Classification of wheat areas and prediction of yields in north-western New South Wales by repetitive Landsat data

Abstract

Radiance data collected by Landsat were used to derive wheat production forecasts. Radiance data at three stages of wheat growth in 1976 for the Tamworth region of the New South Wales wheat-belt were analysed by multivariate statistical methods to discriminate between paddocks that had been sown with wheat and those that had not, and to relate radiance to wheat yield. A discriminant function was derived that classified the paddocks, and a regression equation was derived to predict yields of wheat paddocks. Although data were not available for the green vegetative phases of growth, intrinsic testing of classification and yield prediction indicated high accuracy, although extrinsic testing indicated that they were less accurate. Assuming that data are available in future seasons at the vegetative stages, Landsat has a great potential for improving wheat forecasts in Australia.