3.1. Mass-balance measurements

Mass-balance measurements in GTN-G combine the direct glaciological method (stakes and pits: point measurements with high time resolution) with the photogrammetric/geodetic method (repeated mapping: survey of the entire glacier at lower time resolution) for optimal information and calibration. Field measurements are either carried out with extended stake/pit networks for detailed process understanding and development of numerical models, or use cost-saving approaches with reduced numbers of strategically chosen ‘index’ stakes, providing gross regional indications of change (Reference Haeberli, Hoelzle and SuterHaeberli and others, 1998).

Photogrammetrically/geodetically determined volume changes and glacier mass balances are available in the European Alps back to the late 19th century (Table 1). Uninterrupted in situ measurements since 1967 of mass balance at nine Alpine glaciers are shown in Figure 2. Near-equilibrium conditions until 1981 were followed by very strong and continued if not accelerating mass loss (0.7mw.e. a^–1; trend of increase in mass loss 0.03– 0.04 mw.e. a^–2). During the first 5 years of the 21st century, mean annual mass losses have been close to 1 mw.e. a^–1. The hot, dry summer of 2003 alone caused a record mean loss of 2.45 mw.e., roughly 50% above the previous record loss in 1998.

Fig. 2. Annual (a) and cumulative (b) mass balance of nine glaciers with uninterrupted mass-balance determination (for the entire glacier) in the European Alps (Saint-Sorlin and Sarennes, France; Gries and Silvretta, Switzerland; Careser, Italy; Hintereis, Kesselwand, Vernagt and Sonnblick, Austria). Data from WGMS. The light dashed lines show an average mass loss of nearly 18 m, or 716 mma^–1, (b) and an average increase in mass loss of 34 mma^–2 (a) since 1980. The Silvretta values may be systematically too positive and should be corrected in the near future.

Table 1. Geodetically/photogrammetrically determined secular mass balances of alpine glaciers

It must be kept in mind that climatic conditions which remain constant in time would lead to geometric adjustment of glaciers to new equilibrium states and a corresponding zero balance. Continued non-zero balances indicate ongoing climate forcing (assuming feedbacks from albedo, elevation change, debris cover or dry/wet calving do not affect mass balance for the considered glaciers), and increasing deviations from zero balances reflect accelerating change. True values for continued or accelerated forcing would have to be modelled for glacier areas remaining constant in time (cf. Reference Elsberg, Harrison, Echelmeyer and KrimmelElsberg and others, 2001; Reference Harrison, Elsberg, Cox and MarchHarrison and others, 2005), an important future possibility for distributed energy- and mass-balance models.

3.2. Mass-balance reconstructions

Mass-balance values during earlier periods must be reconstructed. Models of variable complexity are available for this purpose. Information on cumulative glacier length changes thereby plays a key role for calibration and validation. For the time since the end of the Little Ice Age around 1850, reconstructions have been based on (a) a straightforward continuity consideration applied to cumulative length changes (Reference Haeberli and HoelzleHaeberli and Hoelzle, 1995; Reference Hoelzle, Haeberli, Dischl and PeschkeHoelzle and others, 2003), and (b) neural networks using present-day mass-balance measurements together with meteorological data that reach farther back in time (Reference Steiner, Walter and ZumbühlSteiner and others, 2005). The two approaches are completely independent and both provide long-term mass-balance averages of –0.25 to –0.3mw.e. a^–1, i.e. three to four times less than the most recently observed values. The generally observed retreat of Alpine glacier tongues leaves no doubt that these secular trends and values are representative for the entire mountain chain. Backward reconstructions beyond the mid-19th century are much less homogeneous, especially with respect to temporal changes. This effect so far remains unexplained, but may relate to the fact that statistical models are calibrated with direct measurements on glaciers of already reduced size but then extrapolated for periods with considerably larger glaciers and, most likely, different sensitivities and response characteristics.

The continuity approach (a) was also applied to longer-term cumulative length changes of Grosser Aletschgletscher, Switzerland, (Reference Haeberli and HolzhauserHaeberli and Holzhauser, 2003) which was reconstructed in detail from historical documents, moraine dating and fossil trees for the past 3500 years (Reference Holzhauser, Magny and ZumbuhlHolzhauser and others, 2005). Over the past two millennia, characteristic century to half-century mass-balance averages (mean ±0.3 and maximum ±0.5mw.e. a^–1) are comparable to the loss rates since the Little Ice Age and they were far below those observed since 1981. In fact, the mass losses observed since the mid-1980s (about 0.75 mw.e. a^–1) exceed the maximum, and even double the maximum, characteristic long-term loss rates during the past two millennia (Fig. 3). Such comparisons help to confirm the spatial representativeness of half-centennial to centennial mass-balance averages with their uncertainty range or variability in sensitivity (roughly ±50% of the calculated mean) as a function of individual glacier geometry (hypsography) and local to regional differentiation of climate forcing (Reference Haeberli and HoelzleHaeberli and Hoelzle, 1995; Reference Haeberli and HolzhauserHoelzle and others, 2003).

Fig. 3. Measured and reconstructed mass balances in the European Alps for (a) the past 2000 years and (b) the past 500 years. Data from WGMS (measured thickness change 1890–2000 and mass balance 1967–2004), Reference Steiner, Walter and ZumbühlSteiner and others (2005) (neural networks; last 500 years) and Reference Haeberli and HolzhauserHaeberli and Holzhauser (2003) (continuity model; past two millennia); cf. also Tables 3 and 4 in the Appendix. The essential information on rates of mass change is reflected in the slopes of the curves; the vertical position of the curves is somewhat arbitrary.

Distributed mass-balance models taking into account local effects such as those from topography, local albedo or snow redistribution by wind and avalanches not only narrow the uncertainty by providing safer estimates of ablation at glacier margins in complex topography, but also help to explain the variable climate sensitivity of individual glaciers (e.g. Reference Klok and OerlemansKlok and Oerlemans, 2002; Reference Machguth, Eisen, Paul and HoelzleMachguth and others, 2006b; Reference Paul, Machguth, Hoelzle, Salzmann, Haeberli, Orlove, Wiegandt and LuckmanPaul and others, in press). Realistic estimation of local precipitation and accumulation patterns (cf. Reference Machguth, Paul, Hoelzle and HaeberliMachguth and others, 2006a) thereby remains the primary challenge, influencing the results considerably more than the relatively well-established calculation of the summer energy balance and the resulting ablation patterns (e.g. Reference HockHock, 2005).

3.3. Consequences of the recent mass loss

The increasingly fast mass and ‘vertical’ thickness loss clearly points to an accelerating trend in climatic forcing. The corresponding additional energy flux calculated as the latent heat of the disappearing ice (around 10Wm^–2 as an average of the past 5 years) is about twice the estimated present-day radiative forcing alone (severalWm^–2; Reference WildWild and others 2005) and most probably relates to important feedback mechanisms. The albedo feedback from disappearing firn in former accumulation areas has long-term consequences: the extreme summer 2003 essentially eliminated all remaining firn from the surface of many smaller and medium-sized glaciers. In addition, these bare surfaces strikingly darkened because of wind-blown dust from the snow-free and dry mountains around. As a consequence, the melting snow cover in springtime now uncovers a low-albedo surface, which suffers from strongly enhanced melt (Reference Paul, Machguth and KääbPaul and others, 2005).

