Abstract
In the snow cover mapping experiment described, an experiment has been carried out to determine whether wet snow in areas with well-known backscatter characteristics could be detected. Two sets of agricultural areas of three ground cover types were selected, one in a low elevation area (snow free), and the other in a higher elevation area (snow covered). The two sets were covered by an ERS-1 pass in November 1991 when the satellite was in a three day repeat cycle. The areas were covered at a time when the temperature had risen rapidly and was above 0/spl deg/C at both places. Field work was performed in the two test areas in order to determine the exact vegetation cover and tillage at the acquisition time. In addition, temperature and precipitation data were acquired for the time period of interest. Areas of same ground type covered with wet snow and without snow were compared statistically. A significant reduction in the backscatter for snow covered areas was found for the ploughed areas (5.5 dB), while the other agricultural types gave little difference. This indicates that wet snow cover could be detected when the surface is rough, but not when the surface is smooth. An additional experiment was performed to evaluate the performance of a statistical classifier for discrimination between snow-covered and snow-free areas. An accuracy of about 90% correct classification was achieved for snow-covered areas when the speckle noise was reduced by various filters.

This publication has 7 references indexed in Scilit: