Soil classification based on spectral and environmental variables
Abstract. Given the large volume of soil data, it is now possible to obtain a soil classification using spectral, climate and terrain attributes. The idea was to develop a soil series system, which intends to discriminate soil types according to several variables. This new system was called Soil-Environmental Classification (SEC). The spectra data was applied to obtain information about the soil and climate and terrain variables to simulate the pedologist knowledge in soil-environment interactions. The most appropriate numbers of classes were achieved by the lowest value of AIC applying the clusters analysis, which was defined with 8 classes. A relationship between the SEC and WRB-FAO classes was found. The SEC facilitated the identification of groups with similar characteristics using not only soil but environmental variables for the distinction of the classes. Finally, the conceptual characteristics of the 8 SEC were described. The development of SEC conducted to incorporate applicable soil data for agricultural management, with less interference of personal/subjective/empirical knowledge (such as traditional taxonomic systems), and more reliable on automation measurements by sensors.