Influence of simulated vs. satellite-based burned areas on modelled terrestrial carbon fluxes

The Global Carbon Project provides annual updates on anthropogenic and natural components of the Global Carbon Budget. Dynamic Global Vegetation Models (DGVMs) contribute to these estimates and are used to simulate the evolution of terrestrial carbon sinks. However, DGVMs are known to poorly represent disturbances such as fire, leading to uncertainties in estimates of mean, interannual variability (IAV), and trends in land carbon fluxes. To address this issue, we propose a hybrid-process-based assessmentby constraining three DGVMs (OCN, JULES-INFERNO, and ORCHIDEE-MICT) with remotely-sensed burned areas from ESA CCI (FIRECCI51) and climate data from ERA5 reanalysis. We aim to improve the representation of the spatio-temporal variability of regional carbon budgets, namely fire emissions, above-ground biomass carbon (AGC), and vegetation-related variables—leaf area index (LAI) and gross primary productivity (GPP). Prescribing burned area (BA) in DGVMs reveals contrasting patterns between prognostic (model simulations) and diagnostic (simulations with prescribed BA) runs. As prognostic tends to overestimate BA, particularly across tropical and high-latitude regions, diagnostic simulations correct this issue, by reducing bias and improving the IAV and the agreement with satellite-based datasets of BA and fire emissions in these regions. Moreover, enhanced IAV of AGC is simulated by diagnostic runs, essentially due to better representation of biomass carbon in the mentioned regions. Although moderate improvements are found in LAI and GPP, as the differences between the two runs are more limited, the improvements between prognostic and diagnostic are more evident in their IAV, particularly for LAI, rather than on long-term means, indicating that prescribed fire can improve the representation of some variability patterns. Prescribing remotely-sensed BA in models can lead to a better representation of global BA, fire emissions and AGC, particularly improving the IAV, reducing bias and enhancing the agreement with satellite datasets. The moderate improvements in vegetation-related variables underscore the need to better constrain fire impacts and vegetation dynamics in models, to enhance the simulation of spatio-temporal variability and dynamics of regional-scale vegetation and carbon-related fluxes.