Introduction
Glaciers have attracted increasing scientific and public interest as indicators of climate change within the last decade. In many countries, they also play an important economic role (e.g. for tourism (glacier-skiing resorts, mountaineering), for agriculture (irrigation) and for the production of hydropower). Guidelines for compiling a world glacier inventory (WGI), to assess their extent and volume, were published in the 1970s (UNESCO/IASH, 1970; Reference Müller, Caflisch and MüllerMüller and others, 1977). The first Austrian inventory, based on data from 1969, was published by Reference PatzeltPatzelt (1978, Reference Patzelt1980) and Reference Geist, Elvehøy, Jackson and StötterGross (1987). The first Italian inventory was completed in 1983, following an earlier inventory by the Consiglio Nazionale delle Ricerche (e.g. CNR, 1962). Analysis of glacier-length changes worldwide (Reference Gross, Kerschner and PatzeltHoelzle and others, 2003; Reference OerlemansOerlemans, 2005), and recently for the Italian Alps (Reference CitterioCitterio and others, 2007), provides valuable insights into glacier evolution during the past century.
National inventories have been compiled for Austria, based on aerial photos (Reference KuhnLambrecht and Kuhn, 2007), and for Switzerland, based on Landsat Thematic Mapper satellite imagery (Reference Lutz, Geist and StötterMaisch and others, 2000; Reference Hopkinson and DemuthKääb and others, 2002; PReference Paul, Kääb, Maisch, Kellenberger and Haeberliaul and others, 2004, Reference Paul, Kääb and Haeberli2007; Reference Hoelzle, Haeberli, Dischl and PeschkeHoelzle and others, 2007). For South Tyrol, Italy, there are three different glacier inventories: the first, for 1983, was integrated into the WGI database as a table; the second, for 1997, was compiled from aerial photographs; while the most recent inventory, for 2006, is based on airborne laser-scanner (ALS) data and digital orthophotos.
During the last decade, ALS has become one of the standard methods for acquiring topographic data for a variety of applications (Reference Arnold, Rees, Devereux and AmableArnold and others, 2006). The method is characterized by a high degree of automation, in terms of data acquisition and computer-aided analysis, in open-source software (e.g. with GRASS GIS (Geographic Resources Analysis Support System), or Windows-based software such as ArcGIS) (Reference Garvin and WilliamsGeist and Stötter, 2003). Its use is now quite common in glaciology, but still has limitations (Reference Kääb, Paul, Maisch, Hoelzle and HaeberliKennett and Eiken, 1997; Reference EchelmeyerFavey and others, 1999; Reference Garvin and WilliamsGeist and Stötter, 2003; Reference Lippert, Wastl, Stötter, Moran and GeitnerLutz and others, 2003; Reference Höfle, Geist, Rutzinger and PfeiferHopkinson and Demuth, 2006; Reference KoddeKodde and others, 2007). Laser-profiling techniques have been applied in Greenland (Reference Favey, Wehr, Geiger and KahleGarvin and Williams, 1993; Reference Bindschadler, Fahnestock and SigmundBindschadler and others, 1999; Reference AbdalatiAbdalati and others, 2002; Reference Kodde, Pfeifer, Gorte, Geist and HöfleKrabill and others, 2002) and in Alaska, Canada and the Canadian Arctic (Reference Di Lullo, Dinale, Peterlin, Mair and PrinzEchelmeyer and others, 1996; Reference Aðalgeirsdόttir, Echelmeyer and HarrisonAðalgeirsdόttir and others, 1998; Reference Schiefer, Menounos and WheateSapiano and others, 1998; Reference AbdalatiAbdalati and others, 2004; Reference Sithole and VosselmanSchiefer and others, 2008). In Europe, various attempts have been made to utilize ALS for glaciological purposes (Reference Baltsavias, Favey, Bauder, Bösch and PaterakiBaltsavias and others, 2001; Reference Favey, Geiger, Gudmundsson and WehrFavey and others, 2002; Reference Garvin and WilliamsGeist and Stötter, 2003 Reference Geist and Stötter2007; Reference Lippert, Wastl, Stötter, Moran and GeitnerLutz and others, 2003; Sithole and Vosselman, 2004; Reference Würländer, Eder and GeistWürländer and others, 2004; Reference Geist and StötterGeist and others, 2005; Reference Lie, Dahl and NesjeLippert and others 2006; Reference Hoelzle, Chinn, Stumm, Paul, Zemp and HaeberliHöfle and others, 2007).
ALS is a useful and powerful tool that facilitates the compilation of recent and historical glacier topographies, enabling the spatial extent of glaciers to be monitored semi-automatically with its raster data which can be classified into ‘glacier’ and ‘non-glacier’ area classes (Reference KoddeKodde and others, 2007). The single parameters used to derive glacier topographies are topographical smoothness, connectivity, hydrological constraints as well as slope, aspect and the size of individual glaciers. The outlines were calculated from ALS data using Reference KoddeKodde and others‚ (2007) delineation algorithm. The necessary data were acquired for the whole area of the Autonomous Province of Bolzano, South Tyrol, Italy, in 2005.
Data Acquisition and Data Processing
All glaciers in South Tyrol (Fig. 1) were included in the topographic survey used to compile the ALS digital elevation model (DEM), which satisfied the requirement that all the glaciers be captured at the end of the 2005 ablation season. By verifying outlines from the delineation algorithm on orthophotos from 2006, three glaciers previously unaccounted for were found in the Ortler–Cevedale group, resulting in a total glacier cover for South Tyrol of 94.1 km2 for 2006. The South Tyrolean glaciers are in high mountain regions, creating certain problems for the application of ALS. The rugged topography, with large differences in relief, can limit laser range, point density and spectral resolution (Reference Lippert, Wastl, Stötter, Moran and GeitnerLutz and others, 2003; Reference AlbertzAlbertz, 2007), but this problem was overcome using state-of-the-art scanning equipment and by verifying the resulting glacier outlines on the 2006 digital orthophotos. This study comprises the entire glacierized area.
Fig. 1. South Tyrol and the ice-covered area in 2006, including the names of the mountain ranges.
ALS-based digital elevation model
ALS data for the whole South Tyrolean territory (Fig. 1) were obtained by the Compagnia Generale Ripreseaeree (CGR), of Parma, under contract from the Autonomous Province of Bozen, South Tyrol. The initial ALS point dataset was acquired using ‘Optech ALTM 3033’ and ‘TopoSys Falcon II’ scanner systems during late summer 2005. Specifications and recording accuracies of the two scanner systems used are shown in Table 1. Basic information on the ALS data acquisition was recorded during the flight with an integrated global positioning system (GPS) and an inertial measurement unit (IMU). The flight took place at the end of summer 2005; it provided an elevation accuracy of ±0.4m below 2000ma.s.l. and ±0.55m above 2000ma.s.l., giving sufficient accuracy for further calculations. An average point density of 8 points per 25 m2 was achieved for areas below 2000ma.s.l., and 3 points per 25 m2 for areas above 2000ma.s.l. The vertical accuracy over a control area was 0.095 m (G. Zanvettor and others, http://www.provinz.bz.it/raumordnung/download/Beschreibg_Kartogr_2006.pdf).Full information on the collected data, i.e. first pulse, last pulse and intensity, is stored in a separate dataset that has not been released.
Table 1. Area changes for all mountain ranges, 1983–2006
The ALS DEM was calculated from ALS point data (Reference Wack and StelzlWack and Stelzl, 2005); it has a 2.5 m raster resolution. The raster data form the input for the updated algorithms presented below Reference Kodde(Kodde and others, 2007). As no orthophotos were available for the recording year of the DEM, those from 2006 were used to correct the glacier outlines.
Orthophotos
The CGR carried out survey flights during the summer of 2006 to obtain digital aerial photographs of the entire South Tyrolean area using the airborne digital sensor LH Systems ADS40. The resulting orthophotos were compiled at a scale of 1 : 10 000 from scanning strips with a one-pixel ground resolution (pixel length = 0.5 m). All photographs were taken in natural colour with a charge-coupled device (CCD) digital camera. The planimetric accuracy, as stated by the company, is ±2m for elevations below 2000 ma.s.l. and ±6m above 2000ma.s.l. Basic information for these flights was also recorded with a GPS and an IMU. The georeferencing is based on a DEM with 40m raster resolution.