Rapid thickness loss with limited tongue retreat also reduces driving stresses and flow within the ice. In such a situation, effects from mass-balance/altitude feedback become fundamentally important, especially for large, thick glaciers which cannot lose exposed ablation areas fast enough. Thereby, total surface lowering (integrated over time) must be compared with the mass-balance/altitude gradient. Due to glacier dynamics, local surface lowering can be considerably more rapid than local ablation (negative mass balance) alone; it represents a cumulative phenomenon which can lead to irreversible runaway effects (‘downwasting’ or even ‘collapse’ rather than ‘retreat’). Observed and reconstructed mass losses are therefore also a function of glacier size (Reference Hoelzle, Haeberli, Dischl and PeschkeHoelzle and others, 2003), with large glaciers reducing their thickness more rapidly than small ones. Morphological phenomena of downwasting (flat longitudinal and concave transversal surface profiles, abundant debris cover, collapse holes above subglacial drainage channels, lake formation) have now become visible on many glacier tongues (Fig. 4). With continued thickness losses of 1 mw.e. a^–1 or even more, the glaciers with longterm mass-balance time series may disappear within a few decades from now (cf. the experiments with distributed modelling of equilibrium-line altitudes by Reference Zemp, Haeberli, Hoelzle and PaulZemp and others (2006, Reference Zemp, Hoelzle and Haeberli2007) or with dynamically fitted mass-balance/flowmodels by Reference OerlemansOerlemans and others, 1998). This constitutes a serious challenge for long-term monitoring programmes (Reference Paul, Kääb and HaeberliPaul and others, 2007b).

Fig. 4. Roseggletscher in the eastern Swiss Alps, with a collapse hole on its thin tongue. Photo by C. Rothenbühler, summer 2003.

4.1. Assessment of total glacier volume

With the compilation of glacier inventories, treatment of entire mountain ranges became possible by combining time series measured at selected glaciers with statistical information on the whole glacier ensemble at suitable time slices. The first calculations for the European Alps as a pilot study for worldwide applications have already indicated the dramatic overall loss in glacier cover for the past century and even more so for realistic scenarios of atmospheric temperature rise during the 21st century (Reference Haeberli and HoelzleHaeberli and Hoelzle, 1995). A primary challenge with such efforts is the realistic estimation of glacier thickness and corresponding ice volumes. The often-used volume/area scaling (Reference Bahr, Meier and PeckhamBahr and others, 1997) is problematic, as it correlates a statistical variable (area) with itself (area in volume) and, hence, suppresses the large scatter (roughly 30% standard deviation for mountain glaciers) in area/thickness relations, which are statistically more reasonable.

A more process-oriented approach applies geometry-dependent driving stresses in combination with mass turnover as defined by the mass-balance gradient and the altitudinal extent of glaciers (Reference Haeberli and HoelzleHaeberli and Hoelzle, 1995). An even simpler and astonishingly elegant consideration assumes that relative volume change is closely comparable to relative area change, because mean glacier thickness (averaged for the changing glacier area) tends to remain constant (even with pronounced overall changes in glacier mass!): by calculating total volume loss from mass-balance time series and mean area during the investigated time interval, and by assuming that percentages in area loss correspond to percentages in volume loss, total volume can be directly approximated (Reference Paul, Kääb, Maisch, Kellenberger and HaeberliPaul and others, 2004). Results for the total glacier volume in the Alps from the three approaches differ by about ±20% around a mean (100–140km³ for the 1970s; Reference Zemp, Haeberli, Hoelzle and PaulZemp and others, 2006), but have their specific advantages: the area/thickness relation is the easiest to apply; the driving-stress solution enables the best understanding of individual thickness variability; while the ‘area–volume percentage’ solution may provide the safest overall estimate.

4.2. Calculation of ongoing volume changes

Best estimates for total volumes and volume changes (Table 2; cf. Reference HaeberliHaeberli and others, 2004; Reference Paul, Kääb, Maisch, Kellenberger and HaeberliPaul and others, 2004; Reference Zemp, Haeberli, Hoelzle and PaulZemp and others, 2006) show that glaciers in the European Alps lost about half their total volume (roughly 0.5% a^–1) between 1850 and around 1975, another 25% (1% a^–1) of the remaining amount between 1975 and 2000, and an additional 10–15% (2–3% a^–1) in the first 5 years of this century. The latter estimate is obtained from the mean value of the mass-balance observations at nine Alpine glaciers in combination with the new satellite-derived glacier areas from 1998/99 (Reference PaulPaul, 2004) and a simple model of calculating total glacier volume from mean thickness (Reference Maisch, Wipf, Denneler, Battaglia and BenzMaisch and others, 2000).

