We present a synthesis
Bayesian inverse method to optimize one year of daily fluxes at model
resolution (50x50 km over Europe) by inversion of continuous CO2
measurements, daily averaged over Europe (10 sites). Based on a synthetic data
analysis, we studied the impact of three different spatial and temporal
correlations on flux errors. We found that the present network is too sparse to
efficiently constrain European fluxes at model resolution even with the
assumption of perfect transport. However, the agreement between the optimized
fluxes and the true fluxes is improved when aggregated in space and time,
mainly for 8-10 days fluxes over Western Europe.
This region is indeed surrounded by our network. The spatial correlation scheme
used was found to have a negligible impact on this agreement. Adding a white
noise on pseudo-data to simulate transport model errors largely degrades the
agreement. Using real data, European flux variations becomes unreasonably large
due to the inability of our transport model to properly represent the CO2
concentrations at continental sites.
Author: C. Carouge, P. Bousquet, P. Peylin, P. Ciais and P.J. Rayner (claire dot carouge at cea dot fr)
Filesize: 105.99 Kb