This study aimed to evaluate the effectiveness of spectral absorption-feature indices, derived from soil hyperspectral diffuse reflectance spectroscopy, as covariates within a multivariate geostatistical framework to enhance the digital mapping of soil organic carbon (SOC). The approach also incorporated exhaustively measured auxiliary variables derived from topographic and textural attributes. The research was conducted in a 1.39-km2 forested catchment, where 135 topsoil samples (0–0.20 m depth) were collected from soils classified as Typic Xerumbrepts and Ultic Haploxeralfs. All samples were analyzed for SOC concentration, soil texture, and diffuse reflectance spectra across the VIS–NIR–SWIR region (350–2500 nm). The continuum-removal technique was applied to compute radiometric indices associated with absorption features in the visible region and at 1400, 1900, and 2200 nm. Results demonstrated that these indices effectively captured the SOC spatial variability when combined with silt fraction and topographic attributes, which, among the other covariates, actually exhibited the strongest spatial relationships with SOC. Compared to univariate ordinary kriging, the multivariate geostatistical approach yielded improved prediction accuracy in cross-validation, mostly due to the use of hyperspectral indices as auxiliary variables. Moreover, the geostatistical analysis revealed that the multivariate frame of spatial association was characterized by two distinct spatial scales. The findings of this work then support the use of hyperspectral indices as valuable covariates for digital modelling of SOC distribution even in landscapes characterized by heterogeneous topography and pedology.
Buttafuocò et al. (2026) studied this question.