Table 2. Estimated ice volumes in 1850, 1975, 2000 and 2005, and changes in the European Alps. The increase in volume-change rates indicates continued or even enhanced climate forcing, but might also be influenced by the different lengths of the periods considered. Data from Reference Zemp, Haeberli, Hoelzle and PaulZemp and others (2006)

Completely new perspectives are coming up with repeated satellite-derived digital terrain models. Even though problems of resolution and accuracy must be carefully evaluated (Reference Berthier, Arnaud, Vincent and RémyBerthier and others, 2006), relative changes as an expression of variable sensitivities can be investigated and used to assess how representative mass-balance time series from selected glaciers are and how they can be best extrapolated to the entire mountain range. Preliminary results from the Swiss Alps indicate a trend towards more negative balances on large, flat glaciers than on smaller ones (Fig. 5).

Fig. 5. Changes in glacier elevation in the region of Zermatt, Swiss Alps, with Gomergletscher at the lower right, from differencing the SRTM3 DEM (resampled to 25 m) from the year 2000 and the swisstopo DEM25 L1 from about 1985. Irregular black areas indicate data voids in the SRTM DEM, which are usually located outside the glacier perimeter. The grey values denote elevation changes from –80m (darkest grey value) to +10m (brightest grey value) and are plotted in classes of half standard deviations. North is at top and image width is 38 km.

4.3. Calculation of potential future changes

Extrapolation for the future based on scenarios of atmospheric warming and relations between mass balance and temperature/precipitation at the equilibrium-line altitude (ELA) as calibrated with long-term mass-balance measurements and gridded climatologies leave no doubt that in such climate scenarios most of the glacier cover in the European Alps is likely to disappear within decades (Reference Zemp, Haeberli, Hoelzle and PaulZemp and others, 2006, Reference Zemp, Hoelzle and Haeberli2007).

Modern data from satellite images as combined with digital terrain information (Reference Kääb, Paul, Maisch, Hoelzle and HaeberliKääb and others, 2002; Reference Paul, Kääb, Maisch, Kellenberger and HaeberliPaul and others, 2002, Reference Paul, Maisch, Rothenbuühler, Hoelzle and Haeberli2007a) greatly enhance the possibilities for numerical modelling and visualization of large samples. Figure 6 is an attempt to combine hypsographic modelling with digital glacier outlines, a satellite image and a digital elevation model (DEM) to visualize a likely sequence of individual glacier decay for a larger region within the Alps. Thereby, simple geometrical rules like a steady-state accumulation-area ratio (AAR) of 0.6 and parameterizations of the ELA sensitivity to temperature change are assumed (Reference Paul, Maisch, Rothenbuühler, Hoelzle and HaeberliPaul and others, 2007a). Such dramatic changes will have strong and predominantly negative impacts on landscape evolution, water supply, erosion/sedimentation, slope stability and natural hazards. Monitoring and spatio-temporal modelling of glacier changes must help with anticipating and mitigating the consequences of such extreme deviations from our historical–empirical knowledge basis and from Holocene conditions of dynamic equilibria.

Fig. 6. Glacier extent for the Aletsch region in 1973 (black outlines) and modelled extent for six shifts of the steady-state ELA using an AAR of 0.6. The legend gives the greyscale that will disappear due to the corresponding shift in ELA. For an ELA rise of 600 m, only the darkest regions will remain. The background is a Landsat satellite image acquired at 31 August 1998. Landsat data © Eurimage/Swiss National Point of Contact for Satellite Images (NPOC).