Estimating Global Terrestrial Carbon Turnover Times Using Data Assimilation

We estimated carbon turnover times for various terrestrial ecosystems by assimilating estimated CO2 fluxes into the Simple Biosphere Model, Version 2 (SiB2). CO2 fluxes for TRANSCOM3 regions were estimated by inversion of flask measurements using a transport model. We simulated the terrestrial NEE using SiB2 for 1983 to 1993 on a global, 1° by 1° latitude/longitude grid with a 10-minute time step. We iterated for optimal turnover times for 11 ecosystem types using the inverse Hessian assimilation technique. Since the SiB2 response to turnover time is grossly non-linear, we first optimized the linear correlation between the measured and simulated NEE as a first guess to the inverse Hessian technique. The results are reasonable estimates for those carbon pools with relatively short turnover times that are most sensitive to inter-annual variations in climate. Tropical forests and grasslands have shorter turnover times and mid- and high-latitude forests and tundra have longer turnover times.