Dataset supporting SVR–RBF modelling of SOC from Vis–NIR spectra

This dataset contains the raw spectral reflectance data and measured soil organic carbon (SOC) values for 649 soil samples used in the study on SOC prediction from Vis–NIR spectroscopy. The dataset includes the original spectral measurements across the continuous wavelengths at 5 nm intervals, yielding a total of 411 wavelengths and corresponding laboratory-determined SOC values. The data were used to develop and optimise Support Vector Regression (SVR–RBF) models through hyperparameter tuning (C and γ) and multi-level validation strategies, including cross-validation, learning curve analysis, and bootstrap resampling. Preprocessing and modelling procedures are described in the associated publication.