Numerical examination of a field classification of some red, yellow and grey earth profiles

Numerical methods have been used to examine an existing and accepted field classification of 48 profiles of red, yellow, and grey earths (mainly Alfisols) from central north Queensland. The three-dimensional soils data (sites by depths by descriptors, which may be mixed in type) were converted to a form which appeared to the computer as a two-dimensional set of profiles by attributes. The soils data were from independent depth intervals, and no assumptions were made about the inter-relationships among soil layers tvithin a profile; nor were the values of any of the soil attributes weighted. In order to consider shallow profiles on the same total depth basis as the deep profiles, the absence of a soil horizon or sampling interval at depth has been regarded as a positive attribute in the numerical classification. Comparison of the traditional field classification and a numerical classification of the same soils dataset showed that certain soil attributes played an important role in both classifications. The most striking difference between them was the relative importance of soil colour attributes, from which it might be argued that field pedologists have assigned to colour a weight out of proportion to its real importance in soil classification. However, this attribute carries additional information about the mineral constituents and hydrological regimes of the soils, and represents information known to, and used by, the field pedologist but not revealed to the computer. The essential subjectivity in the choice of attributes for soil classification is demonstrated.