Principal Investigator: H. Rott, Univ. of Innsbruck, Austria

Co-Investigator: C. Mätzler, Univ. Bern, Switzerland

Collaborators: D.-M. Floricioiu, T. Nagler, A. Siegel, Univ. Innsbruck

OBJECTIVES

Evaluate the potential of multiparameter SAR for environmental monitoring in high alpine regions with emphasis on hydrological and glaciological applications.

Develop techniques for the extraction of snow and glacier parameters from the spaceborne SAR data.

Derive information on mass balance and dynamics of glaciers in the Austrian Alps and on the Southern Patagonian Icefield from SIR-C/X-SAR data.

INVESTIGATIONS PERFORMED

Snow and Glacier Studies in the Test Site Ötztal, Austria

Extensive field measurements were carried out in the test site Ötztal in the Central Alps of Austria during both SIR-C/X-SAR missions, SRL-1 and SRL-2, to obtain basic data for methodological research. Fig. 1 shows a false color representation of the central part of the test site which extends over altitudes between 1890 m, near the village Vent, and 3770 m. Because of the rugged topography the illumination differences between foreslopes and backslopes are very pronounced.

In Fig. 1 the accumulation areas appear in blue to purple colors because of high backscattering in April. These areas are covering plateaus with comparatively gentle topography at altitudes between 2900 m and 3300 m a.s.l. The glacier tongues, appearing in cyan color, are descending into narrow valleys. Small coniferous forests are found near the villages Vent and Langtaufers. These forests appear in red because of high cross-polarized backscattering at L-band.

Five X-SAR images and three SIR-C images were acquired over Ötztal between 10 April and 14 April 1994 (SRL-1) and five X-SAR and SIR-C images between 1 October and 5 October 1994 (SRL-2). Field measurements were carried out during both missions on ice-free surfaces and on the glaciers Hintereisferner, Kesselwandferner, and Gepatschferner, which is the largest glacier of the Ötztaler Alpen covering 17 km². Corner reflectors were deployed, physical properties of snow, ice, and soil were measured, and meteorological data were recorded at two stations. Runoff gauges are located at Vent for the basin Rofenache, covering 98 km² in area, and just below the terminus of Vernagtferner. Ground-based scatterometer measurements were carried out at 5 GHz and 35 GHz over snow covered ground in April and at 5.3 GHz and 10.3 GHz over bare soil and alpine grassland in October.

The meteorological conditions during the two missions were quite different, though the runoff in both cases was quite low. During SRL-1 the glaciers were covered by dry winter snow, the air temperature at 3000 m was below -10°C, and snowfall was observed during several days. On the glaciers the snowpack was completely dry, whereas at elevations below about 2200 m the lower sections of the snowpack were partly humid. During SRL-2 the mean daily air temperature dropped by 12°C, resulting in significant changes of snow properties and backscattering signatures.

Figures 2 and 3 show examples of the signatures for accumulation areas and for ablation areas of the glaciers. During SRL-1 (in April) all glacier surfaces were covered by dry winter snow of 2 to 3 m depth. Because dry snow is quite transparent, the signals in the accumulation areas originate mainly from the frozen firn below the winter snow. This results in comparatively high ° at X- and C-band, for both the co- and the cross-polarized signals. During SRL-2, when the accumulation areas were covered by wet firn with a frozen crust, ° was lower and shows more pronounced incidence angle dependence. Over the ablation areas the co-polarized signals are similar in April and October. However, C-band HV is clearly higher in April due to diffuse scattering contributions from the dry snow above the ice surfaces.

Seasonal variations of X-band ° are small in the ablation areas, but strongly pronounced in the accumulation areas (Fig. 3). Part of the short-term variability is an effect of the different incidence angles of the various data takes (between 43° and 64° for the accumulation site and between 37° and 58° for the ablation site). In October the change from wet snow surfaces in DT 18 and DT 34 to humid snow with refrozen crusts in DT 46 and 78 results in an increase of ° by several dB. This example indicates high capabilities of multitemporal data for separating accumulation and ablation zones and demonstrates the sensitivity of X-band for short term variations of snow properties.

Target Classification in High Alpine Terrain

Various algorithms for target classification were investigated aiming at hydrological and glaciological applications. Glaciers are of interest as sensitive indicators of climate change and as water resources. A key glacier parameter for climate research and hydrology is the annual mass balance B_n which for Alpine glaciers is usually determined for the period 1 October to 30 September (the mass balance year). B_n is closely related to the ratio of accumulation area to ablation area at the end of the mass balance year. This ratio can be derived from spaceborne imagery. In high mountain regions optical imagery often can not be used because the firn and ice surfaces are obscured by clouds or fresh snow. SAR imagery is much less sensitive in this respect, though humid fresh snow may cause problems for separating accumulation and ablation areas. In the ice-free parts the discrimination of surfaces according to different hydrological characteristics is of interest. In the high Alpine site the following two different surface types could be discriminated: bare surfaces covered by rocks or moraines, and vegetated surfaces covered by sedges, grasses, and dwarf shrubs.

