Bottom-up digital soil mapping. II. Soil series classes

Continuous classification of soil has generally focussed on the classification of soil profiles as single entities. This is because most continuous classification algorithms (for example the fuzzy k-means algorithm) are not able to cope with “composite objects”, i.e. entities that are composed of sub-entities arranged in a specific sequence, as in the case of soil profiles being composed of sequences of soil layers. An algorithm has recently become available that enables the continuous classification of soil profiles as sequences of soil layers. This is the “Outil Statistique dʹAide à la Cartogénèse Automatique” (OSACA) application developed by Carré and Jacobson (Carré, F. and Jacobson, M., 2009. Numerical classification of soil profile data using distance metrics. Geoderma, 148: 336–345). Until now it has been used to carry out the continuous classification of soil profiles using soil properties of the profilesʹ layers, but in this work the soil layers are characterised by their membership to the continuous layer classes created in Part I of this paper. We used OSACA to create what we call “soil series classes”, which are classes of soil profiles whose layers consist of common sequences of membership to the soil layer classes created in Part I of this paper (Odgers, N. P., McBratney, A. B. and Minasny, B. Bottom-up digital soil mapping. I. Soil layer classes. Geoderma, in press). 9 soil series classes were found to be optimal for the set of soil profiles used in our study. The taxonomic distance to the modal profiles of the soil series classes was mapped using regression-kriging, then the membership to each class was calculated from the taxonomic distance.