Multivariate analysis of airborne remote sensing and topographic features for corn yield spatial pattern discrimination

Multivariate discriminant analysis (MDA) was applied to airborne remotely sensed and topographic data to select the best indicators of corn yield spatial variability. In the MDA models, elevation, slope and texture indices (contrast and second angular moment) contribute most to the discrimination of corn yield classes. The classification of yield based on the MDA results in yield classes that are well sorted in a proportion of 99% low and 71% medium yield pixels, while high yield pixels were correctly classified in 88% of the cases. The overall classification rate of 86% demonstrates that, on the basis of MDA, using texture indices and topographic data and, to a lesser extent vegetation indices, the authors can discriminate corn yield spatial patterns