Soil Pattern Recognition with Fuzzy‐c‐means: Application to Classification and Soil‐Landform Interrelationships

Publisher Website
Google Scholar
Abstract
Soil variation is more continuous than discrete and therefore calls for a continuous approach to soil classification. Fuzzy set theory provides such an approach that quantitatively assigns individuals to geographically and taxonomically continuous classes. We hypothesized that the fuzzy‐c‐means algorithm, based on a fuzzy objective function, could be used to quantify pedons into intragrade and extragrade classes and that minimization of the function would allow for the most “precise” classification, which could be validated by mathematically heuristic methods. These methods assume that local extreme (or abrupt changes) in the value of the objective function or its derivatives indicate optimal grouping. This approach was applied on two transects traversing a subcatchment in the central Mount Lofty Ranges, South Australia. The resulting fuzzy soil classes were found to have reasonable spatial continuity and contiguity, which were associated with local variation in slope features and parent material. Thus the approach identified plausible groupings (substructures) among the soil individuals that realistically reflected changes in the external factors (principally landform and lithology) controlling the processes generating the substructures. The structure‐factor relationships could be used for further rapid soil pattern recognition embodied in soil classification and mapping.

This publication has 0 references indexed in Scilit: