Geospatial AI and Remote Sensing for Enhanced Soil Carbon Mapping and Sustainable Agriculture

Soil mapping and carbon assessment are central to understanding agro-environmental change, improving soil health, and supporting sustainable agriculture. Soil organic carbon (SOC) is highly variable across space and time, yet conventional sampling remains costly, spatially limited, and difficult to repeat at the scales required for effective land management. Recent studies demonstrate that satellite observations, particularly Sentinel-2 multispectral data, can support digital soil mapping and SOC prediction, while hyperspectral imaging and LiDAR offer additional information on soil composition, vegetation structure, and terrain. Machine learning models have further improved the integration of spectral, environmental, and field data. However, key challenges persist, including limited model transferability across landscapes and soil types, sensitivity to climatic and seasonal variability, cloud contamination in optical imagery, and the difficulty of producing reliable spatiotemporal mosaics, as well as the need for robust monitoring, reporting, and verification (MRV) frameworks to support emerging soil carbon markets. Further investigation is needed to establish robust, scalable, and interpretable approaches that combine emerging remote sensing technologies with advanced modelling techniques, including geospatial artificial intelligence (GeoAI) and foundation models, for accurate soil carbon monitoring. Additionally, since SOC is often spatially autocorrelated and the relationship between predictors is spatially non-stationary, additional research is needed to estimate SOC using geographically weighted machine learning and deep learning models.This Research Topic aims to advance the integration of remote sensing and machine learning for digital soil mapping, SOC prediction, and agro-environmental monitoring. It will address how multispectral, hyperspectral, LiDAR, and complementary environmental data can be combined to improve model accuracy, spatial resolution, temporal consistency, and transferability across diverse agricultural landscapes. It will also address the role of geographically weighted models for SOC prediction in a variety of environments. Contributions may examine new algorithms, data-fusion strategies, validation frameworks, and uncertainty assessments, as well as the practical application of soil carbon information to sustainable land management, climate mitigation, and agricultural decision-making.This Research Topic welcomes Original Research, Review, Methods, Perspective, and Data Report articles that gather further insights into remote-sensing-based soil mapping and carbon assessment across agricultural and agro-environmental systems, addressing, but not limited to, the following themes:- Digital soil mapping using Sentinel-2 and other satellite observations- Hyperspectral and LiDAR applications for soil and vegetation characterization- Machine learning and deep learning for SOC prediction- Geographically weighted machine learning and deep learning for SOC prediction- Multisource data fusion and spatiotemporal modelling- Temporal mosaicking and cloud-contamination mitigation- Soil carbon stock estimation and monitoring- Digital soil mapping for carbon monitoring, reporting, and verification (MRV) and voluntary carbon markets- Model transferability across regions, soil types, and climatic conditions- Uncertainty quantification, explainable models, and validation strategies- Soil health assessment and sustainable land-management applications- GeoAI, self-supervised learning and foundation models for SOC prediction Topic editors: Diego Urbina Salazar, Ningning Zhu Source: https://www.frontiersin.org/research-topics/85384/geospatial-ai-and-remote-sensing-for-enhanced-soil-carbon-mapping-and-sustainable-agriculture