HYPERSPECTRAL REMOTE SENSING: EVALUATION OF R95/HSS SENSOR FOR SPATIALIZATION AND CHARACTERIZATION OF SOIL IN MANAUS, AM - BRAZIL

Abstract

The HSS (Hyperspectral Scanner System) was evaluated for characterization and spatialization of soils in Manaus, AM, Brazil, with 3 meter resolu- tion images. The images (37 bands between 0,43 and 2,37μm) were converted from radiance values to surface reflectance using the Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes (FLAASH) applicative, based on the MODTRAN +4 (Moderate Resolution Transmittance) a radioactive transfer model. Preceding collec- tion of soil samples, the main soil classes were identified through the SAM (Spectral Angle Mapper) method in order to facilitate analysis of the spectral characteristics of the studied area. Finally, for 13 samples, correlations among soil characteristics and spectral data from the HSS sensor, and spectral curves obtained in laboratory were carried out. Results indicated that: (a) the atmospheric correction accomplished with the FLAASH applicative was appropriate, despite difficulties for total removal of the water vapor feature at 0.94μm and at short-wave infrared band, near to 2.0μm. In general, a spectral coherence of the HSS sensor was confirmed with the information found in literature and the spectrum obtained in the laboratory; (b) a good capacity for spectral identification was observed with the SAM classifier as the results provided an appropriate soil preidentification, helping as well to define sites for field work; (c) the spectral curves representative of the soil classes are comparable both for laboratory and sensor data. Usually, correlations between both data acquisition environments (laboratory and HSS) are high.