Incorporating soil taxonomic distance and decision tree for spatial prediction of soil classes in Ardakan, Yazd

Mapping soil classes digitally generally starts with soil profile description with observed soil classes at a taxonomic level in a particular classification system. At each soil observation location there is a set of co-located environmental variables and a challenge is to correlate soil classes with environmental variables. The current methodology treats soil classes as ‘labels’ and their prediction only considers the minimization of the misclassification error. Soil classes at any taxonomic level have taxonomic relationships between each others. Using classification trees، we can specify an algorithm that minimises the taxonomic distance rather than misclassification error. Therefore، in this research، we have attempted to develop decision tree model for spatial prediction of soil taxonomic classes in an area covering 720 km2 located in arid region of central Iran. In this area، using the conditioned Latin hypercube sampling method، location of 187 soil profiles were selected، which then described، sampled، analyzed and allocated in taxonomic classes according to soil taxonomy of America. Auxiliary data used in this study to represent predictive soil forming factors were terrain attributes، Landsat 7 ETM+ data and a geomorphologic surfaces map. Discriminant analysis was applied to calculate taxonomic distances. Results showed using the taxonomic distances led to achieve overall accuracy up to 70%. Results also showed some auxiliary variables had more influence on predictive soil class model which included: wetness index، geomorphology map and multi-resolution index of valley bottom flatness. General results showed that incorporating taxonomic distance into decision tree model had reliable accuracy. Therefore، it is suggested using of decision tree model with taxonomic distance for spatial prediction of soil classes in the future studies.