From Figs. 1 and 2 it is obvious that the angular dependence of the backscattering parameters causes difficulties for classification. Standardized angular correction terms cannot be applied, because the angular behaviour is different for each target, and even for the same target type may vary considerably in time. In general the incidence angle effects can be significantly reduced for classification purposes by using polarimetric parameters or power ratios.

Fig. 4 shows results of a classification based on polarimetric parameters from DT 14.2 of 10 April 1994. The following 5 parameters contain the main information and therefore were used for classification: C_hv/C_tot, L_hv/L_tot, L_hv/C_hv, C_hhvv, L_hhvv , where tot refers to the total power, and the last terms represent the magnitude of the HH-VV correlation coefficient at C- and L-band, respectively. X-band was found to add only limited information to the classification for April (SRL-1), but was very important for October (SRL-2). For the SRL-2 classification an 8-dimensional feature vector was used, including the 5 parameters specified above, and the ratios C_hv/X_vv, L_hh/X_vv, and L_hv/X_vv.

Supervised classification (maximum likelihood) was carried out after segmentation of glaciers and ice-free areas using high resolution optical data. These data were used because the SAR backscattering signatures of the rough, partly rock-covered ice surfaces near the glacier margins and the signatures of the pro-glacial moraines are similar. Though the test site was fully covered by fresh snow in April, the classification results (Fig. 4) are quite useful. The accumulation areas A_c, determined from the April SIR-C images, agree reasonably well with A_c derived from field work end of September 1993, in particular if uncertainties of field data are also taken into account. Field data are available for the three glaciers Hintereisferner, Kesselwandferner, Vernagtferner, showing the ratios (A_t = total glacier area) A_c/A_t: 0.48, 0.73, 0.32. The corresponding numbers from Fig. 4 are: 0.38, 0.64, 0.37.

Another option for separating accumulation and ablation areas is segmentation by means of multi-temporal ratios of backscattering (April versus October) using X-band or C-band data. In case of X-band, the threshold of the ratio for separating accumulation and ablation areas varies depending on the data take because of backscattering changes due to freezing of the top snow layer (Fig. 3). C-band is less sensitive to short-term variability.

SIR-C/X-SAR Data Analysis and Related Field Activities on the Southern Patagonian Icefield

Extended parts of the Southern Patagonian Icefield (SPI) and of the Northern Patagonian Icefield, were imaged by SIR-C/X-SAR. SPI covers approximately 13.000 km² and stretches for 350 km from 48.3° S to 51.5° S. The main accumulation zones are located on level plateaus at elevations between 1200 m and 2500 m a.s.l. The large outlet glaciers on the western side terminate with calving fronts in the Pacific fjords, the major outlet glaciers on the eastern side are calving into the big Patagonian lakes. Observations of glacier fronts indicate a general trend of glacier retreat, however, with some striking exceptions. Though SPI is the largest contiguous mid-latitude ice mass, very little is known about its glaciers due to the adverse weather conditions, the difficult access, and the large crevasse zones. Thus SIR-C/X-SAR offered the opportunity for advancing the knowledge on glaciers of SPI significantly.

Our analysis of SIR-C/X-SAR data concentrated on Viedma Glacier, which is located in the central part of SPI and covers an area of about 950 km², and on Moreno Glacier covering 250 km². During a field trip to Viedma Glacier in February 1994 ground control points were measured by means of GPS and surface characteristics were studied. An extensive field program on climatological and glaciological data is going on at Moreno Glacier which is known for repeated advances and retreats since 1917, damming up the southern arm (Brazo Sur) of Lago Argentino several times by reaching the opposite bank on Peninsula Magellanes. Since 1990 up to 1996 the glacier has been in an advanced position, touching the opposite bank but not damming the lake because of open channels through the ice. Between March 1994 and November 1996 four field campaigns have been carried out on Moreno Glacier. The field work included the installation of an automatic climate station, measurements of ice ablation and motion at 10 stakes along the traverse profile (Fig. 7) and of 10 additional stakes at lower and higher elevations, ice thickness measurements along the profile and at other points using seismic techniques, GPS measurements of glacier boundaries, echo-sounding of lake-depth in front of the glacier. Together with the interferometric SIR-C/X-SAR data, which have been acquired on 7, 9, and 10 October 1994, this represents the most complete data set ever obtained over any of the glaciers in Patagonia.

