Effects of random and systematic variations in unresolved cloud on recorded radiance and on target discriminability

The purpose of this investigation has been to study the systematic and random effects of unresolved (i.e., smaller than instrument IFOV) cloud and haze on the ability to discriminate, e.g., vegetated targets using digital recorded radiance in preselected bandpasses. Calculations have, for the sake of example, centered on the discrimination of wheat and of soybeans from targets consisting of soybeans at various levels of stress severity. Calculations have mainly centered on nadir values (the only value of view angle for which reflectance data are available for various levels of disease stress severity), but some calculations of target discriminability for various view angles have been made to serve as examples. While means may be found to determine and to correct for systematic dependence of recorded radiance on view angle, it will not be possible to correct for random variations. Thus, while studies of systematic variations of target radiance with scan angle will lead to data calibration (to improve target discriminability), studies of random variation will indicate the limits imposed by fluctuations in physical factors on target differentiation and quantification. It is suggested that a future aim must be to determine which bandpasses or combinations of bandpasses may best be used to minimize random variations in target radiance, so as to optimize the discrimination of selected targets, in this case stressed from unstressed vegetation.