The ice masses of Svalbard cover an area of ca. 36 600 km2, and are thus among the largest glaciated areas in the Arctic. Annual mass balance measurements have been carried out on several Svalbard glaciers over up to 30 yr. However, these glaciers extend over only 0.5% of the total ice-covered area. The measured mean net balance has been negative and no changing trend has been observed. On some glaciers and larger ice caps, the mean net balance has also been measured at different altitudes by detecting radioactive reference layers from nuclear fallouts in 1963 and 1986 in shallow ice cores. The net balance/altitude curves have been estimated for thirteen different regions in Svalbard, and combined with digital elevation models of all Svalbard ice masses used to calculate the net balance in each 100-m altitude interval. The net loss of mass through iceberg calving was estimated and appears to be an important component of the net mass loss from Svalbard ice masses. The overall total net balance is slightly negative, −4.5 ± 1 km3 yr−1, giving a specific net balance of ca. −120 ± 30 mm yr−1 over the archipelago. The contribution of ice caps and glaciers on Svalbard to global sea-level change is, therefore, close to 0.01 mm yr−1 as an average value over the last 30 years, which is less negative than former estimates.
The ice caps and glaciers outside the Antarctica and Greenland ice sheets account for only about 4% of the area and 0.5% of the volume of ice on land, and would yield a global rise in sea level of about 0.5 m if they were to melt completely (Dyurgerov and Meier, 1997a; Meier and Bahr, 1996). However, these ice masses are more significant contributors than the great ice sheets to sea-level rise today, and are likely to remain so over the next century at least (Meier, 1984; Warrick et al., 1996; Church et al., 2001). The uncertainty in identifying the causes of global sea-level rise over the past century, and in forecasting future changes, stems in part from the incomplete knowledge of the mass balance of the world's ice masses. Estimates of the current sea-level rise due to glacier melting rely on mass balance estimates (Cogley and Adams, 1998; Dyurgerov and Meier, 1997b). Lack of long time series of observations is an important cause of uncertainty, but the spatial distribution and representativness of the measured glaciers is also of considerable significance.
Most of the smaller Arctic ice masses exist at higher air temperatures than those of Antarctica. Arctic ice masses are expected, therefore, to show a more rapid response to climatic change than Antarctica ice masses, and they are thus likely to give an early indication of climate change. They will probably lose mass in a warmer climate and cause sea level to rise, in contrast to Antarctica where ice masses are expected to remain stable or even increase in mass over the next century. Climate changes as rapid as those envisaged over the next century are expected to have profound effects on Arctic ice masses, including increased melting at lower elevations and probably increases in precipitation at all elevations.
We consider in this paper the problem of calculating the overall mass balance of ice masses in the archipelago of Svalbard. With its 36 600 km2 of ice caps and glaciers, Svalbard is among the most important glaciated areas in the Arctic (Hagen et al., 1993). Estimated total volume is ca. 7000 km3 or a sea-level equivalent of ca. 0.02 m. Svalbard is at the northern extremity of the warm North Atlantic drift. As a result, the archipelago is particularly sensitive to climate shifts. Estimation of the overall mass balance is clearly a major aim of glacier mass balance investigations, but the spatial distribution of available data is limited. Direct mass balance investigations are often carried out on fairly small glaciers that have been selected partly for logistical reasons rather than the optimum scientific priorities. In total, the glaciers where mass balance measurements have taken place cover less than 0.5% of all glaciated areas in Svalbard. On some glaciers, the mean net balance has also been measured in the accumulation area in shallow cores at different altitudes by detecting radioactive reference layers from nuclear fallouts in 1963 and 1986. The following discussion provides an estimate of the overall glacier net mass balance of the archipelago.
Mass Balance Equation
The overall net balance of Svalbard ice masses, or the annual volume change (∂V/∂t), can be calculated from the mass balance equation:Dowdeswell, 1989), and the bottom melting under grounded ice is so small that it is neglected here, Qb therefore equals zero. Total ice-volume change can thus be found from the net surface mass balance (Qn = Qa − Qm) and the iceberg calving term, Qc. The total yearly volumes are all given in km3 yr−1, and the specific values in mm yr−1 water equivalents.
