Separability of agricultural crops with airborne multiparameter radars

Backscattering measurements were acquired with airborne radar scatterometers over a test site in Cass County, North Dakota on four days in the summer of 1981. Data were acquired at three frequencies (L-, C- and Ku-bands), two polarizations (like and cross) and ten incidence angles (5° to 50°). Crop separability was examined using a two-class separability measure, which compares within-class to between-class variability. Based on this separability measure, a scattcrometer channel most suited for separating each pair of crops was selected and the separability errors were calculated by choosing a subjective decision boundary to separate the two crops. Within the small grains class, higher frequencies were found to be useful, with good to very good separability in most cases. Both L and Ku-bands were helpful for separability within the nonsmall grains class, with mostly very good to excellent results. By discarding mixed pixels, substantial improvement was obtained in both the two-class separability and the separability errors for various crops. These results are helpful in selecting system parameters of future active microwave remote sensors in vegetation discrimination and characterization studies.