Exploring the use of machine learning to detect environmental changes and management options affecting soil carbon storage and greenhouse gas emissions

The amount of carbon locked up in soils dwarfs the amount in the atmosphere and above ground pools. Yet in international dialogues and strategies to tackle climate change, soil carbon is often absent or underplayed. This is partly due to difficulties in accurately assessing the soil organic carbon content of soils at scale.

Measurements taken by hand are now dominant. This means that knowledge of this critical carbon pool is biased towards richer countries with resources to devote to this practice.

With the development of artificial intelligence, and the possibility of combining machine learning and remote sensing via Earth observation, it is possible to develop systems of measuring and monitoring processes affecting soil organic carbon stocks, at scale, in near-real-time.

Our vision is a global monitoring system that can allocate incentives to drive reductions in emissions and sequestration. It would be possible to leverage emissions reductions through such a system and lead towards a more informed climate strategy.

We host a growing portfolio of projects tackling different aspects of this central challenge. As our efforts with machine learning develop, we have brought together a working group, with partner organizations, to tackle aspects of this wicked problem.