Numeric Classification of Subtropical Soils in Veracruz, Mexico

Numeric methods can be applied for the soil class classification because these methods reduce the laboratory analysis and related costs; therefore the method of soil numeric classification precision was evaluated. Study was performed in a humid subtropical zone in the state of Veracruz, Mexico. Three types of soil were selected, where nine soil profiles were taxonomically classified; six of which were used for its analysis with principal components and to select the variables that represent the greatest variance of the soil properties data. The selected variables were used for a discriminant analysis in order to obtain linear equations for the soils class differentiation. Bootstrap technique was used in order to evaluate the accuracy with 1000 replicas with the three remaining profiles. Eleven variables, pH, EC, Total CaCO3, Sand, Slime, Ca (m mol L-1), Ca (cmol kg-1), P2O5, K m (mol L-1), Mg (cmol kg-1) and OM,out of the 21 analyzed by principal components were used for the discriminant analysis. The confusion matrix of the discriminant analysis showed that there was no error in the linear equation test. The global accuracy of thenumeric classification of soils using the 1000 replicates was of 97%.

کلیدواژه ها:

Bootstrap samples ، Soil classification ، Discriminant analysis Principal components