Constraining interannual variability of terrestrial carbon fluxes using proxy data in the CarbonTracker long-window data assimilation system

Interannual variability (IAV) represents a critical aspect of understanding changes in the terrestrial carbon cycle. Climate drivers such as temperature and water availability mainly influence the IAV of terrestrial carbon fluxes. Their contributions vary spatiotemporally across different regions and seasons and are simulated with various bottom-up and AI-based terrestrial ecosystem models. However, significant uncertainties remain in simulating terrestrial carbon flux IAV using such models, particularly in the tropics where correlations between temperature and/or water anomalies and atmospheric CO₂ observations were shown to be large. This study demonstrates a data assimilation system that decomposes net ecosystem exchange (NEE) into components across different timescales, with a specific focus on optimizing the poorly constrained IAV. Instead of directly optimizing NEE fluxes, this framework replaces the IAV component with a regression that links NEE IAV to proxy data, such as temperature and water-related variables, as well as light interception by the canopy. This approach allows the system to optimize the sensitivity of NEE IAV to these proxies, providing a robust method to simulate IAV in NEE also for locations and times where the IAV is not directly observed from atmospheric CO₂, or properly simulated by terrestrial biosphere models. This presentation will demonstrate the selection of proxy data and assess their robustness for use in CTE long-window system. The first results from the data assimilation system will be presented and compared to outputs from the regular Carbon Tracker Europe approach (CTE2024). The comparison will focus on quantifying the IAV of NEE and evaluating ecosystem responses to representative extreme events (e.g., heatwaves and droughts), highlighting differences in the system's ability to capture the impacts of such extreme events.