Surface Mass Balance
Svalbard was divided into 13 regions, conforming to the major drainage basins in the archipelago defined from the glacier inventory of Svalbard (Hagen et al., 1993) (Fig. 1). For each region, a net mass balance per altitude (bn(z)) curve was estimated, representing a mean of the spatial variability within that area. The net balance curves are also time-averaged values, in that they include mass balance and snow accumulation measurements acquired over the last 30 yr or so. Mass balance data from individual glaciers in Svalbard suggest that no temporal trend is present (Hagen and Liestøl, 1990; Lefauconnier et al., 1999), and thus the measurements should represent the present, or recent past, climate. The time series of the individual mass balance components on Brøggerbreen, the longest continuous series in Svalbard, and on Kongsvegen is shown to illustrate this in Figure 2.
The data used to construct the net mass balance/altitude curve in each region of the archipelago are derived from the following data series:
1. Yearly field measurements of winter accumulation, summer melting and net mass-balance at individual glaciers;
2. Point net balance measurements from shallow cores;
3. Snow-distribution maps obtained from depth-probing and ground-penetrating radar;
4. Equilibrium-line altitude (ELA) distribution maps derived from aerial photographs and satellite imagery.
Yearly mass balance components have been measured over periods of different length on a number of glaciers. The set of glaciers where mass balance has been measured systematically (both winter and summer mass balance) includes only about 0.5% of the total ice-covered area of Svalbard (Table 1). Most of these investigations have been carried out on small cirque glaciers at relatively low elevation, and thus they are not fully representative of the area/altitude distribution of the overall ice masses covering higher accumulation areas (Hagen, 1996, Melvold and Hagen, 1998). However, they give reliable estimates of the balance gradients and ELA in their regions and in the altitude interval they cover.
The longest series are from the Kongsfjorden area in the northwestern part of Spitsbergen (Fig. 1), where direct yearly measurements have been carried out continuously since 1967 and 1968 on two small glaciers Austre Brøggerbreen and Midre Lovènbreen, (both ca. 5 km2) by Norwegian scientists (Hagen and Liestøl, 1990). These glaciers cover an altitude range from ca. 100 to 600 m a.s.l. Since 1987, the glacier Kongsvegen (100 km2) has been measured, covering an altitude range from sea level to ca. 1000 m a.s.l., and thus covering higher altitude accumulation area. In south Spitsbergen, Polish researchers have been making mass balance measurements on Hansbreen since 1988/89, and over shorter periods Russian scientists have studied some glaciers in west Spitsbergen (Troitskiy et al., 1975, Jania and Hagen, 1996). The net balance of the measured glaciers has been generally negative during the ca. 30 yr of observations (Table 1), and no changing trend or increased melting has been observed over the last decades (Fig. 2). The glaciers have probably had a stable negative mass balance since about 1920 (Lefauconnier and Hagen, 1990). The glacier area/altitude distribution is important, however, and the measurements on Kongsvegen indicate that glaciers covering higher accumulation areas are closer to zero balance than the lower lying glaciers (Hagen, 1996).
In addition to systematic direct mass balance measurements, there are several point measurements of mean net balance. Shallow cores have been drilled in a number of places throughout the archipelago (Fig. 1) to detect radioactive reference horizons from the fallout of 1962–1963 nuclear bomb tests and from the early May 1986 Chernobyl nuclear accident (Pinglot et al., 1999). The reference layers could be detected in all cores in the accumulation area of the glaciers. Cores were taken at varying altitudes from the highest part of the accumulation area downward to below the ELA (Fig. 3), thus providing average net accumulation values and the net balance gradient for the ice-mass accumulation areas.
On the larger ice caps in eastern Svalbard no mass balance data have been available. For our overall estimates for the whole archipelago it was important to specify the accumulation pattern on at least one of the large ice caps in eastern Svalbard in addition to the smaller ice masses in Spitsbergen. On the ice cap of Austfonna (8120km2), the largest ice cap in the archipelago (Fig. 1), an extensive study of the surface mass balance was carried out in the years 1998–2000. During two following years of spring field campaigns the winter accumulation was mapped and 29 shallow ice cores were retrieved (Fig. 3). The Chernobyl layer was located in 19 ice cores, all drilled in the accumulation area, and the nuclear test layer was located in two deeper ice cores (Pinglot et al., 2001). The temporal variation of the mean annual mass balance shows that no variation occurred for the two time periods, namely from 1963 to 1986, and from 1986 to 1999. Based on the average net accumulation and the altitude of the core sites Pinglot et al., (2001) established net balance gradients in the accumulation area in five transects radiating from the crest of the ice cap (Figs. 3, 4). These gradients have been used to estimate ELA along the transects. Almost no information exists on the ablation rate in the lower part of Austfonna, we have therefore assumed that the shape of the net balance curves bn(z) is similar to those found in other parts of Svalbard and used that to extrapolate the net balance curves found in the accumulation area down into the ablation area. However, the net balance curves are steeper in Nordaustlandet than in west Spitsbergen (Fig. 5) indicating that the conditions are drier and colder in the northeastern part of the archipelago. Limited data exist for region 1,7 and 13, and again the linear model was used based on few points, ELA pattern and information from the nearest regions.
