Department of Earth System Science

 Quantification of an ecosystem's carbon balance and its components is pivotal for understanding both ecosystem functioning and global cycling. Several methods are being applied in parallel to estimate the different components of the CO₂ balance. However, different methods are subject to different sources of error. Therefore, it is necessary that site level component estimates are cross-checked against each other before being reported. Here we present a two-step approach for testing the accuracy and consistency of eddy covariance–based gross primary production (GPP) and ecosystem respiration (Re) estimates with biometric measurements of net primary production (NPP), autotrophic (Ra) and heterotrophic (Rh) respiration. The test starts with closing the CO₂ balance to account for reasonable errors in each of the component fluxes. Failure to do so within the constraints will classify the flux estimates on the site level as inconsistent. If the CO₂ balance can be closed, the test continues by comparing the closed site level Ra/GPP with the Rh/GPP ratio. The consistency of these ratios is then judged against expert knowledge. Flux estimates of sites that pass both steps are considered consistent. An inconsistent ratio is not necessarily incorrect but provides a signal for careful data screening that may require further analysis to identify the possible biological reasons of the unexpected ratios. We reviewed the literature and found 16 sites, out of a total of 529 research forest sites, that met the data requirements for the consistency test. Thirteen of these sites passed both steps of the consistency cross-check. Subsequently, flux ratios (NPP/GPP, Rh/NPP, Rh/Re, and Re/GPP) were calculated for the consistent sites. Similar ratios were observed at sites which lacked information to check consistency, indicating that the flux data that are currently used for validating models and testing ecological hypotheses are largely consistent across a wide range of site productivities. Confidence in the output of flux networks could be further enhanced if the required fluxes are independently estimated at all sites for multiple years and harmonized methods are used.