Investigation of the Seasonal Biases of PM2.5 Concentrations from the Community Multiscale Air Quality Model

RICK SAYLOR (1), Binyu Wang (1), Yunsoo Choi (1), Pius Lee (1), Tianfeng Chai (1), Hyun-Cheol Kim (1), Hsin-Mu Lin (1), Daniel Tong (1), Fantine Ngan (1), Ariel Stein (1), Daewon Byun (1)

(1) NOAA Air Resources Laboratory

     Abstract Number: 160
     Last modified: April 27, 2010

Abstract
The Community Multiscale Air Quality Model (CMAQ) PM$_(2.5) total mass is generally overpredicted during cool months of the year and underpredicted during warm months. Previous evaluations of this seasonal bias have focused on domain-wide averages over limited time spans. In this work we focus on comparisons of model PM$_(2.5) predictions with data from networks with higher temporal and chemical resolution. The goal of this analysis is to focus on model and observed PM$_(2.5) speciation changes over selected time periods at specific locations and from these comparisons gain insight into model processes or inputs that are deficient and in need of improvement.

PM$_(2.5) results from NOAA’s developmental PM track of the National Air Quality Forecasting Capability (NAQFC) have been initially analyzed with respect to available PM$_(2.5) observations from the AIRNow network through all of calendar year 2009 (CY2009). In general, results from CY2009 were consistent with previous evaluations in that total mass was overpredicted during winter months and underpredicted during summer months. Four periods during CY2009 were identified for further analyses, but the initial focus of this work is August 14-16, 2009, during which a rapid transition occurred where the domain-averaged model PM$_(2.5) went from underprediction to overprediction over the span of just a few days.

Continuous PM$_(2.5) mass and speciation data for August 2009 from the Southeastern Aerosol Research and Characterization (SEARCH) study network have been compared against model predictions at the network locations. Initial analysis indicates that the unspeciated components of PM$_(2.5) dominate the model bias prior to the transition while the organic matter (OM) component dominates the model bias afterward. Further data analysis for this time period at additional sites will allow hypotheses to be generated and tested with retrospective model simulations as to the reasons for the observed transition of model bias over this time period.