Linking carbon dioxide variability at Hateruma Station to East Asia emissions by bayesian inversion

著者: Zeng J., Nakajima H., Matsunaga T., Mukai H., Hiraki K., Yokota Y.
雑誌名: Lagrangian Modeling of the Atmosphere, Lin J., Brunner D., Gerbig C., Stohl A., Luhar A., Webley P. (eds.), American Geophysical Union, 163-172
DOI: 10.1029/2012GM001245
概要: We introduce a new approach to use a Lagrangian particle dispersion model (LPDM) and a Bayesian inversion method to study the variability of atmospheric carbon dioxide (CO₂) observed at the monitoring station (24.05°N, 123.8°E) on Hateruma Island, Japan. The 3 hourly means of CO₂ mixing ratio in the winters of 2006–2010 were used to infer fossil fuel carbon emissions from East Asia. The LPDM was used to compute the back trajectories of 10,000 particles for 10 days for each CO₂ data point. The residence times of particles in 1° × 1° grid cells within 500 m above the surface were used as the model parameters that establish the source-receptor relationship between the emissions and the increment of CO₂ over the baseline. The results from Bayesian inversion indicate larger emissions from some regions in southern China than what the a priori emission data suggest; however, the inversion results are sensitive to the parameterization of covariance matrices for data and model parameters. Our analysis shows that using small uncertainties for the CO₂ increments and large a priori uncertainties for carbon emissions generally results in good correlations between modeled and observed CO₂, but the corresponding a posteriori emissions may show unrealistic emission patterns; therefore, it is important to estimate a priori errors properly in using Bayesian inversion.