Estimativa de área da cultura de soja com imagens LANDSAT 5/TM por meio de matriz de erros

Abstract

Crop area is fundamental importance in crop yield forecasting in Brazil. Remote sensing data can be used for monitoring strategic agricultural areas. These work it has as objective the soybean crop area estimate for eight municipalities of the Paraná State, in two harvest 2003/2004 e 2004/2005, based on error matrix generated from supervised classification of Landsat 5/TM images. The classifications were realized by Parallelepiped and Maximum Likelihood algorithms, producing that called as soybean mask. The area estimation obtained from soybean mask corrected by error matrix, presented high correlation with official estimations and those obtained by direct expansion. Hence, a soybean crop area estimate using orbital data show to be viable that is realized objective and direct form, present less cost and can be made prior to the conventional methods.