On Storøyjøkulen (ca. 30 km2) on the island of Storøya, east of Nordaustlandet, Swedish glaciologists obtained mass balance information and the accumulation pattern in the years 1976 to 1978 (Jonsson, 1982). They found that the ELA was as low as between 100 and 150 m a.s.l., the accumulation pattern was asymmetrical as on the Austfonna ice cap with highest accumulation in southeast. The highest part of the ice cap is only ca. 240 m a.s.l., and superimposed ice was formed all over the ice cap.
Superimposed ice and internal refreezing is formed all over Svalbard. However, when using the net balance data obtained from core data the average net balance should be correct when assuming that the density below the reference horizon is constant over time.
Snow accumulation distribution on Spitsbergen, Svalbard, has been mapped by depth-probing and ground-penetrating radar in some selected east-west profiles following many of the large glaciers in Spitsbergen (Winther et al., 1998) and on Austfonna (Pinglot et al., 2001). The results from Winther et al. confirmed the precipitation pattern with higher accumulation rate and thus lower ELA in the eastern part of Spitsbergen compared to the western part as shown in Figure 6.
The ELA can be derived from aerial photographs and satellite imagery in addition to the altitude given from the direct measured glaciers and the shallow cores. In the Atlas of Glaciers in Svalbard and Jan Mayen, Hagen et al. (1993) published a map of the distribution of the ELA in Svalbard. However, very little information was available from the eastern part of the archipelago. The satellite imagery mainly gives information about the snowline, which is at higher altitudes than the actual ELA due to the formation of superimposed ice. No general function has been established between the snowline and the ELA, but the pattern of the ELA distribution is likely to have the same shape as the snowline distribution. This linkage was used by Hagen et al. (1993) for Spitsbergen and by Dowdeswell and Bamber (1995) on the islands Barentsøya and Edgeøya in eastern Svalbard to give information about the ELA pattern. The ELA map in Figure 6 shows the pattern of the current ELA. The map reflects the precipitation pattern in Svalbard; the higher the ELA, the lower is the snow accumulation.
In the calculations of the surface net balance we estimated the average net balance gradient (δbn/δz) or change in net balance per unit change in altitude in each of the 13 regions of Svalbard (Fig. 5). Hypsometric data for each region were taken from digital elevation models with a 100 m horizontal resolution, produced from Norsk Polarinstitutt topographical maps at a scale of 1:100,000 (contour interval 50 and 100 m). From the net balance curves bn(z) and the altitude models A(z) we obtained the specific net balance bni and the surface area Ai for each altitude interval. Thus, the total net balance per year (Bn) in each region can be given from:Qn = ∑ Bnj, j = 1, … ,13), was found to be slightly negative: Qn = −0.5 ± 0.1 km3 yr−1. The mean specific surface net mass balance can then be given for the overall ice masses by: bn = Qn/A, giving bn = −14 ± 3 mm yr−1. The overall area/altitude distribution of specific (bn(z)) and total net balance (Qn(z)) is shown in Fig. 7.
The result depends of the accuracy of the net balance curves bn(z). The error bars are only based on a qualified assessment. It is difficult to give quantified estimates of the error in these curves. The altitude of the ELA is fairly good. The error of individual mass balance measurements on monitored glaciers can be given by statistical methods and error analysis to assess and quantify the accuracy, but no consensus appears on error estimates even for glaciers on which traditional mass balance measurements are carried out. The reason for that is that there are a number of possible error sources. In our case, where we extend data from few areas and points to cover large areas, it is the representativness of the points that brings in the largest error. It is a matter of subjective evaluations and experience when we combine all available data to obtain the net balance curves. The result from many shallow cores taken in different altitudes is that they can be used to give the net balance curve with fairly good accuracy. This has been shown on Kongsvegen where good correlation was found between core data and direct measurements of the net balance per altitude, bn(z), based on annual measurements of accumulation and ablation (Lefauconnier et al. 1994a). The errors in the lower ablation areas are probably larger than in the accumulation areas since they are based on fewer points and in many parts just on the linear model. However, we believe that this crude estimate and extrapolation of data to obtain the overall net balance is far better than former estimates based on only individual glaciers. Since the curves bn(z) are based on direct measurements of the net balance the uncertainty in total overall net surface balance is less than in the estimates of the individual accumulation and ablation volumes.