Calculation of glacier outlines
The southern Stubai Mountains, in the north of South Tyrol, were chosen for testing the original delineation algorithm by Reference KoddeKodde and others (2007) for calculating glacier outlines based on the ALS DEM. Several different types of glacier (e.g. valley, mountain and glacierets) can be found there. Individual glaciers were delineated on the basis of smoothness, connectivity and hydrology. In the original delineation rule described in Reference Kennett and EikenKodde (2006), the laser intensity map was used as well. After checking the resulting delineations (Fig. 2), some improvements were made in the operation of this calculation rule, as the original method was developed for a completely different dataset based on different raster resolutions.
Fig. 2. Work of delineation algorithm (modified after Kodde and others, 2007).
Improvement to Kodde’s algorithm is based on: a thorough testing of several different raster datasets as input data (1, 2.5 and 5m raster resolution); the effects of the different raster resolutions; adaptation of default values (threshold and window size) to the 2.5 m ALS raster DEM or even bigger; enlargement of the script for calculating all possible glaciers, not just the largest one; and setting of the minimum glacier area to 1 ha. In a second step, the improved algorithm was used to identify all possible glaciers in South Tyrol. The main difference between the results of the two methods is that the accuracy for the new algorithm is limited by the raster size and resolution of the DEM, but this loss of accuracy could be corrected using digital orthophotos.
Execution of the delineation algorithm
The glacier delineations were determined with GRASS GIS. Initially, the raster datasets of the DEM were imported into the GRASS database, and the regions of the raster data were set for each raster tile to improve calculation speed.
The results of the improved algorithm are good, but not accurate enough for the mostly small South Tyrolean glaciers, 59.8% of which are smaller than 0.1 km2. The method also has limitations in the highly debris-covered ablation area of Suldenferner, Ortler–Cevedale Group. The identification of the slightly debris-covered parts of glaciers was done on the basis of the surface shape and roughness of the ALS DEM; for verification purposes orthophotos and a hillshade of the ALS DEM were used. For the extensively debris-covered tongue of Suldenferner, orthophotos from 2000, 2003 and 2006 were used to define the glacier tongue boundary. An estimation of the maximum error of the delineation algorithm in a sample of ten glaciers (two glaciers of each area class, analogous to Reference Paul, Kääb, Maisch, Kellenberger and HaeberliPaul and others (2002)) resulted in an average maximum surface error of ±3.7%. Because of the small size of most of the South Tyrolean glaciers (mean of 0.39 km2), the results of the derivations were verified and corrected with the 2006 digital orthophotos. The corrected glacier boundaries were transformed to the ESRI polygon shapefile format, and ArcGIS was used to calculate additional information such as drainage area, identification number, glacier name, glacier area, aspect of the ablation area, etc. Once the glacier boundaries had been determined, additional information such as area–altitude distributions, minimum, maximum and mean elevation of the glaciers was derived.
Comparing different inventory datasets
The data from the 1997 inventory are based on shapefiles for the glacier outlines and a DEM with a 20 m raster resolution from 1999 with a vertical accuracy of ±5m, calculated from aerial photogrammetry data (G. Zanvettor and others, http:// www.provinz.bz.it/raumordnung/download/Beschreibg_Kartogr_2006.pdf). The ALS-based DEM from 2005 has a resolution of 2.5 m. In order to be able to compare the two different raster datasets, to assess the relative volume change, it was necessary to resample the 2.5 m 2005 DEM with a bilinear interpolation to match the coarser 20 m 1999 DEM. Subsequently, the resampled 2005 DEM was subtracted from the original 1999 DEM, resulting in a new raster with elevation changes. The extent of these changes depends on slope and aspect, as well as the flight direction of the ALS DEM Reference Krabill(Kraus, 2005). A mean elevation difference of 5.7m was calculated for all raster cells, both glacier and non-glacier, in the area of investigation. However, the most significant differences, attributable to inaccuracies in the 1999 DEM as well as in the 2005 DEM, occur in steep, rocky regions with a highly heterogeneous relief Reference Krabill(Kraus, 2005). Such abrupt changes in elevation are negligible for the comparatively flat glacierized areas in South Tyrol.
