Sensitivity of a Chemical Mass Balance model to different molecular marker traffic source profiles

Use of the Chemical Mass Balance (CMB) model for aerosol source apportionment requires the input of source profiles of chemical constituents. Such profiles derived from studies in North America are relatively abundant, but are very scarce from European studies. In particular, there is a lack of data from European road vehicles. This study reports results from a comparison of road traffic source profiles derived from (1) US dynamometer studies of individual vehicles with (2) a traffic profile derived from measurements in a road tunnel in France and (3) new data derived from a twin-site study in London in which concentrations at an urban background site are subtracted from those measured at a busy roadside to derive a traffic increment profile. The dynamometer data are input as a diesel exhaust, gasoline exhaust and smoking engine profile, or alternatively as just a diesel exhaust and gasoline exhaust profile. Running the CMB model with the various traffic profiles together with profiles for other sources of organic carbon gives variable estimates of the contribution of traffic to organic carbon and to PM2.5 concentrations. These are tested in two ways. Firstly, unassigned organic carbon in the output from the CMB model, assumed to be secondary organic carbon, is compared to secondary organic carbon estimated independently using the elemental carbon tracer method. Secondly, the estimated traffic contribution to organic carbon and PM2.5 is compared with an estimate derived simply from the measured elemental carbon concentrations, and the effect on aerosol mass closure is investigated. In both cases the CMB model results correlate well with the independent measures, but there are marked differences according to the traffic source profile employed. As a general observation, it appears that the use of dynamometer data with inclusion of a smoking engine profile has a tendency to over-estimate traffic emissions at some sites whereas the tunnel profile shows a tendency to under-estimate. Overall, the traffic profile derived from the twin-site study gives probably the best overall estimate, but the quality of fit with independent estimates of secondary organic carbon and traffic particle mass depends upon the site and dataset for which the test is conducted.