Integrating factor analysis and support vector machine for digital mapping of soil quality using DEM-derived...

Soil is one of the most important natural resources, playing a fundamental role in ecosystem sustainability, food security, and agricultural productivity. However, its quality and functions have been increasingly threatened by soil erosion, land-use change, and inappropriate management practices. Therefore, the development of accurate, rapid, and cost-effective approaches for digital soil quality mapping has become increasingly important. The aim of this study was to evaluate the potential of geomorphometric variables derived from the ALOS-PALSAR Digital Elevation Model (DEM) for predicting the Soil Quality Index (SQI) using the Support Vector Machine (SVM) algorithm, as well as to identify the most influential geomorphometric factors controlling soil quality variations in the Loshan region, northern Iran. A total of 76 surface soil samples (0-30 cm) were collected, and their physical, chemical, and biological properties were determined. Subsequently, the Soil Quality Index was calculated using both the Total Data Set (TDS) and the Minimum Data Set (MDS) approaches. Geomorphometric variables, including slope, aspect, elevation, curvature, Compound Topographic Index (CTI), Stream Power Index (SPI), and Terrain Characterization Index (TCI), were extracted from the DEM. The SVM algorithm was then developed to model and predict the spatial distribution of SQI, and its performance was evaluated using the coefficient of determination (R²) and the root mean square error (RMSE). Furthermore, factor analysis was conducted to clarify the contribution of geomorphometric variables to soil quality variation. The results indicated that the TDS and MDS approaches produced similar spatial patterns of soil quality, with mean SQI values of 0.63 and 0.60, respectively, indicating that the majority of the study area fell within the moderate to low soil quality classes. The SVM model achieved satisfactory predictive performance, with validation R² values of 0.73 and 0.72 for the TDS and MDS models, respectively, while the RMSE ranged from 0.047 to 0.054. Factor analysis identified two dominant factors, namely the topographic factor and the hydrological factor, which together explained more than 70% of the total variance in the geomorphometric variables. Among the investigated variables, slope, elevation, CTI, TCI, and SPI exerted the greatest influence on soil quality variation, with slope being identified as the primary controlling factor. The findings demonstrate that DEM-derived geomorphometric variables provide strong predictive capability for soil quality assessment, while factor analysis serves as an effective approach for identifying the key geomorphometric factors controlling soil quality and for improving the interpretation of machine learning prediction models.