Classification of land-cover and forest types using multidate ERS tandem data acquired over hilly terrain

Repeat-pass interferometry with ERS SAR satellite appears to be very useful for land cover mapping. Using 4 ERS tandem pairs, this paper studies the results of forest mapping over hilly terrain. For a given tandem pair, it appears that coherence is far better than backscatter to discriminate forest from non-forest. Using additional pairs plus seasonal backscatter change images at appropriate seasons allows one to discriminate various classes within the forest theme with satisfactory accuracy. However, it was found that local weather effects (snow, freezing) and strong slopes degrade the classification accuracy, which should be further addressed in a global classification strategy