Mapeamento de areas de cafe no municipio de Guaxupe/MG por meio de processamento digital de imagens Landsat / Coffee crop areas mapping in mountain relief through Principals Components Analysis

AUTOR(ES)
DATA DE PUBLICAÇÃO

2009

RESUMO

The economic importance of Brazilian coffee growing in the world market is notorious and makes up significant portion of the country s foreign trade exports. Minas Gerais stands out as the core of Brazilian coffee growing, with most of the planting areas (Coffea arabica) concentrated in the south, where it is grown in small plots widely spread throughout the hills. The need to adequate coffee agriculture by planning, cost control and productivity improvement has increased the search for techniques and tools for the prediction of agricultural production, necessarily involving the location and quantification of cultivated areas. In this context, the goal of this research has been to evaluate data from the TM/Landsat-5 remote sensor, providing information about coffee growing areas in hilly regions. The city of Guaxupé/MG/Brazil was chosen for this study due to its strong coffee growing, kept under an organized cooperate system. Images from the Landsat-7 and Landsat-5 satellites and from the MODIS sensor have been employed for the purpose of using digital processing tools for atmospheric correction and radiometric normalization, in order to analyze coffee crops in three dates: 08/15/2001, 12/05/2001 and 04/12/2002, characterizing phenological stages such as harvesting periods (rest and senescence of boughs), flowering and beginning of flower bud growing, respectively. The use of a digital elevation model generated from ASTER/TERRA sensor enabled the use of a lighting factor determination technique, consisting in the creation of lighting classes that contributed in the identification of crop areas and terrain-shadowed vegetated areas. Field data were gathered to help identifying coffee plantation separated by sample classes of dense coffee and adult coffee as a function of the spacing of the field foods and lines used in planting. PCA (Principal Component Analysis) was applied in order to reduce the redundancy of the data obtained from orbital imaging in a way that allows the selection of training samples for use in supervised classification. Using the Mahalanobis Distance as a classifier, the images in the selected dates showed highly positive result when the classification was compared to the coffee area mask extracted from Ikonos images. The results of these classifications were validated through the determination of Global Accuracy and Kappa Index, which showed values of GA= 0.78 and K= 0.40 for the 08/15/2001 image, GA= 0.81 and K= 0.29 for the 12/05/2001 image, and GA= 0.76 and K= 0.24 for 04/12/2002, confirming that the dry season (May through October) is favorable for the classification of coffee, which is under the harvesting process in this period, during which the falling of leaves and remotion of fruit separates it from other ground cover such as vegetation. The spectral data obtained from satellite imaging through digital processing have proven themselves adequate for the location of coffee-growing areas in hilly regions when aided by a digital elevation model. The value of the area as being coffee crop, calculated by sum of the areas found from each date classification, produced 73,06% of the agreement with coffee mask considered as a reference data. Due this the methodology showed very suitable to quantify coffee areas in hilly region

ASSUNTO(S)

landsat (satelites) principals components analysis modis remote sensing aster analise por componentes principais cafe - guaxupe (mg) sensoriamento remoto radiometric normalization

ACESSO AO ARTIGO

Documentos Relacionados