Mapping and Monitoring of Soil Organic Carbon Using Regression Analysis of Spectral Indices

The soil carbon sinking ability is dominantly controlled by local topographical settings, soil–crop management and traditional farming practices on which the food demand of the major population is dependent. The degradation of natural resources causing poor soil health is likely to strain the hilly and mountain ecosystem. This study aims to map soil organic carbon (SOC) of rice–fallow system under varying slopes and its changes during the past 20 years under traditional management practice using geospatial tools and techniques. Regression models of SOC were derived from remote sensing (RS)-based indices using multiple linear regression-stepwise (MLR-stepwise), partial least square regression (PLSR) and principal component analysis-regression (PCA-R). The MLR-stepwise model was found to be superior in performance with high R² (0.87) and least RMSE (0.026) compared to PLSR (R² = 0.71 and RMSE = 0.05) and PCA-R (R² = 0.27 and RMSE = 0.11) models for SOC prediction.