Visible-NIR reflectance:: a new approach on soil evaluation

The aim of this work was to develop a spectral reflectance (SR)-based methodology to evaluate soil types and soil tillage systems. Soil samples, located along nine toposequences with basalt and shale, were collected in Sao Paulo State, Brazil. SR of soil samples was measured with a laboratory spectroradiometer between 450 and 2500 nm. SR curves were evaluated at different depths to determine soil classes, as for conventional methods of soil classification. Soil survey maps were developed by descriptive interpretation of the spectral curves and statistical analysis as well. These maps were compared with detailed and semidetailed soil maps based on chemical, physical and mineralogical soil attributes (conventional method). Organic matter, total iron, silt, sand and mineralogy (quartz, magnetite, kaolinite and smectite) were the most important attributes influencing reflectance intensity and spectral features and allowed characterization and discrimination of soils. Strong absorption features centered at 1900 nm were related to OH molecules in free water of 2:1 minerals. Identification of soil classes was most precise when the entire spectral curve of different depths was analysed simultaneously, including features and intensities. Parent material variation along toposequences and weathering variability modified soil characteristics and its respective SR data. The detection of these alterations was essential to determine soil line demarcation. Soil line demarcation and number of soil classes detected by SR analyses were similar to the detailed soil map determined by conventional methods. SR laboratory measurements can be used as a methodology to assist soil surveys. Spectral data can also group soils under similar tillage systems. (C) 2003 Elsevier B.V. All rights reserved.