Soil Organic Carbon Assessment Using Remote-Sensing Data and Machine Learning: A Systematic Literature Review

In the current global change scenario, valuable tools for improving soils and increasing both agricultural productivity and food security, together with effective actions to mitigate the impacts of ongoing climate change trends, are priority issues. Soil Organic Carbon (SOC) acts on these two topics, as C is a core element of soil organic matter, an essential driver of soil fertility, and becomes problematic when disposed of in the atmosphere in its gaseous form. Laboratory methods to measure SOC are expensive and time-consuming. This Systematic Literature Review (SLR) aims to identify techniques and alternative ways to estimate SOC using Remote-Sensing (RS) spectral data and computer tools to process this database. This SLR was conducted using Systematic Review and Meta-Analysis (PRISMA) methodology, highlighting the use of Deep Learning (DL), traditional neural networks, and other machine-learning models, and the input data were used to estimate SOC. The SLR concludes that Sentinel satellites, particularly Sentinel-2, were frequently used. Despite limited datasets, DL models demonstrated robust performance as assessed by R2 and RMSE. Key input data, such as vegetation indices (e.g., NDVI, SAVI, EVI) and digital elevation models, were consistently correlated with SOC predictions. These findings underscore the potential of combining RS and advanced artificial-intelligence techniques for efficient and scalable SOC monitoring.