Soil classification using visible/near-infrared diffuse reflectance spectra from multiple depths

• Spectroscopy and logistic regression can be used in combination to classify soils.
• Soil spectra from multiple depths can be integrated in the classification method.
• Important predictors in the models are related to main soil mineral constituents.
• Models capture soil characteristics used in the definition of orders and suborders.

Visible/near-infrared diffuse reflectance spectroscopy (VNIRDRS) offers an alternative to conventional analytical methods to estimate various soil attributes. However, the use of VNIRDRS in soil survey and taxonomic classification is still underexplored. We investigated the potential use of VNIRDRS to classify soils in a region with variable soils, geology, and topography in southeastern Brazil. Soils were classified in the field according to the Brazilian Soil Classification System, and visible/near-infrared (400–2500 nm) spectra were collected from three depth intervals (0–20, 40–60 and 80–100 cm) and combined in sequence to compose a pseudo multi-depth spectral curve, which was used to derive the classification models. Principal component (PC) analysis and multinomial logistic regression were used to classify 291 soils (202 in calibration and 89 in validation mode) at the levels of order (highest), suborder (second highest) and suborder plus textural classification (STC). Based on the validation results, best classification was obtained at the order level (67% agreement rate), followed by suborder (48% agreement) and STC (24% agreement). The inherent complexity and variability within soil taxonomic groups and in contrast the strong similarity among different groups in terms of soil spectra and other attributes cause confusion in the classification model. This novel approach combining spectral data from different depths in multivariate classification can improve soil classification and survey in a cost-efficient manner, supporting sustainable use and management of tropical soils.