The annual surface accumulation (Qa) and the annual surface melting (Qm) roughly estimated over the Svalbard archipelago and based on a subset of the data for the net balance calculations, give a total of close to 25 ± 5 km3 yr−1 for both the accumulation and the ablation without calving. This corresponds to a specific value of 680 ± 140 mm yr−1.
The ice flux at the marine glacier margins of Svalbard is the product of terminus velocity, ice thickness at the calving front, and the horizontal width of the ice front. The average thickness of the 1000-km-long calving front in the archipelago is estimated to be about 100 m, although many outlet glaciers have no ice-penetrating radar or marginal bathymetric data. All margins are grounded. Many of these ice fronts are relatively slow-moving or stagnant with velocities of less than 10 m yr−1, although some relatively fast-flowing outlet glaciers flow at 50 to 100 m yr−1 and sometimes a little more (e.g., Dowdeswell and Collin, 1990; Lefauconnier et al., 1994b). The bigger iceberg producers are dominant. There are few glaciers where surface velocity measurements and calving estimates have been carried out. Lefauconnier et al. (1994b) tracked ice-surface features on sequential SPOT imagery to calculate ice velocity, which was combined with field measurements of terminus thickness of Kronebreen (ca. 700 km2) in Kongsfjorden in northwest Spitsbergen to yield a calving rate of 0.25 km3 yr−1. Kronebreen is by far the fastest-flowing glacier in Svalbard with a velocity of about 2 m d−1 at the front. Field surveys of changing ice-cliff height, calving events and surface velocity at its marine margin have allowed the rate of mass loss to be estimated at Hansbreen (57 km2), in south Spitsbergen (Jania and Kaczmarska, 1997). A calving rate of ca. 0.02 km3 yr−1 was found, which is an important part of the mass loss (ca. 30%), equivalent to ca. 0.35 mm yr−1 water equivalent on the glacier surface. In Basin 5, a 670 km2 drainage basin of the ice cap Austfonna, Dowdeswell and Drewry (1989) found a velocity of ca. 40 m yr−1 at the calving terminus. On Austfonna, flow velocities have also been measured from remote sensing data by interferometry (Dowdeswell et al., 1999), but a full calculation of mass loss by calving awaits the associated marginal ice-thickness data.
Over the last nearly 100 yr the glacier fronts in Svalbard have been generally retreating, thus giving larger calving rates than in a steady state situation. In the Hornsund fjord (Fig. 1) the tidewater glacier fronts have retreated and an area of about 90 km2 have been deglaciated over the last 90yr (0.1 km3 yr−1) (Jania, 1988). Any advance or retreat of the ice front must also be considered when calving rates are estimated.
Compiling the available information together, an initial estimate of the average velocity of calving fronts through the archipelago is about 20 to 40 m yr−1. Thus, with an average margin thickness of 100 m and 1000 km length, the current calving ice flux is estimated to be about 3 ± 1 km3 yr−1. In addition to this the retreat of the calving glacier fronts is estimated to give 1 km3 yr−1. This estimate does not take into account the annual height change from melting. The total calving loss Qc is thus estimated to be about 4 ± 1 km3 yr−1.
An additional complication to the assessment of the rate of iceberg calving within Svalbard, is that a number of glaciers and ice-cap drainage basins undergo periodic surges (Hagen et al., 1993; Hamilton and Dowdeswell, 1996). Between the active surge advances, the termini of these ice masses are largely stagnant, contributing little to mass loss through iceberg production (Melvold and Hagen, 1998). In years when major Svalbard tidewater outlet glaciers surge, several cubic kilometers of ice can be released to the adjacent seas, and thus in years with such events the above estimate of calving rate can be far too low. The general retreat of calving and surging glaciers over the last hundred years has been mapped along the east coast of Spitsbergen by Lefauconnier and Hagen (1991). The largest surge observed was at Bråsvellbreen in 1936 to 1938 (Liestøl, 1969). This outlet from the Austfonna ice cap is about 1100 km2 and is grounded in water in a 30 km long glacier front. During the surge the front advanced up to 20 km (Schytt 1969). Negribreen in the inner part of Storfjorden, east Spitsbergen, surged in 1935 to 1936 and advanced about 12 km in less than a year along a 15-km-long section of the front. That is an average speed of 35 m d−1 (Liestøl, 1969). The 1250 km2 Hinlopenbreen surged in 1970 and calved about 2 km3 of icebergs in a single year (Liestøl, 1973). Synoptic satellite SAR investigations of ice-surface velocities on Svalbard ice caps and glaciers will eventually provide, together with airborne ice-penetrating radar measurements of ice thickness, a more complete dataset on iceberg calving (Dowdeswell et al., 1999).
