Variabilities of atmospheric CO₂ in Austral regions: comparison between simulations and measurements

Strong impetus exists to improve the network of stations monitoring atmospheric CO₂, because of the need to better understand regional sources and sinks of CO₂. 1 helped optimize the Austral network (south of 30 deg. S) by (1) installing a new monitoring station at Kerguelen, and (2) assisting in intercalibration campaigns whose objective is to achieve a global network of CO₂ measurements precise to 0.1 ppmv. From this network of CO₂ measurements the objective of my thesis was to derive the related net fluxes of CO₂ that vary both in space and time. For 1990, 1 linked atmospheric CO₂ concentrations and net fluxes with the aid of a 3-D atmospheric transport model driven by meteorological analyses of ECMWF. Furthermore linkage requires a prior estimation of CO₂ net fluxes. Net fluxes from the terrestrial biosphere are derived from remote sensing measurements of its reflectance; and for the ocean I employed recent apCO₂ measurements in subtropical and sub-antarctic areas. For such analysis, one must take into account three kinds of variability: spatial gradients of CO₂, seasonal cycles, and short term variations (synoptic scale). Atmospheric CO₂ gradients between stations over the Indian Ocean reveal low values near 30 deg. S, and can be explained by the oceanic sink simulated by the model at this latitudes. On the contrary the ocean flux predicted from an extrapolation based on the correlation of ΔpCO₂ with surface temperature in this region, cannot explain the lower atmospheric CO₂ concentrations observed at Cape Grim relative to that at Amsterdam Island (-0.5 ppmv in 1990). Hence further ΔpCO₂ measurements are needed in the Eastern Indian Ocean to confirm a potentially enhanced ocean sink there. Although the distinct seasonal cycle of atmospheric CO₂ observed at Amsterdam Island is partially explained by oceanic fluxes, this contribution plays quite a minor role. Terrestrial biosphere appears to be the main contributor to the seasonal cycles observed south of 30 deg. S. The good agreement between measurements and model derived values (within 0.3 ppmv) indicates the necessity to properly select air masses in the simulations, as it is done in the monitoring stations. For such selection here, 1 used back trajectories calculated with the model. On the other hand the in-depth study of seasonal cycles demonstrates the great influence of interhemispheric transport on the CO₂ variations in the middle and high southern latitudes. To help constrain the model at synoptic scale, 1 also used another tracer, ²²²Rn, having a short half-life (3.8 d), and a continental source. Radon-222 simulations demonstrate that the transport model is able to reproduce synoptic events, both near the surface and in troposphere. For further interpretation of measurements and simulations, it will be necessary in the future to focus not only on the simulation of vertical transport, but also on the precision of measurements, and the simulation of Radon source. Applied to the CO₂, this study demonstrates for the first time the ability of a model to simulate short term variations of CO₂. At Amsterdam and Kerguelen islands, synoptic variations are mostly induced by CO₂ exchange at the sea surface. Hence synoptic variations are sensitive to air-sea gas exchange rate; which remains poorly constrained. The amplitude of the synoptic variations are best matched when the exchange coefficient is calibrated on the mean value given by oceanic measurements of ¹⁴CO₂, which is about 70% higher than estimates made with other trace gases. (author)