Global Soil Organic Carbon Storage and Dynamics: Current knowledge and Machine Learning Potentials

Soil organic carbon (SOC) determines multiple ecosystem services that soils provide to humanity, serving as a critical component in maintaining soil health, fertility, and climate regulation. However, changes in land use and climatic conditions may alter the current soil carbon balance, potentially converting soils from carbon sinks into sources of atmospheric CO2. Such shifts can also alter soil properties and ecosystem functions impacting environmental stability and human well-being. Using a large number of global soil profile observations, environmental datasets, and different modeling techniques, we (1) quantified the magnitude and uncertainty associated with global and regional SOC estimates, (2) evaluated projections of future SOC stock changes based on Coupled Model Intercomparison Project Phase Six (CMIP6) Earth System Models, and (3) explored the potential of machine learning (ML) techniques to address existing knowledge gaps in SOC storage and dynamics. Our findings highlight significant variability in global SOC stock estimates, both for surface soils (0–30 cm) and deeper soil profiles (0–1 m), with predictive accuracy varying across depth intervals and biomes. Projections from CMIP6 Earth System Models indicate a potential increase in global soil carbon stocks under high-emission scenarios. Meanwhile, recent advancements in ML approaches show promise in reducing uncertainties surrounding SOC storage and dynamics, offering new pathways for improved understanding and modeling. Despite these advances, critical knowledge gaps persist regarding the current distribution and future fate of global SOC stocks in the context of changing climate and land-use patterns. Addressing these uncertainties will require a coordinated and multidisciplinary effort, encompassing: (1) harmonizing SOC profile observations and collecting samples from under-represented biomes, (2) improved representation of soil-forming processes and pedogenic feedbacks within Earth System Models, and (3) leveraging advanced data-driven approaches to enhance predictive capabilities. These activities will refine our understanding of the magnitude and trajectory of SOC stocks, enabling more accurate predictions and informing sustainable management strategies for global soil resources.