Downscaled GCM projections of winter and summer mass balance for Central European glaciers (2000–2100) from ensemble simulations with ECHAM5-MPIOM

Authors


Abstract

This study is based on the study from Matulla et al. (2009) where the glacier under estimation has been Peyto Glacier in Canada. The same methods have been used for five Austrian glaciers; projections of glacier mass balance are generated from ensembles of general circulation model (GCM) simulations by the use of direct statistical downscaling. Thereby, the general features of the atmospheric circulation over an expanded geographical region covering the European Alps are linked empirically to winter and summer mass balance records measured at five glaciers in Austria. The projections are taken from an ensemble of ECHAM5-MPIOM simulations forced with the IPCC-SRES scenarios A1B and B1. Results based on the statistical downscaling indicate decreasing balances for both winter and summer. These results suggest continued frontal recession and downwasting of the alpine glaciers in this region until 2100. For Jamtalferner, these suggestions reach reductions of about 1000 mm water equivalent in summer. Copyright © 2012 Royal Meteorological Society

1. Introduction

Glaciers are essential reservoirs of fresh water for mountain regions and downstream across the Earth. Another important aspect of glaciers is their potential to generate geohazards—in particular, outburst floods from dammed lakes and related overflows of mountain valleys. Glaciers are key indicators of global climate change because of their sensitive reaction to climate change, which in turn offers the possibility to reconstruct climate changes from glaciers changes (Oerlemans, 2005). However, only a small number of the glaciers worldwide are under regular monitoring (Matulla et al., 2009). These monitored glaciers clearly show a general reduction of glacier mass worldwide (Zemp et al., 2009).

The European Alps are the most prominent rock mass nestled in the mountainous region of Central Europe. The alps range from France in the west, over Southern Germany Switzerland, Northern Italy, Austria and Slovenia in the east. Austria covers most of the so-called Eastern Alps where the glaciers—on which this study focuses on—are located. Compared to the Eastern Alps, the Western Alps are higher and their central line is curved. Thus, the Eastern Alps are oriented more parallel to the prevailing zonal air flow from the North Atlantic towards the European continent making precipitation generation more complex.

Mass balance is the direct signal of a glacier to climate forcing. In winter, glaciers accumulate precipitation as snow, which in the case of a positive mass balance, is transformed to firn and ice over time. Strong winds may cause significant amounts of snow drifts across or even away from the glacier (Mott et al., 2008). The ablation (mass loss) of the glacier during summer is predominately from snow and ice melt which is closely related to the incoming shortwave radiation balance at the glacier surface. The melt-water generated is released into streams and lakes providing an important source of water intake to the ecosystems located downstream. It was shown in several studies that glacial melt-water plays an important role in Alpine river-flow conditions (Huss et al., 2008; Koboltschnig et al., 2007, 2008). Besides their effect on the amount of river discharge, which is also important in the context of Alpine power production, glaciers are important for the low temperature of glacial rivers and as such in turn to the quality of the fish distribution, for instance (Füreder et al., 2001).

Glacier dynamics are sensitive to variations of the components of the energy balance at their surfaces which are closely linked to incident shortwave radiation. Energy balance at the glacier surface is well parameterized by air temperature and precipitation totals (Ohmura, 2001) such that simple glacier models forced by summer air temperature (to model ablation) and winter precipitation (to simulate accumulation) are rather successful when simulating the mass balances of Alpine glaciers (Hock, 1999).

At long timescales, changes from stadial to interstadial periods are ruled by orbital forcing and feedback processes (e.g. the ice-albedo feedback). At short timescales, glacial variability is controlled by internal and external forcings of the climate system (e.g. changes in the ocean dynamics and related to the fluctuations of atmospheric circulation patterns (e.g. the North Atlantic Oscillation, NAO) or changes in the radiative forcing (e.g. from changes of the level of anthropogenic greenhouse gases (GHGs), water vapour, aerosols, clouds). As with the long-term scale, these effects are well captured by local air temperature variations which enables the modelling of glacier mass balance from observations of air temperature and precipitation.

