Land cover classification methods for multiyear AVHRR data

Abstract

AVHRR data have been extensively used for global land cover classification, but few studies have taken direct and full advantage of the multiyear properties of AVHRR data. We generated three types of signatures from 12-year monthly composite NDVI (normalized difference vegetation index) and channel 4 brightness temperature (T4) of NOAA/NASA Pathfinder AVHRR Land data for land cover classification. Both quadrature discriminate analysis (QDA) and linear discriminate analysis (LDA) are explored for classification. A global land cover training database created from Landsat TM and MSS imagery is used for training and validation. It turns out that QDA performs much better than LDA, and the overall classification rate is as high as 95.9%.