Evaluation of CMAQ O 3 and PM 2.5 performance using Pacific 2001 measurement data

Abstract

Measurement data from the Pacific 2001 air quality field study was used to evaluate the performance of Models-3/CMAQ for the period of 9–20 August in 2001, for a domain centered about Vancouver and the Lower Fraser Valley in British Columbia, Canada. The model predicted O 3 concentrations reasonably well with a normalized mean bias (NMB) and normalized mean error (NME) of 13.3% and 51.2% over all measurement sites and hours. It showed an excellent performance in predicting daily O 3 peaks with a NMB and NME of −2.2% and 24.3%. The model also did well in predicting the time at which the peaks occurred. The model captured the diurnal variations and spatial distributions of O 3 concentrations very well, with some difficulty at locations deep in the valley. Model predictions of PM 2.5 concentrations over all sites and hours had a moderate NMB and NME of 30.9% and 66.2%. The model did well in predicting the general temporal trend of PM 2.5 concentration levels. However, the high hourly variability in PM 2.5 measurements was not possible for the model to capture. In addition, the temporal variations of modelled concentrations change more dramatically than the measured. The model performed poorly in predicting PM sulphate, ammonium, and nitrate concentrations. The extreme over-prediction in nitrate concentrations was partially related to an overabundance of total ammonia in the system. Sensitivity tests of NH 3 emission reductions showed that reducing NH 3 supply to the system could help reduce the over-predictions of both ammonium and nitrate concentrations, although the model science also needs to be improved. To evaluate the PM organics performance, carbon multiplication factors (CMFs) of 1.4 and 1.9 were applied to an urban site and an urban-to-rural transition site, respectively. The model showed satisfactory performance for PM 2.5 organics with a NMB and NME of −6.0% and 50.5%.