Machine learning-based drivers of present and future inter-annual variability in air-sea CO2 fluxes

The inter-annual variability of the air-sea CO2 flux, is non-negligible, can modulate the global warming signal, yet it is poorly represented in Earth Systems Models (ESMs). ESMs are highly sophisticated and computationally demanding, which makes it challenging to perform dedicated experiments to investigate the key drivers of the CO2 flux variability across different spatial and temporal scales. Machine leaning methods can objectively and systematically explore large datasets, ensuring physically meaningful results. Here, we show that a Kernel Ridge Regression can reconstruct the present and future CO2 flux variability in five ESMs. Surface concentration of dissolved inorganic carbon (DIC) and alkalinity emerge as the critical drivers, but the former is projected to play a lesser role due to decreasing vertical gradient. Our results demonstrate a new approach to efficiently interpret the massive datasets produced by ESMs and at the same time offer guidance into future model development and monitoring strategies to constrain the CO2 flux.