The atmosphere integrates surface–atmosphere exchange across spatial and temporal scales. Monitoring and interpreting atmospheric greenhouse gases (GHGs), such as carbon dioxide (CO2) and methane (CH4), therefore provide fundamental information on carbon fluxes from natural and anthropogenic sources and their links to energy and water cycling. Combining observations with process-based and data-driven modelling can constrain poorly understood processes, particularly in the terrestrial biosphere, and reveal the mechanisms and climatic drivers governing GHG exchange and carbon–climate feedbacks. Such process-level understanding is essential for predicting how these feedbacks will evolve in a changing climate. Machine-learning (ML) approaches provide additional opportunities to extract information from increasingly diverse and extensive datasets.

This session focuses on top-down, bottom-up, and hybrid modelling frameworks that investigate GHG exchange using observational platforms such as surface and ecosystem networks (e.g. ICOS Atmosphere and Ecosystem, FLUXNET, NOAA), aircraft campaigns (e.g. MAGIC and COMET), and active and passive remote-sensing missions (e.g. ECOSTRESS, OCO-2/3, TROPOMI, and GOSAT). We particularly welcome studies that combine data sources not traditionally used together to constrain GHG fluxes and the underlying processes. Contributions may explore how the integration of diverse observations with process-based models, machine learning, or novel data-assimilation approaches can identify climatic controls, improve process representation, and advance understanding of carbon–climate feedbacks.