Melhorias na acurácia do mapeamento por NDVI

Abstract

This work proposes a supervised classification method by the Bhattacharya distance tool using the Normalized Difference Vegetation Index of croplands and forest remnants of the Atlantic forest and other land use classes from the Mourão river watershed, Paraná, aiming to mitigate the effects of seasonality in the classification. The overall accuracy and Kappa index of the product generated by the supervised classification were 73.92% and 0.53, respectively. We also investigated the user’s accuracy, the producer’s reliability and the conditional Kappa index. The results showed that the method presented an excellent performance for the differentiation of forest cover (k = 76.94), temporary crops (k = 0.89) and water (k = 0.87).