multinomial logistic regression method

Limitations in conventional soil identification methods and the advances made in information technology in soil science have attracted more attention to new approaches to soil mapping in order to improve the purity of soil maps. The present study was carried out in a part of Ilam province to identify and prepare soil maps of this region. At first, 46 profiles were identified. Then, based on the morphological characteristics of each profile, soil samples were taken from all genetic horizons and analyzed for chemical and physical properties. Then, the soils were classified based on the Soil Survey Staff keys (2014). A multinomial logistic regression model was used for spatial prediction of soil taxonomic classes. The geomorphometric features were extracted from digital elevation model with a resolution of 30 m2 by SAGAGIS2.2 software. The classification results of each soil control profile in the studied area showed that, in general, the soils were in three order categories: Mollisols, Inceptisols, Entisols, and six classes at the family level. The correlation between the features of digital elevation model showed that the parameters of the mid slope position, spatial solar radiation, index of moisture content, ground roughness index, surface curvature, and profile curvature had the most effect on the formation of soil family classes. The overall accuracy and Kappa index of spatial prediction map from the regression model was 60% and 0.38 at the familial level, respectively. Finally, the results of this study showed that geomorphometric variables had a significant influence on the prediction of soil classes. Therefore, it is suggested that in future studies, other covariates derived from remote sensing data should also be used to improve the quality and accuracy of soil maps.