Mapping basic properties of Danish sandy soils using on-the-go proximal sensors and terrain attributes

• Published in: Geoderma Regional Vol. 42; p. e00981
• Main Authors: Khatkar, Ameesh, Beucher, Amélie, Koganti, Triven, Munkholm, Lars Juhl, Lamandé, Mathieu
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.09.2025
• Subjects: Bulk density; Clay content; Digital soil mapping; Electromagnetic induction; Gamma ray spectroscopy; Soil water content; Total carbon; Total carbon; PTF; Electromagnetic induction; CPS; RTK; GRS; GPS; LOOCV; nRMSE; PRP; MLR; Soil water content; Gamma ray spectroscopy; Clay content; DOE; OK; Bulk density; EMI; Digital soil mapping; CI; GNSS; DEM; TA; CR; PA; PSS; HCP; LCCC; cLHS
• ISSN: 2352-0094; 2352-0094
• DOI: 10.1016/j.geodrs.2025.e00981

On-the-go proximal soil sensing based on geophysical sensors is increasingly recognized as the ‘gold standard’ in digital soil mapping due to its capacity to generate high-resolution maps of soil properties at the field scale. However, studies of their limitations are scarce. In this study, we evaluated the suitability of electromagnetic induction (EMI) and gamma-ray spectroscopy (GRS), along with terrain attributes (TA), to predict four soil properties, i.e. clay content, total carbon, bulk density, and soil water content. Soil samples were collected from the top (15 cm depth) and subsoil (40 cm depth) at 69 points distributed in three sandy arable fields. Soil properties were estimated through multiple linear regression (MLR) and cross-validated using the Leave-One-Out Cross-Validation (LOOCV). The MLR models were then filtered based on Lin's concordance correlation coefficient (LCCC), coefficient of determination (R2) and normalized root mean square error (nRMSE). The results indicated that estimating the soil properties in sandy soils is challenging, specifically in subsoil, as reliable models were achieved only for topsoil in two fields. Inverting the EMI data improved modelling results compared to using raw EMI data. Despite the challenges encountered, predictors from the EMI and GRS emerged as key contributors to models with the highest performance, indicating the potential of on-the-go geophysical sensors for generating high-resolution digital soil maps.