Another important issue relevant for the mass balance of a glacier relates to local effects acting on different glaciers, depending on their size, area and geographical location, the dynamics of surface flow (e.g. temperate or cold basal) as well as the influence from debris-cover. These effects can cause different responses to climate even for two glaciers which are geographically nearby and have to be considered for the subsequent interpretation of glacier mass balance modelling. In order to portray these complex relationships, it is necessary to calibrate the downscaling models separately for each glacier under consideration (see below for a list).

Since glaciers are of great socio-economic-ecological importance (tourism, water supply to ecosystems, generating plants, people, agricultural production, etc.), it is of particular interest to derive estimates of how glaciers may react to possible future climate conditions.

General circulation models (GCMs) constitute the main instrument to analyse possible changes in the global climate triggered by changes in forcing parameters. Conceptual and computational limitations are the main reasons why the spatial resolution of GCMs is still limited to a few hundred kilometres. Moreover, it is estimated that the skilful spatial scale of GCMs is about eight times the grid scale. Therefore, output on spatial scales smaller than the skilful scale should not be interpreted (von Storch et al., 1993; Joannesson et al., 1995). However, the assessment of changes in the forcing parameters on glacial mass balances requires information on a much finer scale than the GCM spatial scale. In order to bridge this gap, a technique called downscaling is to be used. Downscaling is the process of cascading down output from GCMs to the local scale.

Overall, two major downscaling approaches can be distinguished: (1) process-based techniques (called dynamical downscaling) with regional circulation models (RCM) nested into a GCM (Giorgi, 1990; Grotch and MacCracken, 1991; Giorgi et al., 1994) and (2) the use of statistical established transfer functions (called empirical downscaling) between the scales (von Storch et al., 1993; Zorita et al., 1995; Zorita and von Storch, 1999; Matulla et al., 2002). The procedure used in the present study belongs to the latter group. In principle, dynamical downscaling should generate the more reliable regional results, because topography, land-use patterns etc. are accounted for and the physics of the system is regarded at a high resolution. However, dynamical downscaling requires detailed surface information and is very computer intensive. Empirical approaches offer an alternative that do not need high-end computing facilities. However, empirical techniques rely on the stability of the statistical relationship between the predictors and the predictand, which is not necessarily the case under changed climate conditions. In any case, scenarios produced by both dynamical and statistical downscaling approaches are ultimately constrained by the realism of the boundary forcing of the host GCM.

If a local-scale variable like a biosphere phenomenon or for instance glacier mass balance and large-scale atmospheric variables can be linked directly through an empirical relationship, statistical downscaling techniques might be applied without the effort of modelling between the local-scale phenomenon (e.g. the flowering stages of a plant or the mass balance of a glacier) and the local-scale atmospheric variables (Maak and von Storch, 1997; Matulla et al., 2003; Matulla et al., 2009). Shea and Marshall (2007) and Matulla et al. (2009) showed the possibility of modelling the mass balance of glaciers directly from flow indices or the large-scale state of the atmosphere, respectively. This is the approach that is applied here.

The set-up of the present study follows Matulla et al. (2009), whereby a direct downscaling approach was used to generate ensembles of local-scale glacier mass balance for Peyto Glacier (Alberta, Canada) from large-scale atmospheric fields. This method is applied here to selected glaciers located in the Austrian Eastern Alps region.

2. Data

2.1. Mass balance observations at Austrian glaciers

Glaciers under consideration (see Figure 1) include the Jamtalferner, located in the Silvrettagruppe, the Hintereisferner and Vernagferner, both located in the Ötztaler Alpen, and the Goldbergkees and Wurtenkees, located in the Hohe Tauern region. Table I gives some characteristics of the glaciers used for the analysis applied in the present study. The length of the time series varies from 19 years for Jamtalferner and Goldbergkees to 53 years for Hintereisferner. The time series of Vernagtferner are 42 years long and for Wurtenkees 25 years. An overview on glaciers measured for mass balance in Austria can be found in Schöner et al. (2000).

Figure 1.

