■HHHHHi
were acquired close to two of the SAR image acquisitions. In addition, the Otztal test site was surveyed by the
NASA/JPL-AIRSAR instrument in August 1989 and June 1991. Auxiliary data available includes a high
resolution DEM obtained from 1:10000 topographic maps and extensive field measurements. For the polar ice
sheets sub-application, 10 ERS-1 SAR scenes were available between the period January and July 1992 covering
different regions of the Antarctic. No coincident optical imagery was obtained over these study areas; however,
many processes of the ice sheet dynamics reveal characteristic timescales of years to decades. Thus, a full
Landsat TM scene from 1985 of the region Heimefrontfjella / Amundsen Ice was used to investigate the
synergistic use of SAR and optical data for this application area.
For the geology / pedology application, the selected test site is located in the western part of
Mali (West Africa) covering an area of some 25 km x 25 km around the village of Diamou. Two Landsat TM
images from January 1987 and November 1992 were acquired, complemented by two ERS-1 SAR images
recorded in June and October 1992. Both Landsat TM images were rectified on the basis of 1:200000
topographic maps to a pixel size of 25 x 25 metres. The ERS-1 SAR images were both rectified to the
geocoded November 1992 TM image. The very limited amount of rainfall (between 1968 and 1985, the average
annual precipitation was only about 650 mms) restricts the agricultural activities, the one rainy season starting
in June. Due to limited ground-truth data, auxiliary data was restricted to meteorological measurements,
1:200000 geological maps (showing geological units and faultlines, rock type, etc.), 1:50000 aerial photographs
and soil and sub-soil samples.
3 - INVESTIGATION OF CONTEMPORALITY
3.1. Vegetation Application
The vegetation application covers the two main areas of agriculture and forestry, the key parameters for which
the sampling requirements were studied being for agriculture : Leaf Area Index (‘LAI’), crop height, crop cover,
and crop structure (the Normalised Difference Vegetation Index, ‘NDVI’, is one measure of canopy
development); and for forestry, the biomass. For the optical satellite data, for agriculture, a positive correlation
overall was found between both LAI and crop height vs. NDVI. Significant changes in NDVI were observed to
arise on a timescale of approximately 14 days. By using bands 5, 7 and 9 of the ATM 1989 data, it is possible
to discriminate between all the crops of interest. For forestry, a negative correlation was found between NDVI
and age due to the decreasing proportion of green vegetation in older trees, the timescale for significant change
being of the order of 10 years.
For the microwave (Maestro 1989) agriculture data, strongest relationships between
backscatter and LAI (negative correlation) were found at C-band / HV and L-band / HH frequency / polarisations,
between backscatter and NDVI (negative correlation) at C-band / HV and L-band / HV, and between backscatter
and crop height (positive correlation) at C-band / HH and C-band / VV. However, the backscatter from
vegetation cover types is complex, and more than one factor alone will influence the response. Results from the
ERS-1 data show that with single-date imagery, it is not possible to discriminate between more than two crop
types. Better classification could be attained, however, by using multi-temporal composites of ERS-1 imagery.
Analysis of the temporal pattern of development of microwave backscatter in the ERS-1 and Agriscatt ‘87 data
sets suggests that a monthly sampling frequency should be adequate in order to observe general trends. For
forestry, as with the optical data, a fairly strong relationship was found between backscatter and age, the main
contribution being from the Scots Pine and Grand Fir stands. A positive correlation was also found between
backscatter and canopy height.
Synergistic analysis was performed based on an IHS (Intensity, Hue, Saturation)
transformation of three bands of the optical data, substituting the Intensity band by one of the microwave data
channels, and then performing the reverse transformation back to RGB (Red, Green, Blue). For agriculture,
using a combined ATM-Maestro dataset, no significant improvement in correlation was found with any of the
parameters LAI, crop height or NDVI. However, the SAR and optical images used in the combined data set
were recorded 5 weeks apart, and data from only 6 fields were included in the analysis. The synergistic sampling
interval is estimated to be within the timescale of observable change in the individual data sets, i.e. between 7
and 14 days. For forestry, using a combined SPOT 1986 - ERS-1 1992 data set, the correlations obtained for
parameters such as species, age, canopy height and biomass are significantly higher than those observed in either
of the individual data sets, with an estimated synergistic sampling requirement of once a year.
