CORRELATIONS BETWEEN X-BAND, C-BAND, AND L-BAND IMAGERY WITHIN AN AGRICULTURAL ENVIRONMENT

The similarities and differences between imagery acquired at X-, C- and L-bands are examined by calculating the correlations between X-, C-, and L-band SAR data acquired over an agricultural area near Melfort, Saskatchewan in western Canada with the CCRS SAR-580. The SAR data were acquired on two separate occasions during the growing season (July and August, 1983) with vertical polarization in both transmit and receive. It was found that the correlations between the C- and L-band and between X- and L-band data were very low indicating that the radar backscatter at the different frequencies is caused by different mechanisms. The correlation between X- and C-band data was low for grains, which have a well defined vertical structure. For canola and field peas, which are both broadleaved plants with little definite structure, there was a higher correlation between the measured backscatter values. For L-band data the primary cause of the radar backscatter is probably from the underlying soil and thus largely influenced by soil properties. For the C-band and X-band data, the vegetation parameters are probably the predominant factors, particularly for the July data set. As an additional evaluation of the differences between the data at these three frequencies, the imagery was classified and the pixel classification accuracies for each band were compared. It was found that for a four-class classification, taken one band at a time, that similar classification accuracies were obtained with the X- and C-band data (with the C-band being slightly better) but the classification accuracies for the L-band data were considerably lower. This analysis supports the previous conclusion that crop type (vegetation cover) is the dominant geophysical parameter driving the radar backscatter at X- and C-bands, but not at L-band.