Continued from previous page
The interpretation of the comparison of observed and model-predicted concentrations for both organic carbon and black carbon is more difficult because of both inaccuracies in the observations (Section 5.1.2) and the fact that most measured concentrations are only available on a campaign basis. In addition, the source strength and atmospheric removal processes of carbonaceous aerosols are poorly known. Most models were able to reproduce the observed concentrations of BC to within a factor of 10 (see Figure 5.10) and some models were consistently better than this. Both modelled and observed concentrations varied by a factor of about 1,000 between different sites, so agreement to within a factor of 10 demonstrates predictive capability. However, there are still large uncertainties remaining in modelling carbonaceous aerosols.
Figure 5.10: Observed and model-predicted concentrations of black carbon (in ng C m-3) at a number of locations. The models are listed in Table 5.8. Observations refer to those summarised by Liousse et al. (1996) and Cooke et al. (1999). Symbols refer to: circle, Liousse Atlantic; square,Liousse Pacific; diamond, Liousse Northern Hemisphere rural; plus, Liousse Southern Hemisphere rural; asterisk, Liousse Northern Hemisphere remote; cross, Liousse Southern Hemisphere remote; upward triangle, Cooke remote; left triangle, Cooke rural; downward triangle, Cooke urban.
Table 5.9 presents an overview of the comparison between observed and calculated surface mixing ratios. Table 5.9a gives the comparison in terms of absolute mass concentrations while Table 5.9b gives the comparison in terms of average differences of percents. The average absolute error for sulphate surface concentrations is 26% (eleven models) and the agreement between modelled concentrations and observations is better for sulphate than for any other species. The largest difference with observed values is that of carbonaceous aerosols with an average absolute error (BC: nine models, OC: eight models) of about 179%. This may be partly due to the large uncertainties in the estimated strength of biomass burning and biogenic sources. The average absolute error for the dust (six models) and sea salt (five models) simulations is 70 and 46%, respectively.
Continued on next page
Other reports in this collection