Using CarbonTracker carbon flux estimates to improve a terrestrial carbon cycle model

Estimates of net ecosystem exchange (NEE) from NOAA’s CarbonTracker CO2 data assimilation system show patterns of annual net uptake not represented in most terrestrial carbon cycle models. This is mainly because such models lack information on the land-use history of individual ecosystems, which is the main driver of long-term mean carbon exchange. Instead, they assume the biosphere to be in steady-state, with annual gross photosynthesis equalling ecosystem respiration everywhere. This limits their use in interpreting observations of carbon dynamics such as with eddy-covariance techniques or through atmospheric CO2 records. We have implemented a method that takes the long-term mean NEE estimates from CarbonTracker to derive the size of the dominant carbon pool in each ecosystem of the SIBCASA biosphere model. With the new pool sizes, the SIBCASA model is no longer in steady-state and reproduces annual carbon uptake patterns from CarbonTracker. We will show that the non steady-state SIBCASA model is not only much more consistent with the atmospheric CO2 record, but also with independent data on standing wood biomass and forest age from the Forest Inventory and Analysis (FIA) Program of the U.S. Forest Service. Four years of CarbonTracker NEE are needed to reliably derive a long term mean for this process, and we use three other years from CarbonTracker to evaluate the non steady state SIBCASA NEE. We will furthermore show that the non steady-state SIBCASA NEE is a much better first-guess for the CarbonTracker data assimilation process, allowing more confidence in its final NEE estimate, and reducing a systematic bias in CarbonTracker modeled atmospheric CO2. This overcomes a long standing issue in inverse modeling, and opens the way for further assessment and improvement of carbon cycle models such as SIBCASA.