Remote sensing has been used in the evaluation of physical and chemical characteristics of soils as a source of rapid, non-destructive, and low cost information, which may assist in management of environmental liabilities. The aim of this study was to investigate the potential for use of hyperspectral data in determining some properties of soil under different levels of application of sugarcane residue (vinasse) using spectral data obtained from samples in field conditions and laboratory oven-dried soil. A randomized block design was used with six treatments and four replications, the treatments being: 1. without vinasse; 2. fertigation with chemicals; 3, 4, 5, and 6 received the application of 150, 300, 600, and 900 m³ ha-1 of vinasse, respectively. Models were generated for prediction of chemical and physical soil properties for both types of samples from spectral curves of visible and near-infrared wavelengths. For determination of soil particle size, models classified as acceptable to good (R² from 0.7 to 0.9) were generated, and these models were not affected by the type of sample used. In regard to chemical properties, models were generated with capacity only for differentiation between high and low concentrations (R² from 0.5 to 0.65) for the properties of pH, organic matter, H+Al and cation exchange capacity. The models generated for laboratory oven-dried soil were in some cases better than those generated for samples in field conditions.
spectroradiometry; multivariate analysis; soil properties
