Decoupled direct 3D sensitivity analysis for particulate matter (DDM-3D/PM)

Abstract

The decoupled direct method (DDM) and DDM-3D have been implemented in air quality models in order to efficiently compute sensitivities. Initial implementation of DDM/DDM-3D in models was confined only to gas-phase species as the treatment of sensitivities in the dynamics of secondary aerosol formation is more complex. Here, it is extended to calculate particulate matter sensitivities. DDM-3D/particulate matter (PM) results compare well spatially and temporally with the traditional brute-force approach, particularly for species responses to emissions of their “parent” precursor (e.g., sulfate to SO₂ emissions.) Correlations of more indirect relationships between aerosols and gaseous emissions (e.g., nitrate to SO₂ emissions) are worse, but these sensitivities are usually small. DDM-3D/PM appears to work better than the brute-force approach in some cases due to numerical noise and other factors, as identified from the application on a southeastern US domain for a summer episode. DDM-3D/PM is also computationally efficient. While CPU usage was found to scale linearly with the number of sensitivity parameters of interest (for a given domain size), it was significantly less than using the brute-force approach.

Introduction

Policy makers are tasked with developing control strategies to lower levels of air pollutants, including particulate matter (PM). In order to simulate various different strategies, the ability to quickly and efficiently calculate the response of ambient concentrations of various gaseous and aerosol pollutants in the atmosphere to changes in emissions is increasingly more important. Due to the non-linear relationship between ozone and secondary particulate matter components with their precursors, several directions might exist to lower final ambient concentrations, and the levels of control required are not immediately known. The decoupled direct method in 3D (DDM-3D) has proven to be a powerful and efficient approach to identifying how sources impact ozone and PM air quality for use in policy development (Dunker, 1984; Mendoza-Dominguez et al., 2000; Dunker et al., 2002; Odman et al., 2002). DDM-3D operates with an underlying atmospheric model to simulate pollutant concentrations and simultaneously compute local sensitivities of pollutant concentrations to perturbations in input parameters such as emission rates and initial and boundary conditions.

During the Southern Appalachians Mountains Initiative (SAMI), DDM-3D was extended in the urban-to-rural multiscale model (URM) to treat secondary particulate matter formation in addition to gas-phase species (Odman et al., 2002). Later, Dunker et al. extended the CAMx air quality model to include DDM for treatment of gas-phase pollutants (Dunker et al., 2002). Recently, Cohan et. al. implemented the higher-order DDM-3D for gas-phase species in the Community Multiscale Air Quality (CMAQ) model to develop an optimized air quality strategy in the state of Georgia (Cohan et al., 2005) following the work of Hakami et. al. (Hakami et al., 2003). CMAQ is a widely used air quality model with detailed physical and chemical treatment of gaseous- and condensed-phase pollutant dynamics (Byun and Ching, 1999). Here we describe the extension of CMAQ (version 4.3) DDM-3D to include PM formation and transformation (DDM-3D/PM), evaluation of its accuracy and computational performance.