Estimates of crop development stage by remote sensing techniques are potentially more cost effective for large areas than present labor intensive estimates. Here, discriminant analysis for identifying crop development stage using multispectral data is discussed. The data analyzed were Exotech 100 measurements in four wavelength bands (.5-.6, .6-.7, .7-.8, .8-1.1 um) over field plots of soybeans and corn in West Lafayette, Indiana. Different row spacings, soils, plant populations, planting dates, soybean cultivars and crop years (1978, 1979, 1980) increased the generality of the data set.

Results show that development stage classes at the beginning and end of the growing, season have relatively high classification accuracy. Midseason classes have moderate to low classification accuracies. Overall accuracy increases as development stages are pooled to form fewer classes per growing season. Applying discriminant analysis to a subset of development stages in a season increases classification accuracy.

Date of this Version