Statistical Separability of Spectral Classes of Blighted Corn

The purpose of this study was to determine the statistical separability of multispectral measurements from corn having varying levels of southern corn leaf blight severity. Multispectral scanner data in twelve spectral channels in the wavelength range 0.4 to 11.7 µm were analyzed for ten selected flightlines of the 1971 Corn Blight Watch Experiment. A total of 168 corn fields having 18,804 sample points were analyzed. The blight rating information for these fields were available. Maximum average transformed divergence between spectral classes (found by LARSYS Cluster Algorithm) of all possible pair of blight levels, maximized over a subset of channels, was computed in each of one, two, three, and four spectral channels for each of ten flightlines. From the statistical analysis of the values of average transformed divergence, it was concluded that the greater the difference between the blight levels, the more statistically separable they are. This result is encouraging considering the fact that there are variables other than the blight severity within and between flightlines. Although the analysis was done for corn blight only, the conclusions obtained from this analysis may well be applicable to other crop stresses because corn blight is representative of the problems, of many plant stresses, especially non-systemic stresses