Autoria: SANCHES, I. D. A.; ANDRADE, R. G.; QUARTAROLI, C. F.; RODRIGUES, C. A. G.

Resumo: The present study has compared three methods of atmospheric correction, two dark object subtraction (DOS) methods and one method based on the fourth component of the tasseled cap transformation (TC4), to obtain reflectance and NDVI values of Landsat 5 ? TM satellite images. For each atmospheric correction method, reflectance and NDVI values of five targets (crop/pasture, forest, straw, bare soil and water) were analyzed in three images (Landsat path-row 222-75). The study area is localized in the west of Sao Paulo State, Brazil. The images were acquired on 29/10/2009, 02/02/2010 and 26/06/2010 with solar zenith angles of 28°, 34° and 55°, correspondingly. The TC4 was the only method which eliminated the haze in the image, but it is not recommended if the objective is to obtain reflectance values. The DOS methods did not removed the image haze but performed reasonably to obtain reflectance values. For the visible bands, both DOS methods analyzed performed well. The DOS method which considered the atmospheric transmittance (DOS2), although more indicated to correct the infrared bands, it is dependent of the image solar zenith angle. The image acquired under a zenith angle of 55°, after been corrected using the DOS2 method, presented reflectance values overestimated. Without the atmospheric correction the NDVI of vegetation targets tended to be underestimated.

Ano de publicação: 2011

Tipo de publicação: Artigo em anais e proceedings

Unidade: Embrapa Territorial

Palavras-chave: Estudos multitemporais, Sensoriamento Remoto, Subtração do pixel escuro, Transformação tasseled cap, Ângulo zenital