Soil discrimination analysis by spectral response in the ground level

Abstract

The objective of this study was to develop and test a discrimination method for soil classes by their spectral response (SR), using a laboratory sensor. Spectral data were used to develop statistical model for discriminating soil classes in an area at the southwest of São Paulo State, Brazil. Discriminant equations were developed for 18 soil classes. The spectral data were obtained in superficial and subsuperficial soil samples in the study area, with a total of 370 samples. The samples were collected in 185 ha, with one borehole per ha. The results showed that soil classes can be separated and delimitated by discriminant analysis. The analysis presented a classification index higher than 80% for each soil class. The global classification index was 90.71%, when all soil classes were used to develop the model, and 93.44% when most individuals classes were used. The simulated statistical test was efficient in the discriminant analysis, presenting a classification index higher than 91%, with a global error of 8.8%. The analysis demonstrated a reduction of the model quality when applied for 20% sub-group of the samples with global error of 33.9%. The method helped in the soil classes discrimination by their spectral reflectance, based on their physical interaction with electromagnetic energy.