CO₂ net flux optimization: atmospheric CO₂ and ¹³CO₂ measurements assimilation in a three dimensional transport model

A primary goal of developing the CO₂ and δ¹³C atmospheric measurement network is to characterize better the sources and sinks of atmospheric CO₂. In this thesis, we present a three dimensional assimilation of CO₂ and δ¹³C measurements within an atmospheric transport model in order to infer sources and sinks of CO₂ at a continental scale (continents and ocean basins) for the period 1985-1995. Solving this inverse problem requires 1) a data space representing monthly CO₂ and δ¹³C measurements (103 sites: surface, ships, planes), 2) a flux space describing a priori fluxes between carbon reservoirs, and 3) a 3-d transport model linking the flux space to the data space. Knowledge of these three elements, together with their associated errors, allows one to reduce the uncertainties in the CO₂ sources and sinks. Modeling errors are not taken into account explicitly but are evaluated by using two different transport models. In the 1985-1995 period, the global continental sink is found to be -2.5±1.5 GtC yr^-1 for an optimized deforestation of +1.3±0.6 GtC yr^-1, yielding a net continental sink of -1.2±1.6 GtC yr^-1. The continental partition of this budget is (in units of GtC yr^-1): North America -0.5±0.6, Europe -0.3±0.7, Northeast Asia -1.3±0.5, Tropics (except Asia) +0.2±1.0, Tropical Asia +0.8±0.4, Southern hemisphere -0.1±0.3. For oceans we find a net global sink of -1.5±0.5 GtC yr^-1 with the following partition (in units of GtC yr^-1): North Pacific -0.3±0.2, North Atlantic -0.7±0.3, Equator +0.5±0.1, 20 deg. S-50 deg. S Oceans -0.8±0.2 and Austral Ocean -0.2±0.1. Several sensitivity tests have been performed on this solution to determine the influence of the a priori flux scenario, atmospheric transport model, measurement network, inter-annual variability in measurements, and δ¹³C atmospheric measurements. All these tests have confirmed the global stability of the control solution within a posteriori error bars, with the exception of the atmospheric transport model that appears to be the major source of uncertainty in determining the carbon budget. Furthermore, δ¹³C atmospheric measurements play only a rather limited role in constraining net CO₂ fluxes within the framework of this study. (author)