Evaluating Calibration and Spectral Variable Selection Methods for Predicting Three Soil Nutrients Using Vis-NIR Spectroscopy

Soil nutrients, including soil available potassium (SAK), phosphorous (SAP), and organic matter (SOM), play an important role in farmland productivity, food security, agricultural management. Spectroscopic analysis has proven to be a rapid, nondestructive, effective technique for predicting properties general potassium, phosphorous, particular. However, the successful estimation of nutrient content by visible near-infrared (Vis-NIR) reflectance spectroscopy depends on proper calibration methods (including preprocessing transformation multivariate regression analysis) selection appropriate variable techniques. In this study, raw spectrum 13 transformations combined with 2 (competitive adaptive reweighted sampling (CARS) successive projections algorithm (SPA)) algorithms (support vector machine (SVM) partial least squares (PLSR)), total 56 methods, were investigated modeling above three nutrients using hyperspectral Vis-NIR data (400–2450 nm). The results show that first-order derivatives based logarithmic inverse (FD-LGRs) can provide better predictions best form is SG+MSC. CARS was superior SPA selecting variables, PLSR model outperformed SVM models. accuracies (R2, RMSE) 0.7532, 32.3090 mg/kg; 0.7440, 6.6910 0.9009, 3.2103 g/kg, respectively, their corresponding (FD-LGR)/SPA/PLSR, SG+MSC/CARS/SVM, respectively. Overall, prediction content, followed potassium. It concluded application images (Vis-NIR data) efficient method mapping monitoring at regional scale, thus contributing development precision agriculture.