Application of the Kolmogorov-Zurbenko filter and the decoupled direct 3D method for the dynamic evaluation of a regional air quality model

Regional air quality models are being used in a policy-setting to assess the changes in air pollutant concentrations from changes in emissions and meteorology. Dynamic evaluation entails examination of a retrospective case(s) to examine whether an air quality model has properly predicted the air quality response to known changes in emissions and/or meteorology. In this study, the Kolmogorov-Zurbenko (KZ) filter has been used to spectrally decompose pollutant time series into different forcings that are controlled by different atmospheric processes influencing the predicted and observed pollutant concentrations. Through analyses of the different components influenced by different forcings as part of dynamic model evaluation, we can discern which of the component(s) or scale(s) of forcing are simulated well by the model and the component(s) or scale(s) of forcing needing further improvement in the model. The KZ filter has been applied to both the observed and Community Multiscale Air Quality (CMAQ) model-predicted summertime ozone (O-3) time series in years 2002 and 2005. The 2002-2005 time period is a good candidate for the dynamic evaluation case study because of the large changes in NOx emissions as a result of the U.S. Environmental Protection Agency's (USEPA) NOx State Implementation Plan (SIP) call together with a gradual decline in mobile emissions. Results suggest that the CMAQ model performs similarly for both years in terms of capturing the observed synoptic-scale forcing. However, the changes in the observed ozone baseline component (i.e. longer-term variations) are not properly captured by the model at some locations. The factors contributing to the ozone baseline include emissions loading, boundary conditions, and other parameters that vary slowly over time. Analysis using a reduced form model developed from the sensitivity coefficients calculated from the decoupled direct method in three dimensions (DDM-3D) reveals that ground-level NOx emissions, especially those from mobile and area source sectors, may be overestimated in 2005 as there is an increase in the mean bias for the 2005 ozone simulations from its 2002 level at the majority of AQS sites located within the states affected by the NOx SIP Call. (C) 2013 Elsevier Ltd. All rights reserved.