The actual total net surface balance, Qn, combining all regions of Svalbard, was found to be slightly negative: Qn = −0.5 ± 0.1 km3 yr−1. The calving loss, Qc, was estimated to be −4 ± 1 km3 yr−1. The overall net balance of Svalbard ice masses is then:bn = −120 ± 30 mm yr−1 in water equivalent. The contribution of the ice caps and glaciers on Svalbard to global sea-level change is then ∂h/∂a ≈ ∂V/(∂a Asea) ≈ 0.01 mm yr−1 as an average value over the last 30 yr. This is less negative than earlier estimates based on net mass balance data for individual glaciers where stake measurements of accumulation and ablation have taken place (Dowdeswell et al., 1997).
The use of regionally derived net balance gradients and hypsometries is of particular importance because the overall mass balance of Svalbard can then be calculated in regionally representative height intervals of 100 m instead of using the mean specific values given for a small number of individual glaciers. Due to the area/altitude distribution of the measured glaciers, the negative mass balance values of these glaciers have usually given an overestimated sea-level contribution. Recent 10-yr- interval repeated GPS profiling over some large ice fields in Spitsbergen has also indicated that the higher accumulation areas are more stable and do not show the same shrinking as the smaller cirque and valley glaciers at lower altitudes (unpublished data).
Our analysis shows how important the area/altitude distribution within the archipelago is for the sensitivity of the glacier mass balance to climate changes. A diagram of the total net balance and area per altitude increment shows that the present equilibrium line is very close to the bulk of the glacier area (Fig. 7). Thus, the net surface balance of Svalbard glaciers and ice caps is sensitive to quite small changes in the equilibrium line altitude, with a shift of a few tens of meters up or down having a large effect on the total mass balance due to the nature of the hypsometric distribution. The analysis also indicates that it is important to monitor the mass balance per altitude and give these values in addition to the mean specific mass balance of a glacier.
The net loss of mass through iceberg calving appears to be a very important component of the net mass loss from Svalbard ice masses, even though the ice velocities at the 1000-km-long calving front are for the most part quite low and are certainly rather poorly specified. The total surface accumulation and melting was estimated to be close to 25 ± 5 km3 yr−1. This corresponds to a specific runoff value of 680 ± 140 mm yr−1. The yearly calving volume of 4 ± 1 km3 is thus only about 16% of the volume lost by melting. In spite of that, it is an important part of the net balance calculations. In the colder conditions of Severnaya Zemlya, 1600 km east of Svalbard, iceberg calving accounts for 35 to 40% of total mass loss from the 5500 km2 Academy of Sciences Ice Cap (Dowdeswell et al., 2002). Calving is likely to be increasingly important in colder glacial environments.
We derived a value of 0.01 mm yr−1 as our best estimate of the contribution of Svalbard ice masses to sea-level rise. Svalbard is about 15% of the total for the Arctic islands, as calculated by Dyurgerov and Meier (1997a). However, we have indicated that although our value is derived not just from net mass balance observations at a set of glaciers, but also from other measurements of accumulation and the way that it varies spatially across the archipelago and with altitude, it is still an imperfect estimate. We need better spatial information about the ablation rate in lower altitudes where ice core measurements cannot give any net balance data. Almost all Svalbard glaciers have been observed to retreat in the time since the end of the LIA about 80 to 90 years ago. This is a convincing indication that the glacier mass balance over the archipelago has been generally negative over this period. The mass loss from iceberg calving is clearly a very important part of the overall mass balance equation for the very many Arctic glaciers and ice-cap outlets that end in the adjacent seas. It has, however, been calculated for a very small number of Arctic glaciers and ice caps.
This work was supported in the years 1998–2001 by the EU-project ICEMASS, ENV4-CT97-0490, “The Response of Arctic Ice Masses to Climate Change.” The Norwegian Polar Institute provided the digital elevation models of Svalbard and Jostein Amlien masked out the glacier area in each altitude from these models. We are also grateful to J. G. Cogley and M. Dyurgerov who reviewed the manuscript and gave valuable suggestions for improvements.