Location of the Austrian glaciers under estimation: Jamtalferner, JTF; Hintereisferner, HEF; Vernagtferner, VNF; Goldbergkees, GOK and Wurtenkees, WUK. The main chain of the Alps is printed in black

Table I. Characteristics of the glaciers
GlacierCoordinatesExpositionHeight (m)AreaLength
 NOacabMaximumMinimum(km2)(km)
  1. JTF, Jamtalferner; HEF, Hintereisferner; VNF, Vernagtferner; GOK, Goldbergkees and WUK, Wurtenkees. The exposition is separated into accumulation (ac) and ablation (ab) areas. From fluctuations of the glaciers from 2000 to 2005, the minimal height values of GOK and WUK are gained on a personal communication (Hynek, ZAMG, 2009).

JTF46.8710.17NN312023703.482.8
HEF46.8010.77ONO372724007.47.1
VNF46.8810.82SSO362727658.363.15
GOK47.0312.47SONO308023701.492.8
WUK47.0313.01SWS312025200.973

The glacier Jamtalferner is located north of the main Alpine chain in the most western part of the Austrian Alps, the Silvrettagruppe. This region is characterized by high-precipitation sums. The accumulation area of Jamtalferner is structured into three parts. Two of them are exposed northward and the third one westward. The weakly developed tongue is directed northward.

The Hintereisferner and Vernagtferner glaciers are located in the Ötztaler Alpen, across the main Alpine crest, at an altitude between 2400 and 3727 m (a.s.l.). The Ötztaler Alps constitutes a comparatively dry (shielded from precipitation) inner-alpine climate region of the Austrian Central Alps. The Hintereisferner is a valley glacier, with a distinctive glacier tongue, flowing down from Weißkugel (3.738 m) eastward at the beginning and northeastward afterwards. Vernagtferner glacier is located north of Hintereisferner exposed towards the Southeast. The glaciers of the Ötztaler Alps are exceptional in that long-term mass balance measurements date back to 1951 and numerous studies have been published that document the climate–glacier relationship (Kuhn et al., 1999; Escher Vetter et al., 2005).

The glaciers Goldbergkees and Wurtenkess lie in the most eastern glacierized part of the Austrian Alps with Goldbergkees on the northern side and Wurtenkees at the southern side of the main Alpine crest within the Hohen Tauern. The Goldbergkees and Wurtenkees are both separated into three parts covering distinct altitudinal bands. The glacier Goldbergkees flows first southwards from Sonnblick peak (3105 m) and then turns towards the north at the lower parts of the glacier. A ski resort on Wurtenkees glacier indicates direct anthropogenic influence at this site through artificial snow production, for instance. The extensive mass balance measurements of Goldbergkees and Wurtenkees were described by Auer et al. (2002), Schöner et al. (2000) and Koboltschnig et al. (2008). Studies on climate–glacier relationship for Goldbergkees and Wurtenkees benefit from the extensive meteorological observations at Sonnblick observatory back to 1886 (Hammer, 1993; Schöner et al., 2007).

Separate winter and summer mass balance records are available for the glaciers used in this study (see Figure 2). All balance measurements are expressed in metres of water equivalent (m w.e.), where bw is the area-average winter balance and bs the area-average summer balance. For all glaciers, bw refers to the period from 1 October to 30 April and bs refers to the period from 1 May to 30 September. It is, however, important to notice that mass balances data from Jamtalferner, Goldbergkees and Wurtenkees are based on observations carried out for winter and summer separately, whereas data from Hintereisferner and Vernagtferner are summer and winter mass balances reconstructed from annual net balances and available observations on winter snow cover (Escher Vetter et al., 2005; Prinz, 2007). Consequently, data quality varies for the different glaciers used in this study.

Figure 2.

Measured winter (top) and summer (bottom) mass balance for the glaciers Jamtalferner (JTF), Hintereisferner (HEF), Vernagtferner (VNF), Goldbergkees (GOK) and Wurtenkees (WUK)

The glaciers used in this study are monitored by a number of different institutions (see the acknowledgements). However, there are no details on hand regarding the accuracy of the measurements. From the experience of high quality mass balance measurements carried out at Goldbergkees and Wurtenkees the error of the ablation measurements can be quantified to ± 0.05 m w.e. and also ± 0.05 m w.e. for accumulation measurements. Taking into account randomness of the error and the uncertainty inherent to the spatial interpolation from point measurements onto the entire glacier area (Kaser et al., 2006) a total error of about ± 0.10 m w.e. has to be assumed for a proper uncertainty estimate. Figure 2 shows the winter and summer balances for the glaciers under investigation.

2.2. Climate data

In this study, the NCEP/NCAR reanalysis data and two GHG emission change scenarios [A1B and B1 (IPCC 2001)] which extend until the end of the 21st century which are used. This climate data is used in a monthly resolution. The large-scale scenarios are realized with the ECHAM5-EMPIOM–GCM (Roeckner et al., 2006a, 2006b, 2006c). For each of these scenarios, three GCM simulations are analysed to assess the potential bandwidth to which extent a possible future release of GHG into the atmosphere may take place.

The atmospheric parameters—thickness (500–850 hPa difference of geopotential height), specific humidity (SPCH; average of 700 and 850 hPa) and mean sea level pressure (SLP) have been selected for the geographical region from 70°N/50°W to 30°N/40°E. SPCH is spatially averaged as source of thermodynamic flow conditions around the height of mountain summits where the glaciers are located. Thickness is used as a parameter characterizing the air temperature. These large-scale predictors are used to set-up the connection between precipitation, temperature, flow conditions and glacial mass balances.

The atmospheric parameters are given on a rather large geographical region with a horizontal grid distance from 2.5° × 2.5° (approx. 250 × 250 km). The region therefore includes 629 grid points and as such reasonably enough information which is necessary to capture the temporal integrated atmospheric processes over a month or a season.

3. Methods

Shea and Marshall (2007) conclude that circulation indices can be used for modelling glacier mass balances. However, their conclusions are based only on a few grid points of reanalysis data. This appears reasonable for the recent past as reanalysis data are generated by assimilating the best observational knowledge of the atmospheric state. Moreover, there are sophisticated methods applied to reflect the physics of the atmosphere and the observations in greater detail. In this respect, data assimilation has become an important tool in numerical weather prediction and the generation of reanalysis data (Compo et al., 2006).

However, using the output fields of GCMs it is necessary to take into account vast geographical sectors and as such large numbers of grid points. In other words, spatial scales where GCMs produce meaningful results (von Storch et al., 1993) are rather large involving several hundreds of grid points. Besides the geographical extent, it is also of vital importance that the main centres of variability potentially exerting an influence on the glacier mass balance variability are included.

These inclusion of the main centres of variability were the reason for Matulla et al. (2009) to extend Shea and Marshall's (2007) idea to larger regions rather than to use just a small number of grid points. One way to link hundreds of time series at the grid points (frequently named a ‘field’) to local scale series (here, the glaciers' mass balances) is based on the so-called Empirical Orthogonal Function (EOF) analysis (von Storch and Zwiers, 1999). EOFs describe areas that display much variability and they are constructed in a way that a few of them are sufficient to cover a reasonable fraction of the field's total variability. The appendant time series are then entered into a multiple linear regression (MLR) models to simulate the evolution of the winter and summer mass balances. This approach is well established in climate research. In MLR models, one variable is predicted by various influencing variables. In our case, the predicted variable is the glacier mass balance and the influencing variables are the PCs of the atmospheric parameters. In the calibration time, the measured mass balance is used to estimate the regression coefficients of the atmospheric parameters from the NCEP/NCAR reanalysis data. With these estimated coefficients and the atmospheric parameters of ECHAM5-EMPIOM-GCM the possible mass balance until 2100 is estimated.

In order to avoid correlations that are only due to the same sign of the trend in the mass balance series and the time coefficients, we have detrended and standardized the mass balance data of the Austrian glaciers and standardized the atmospheric parameters.

This method has been used in Matulla et al. (2009) for Peyto Glacier in Canada as well. Furthermore, in this study here, a set of different MLR models were tested in a temporal cross-validation set-up to identify the best performing model, i.e. those models with combinations of predictors that show highest correlations with observed mass balances. In addition to different atmospheric fields taken into account, the empirical models vary regarding the number of time coefficients used and the months included for winter and summer (for instance, November–March or December–February). To prevent the models from being over-fitted, we took into account only three time coefficients in case of SLP and thickness and four time coefficients in case of SPCH. Concerning winter (summer) balances at least months from December to February (July–August) were regarded.

The performances of the different models are measured by the Pearson's correlation coefficient and the reduction of error (RE) (Table II). RE compares the root mean squared error (RMSE) of the downscaling model with the climatological mean. The range of RE values is between infinity and 1. Values close to 1 indicate a perfect reconstruction, values close to 0 indicate that the reconstructions are as good as the climatological mean and negative values indicate reconstructions are even worse than using the climatological mean. Based on these skill measures the best performing models are then used to generate scenarios for winter and summer balances until 2100.

Table II. Statistical indices obtained from the validation experiments of the calibration time with NCEP/NCAR reanalysis data
GlacierBalanceCross validationValidation—all years
  RR2RERR2RE
  1. Correlation coefficient R, simulated variance R2 and reduction of error RE for winter (bw) and summer balance (bs).

JTFbw0.580.340.330.750.560.56
JTFbs0.550.300.290.740.550.54
HEFbw0.340.120.100.470.220.21
HEFbs0.530.280.230.630.400.30
VNFbw0.530.280.230.640.410.41
VNFbs0.320.100.090.480.230.22
GOKbw0.340.110.090.560.320.32
GOKbs0.710.500.480.880.770.77
WUKbw0.280.080.070.460.210.20
WUKbs0.640.410.410.750.560.55

4. Results

Table II contains statistical measures characterizing the performance of the downscaling models. Two validation approaches have been carried out to assess the skill of the MLR-based regression technique to estimate the intraseasonal mass balance variability of the glaciers. Due to the short time-interval of the mass balance measurements, a temporal cross-validation was conducted (cf. Michaelson, 1997). For the generation of future estimates the models showing best results in the cross-validation period were selected and recalculated based on all measurements (‘best guess models’). The skill of these models is shown in the three rightmost columns of Table II.

Figure 3 shows the three leading EOFs of winters' (DJF) monthly mean SLP, relative topography (RELTOP) and SPCH over the North Atlantic and Europe. The EOFs were also calculated for summer. They show a rather similar pattern (not shown).

Figure 3.

First three EOF patterns of sea level pressure (top row), thickness (middle row) and specific humidity (bottom row) over Europe and the Atlantic Ocean for the winter season from December until February

In winter, the leading EOFs for each variable simulate about 62, 49 and 79% of the total variance regarding RELTOP, SPCH and SLP, respectively. The leading SLP EOF resembles the basic structure of the NAO pattern (Lamb and Peppler, 1987; Hurrell, 1995). The second pattern shows a positive SLP anomaly over the British Isles. Albeit the EOFs simulate a different amount of variability it seems that they are both connected to the NAO which perhaps makes them span a degenerated subspace that accounts for a rather large fraction of variability. The third SLP EOF indicates a longitudinally oriented dipole leading to a meridional-oriented flow.

The first RELTOP EOF pattern (second row, first column of Figure 3) shows a pronounced dipole structure between the European continent and the North Atlantic Ocean that stands for a warm North Atlantic versus a cold Europe. This may be related to some extent to a meridional flow as reflected by the third SLP EOF. The second RELTOP EOF depicts a south–north gradient of temperature indicating a zonal-oriented frontal band that reasons a warm continent. This is to be seen in the third EOF too, but combined with a blocking situation across Europe. The patterns are in accordance to those shown in Matulla et al. 2002, who discussed geopotential height in 500 hPa and temperature in 850 hPa.

As can be seen in Figure 3 (third row) SPCH is far more variable in space and time than the temperature field or SLP. As such, the space spanned by the first three SPCH EOFs represents less variability than those in case of the other large-scale fields. However, the first two patterns indicate a south–north, west–east dipole structure reflecting a separation of the SPCH field into two states of the atmosphere that may be connected to the positive and negative phases of the NAO.

