Mapping of coffee lands (Coffea arabica L.) in the zona da mata region, Minas Gerais state, using remote sensing

Abstract

The aim of this work was to map coffee lands in the Zona da Mata region, in Minas Gerais state, using non-conventional aerial photographs and satellite images. A pilot area, representative of the regional coffee lands, was chosen. A non-conventional aerophotogrammetric survey of the study area was carried out (scale 1:10000) and an ETM⁺Landsat7 satellite image was acquired. This image was registered and transformed into surface reflectance data. Photointerpretation of the limits of land use classes was done over a digital mosaic. These limits were overlaid onto the image, providing reflectance sampling of each land use type for statistical analysis and assessment of the vegetation’s spectral response. Statistical analysis showed that bands 3, 4, 5 and 7 were the most representative in the discrimination of vegetation canopies. Although statistical analysis showed a significant difference between the bands for the different land use/land cover types, the classifications did not provide good target discrimination due to shading, to the region’s very steep landscape and to the spectral signature similarity between coffee and forest. The mapping accuracy between the classified image and photointerpretation was considered regular to weak and the best results were obtained through a combination of bands. The use of ETM/Landsat7 images to map coffee lands presented limitations, despite the few types of land use. This is due to the shading of the images, owing to the steep topography, and to the fragmentation of most of the coffee lands into small fields.