Results
Changes in glacier area
Three different inventories, from 1983, 1997 and 2006, have been analyzed. Results show a reduction in the glacierized area between 1983 and 2006 from 136.6 km2 to 93.4 km2 (31.6%). From 1983 to 1997 the mean reduction was 19.7%, and from 1997 to 2006 it was 11.9%. The largest ice-covered region in South Tyrol is the Ortler–Cevedale Group, with a total glacier area of 37.5 km2 (2006). Within the investigated area, the area loss of 24.9% is the lowest for this region.
In general, mountain ranges with an extensive glacier cover in South Tyrol show different amounts of change (Table 2). The absolute area loss is greatest for the Ötztal Alps (7.9km2) and the Ortler–Cevedale Group (12.4km2). Between 1983 and 2006, 19 glaciers disappeared, but the number of glacier parts increased from 205 to 302. This phenomenon is due to the disintegration of individual glaciers, mostly in the Zillertal Alps (+37) and the Ortler–Cevedale Group (+26).
Table 2. Specification and recording accuracies of the scanner systems used
Figure 3a shows the relative shift of glaciers between area classes. In 2006 the number of glaciers <0.1km2 is 4.5 times higher than in 1983 and comprises 60% of the total glacier number, while the other area classes lost at least 34% of their glaciers. Figure 3b shows the loss of area in the corresponding area classes between the inventories of 1983 and 2006. The higher number of glaciers <0.1 km2 demonstrates the impact of glacier disintegration in South Tyrol. Many glaciers <0.5km2 have melted and are now in the <0.1km2 class, accounting for the increased glacier area in this class.
Fig. 3. Glacier number and area size change by different classes as (a) percentage change of the total number, and (b) total area (km2) between 1983 and 2006.
Changes in area–altitude distribution
An analysis of the area–altitude distribution was only possible for the 1997 and 2006 inventories; no DEM or contour lines were available for the 1983 inventory. A comparison of the area–altitude distribution for both dates (Fig. 4) shows a maximum glacierized area of 3000–3100ma.s.l. in 1997. In 2006, this maximum has shifted upwards by 100m (one elevation band) to 3100–3200m a.s.l. The largest change in glacier area, a loss of almost 8km2, is observed between 2700 and 2900ma.s.l. The slight increase in glacier area within the highest elevation band (3800ma.s.l.) is a result of an extending firn cover on the plateau of Oberer Ortlerferner. In 1997, no firn remained up to the peaks of the highest mountains, but in 2006 firn and old snow survived at the highest elevations of the Ortler–Cevedale Group. The mean glacier elevation in 2006 was 3017ma.s.l., meaning an upward shift of 39m since 1997 (2978ma.s.l.). This considerable change over 9 years underlines how the loss in area has affected almost all elevation bands, most of all at lower elevations.
Fig. 4. Area–elevation distribution and area change by 100m elevation bands between the two inventories.
Volume changes
The total change of ice and firn (volume change, △V) in the glacierized area from 1997 to 2006 is –1.037 km3 (Fig. 5). This △V is calculated from the difference between the two DEMs and is not corrected for the density of firn and glacier ice. The most extensively glacierized areas show the most volume loss. For example, in the Ortler–Cevedale Group almost 0.49 km3 of the ice and firn volume was lost (Table 3). Reductions in glacier volume show large variations, especially for small glaciers.
Fig. 5. Volume-loss–elevation distribution between the two inventories within 100m elevation bands.
Table 3. Volume change for all mountain ranges between 1997 and 2006
A combination of area and volume changes during the last 9 years resulted in a mean elevation change of all glacierized areas of –7.4m (change in ice volume 1997–2006 divided by the 1997 glacier area). This corresponds very well with the results for the Austrian Ötztal Alps further to the north for the same period, where –8.2m of mean elevation change was observed (J. Abermann and others, http://www.phys.uu.nl/~boer0160/karthaus08/lecturenotes/masschanges_oetztal_ja.pdf).
