1162 
3.2. Hydrology Application 
For the hydrology application, the chosen sub-application is évapotranspiration, the key parameters of the 
évapotranspiration model selected for investigation being LAI, soil moisture and canopy height. For LAI, the 
most important of the selected parameters in the évapotranspiration model, a positive correlation was determined 
with SAR amplitude during the senescence period of the canopy, the closest relationship being observed for 
wheat and oats. In the turnip rape and wheat canopies, LAI was observed to double within two weeks in June. 
A strong positive relationship was also found between LAI and AVHRR-derived NDVI values. 
No correlation was observed between SAR amplitude and soil moisture, though this is based 
on SAR measurements from only 3 dates and soil moisture measurements only from the barley field. A 
negative correlation was measured between NDVI and soil moisture, perhaps due to reduced surface reflectance 
arising from moisture in the vegetation. 
A positive correlation was found between SAR amplitude and canopy height during the 
growing period, though some of this correlation may be due to the influence of LAI. No relationship was found 
between Landsat TM intensities and canopy height for the different crop species. 
3.3. Cryosphere Application 
For the cryosphere application, three sub-application areas were selected : seasonal snow cover, mountain 
glaciers and polar ice sheets. In analysing the synergistic content of the data for snow and ice applications, 
inversion of the remote sensing data is first performed independently for each sensor. Final products may then 
be derived either by merging the classification results of the two sensor types, or by a re-analysis of the results 
for one sensor under consideration of information derived from the other sensor. 
The mapping of snow and glacier areas is based on Landsat TM and AIRS AR data acquired at 
the ôtztal test site. The classification of the TM data was based on thresholding the ratio of the surface 
reflectivities in TM bands 3 and 5 after elimination of pixels ‘contaminated’ by cloud. The classification with 
microwave data was performed with AIRSAR C-band data acquired in August 1989,6 days prior to the Landsat 
acquisition. The data was geocoded using digital elevation data to eliminate terrain-induced distortions. For the 
classification of snow and ice areas in complex terrain using microwave data, it is necessary to compare the real 
SAR image with a simulated image to compensate for the angular variations of backscatter due to topography. 
If the simulated image is generated with backscatter functions for snow-free terrain, a threshold function can be 
applied to separate snow-covered and snow-free areas (N.B. if multi-temporal SAR data are available, the 
threshold function can be applied to the ratio of a snow-covered to a snow-free image). 
Comparing the classification results obtained from the TM and AIRSAR data for the same 
classification types with the field data, the TM results are accurate to 1 % in each of the 3 classes (snow, glacier 
ice and snow / ice free). An overall agreement of 71.7 % was found on a pixel-by-pixel basis between the TM- 
and AIRSAR-derived classification, mis-classification by the SAR data being primarily attributable to : very 
rough ice-surfaces (such as ice-falls) being mis-classified as rock; smooth ice surfaces in the upper parts of 
ablation areas being mis-classified as snow; unusually rough snow-surfaces below 3000 metres being mis- 
classified as ice. The results are improved (to give a pixel-by-pixel correspondence of 82.5 %) by including a 
layover mask of glacier area derived from the TM data in the AIRSAR classification. 
From an analysis of the temporal dynamics of snow cover from the ERS-1 and TM data for 
the ôtztal test site, the average retreat of snow extent in ice-free mountain areas was determined to be 1.5 % per 
day at the start of the melt period. When less snow cover is left, the rate of the retreat slows. For the 
Hintereisfemer glacier, the average retreat of snow cover was 0.8% of glacier area per day during the unusually 
warm period in July and August 1992, and from TM data of a mountain basin region (Fotsch) in the Tyrol 
recorded in 1984 a mean decrease of snow extent of 1.4 % of basin area per day was obtained for the period 
between April and May, and 0.9 % of basin area per day between May and July. Weekly SSM/I multi-spectral 
radiometer measurements of Central and Eastern Europe recorded in the winter of 1988 suggest that much faster 
temporal dynamics of snow extent arise in lowland areas, emphasizing the requirement for short observation 
intervals. 
Finally, as an example of the inversion of SAR data in respect to ice sheet properties, the 
motion of the ice edge was determined from two ERS-1 images of the Ekstroem Ice Shelf, Antarctica, recorded 
in February and July 1992. The two images were co-registered by means of control points from parts of the 
image where no ice motion is expected (e.g. where the ice is resting on rock). The maximum shift of the ice 
edge between the two dates was found to be 100 metres.