Comparing the Interannual Variability of Forward and Inverse Models

It is commonly found that inverse models of carbon dioxide exchange from terrestrial ecosystems tend to estimate much more interannual variability (IAV) in CO 2 flux than forward, or bottom-up, models. For instance, the bottom-up model used in CarbonTracker (CASA-GFED2 of van der Werf et al., 2006), predicts a peak-to-peak IAV of 0.2 PgC/yr for North America over the period 2000-2005, whereas after optimizing to agree with atmospheric CO2 observations, the CarbonTracker inverse model finds about four times more IAV. The peak-to-peak variability of 0.8 PgC/yr in North American flux from CarbonTracker is in fact as large as its estimate of the long-term mean uptake over the same region (-0.8 PgC/yr). In part to investigate this difference, the North American Carbon Program has organized a synthesis project to compare inverse and forward models' estimates of North American CO2 exchange over the period 2000-2005. We will report on early results from this effort, using a collection of flux estimates from diverse modeling groups around the world.