Multitemporal classification of natural vegetation cover in Brazilian Cerrado

Abstract

Spectral-Temporal Analysis by Response Surface (STARS), which utilizes surface response to represent time series spectral-temporal behavior of pixels in satellite images, was used to map and discriminate savanna vegetation classes in portion of Cerrado biome of Mato Grosso State, Brazil, using MODIS data. STARS utilized 16 daily MODIS, cloud-free, images that were collected from September 1st 2008 to August 31st 2009. The Multi-Coefficient Image (MCI) resulted from the STARS was used as input attributes for the three tested classifiers: (i) ML — maximum likelihood; (ii) SVM — support vector machine; and (iii) NN — neural network. The results showed that the NN classifier presented higher kappa coefficient (0.58) and overall accuracy of 68.6%.