EXCITED

The exchange of carbon dioxide between atmosphere and vegetation is currently a poorly constrained part of the global carbon budget. In particular, the exchange of CO2 for tropical ecosystems, which dominates the interannual variability of the atmospheric CO2 growth rate, is uncertain. To model the biospheric CO2 exchange, data-based models have been developed using half-hourly training data from eddy-covariance measurement sites. Although the state-of-the-art models (e.g.,. the FLUXCOM approach) are widely used and successful in the prediction of CO2 fluxes for short time scales in temperate regions, the annual uptake of CO2 by tropical ecosystems is severely overestimated (a net uptake of ~10 PgC/yr for tropical ecosystems). These values are outside a reasonable range, even exceeding the amount of CO2 that is released annually by fossil fuel combustion. In this project, we aim to better constrain the CO2 exchange of tropical ecosystems on longer time scales, by using CO2 flux estimates from inverse models (e.g. CarbonTracker) as additional training data. We will explore different ways of combining the high-frequency and low-frequency training data streams such that the resulting CO2 fluxes are consistent across temporal scales, ranging from hours to years.