The interferometric analysis of SIR-C/X-SAR represents a most valuable supplement to the field data. Interferograms were calculated for X-, C-, and L-band data, but only the L-band data showed reasonable coherence over the glacier surfaces. Coherence was lost at X- and C-band within one day because the snow and ice surfaces were melting up to high altitudes.

One color sequence (one fringe) in the L-band interferogram (Fig. 7) corresponds to a phase shift of 2 between the two images, the fringes are superimposed to the SAR amplitude image. In the ice-free regions the interferometric fringes correspond to an altitude difference of 550 m. On the glaciers the fringes are related to the ice velocity component in direction of radar illumination (the range direction) and to altitude. According to the radar wavelength and the imaging geometry, at a given altitude one fringe corresponds to a difference in ice motion of 21.7 cm in 24 hours. The ice velocities can be derived interferometrically over the main parts of the terminus, but coherence is lost near the heavily crevassed calving front. For this area amplitude correlation was applied to derive motion vectors. The results of the motion analysis are discussed in the next section.

SIGNIFICANT RESULTS

Glacier Mass Balance and Runoff in the Alps

Glaciers are important sources of fresh water in many mountain regions. The glacier contribution to runoff depend on the climatic conditions and show considerable interannual variations. Thus in a year of positive mass balance the runoff contribution from glaciers is smaller than the precipitation falling on the glacier surfaces. During years of negative mass balance the runoff is increased due to the glacier contribution. Because negative mass balance is often linked with reduced precipitation and with increased evaporation in the ice-free zones, the glaciers are balancing the runoff regime. These aspects are of importance for water management. If the relation between the extent of the accumulation area A_c and the mass balance B_n of a glacier is known, B_n can be estimated by means of satellite imagery. The accumulation areas in Fig. 4, based on data of SRL-1, refer to the end of the mass balance year October 1992 to September 1993. According to this classification the mean ratio for all glaciers of the basin Venter Ache was A_c/A_t = 0.33. This points out that the glacier mass balance was clearly negative, because A_c/A_t = 0.7 corresponds to B_n of about zero (no change of glacier mass).

Fig. 5 shows the extent of accumulation and ablation areas derived form X-SAR data of DT 46.1, acquired on 3 October 1994. The backscattering ratio between April and October was the basis for this classifications, using the threshold of -7 dB for separating accumulation and ablation areas. The ratios A_c/A_t from Fig. 5 refer to the mass balance year 1993/94. For the drainage basin Venter Ache the mean ratio was A_c/A_t = 0.20. For the three main glaciers Hintereisferner, Kesselwandferner, Vernagtferner we obtain from the SAR classification A_c/A_t: 0.28, 0.41, 0.12, which compares well with the field data (0.31, 0.33, 0.22).

Applying the mean relation between A_c/A_t and B_n, (known from field investigations on three glaciers) to the glacier area of 40.0 km² in the basin Rofenache, we come up with the following runoff contributions due to decrease of glacier mass: 30 Million m³ for the mass balance year 1992/93 and 44 million m³ for 1993/94. Compared to the overall runoff of 137 Million m³ for 1992/93 and 177 Million m³ 1993/94 this is a considerable percentage.

Glacier Retreat in Patagonia

As the glaciers in the Alps, most of the glaciers of SPI showed significant retreat during the last decades. A pronounced sign of retreat, the outbreak of a lake dammed by Viedma Glacier, could be detected by means of SIR-C/X-SAR images from the two flights. The lake, Laguna Viedma, has been stable at least during the last century. Fig. 6 shows the lower terminus of Viedma Glacier with the ice front calving into Lago Viedma, one of the big Patagonian lakes. The comparison of SIR-C/X-SAR images from SRL-1 and SRL-2 showed that the lake area decreased from 5.5 km² in April 1994 to 1.5 km² in October 1994 and the water level lowered by about 100 m. The total drainage due to this event is estimated at 400 Million m³ of water. This outbreak of the lake, after a long period of only minor frontal retreat, indicates changes in the hydraulic regime of the glacier and possibly the start of a major retreat.

Dynamics and Mass Fluxes of Moreno Glacier

The synergistic data set of SIR-C data and field observation for Moreno Glacier enables the calculation of mass fluxes. Fig. 8 shows a map of ice motion of the glacier terminus. The range component of the velocity vector was derived from the L-band interferogram (Fig. 7) after phase unwrapping and subtraction of the topographic phase ramp. The motion close to the calving front was derived by means of amplitude correlation, using L-band data from 7 and 10 October 1994. The flowlines, derived from Landsat images and aerial photography, were required to calculate the magnitude of the velocity vector from the interferometrically derived range component. Selected flowlines are shown in Fig. 8. The interferometric analysis shows good agreement with the field data of ice motion measured over a year.