5. Discussion and conclusion

The Alpine region in Europe is to a large degree influenced by synoptic scale variability coming from changes in large-scale hemispheric Rossby waves and the appendant position of the high tropospheric jet stream. The regional manifestation of these hemispheric scale waves positioning and strength of high and low-pressure systems caused the resulting advection of air masses into the alpine region. Under the influence of high-pressure systems during summer, the Alpine region is influenced by descending air masses in conjunction with a drying of the air, few clouds, only little precipitation, and a high solar energy input resulting in above-normal temperatures. So high-pressure systems cause glaciers' mass balances to be strongly negative. Particularly during late summer, when the ice of the ablation area is directly exposed to the shortwave radiation and the snow that covers the accumulation area is already old and filthy the reduced albedo prompts high decreases. During winter, however, high pressure over Eastern Europe is usually connected with the advection of cold and dry air masses.

Low-pressure systems may also affect the Alpine region quite differently. During winter, for example, low-pressure systems over the British Isles and the North Sea are connected with a southerly advection of air that is warm and humid. A more easterly centred position of the low-pressure systems over Eastern Europe does, however, often lead to the advection of cold and humid air masses stemming from Polar Regions especially into the northern Alpine region. In such situations, the Alpine region receives more clouds than on average and precipitation, little incoming solar radiation and low temperatures that give rise to positive mass balances as they mean mass growth by snow and enhanced albedo as well as less insolation.

If we want to discuss the EOF patterns from a synoptic-climatological standpoint it is important to note that the first and third EOFs of SLP and RELTOP do not coincide. The first EOF of SLP is rather similar to the third EOF of RELTOP and vice versa, whereas the second EOFs seem to coincide rather well. This is due to the independent determination of the different parameters. Dealing with EOF patterns one should keep in mind that the sign of the field does not matter. Hence, a more meridional flow direction as derived from the MSLP field, foe example, in EOF 3 means either northerly or southerly wind in the Alpine area.

To sum up, the position, strength of high and low-pressure systems over the North Atlantic European region and the timing in the seasonal cycle controls the character of air masses advected into the alpine region. Thus, besides changes in the mean background temperatures these changes related to alterations in the atmospheric circulation are important for analysing changes in Alpine glaciers mass balances.

In the following, the focus of the analysis is placed on the winter season, because the modelling of winter balances is more difficult than the simulation of summer balances (see Table II). Therefore, results based on the winter season may be seen as benchmarks regarding the skill of the downscaling models for the summer season. The summer mass balance is, besides others, mainly controlled by the temperature field showing a comparatively homogenous pattern. Winter balances on the other hand are dominated by SPCH—a rather inhomogeneous variable showing a great amount of localized variability.

For Jamtalferner and Vernagtferner, which are located north of the main Alpine chain the performance of the models for winter is better than those of summer. This may be related to large-scale synoptic processes advecting moist air masses from the North Sea towards Central Europe. Due to orographically induced uplift processes this leads to precipitation within those regions.

The difference in model skills (see Table II) amongst the models of adjacent glaciers, as for instance the glaciers Hintereisferner and Vernagtferner, is mainly controlled by the impact of synoptic weather patterns differently influencing adjacent glaciers. For example, in case of the glaciers Hintereisferner and Vernagtferner the weather patterns may be those describing the influence of southern air masses that affect Hintereisferner, but not Vernagtferner. For Hintereisferner it is known that the glacier gets a significant amount of precipitation (especially towards higher altitudes) from air masses coming from south of the Alps, whereas, the Vernagtferner receives precipitation mainly by the advection of northern air masses. Besides the difference regarding the dynamical synoptic situation, the two glaciers have accumulations and ablations areas of different orientation. Moreover, Hintereisferner flows about 400 m deeper down into the valley than Vernagtferner.

The model skill at Goldbergkees and Wurtenkees are quite similar. Winter balances at the Goldbergkees glacier are reproduced somewhat better by the downscaling models than those of Wurtenkees. Table III displays for each glacier the months taken into account to simulate the appendant balances, which may differ among the glaciers, as discussed above. Moreover, Table III also displays the atmospheric predictors used as well as the number of time coefficients that are entered into the best performing downscaling models.