Major losses in area and volume are observed for valley glaciers like Suldenferner, Langtauferer Ferner (Ötztal Alps) and Übeltalferner (Stubai Alps), while small glaciers (<1 km2) show a large variation in losses, from a small expansion at high elevations to a complete downwasting, especially in lower ranges like the Rieserferner Group. Large valley glaciers show up to –18.0m in mean elevation change because their tongues reach very low areas with highly negative mass balances. Small glaciers are mostly situated at high altitudes, with a mean elevation of 3013 ma.s.l., and, with a few exceptions, cover only one to three elevation bands.
The elevation band with the highest absolute volume change lies 100 m above the elevation band with the highest absolute area change (2900–3000ma.s.l.) and shows that very large areas are affected by net melting. More volume than area loss is observed, because of an increasing net melt rate at lower altitudes.
Changes in ELA
The equilibrium-line altitude (ELA) on glaciers is the altitude at which annual accumulation equals annual ablation. Changes in climatic conditions can be inferred from variations in the ELA (Reference KrausKuhn, 1981; Reference Ohmura, Kasser and FunkOhmura and others, 1992; Reference Lambrecht and KuhnLie and others, 2003). Reference GrossGross and others (1977) suggested a steady-state accumulation-area ratio (AAR) of 0.67 for glaciers in the Alps for the approximation of the ELA. This approximation was used to derive the ELA here as well, despite the fact that glaciers were clearly not in a steady-state condition. Compared with measured ELAs on Langenferner and Weissbrunnferner based on mass-balance studies, the approximated ELA derived using the AAR method was significantly lower (Di Lullo and others, 2008a,Reference Di Lullo, Dinale, Mair, Oberschmied and Peterlinb). From 1997 to 2006 the calculated mean ELA for all mountain ranges in South Tyrol rose by 54 m, from 2938 to 2992ma.s.l., but there is large variability within the individual ranges. The most prominent rise of the mean ELA, 89 m, occurred in the Ortler–Cevedale Group, while the mostly southerly-exposed Zillertal Alps show a rise of only 9m (Fig. 6).
Fig. 6. ELA change between the two inventories of 1997 and 2006.
Conclusions and Perspectives
Modern Earth-observation technologies (e.g. laser-scanning or radar remote sensing) enable glacier monitoring of extensive areas with both a high spatial and temporal resolution. The new South Tyrolean glacier inventory, based on ALS data, offered an opportunity to investigate glacier changes in areas with many small glaciers (<0.1km2). It also provides a baseline for future studies of historical changes and of the development of the glacierized areas of the South Tyrolean Alps. Conclusions drawn from the two fully digital inventories of 1997 and 2006, and the table-based inventory of 1983, correlate well with the 1996 Austrian glacier inventory (Reference KuhnLambrecht and Kuhn, 2007) and the 2000 Swiss glacier inventory (Reference Lutz, Geist and StötterMaisch and others, 2000; Reference Paul, Kääb, Maisch, Kellenberger and HaeberliPaul and others, 2002; Reference Bauder, Funk and HussBauder and others, 2007). The 23 years between the 1983 and 2006 inventories have shown a strong decrease in the glacierized area and ice volumes for South Tyrol. Collapse and disintegration of many glaciers can be observed, especially in the Ortler–Cevedale Group and the Zillertal Alps, but there is also large spatial variability in the degree of loss between the individual mountain ranges. Strongly glacierized ranges, like the Ortler–Cevedale Group and the Ötztal Alps, are dominated by large glaciers (>1 km2) where area losses are below average, ranging from 24.9% to 27.5%. The largest relative area losses occur in areas with little ice cover. The Sesvenna Group lost its only glacier, and there were extensive losses in the Texel Group (60%). The loss of glacier area has affected almost all elevation bands; only the top band (3800ma.s.l.) shows a slight increase. For all other ranges, losses appear to represent the reaction of glaciers to contemporary climatic conditions that have induced net ablation to the top of the highest peaks.
Acknowledgements
The present project is funded by a PhD scholarship at the University of Innsbruck. We thank M. Kodde for help in improving the glacier delineation algorithm, and G. Kaser, M. Kuhn and the Glaciological Seminar Innsbruck for their support. We also thank S. Ommanney (scientific editor), A. Arendt and an anonymous reviewer for helpful comments.
