Analysis of urban pollution episodes by inverse modeling

Abstract

Urban pollution episodes pose two relevant issues: a) was the episode controlled by specific meteorology, a rise of emissions or both? b) Were mitigation measures effective in curbing down pollution? A methodology for answering those questions comes from an inverse modeling approach. In this work we have applied the methodology to the city of Santiago, Chile for which the required input data are available. We use a Kalman filter and ambient observations to constrain sources of tracers such as CO, elemental carbon and suspended street dust. The period analyzed is the week from May 20th till May 26th 2005. We find that a posteriori CO emissions were 76% of the a priori estimates. For suspended street dust a posteriori values are 36% and 21% of the prior values for coarse and fine fractions,
respectively. Elemental carbon emissions are underestimated in the prior inventory – we find a correction factor of 1.53 for the whole week. Sensitivity analyses tested the robustness of a posteriori estimates, generating ensembles of simulations for different modeling options. For different initial prior estimates, the ratio of standard deviation to mean values was below 0.20 for 75% of the a posteriori, estimated emissions. For different choices of the error covariance matrices and model errors those ratios were below 0.30 for 75% of a posteriori emissions, which shows the robustness of results for different parameter choices – only a small fraction of results were not significant. The high pollution peaks on May 21st are due to specific meteorological conditions and increased traffic emissions as well. Contingency measures taken on Sunday May 22nd and better dispersion conditions on Monday May 23rd stopped the
accumulation of those pollutants, showing the effectiveness of short term strategies such as traffic bans and street sweeping operations in curbing down traffic pollution at Santiago.