TY - JOUR
T1 - High-resolution 3D soil texture mapping in Denmark using satellite time series and bare soil composites
AU - Møller, Anders Bjørn
AU - Nyborg, Lotte
AU - Grogan, Kenneth
AU - Druce, Daniel
AU - Svane, Simon Fiil
AU - Greve, Mette Balslev
AU - Gutierrez, Sebastian
AU - Styczen, Merete
AU - Greve, Mogens H.
AU - Knudsen, Leif
AU - Beucher, Amélie
PY - 2025/1
Y1 - 2025/1
N2 - Accurate soil texture and soil organic carbon (SOC) maps are fundamental for agronomic decision making, land management, and climate change mitigation. This study presents a high-resolution (10 m) digital soil mapping approach for Denmark, integrating multi-temporal satellite imagery with extensive georeferenced soil observations (>50,000 samples). Using Sentinel-1 and Sentinel-2 bare soil composites alongside a time series of images with vegetation, we modeled soil texture fractions, SOC, and calcium carbonate (CaCO3) across multiple depths. The results highlight that bare soil composite images were crucial for predicting mineral texture fractions. The images from Sentinel-1 were especially important in this regard, although they have rarely been used in previous studies. For SOC and CaCO3 predictions, images with vegetation had greater importance than bare soil products. Despite their high importance, satellite-derived covariates did not replace conventional predictors such as parent materials, spatial position, and topography. Our findings thereby demonstrate the complementary nature of Sentinel-1 and Sentinel-2 data and show that integrating remote sensing products with covariates from other sources enhances prediction accuracy. Compared to previous Danish soil maps, the proposed approach significantly improves spatial soil characterization, supporting sustainable agricultural practices and soil resource management. These results underscore the value of modelling frameworks based on multiple complimentary sources of information, to create soil maps for agronomic and environmental applications.
AB - Accurate soil texture and soil organic carbon (SOC) maps are fundamental for agronomic decision making, land management, and climate change mitigation. This study presents a high-resolution (10 m) digital soil mapping approach for Denmark, integrating multi-temporal satellite imagery with extensive georeferenced soil observations (>50,000 samples). Using Sentinel-1 and Sentinel-2 bare soil composites alongside a time series of images with vegetation, we modeled soil texture fractions, SOC, and calcium carbonate (CaCO3) across multiple depths. The results highlight that bare soil composite images were crucial for predicting mineral texture fractions. The images from Sentinel-1 were especially important in this regard, although they have rarely been used in previous studies. For SOC and CaCO3 predictions, images with vegetation had greater importance than bare soil products. Despite their high importance, satellite-derived covariates did not replace conventional predictors such as parent materials, spatial position, and topography. Our findings thereby demonstrate the complementary nature of Sentinel-1 and Sentinel-2 data and show that integrating remote sensing products with covariates from other sources enhances prediction accuracy. Compared to previous Danish soil maps, the proposed approach significantly improves spatial soil characterization, supporting sustainable agricultural practices and soil resource management. These results underscore the value of modelling frameworks based on multiple complimentary sources of information, to create soil maps for agronomic and environmental applications.
KW - Digital soil mapping
KW - Machine learning
KW - Remote sensing
KW - Soil organic carbon
KW - Soil texture
UR - https://www.scopus.com/pages/publications/105011745993
U2 - 10.1016/bs.agron.2025.07.001
DO - 10.1016/bs.agron.2025.07.001
M3 - Journal article
AN - SCOPUS:105011745993
SN - 0065-2113
VL - 194
SP - 133
EP - 186
JO - Advances in Agronomy
JF - Advances in Agronomy
ER -