Table III. Considered atmospheric parameters and seasons—left for winter balances and right for summer balances—in the best performing models for all glaciers
GlacierMonthsPCsGlacierMonthsPCs
 ThickSpchMslp  ThickSpchMslp
JTF10–3013JTF6–9300
HEF10–2113HEF5–8201
VNF11–4032VNF6–9242
GOK12–3030GOK6–9320
WUK10–4010WUK6–9300

To derive the future evolution of glaciers mass balances the scenario PCs were estimated by subtracting the scenario run from the control run and by projecting the resulting anomaly patterns on the EOFs deduced from the reanalysis data entering the MLR models. This approach allows for proper results as long as the circulation–climate relationship between the synoptic scale of the large-scale weather patterns and the glaciers mass balance derived from recent meteorological observations is constant in time, i.e. it does not change fundamentally in the future (von Storch et al., 1993). A second precondition that has to be fulfilled by the statistical downscaling model relates to the synoptic scale development that should not exit the linear space used to successfully picture the time evolution of the observed weather.

Based on the temporal cross-validation experiments the downscaling models which show highest skills were applied to simulate the winter and summer mass balances until the end of this century. Furthermore, the presented simulations are driven by different experiments (whereby, each emission scenario was realized by three simulations) carried out with the ECHAM5-MPIOM GCM forced with IPCC emission scenarios. (Roeckner et al., 2006a, 2006b, 2006c)

The use of ensemble model simulations allows for a better assessment of the bandwidth of potential future changes of glacier mass balances compared to just using one experiment per scenario. Different initial states wherefrom the GCM simulations start from give reason to diverse climate picturing evolutions. This is possible as climatic evolutions also influenced by internal links between the climate spheres of the Earth system.

For this reason, we emphasize that our results for the potential evolution of glacier mass balance in the Alps are connected to the GHG emission scenarios. The possible evolution of GHG changes might be different from one scenario run alone and thus ensembles simulations for two distinct emission scenarios were taken into account.

Figure 4 displays the temporal evolution of the winter and summer mass balances for the glaciers under investigation. Winter and summer balances show decreasing trends across all investigated scenarios. These trends are more pronounced during winter than summer. Regarding the two scenarios (A1B or B1), little differences in the mass balances are to be seen until 2050. Differences however get larger during the second half of the century under the A1B scenarios which indicate larger mass losses.

Figure 4.

Observed mass balance (black), scenario output is plotted in grey with the mean for A1B in red and B1 in green. The left column shows winter mass balance and the right column summer mass balance

In order to remove the offset between measured and scenario-based mass balance the difference between the mean of the past 5 years of the measurements and the mean of the first 5 years of the scenario-based mass balance is subtracted from the scenario-based mass balance. Consequently, the scenario runs are attached to the measurements.

Another difference between the observed summer and winter balance behaviours relate to differences in the variances of the time series. Winters' time series depict variances reduced by a factor of 10 compared to the variances of summer mass balances. A reason explaining this phenomenon relates to the fact that the year-to-year variation of winter precipitation exhibits a reduced variability compared to the one of summer solar radiation.

Another already well-documented fact in the literature is the reduction of the interannual variability between observations and scenarios. This can be understood by the reduction of variance introduced by linear models that are used to link the large scale of GCMs output and the local scale of the mass balance observations. The amount of the reduction of the variance is different for each glacier which can be seen in the observational period too (not shown here). When interpreting changes in the variability between observed and estimated variability, this fact has to be taken into account.

Approaches to overcome this problem should however not be overused, as for instance pointed out in the paper of von Storch (1999) discussing methodological shortcomings of so-called inflation method for artificially increasing the estimated amount of variability.

Summer balances at the Vernagtferner show a break around 1980 that may be addressed to a change in the employed gauges. Summer balances of Hintereisferner show a rather distinctive trend, which can also be seen across the observed time series. Winter balance scenarios do not show pronounced trends throughout the whole scenario period. The scenario for summer at Jamtalferner shows reductions of about 1000 mm until 2100. Reductions at Hintereisferner are larger; over 50% more mass loss than today. Based on our estimations, the A1B emission scenario would give reason for reductions that are twice as large as those caused by the B1 emission scenario. The Vernagtferner reaction is apparently less sensitive to changes in GHG for both summer and winter.

