Selection of significant raster images for digital soil mapping using data reduction technique

The study was conducted to select the significant environmental covariates for mapping the soil through the computer-assisted soil mapping method. A total of 340 soil profile information was intersected against the environmental covariates to reduce the dimension of the data by which the quality of the spatial soil predictions can be improved. Robust PCA is known for its supremacy in handling image drive data as it can even process the data extracted from raster image series, which are high in outliers. The selection of significant covariates for digital soil mapping was done through robust Principal Component Analysis (PCA) and conventional PCA. The scree plot indicates that four principal components are to be considered from both methods. The selected principal components of robust PCA cumulatively contribute 63.17% of the total information to the original dataset, whereas conventional PCA contributes 52.46% only. Contribution charts were employed for extracting the significant environmental covariates in which 26 out of 33 covariates are obtained from the selected principal components of robust PCA. Soil mapping would be efficient if the covariates are selected through this process and employed for the mapping process.