies, and suggest the possibility of accelerating mass loss. Use of the more spatially restricted Davis et al. (2005) SRALT data rather than the Zwally et al. (2006) results illustrates the persistent uncertainties; depending on the assumed density structure of the changes, Davis et al. (2005) combined with the Rignot et al. (2005) estimate for the Antarctic Peninsula would suggest near-balance or antarctic growth.

Interpretations of GRACE satellite gravity data indicate mass loss from the Antarctic Ice Sheet, including the Antarctic Peninsula and small glaciers and ice caps nearby, of 139 ± 73 Gt yr–1 between April 2002 and July 2005 (Velicogna and Wahr, 2006). Near-balance was indicated for East Antarctica, at 0 ± 51 Gt yr–1, with mass loss in West Antarctica of 136 ± 21 Gt yr–1. Independent analyses by Ramillien et al. (2006) found, for July 2002 to March 2005, East Antarctic growth of 67 ± 28 Gt yr–1, West Antarctic shrinkage of 107 ± 23 Gt yr–1 and a net antarctic loss of 40 ± 36 Gt yr–1.

Assessment of the data and techniques suggests overall Antarctic Ice Sheet mass balance ranging from growth of 50 Gt yr–1 to shrinkage of 200 Gt yr–1 from 1993 to 2003. As in the case of Greenland, the small number of measurements, lack of agreement between techniques, and existence of systematic errors that cannot be estimated accurately preclude formal error analyses and confidence limits. There is no implication that the midpoint of the range given provides the best estimate. Lack of older data complicates a similar estimate for the period 1961 to 2003. Acceleration of mass loss is likely to have occurred, but not so dramatically as in Greenland. Considering the lack of estimated strong trends in accumulation rate, assessment of the possible acceleration and the slow time scales affecting central regions of the ice sheets, it is reasonable to estimate that the behaviour from 1961 to 2003 falls between ice sheet growth of 100 Gt yr–1 and shrinkage of 200 Gt yr–1.

Simply summing the 1993 to 2003 contributions from Greenland and Antarctica produces a range from balance (0 Gt yr–1) to shrinkage of 300 Gt yr–1, or a contribution to sea level rise of 0 to 0.8 mm yr–1. Because it is very unlikely that each of the ice sheets would exhibit the upper limit of its estimated mass balance range, it is very likely that, taken together, the ice sheets in Greenland and Antarctica have been contributing to sea level rise over the period 1993 to 2003. For 1961 to 2003, the same calculation spans growth of 125 Gt yr–1 to shrinkage of 260 Gt yr–1, with 1993 to 2003 likely having the fastest mass loss of any decade in the 1961 to 2003 interval. Geodetic data on Earth rotation and polar wander provide additional insight (Peltier, 1998). Although Munk (2002) suggested that the geodetic data did not allow much contribution to sea level rise from ice sheets, subsequent reassessment of the errors involved in some of the data sets and analyses allows an anomalous late 20th century sea level rise of up to about 1 mm yr–1 (360 Gt yr–1) from land ice (Mitrovica et al., 2006). Estimated mountain glacier contributions do not supply this, so a contribution from the polar ice sheets is consistent with the geodetic constraints, although little change in polar ice is also consistent. Ice shelves

Changes in the mass of ice shelves, which are already floating, do not directly affect sea level, but ice shelf changes can affect flow of adjacent ice that is not floating, and thus affect sea level indirectly. Most ice shelves are in Antarctica, where they cover an area of about 1.5 × 106 km2, or 11% of the entire ice sheet, and where nearly all ice streams and outlet glaciers flow into ice shelves. By contrast, Greenland ice shelves occupy only a few thousand square kilometres, and many are little more than floating glacier tongues. Mass loss by surface melt water runoff is not important for most ice shelf regions, which lose mass primarily by iceberg calving and basal melting, although basal freeze-on occurs in some regions.

Developments since IPCC (2001) include improved velocity and thickness data to estimate fluxes, and interpretation of repeated SRALT surveys over ice shelves to infer thickening or thinning rates. Melting of up to tens of metres per year has been estimated beneath deeper ice near grounding lines (Rignot and Jacobs, 2002; Joughin and Padman, 2003). Significant changes are observed on most ice shelves, with both positive and negative trends, and with faster changes on smaller shelves. Overall, Zwally et al. (2006) estimated mass loss from ice shelves fed by glaciers flowing from West Antarctica of 95 ± 11 Gt yr–1, and mass gain to ice shelves fed by glaciers flowing from East Antarctica of 142 ± 10 Gt yr–1. Rapid thinning of more than 1 m yr–1, and locally more than 5 m yr–1, was observed between 1992 and 2001 for many small ice shelves in the Amundsen Sea and along the Antarctic Peninsula. Thinning of about 1 m yr–1 (Shepherd et al., 2003; Zwally et al., 2006) preceded the fragmentation of almost all (3,300 km2) of the Larsen B Ice Shelf along the Antarctic Peninsula in fewer than five weeks in early 2002 (Scambos et al., 2003).

Box 4.1: Ice Sheet Dynamics and Stability

The ice sheets of Antarctica and Greenland could raise sea level greatly. Central parts of these ice sheets have been observed to change only slowly, but near the coast rapid changes over quite large areas have been observed. In these areas, uncertainties about glacier basal conditions, ice deformation and interactions with the surrounding ocean seriously limit the ability to make accurate projections.

Ice sheets are thick, broad masses of ice formed mainly from compaction of snow (Paterson, 2004). They spread under their own weight, transferring mass towards their margins where it is lost primarily by runoff of surface melt water or by calving of icebergs into marginal seas or lakes. Water vapour fluxes (sublimation and condensation), and basal melting or freezing (especially beneath ice shelves) may also be important processes of mass gain and loss.

Ice sheets flow by internal deformation, basal sliding or a combination of both. Deformation in ice occurs through solid-state processes analogous to those involved in polycrystalline metals that are relatively close to their melting points. Deformation rates depend on the gravitational stress (which increases with ice thickness and with the slope of the upper surface), temperature, impurities, and size and orientation of the crystals (which in turn depend in part on the prior deformational history of the ice). While these characteristics are not completely known, model tuning allows slow ice flow by deformation to be simulated with reasonable accuracy.

For basal sliding to be an important component of the total motion, melt water or deformable wet sediment slurries at the base are required for lubrication. While the central regions of ice sheets (typically above 2,000 m elevation) seldom experience surface melting, the basal temperature may be raised to the melting point by heat conducted from the earth’s interior, delivered by melt water transport, or from the ‘friction’ of ice motion. Sliding velocities under a given gravitational stress can differ by orders of magnitude, depending on the presence or absence of unconsolidated sediment, the roughness of the substrate and the supply and distribution of water. Basal conditions are well characterised in few regions, introducing important uncertainties to the modelling of basal motion.

Ice flow is often channelled into fast-moving ice streams (which flow between slower-moving ice walls) or outlet glaciers (with rock walls). Enhanced flow in ice streams arises either from higher gravitational stress linked to thicker ice in bedrock troughs, or from increased basal lubrication.

Ice flowing into a marginal sea or lake may break off immediately to form icebergs, or may remain attached to the ice sheet to become a floating ice shelf. An ice shelf moves forward, spreading and thinning under its own weight, and fed by snowfall on its surface and ice input from the ice sheet. Friction at ice shelf sides and over local shoals slows the flow of the ice shelf and thus the discharge from the ice sheet. An ice shelf loses mass by calving icebergs from the front and by basal melting into the ocean cavity beneath. Estimates based on available data suggest a 1°C ocean warming could increase ice shelf basal melt by 10 m yr–1, but inadequate knowledge of the bathymetry and circulation in the largely inaccessible ice shelf cavities restricts the accuracy of such estimates.

Ice deformation is nonlinear, increasing approximately proportional to the cube of the applied stress. Moreover, an increase in any of the six independent applied stresses (three stretching stresses and three shear stresses) increases the deformation rate for all other stresses. For computational efficiency, most long simulations with comprehensive ice flow models use a simplified stress distribution, but recent changes in ice sheet margins and ice streams cannot be simulated accurately with these models, demonstrating a need for resolving the full stress configuration. Development of such models is still in its infancy, with few results yet available.

Ice sheets respond to environmental forcing over numerous time scales. A surface warming may take more than 10,000 years to penetrate to the bed and change temperatures there, while a crevasse filled with melt water might penetrate to the bed and affect the temperature locally within minutes. Ice velocity over most of an ice sheet changes slowly in response to changes in the ice sheet shape or surface temperature, but large velocity changes may occur rapidly on ice streams and outlet glaciers in response to changing basal conditions or changes in the ice shelves into which they flow.

The palaeo-record of previous ice ages indicates that ice sheets shrink in response to warming and grow in response to cooling. The data also indicate that shrinkage can be far faster than growth. Understanding of the processes suggests that this arises both because surface melting rates can be much larger than the highest snowfall rates, and because ice discharge may be accelerated by strong positive feedbacks (Paterson, 1994; P.U. Clark et al., 1999). Thawing of the bed, loss of restraint from ice shelves or changes in melt water supply and transmission can increase flow speed greatly. The faster flow may then generate additional lubrication from frictional heating and from erosion to produce wet sediment slurries. Surface lowering as the faster flow thins the ice will enhance surface melting, and will reduce basal friction where the thinner ice becomes afloat. Despite competition from stabilising feedbacks, warming-induced changes have led to rapid shrinkage and loss of ice sheets in the past, with possible implications for the future.