Where in the world are we confident in terrestrial carbon balance?

Terrestrial ecosystems play a major role in the global carbon (C) cycle. However, our ability to quantify where in the world is a net C source or sink, and to what extent this is changing continues to be a critical challenge. Terrestrial ecosystems are responsible for the largest C fluxes in the world dwarfing anthropogenic emissions from fossil fuels. These processes are sensitive to climatic and anthropogenic disturbances on varied scales in time and space. This complex interconnection of internal ecosystem processes and external exchanges, mediated by ecosystem properties, challenges both observation and process-based modelling efforts to understand and quantify ecosystem C exchanges.

The expansion of satellite-based Earth Observation (EO) has provided unprecedented information at global scales on the state and evolution of terrestrial ecosystems. Increasingly, these data are provided with more robust estimates of their uncertainties and their variation in space and time. Process-models of terrestrial ecosystems have advanced with our growing ecological understanding derived from in-situ information. However, while there is great potential for EO to contribute to model calibration and validation, helping diagnose ecological function and improve model predictive skill, at present the connections between EO and process models are weakly developed.

Bayesian model-data fusion (data assimilation) approaches offer a powerful opportunity to integrate EO and process-models by informing the model parameter calibration with a diverse, location-specific array of complementary ecologically relevant observations, fully propagating their uncertainties. In this study, will use the state-of-the-art CARDAMOM Bayesian calibration framework to retrieve parameters for a process-based model of the terrestrial ecosystem (DALEC).

We will present a global (0.5 x 0.5 degree) analysis of the global carbon and water cycles for a 21-year period (2003-2023). CARDAMOM is applied uniquely at each 0.5 degree pixel, retrieving uncertainty bounded estimates of DALEC parameters as a function of information available for that location. From these ‘local’ parameters we estimate the state and dynamics of terrestrial ecosystems with fully realised uncertainties in space and time.

Our analysis will identify where in the world we have confidence in source / sink dynamics and diagnose environmental relationships driving current trajectories, including the large growth in atmospheric CO2 concentrations in 2023. Our preliminary analyses suggest this increase is driven by elevated fire activity, particularly in south west Amazon and Canadian boreal forests, and broad enhancement of heterotrophic respiration driven by warming.