Improvement of predicted fine and total particulate matter (PM) composition by applying several different chemical scenarios: A winter 2005 case study

A new method using several different chemical scenarios is developed to predict chemical composition of fine (PM2.5) and total (PM10) aerosol. This method improves the accuracy of predicted PM concentrations. The Mesoscale Model version 5 (MM5) and a 3-dimensional Eulerian chemical model (CAMx4.2) are used to predict PM2.5 and PM10 concentrations using gridded input emissions (from the “Total” group) over a 48–72 h time period for Christchurch (New Zealand) for winter 2005. The aerosol concentrations are obtained for four different chemical compositions (chemical scenarios) of the input aerosol emissions. PM2.5 chemical compositions are based on previous Christchurch winter studies and from observations in other countries with similar winter pollution problems, and used in CAMx4.2 to model seven winter 2005 heavy pollution episodes. The error between observed and modelled PM2.5 concentrations is based on predictions of fine aerosol that are derived from linear regression with PM10. It is used to find the minimum difference between modelled and observed PM2.5 for an observation site located in the Christchurch residential area. Combination of the chemical scenarios with analysis of the minimum error is used to create a new complex chemical scenario. The new complex scenario is used to re-calculate all pollution episodes to obtain new values of PM with minimum error compared with observed aerosol concentrations. Mean Absolute Error of the calculated PM2.5 (for all pollution episodes) decreased from 21–24 μg m− 3 to 14–16 μg m− 3 compared with observations. The chemical composition of the modelled PM2.5 is also discussed.