The motion analysis shows that the ice stream on the orographically right side provides the main contribution of mass to the ablation area. Near the equilibrium line, in a heavily crevassed zone, this ice stream reaches velocities up to 4 m per day, as derived from the interferogram. Based on the ice motion analysis and on seismic ice thickness measurements an annual ice transport of 0.95 x 10⁹ m³ was calculated for the transverse profile. The maximum depth near the centre of the profile is 720 m. From the motion field it has to be concluded that the ice thickness decreases considerably towards the calving terminus. About half of the ice flowing through the profile reaches the calving front, the other part is lost due to melting. The combination of SIR-C interferometry and field measurements enabled for the first time to calculate the mass fluxes of one of the main glaciers in Patagonia, a region of significant interest for global change research.

PUBLICATIONS

Mätzler C., T. Weise, T. Strozzi, D.-M. Floricioiu, H. Rott - 1996: Microwave snowpack studies in the Austrian Alps during the SIR-C/X-SAR experiment in April 1994. Inst. of Applied Physics, Univ. Bern, Research Report No. 96-3: 38 pp.

Mätzler, C., T. Weise, T. Strozzi, D.-M. Floricioiu, H. Rott - 1997: Microwave snowpack studies during the SIR-C/X-SAR experiment. Int. J. of Remote Sensing, in press.

Nagler T. - 1996: Methods and analysis of synthetic aperture radar (SAR) Data form ERS-1 and X-SAR for snow and glacier applications. Ph. D. thesis, Univ. of Innsbruck.

Rott H.: Glacier studies by means of SIR-C/X-SAR. In: The X-SAR Picture Book, W. Noack (Editor), Springer Verlag, in press.

Rott H., T. Nagler and D.-M. Floricioiu - 1996: Snow and glacier parameters derived from single channel and multi-parameter SAR. Proc. of Int. Symp. on Retrieval of Bio- and Geophysical Parameters from SAR Data for Land Applications, CNES, Toulouse: 479-488.

Skvarca P., H. Rott and M. Stuefer - 1995: Synergy of ERS-1 SAR, X-SAR, Landsat TM imagery and aerial photography for glaciological studies of Viedma Glacier, southern Patagonia. Proceedings, VII Simposio Latinoamericano de Percepciòn Remota, SELPER, Puerto Vallarta, México, Nov. 1995: 674-682.

FIGURES

Figure 1: Geocoded, terrain corrected SIR-C image, DT 46. of the test site Ötztal, Austria, channels Lhv (SRL-2) in red, Chh (SRL-2) in green and Chh (SRL-1) in blue. The yellow line corresponds to the boundary of the drainage basin Rofenache. Glaciers: H-Hintereisferner, G-Gepatschferner, K-Kesselwandferner, V-Vernagtferner, R-the runoff gauge near Vent, L-Langtaufers.

Figure 2: Angular dependence of backscattering at C-band VV and HV polarizations in dependence of the local incidence angle for the accumulation and ablation areas on the glaciers, from SIR-C data of 12 April 1994 (SRL-1) and 3 October 1994 (SRL-2).

Figure 3: Seasonal and short-term variations of s⁰ from X-SAR data for sites in the accumulation area and in the ablation area.

Figure 4: Thematic map of the test site Ötztal, based on classification of SIR-C data, DT 14.2, 10 April 1994. White - accumulation areas, blue - ablation areas, grey - bare rock and soil, green - vegetation, black - layover and foreshortening.

Figure 5: Extent of accumulation areas (white) and ablation areas (blue) derived from X-SAR data of DT 46.0, 3 October 1994; black - layover and shadow.

Figure 6: Section of X-SAR image, DT 13.06, from 10 April 1994 (in red) and 1 October 1994 (in blue) showing the terminal part of Viedma Glacier. The green line represents the boundary of Laguna Viedma on 10 April 1994, the magenta line on 1 October 1994.

Figure7: Interferogram of the Moreno Glacier region, based on SIR-C data in L-band from 9 and 10 October 1994. The dotted line across the terminus corresponds to the transverse profile of ice thickness

and motion. B.S. - Brazo Sur; L.A.- Lago Argentino.

Figure 8: Magnitude of the velocity vector (in colors) on the terminus of Moreno Glacier, derived by means of SIR-C interferometry and amplitude correlation. The lines of flow direction were analyzed from optical images.