This study was carried out to test the "decision tree" classifier via remote sensing (RS), to identify planted areas with sugarcane, at different planting dates in Boa Fé, located in the Triângulo Mineiro, more specifically in the town of Conquista, Minas Gerais, Brazil. RS techniques, integrated into a Geographic Information System (GIS), allow a temporal analysis of land use and occupation, especially in order to identify and monitor agricultural areas. Based on the calculation of mean bias (VM), this study showed that in areas of sugarcane, where irrigation is frequent and significant rainfall occurring prior to the passage of Landsat-5, the estimated values were slightly underestimated, with the value of this indicator equal to -0.13 ha. It was also verified that the highest values of NDVI provided a slight overestimation of the results, with values of mean bias of 0.04 to 0.23 ha. According to the results, the decision tree classifier had a great potential for mapping the areas cultivated with sugar cane.
decision tree; remote sensing; GIS; sugarcane
