Soil Organic Carbon Prediction Based on Different Combinations of Hyperspectral Feature Selection and Regression Algorithms

Cropland soil organic carbon (SOC) is crucial for global food security and mitigating the greenhouse effect. Accurate SOC prediction using hyperspectral data essential dynamic monitoring of pools in croplands. However, effective methods to reduce dimensionality integrate it with suitable regression algorithms reliable models are poorly understood. In this study, we analyzed 108 samples from Changting County, Fujian Province, China. Our objective was evaluate performance various combinations six feature selection four prediction. findings as follows: combination Successive Projections Algorithm (SPA) Partial Least Squares (PLS) yielded most favorable results, R2 (0.61), RMSE (1.77 g/kg), MAE (1.48 g/kg). Moreover, determined relative importance variables, following ranking: 696 nm > 892 783 1641 1436 396 392 2239 2129 nm. Notably, exhibited highest SPA-PLS model, Variable Importance Projection (VIP) value 1.22. This study provides profound insights into prediction, highlighting superiority optimal combination.