Glaciers Goldbergkees and Wurtenkees feature a larger variability of the summer balance observations compared to the already mentioned glaciers. These glaciers are expected to lose increasingly more mass towards the end of the 21st century.

In comparison to findings obtained for the Austrian glaciers, Peyto glacier in Alberta, Canada shows slightly increasing mass balance for the same scenarios during winter, which is however outweighed by the pronounced negative trend in summer (Matulla et al., 2009). Summer balances are rather similar to those found in this study for the Eastern Alps. The different behaviour of the winter balances may be reasoned by different distances of the glaciers to neighbouring oceans. Peyto glacier is just about 800 km away from the Pacific Ocean, whereas the Alpine glaciers are located 1200 km off the Atlantic coast. However, the Mediterranean Sea is an important source region of moisture, latent and sensible heat. This is especially important in the winter half year when low-pressure systems located south of the Alps (Genua low-pressure system) advect moist and mild air masses towards and around the main Alpine crest.

These circumstances may be explained by significantly different synoptic scale processes such as the builtd up of cyclones and the advection of moist air from the Atlantic and the Mediterranean shaped by the land and the topography which is not so much influenced in the case of the glacier Peyto but much more in the case of the Austrian glaciers.

Another crucial factor is the orientation of the mountain crest against the main airflow. The Rocky Mountains run perpendicular to the westerly stream flow. The Austrian Alps face the air masses coming from the Atlantic in parallel, which makes effects like the orographic lifting more diverse.

The downscaled scenarios show a pronounced retreat of the investigated Austrian glaciers until the end of the 21st century. During winter, the glaciers are projected to gain less mass than during the observation period. In summer, projections suggest large losses.

Taking into account the ice volume and the glaciers' areas displayed in Table IV together with the mass balances as modelled under the considered climate change scenarios the extension of the glaciers are assessed. An inherent shortcoming that comes along with the rather simple analysis regards the mass balance that is assumed to be stationary. That means that the area and the ice thickness as well as the ice flow velocity are assumed to behave as during the past. Accordingly, the glaciers stay in a dynamical equilibrium. During recent years, glaciers mass balances are negative. A retreat of the glaciers implies a shrinking of the area where the solar insolation provides energy for the melting process. For a smaller area, this means that in fact less melting as simulated by the models that are based on larger glacial areas. The same holds for accumulation, which is triggered by snow flow, firn and eventually ice that can be accumulated on a smaller area only. Based on the scenarios used here a mass balance of–500 mm is to be seen towards the end of the 21st century. At the end of the century, the glaciers shall show pronounced reduced mass and area.

Table IV. Mean thickness, maximal thickness, area (data from 2003) and ice volume of the glaciers Goldbergkees (GOK) and Wurtenkees (WUK)
GlacierMean thickness (m)Maximum thickness (m)Area (km2)Volume (km3)
  1. WUK is separated into an upper and lower part (data from Binder, 2009).

GOK42 ± 10162 ± 361.430.057 ± 0.014
Upper WUK13 ± 540 ± 150.430.006 ± 0.002
Lower WUK40 ± 879 ± 150.380.015 ± 0.003

Moreover, the debris layers on top of the glacier change both in extent and thickness. Those debris layers protect the glacier from incoming solar insolation and therefore reduce the energy balance on top of the glacier. This implies a reduced melting of glaciers since they are covered by debris layers.

Acknowledgements

The mass balance data of Vernagtferner were provided by H Escher Vetter (Commission of Glaciology, Bavarian Academy of Sciences), data of Jamtalferner and Hintereisferner by A Fischer (University of Innsbruck, Institute of Meteorology and Geophysics) and data of Goldbergkees and Wurtenkees by Dr Schöner (ZAMG). We are grateful not only for providing us with the data but also to give us several advices on data quality of mass balance observations.

Ancillary