You are here:
Incorporating principal component analysis into air quality model evaluation
Citation:
Eder, B., J. Bash, K. Foley, AND Jon Pleim. Incorporating principal component analysis into air quality model evaluation. ATMOSPHERIC ENVIRONMENT. Elsevier Science Ltd, New York, NY, 82:307-315, (2014).
Impact/Purpose:
The National Exposure Research Laboratory’s Atmospheric Modeling Division (AMAD) conducts research in support of EPA’s mission to protect human health and the environment. AMAD’s research program is engaged in developing and evaluating predictive atmospheric models on all spatial and temporal scales for forecasting the Nation’s air quality and for assessing changes in air quality and air pollutant exposures, as affected by changes in ecosystem management and regulatory decisions. AMAD is responsible for providing a sound scientific and technical basis for regulatory policies based on air quality models to improve ambient air quality. The models developed by AMAD are being used by EPA, NOAA, and the air pollution community in understanding and forecasting not only the magnitude of the air pollution problem, but also in developing emission control policies and regulations for air quality improvements.
Description:
The efficacy of standard air quality model evaluation techniques is becoming compromised as the simulation periods continue to lengthen in response to ever increasing computing capacity. Accordingly, the purpose of this paper is to demonstrate a statistical approach called Principal Component Analysis (PCA) with the intent of motivating its use by the evaluation community. One of the main objectives of PCA is to identify, through data reduction, the recurring and independent modes of variations (or signals) within a very large dataset, thereby summarizing the essential information of that dataset so that meaningful and descriptive conclusions can be made. In this demonstration, PCA is applied to a simple evaluation metric – the model bias associated with EPA's Community Multi-scale Air Quality (CMAQ) model when compared to weekly observations of sulfate (SO42−) and ammonium (NH4+) ambient air concentrations measured by the Clean Air Status and Trends Network (CASTNet). The advantages of using this technique are demonstrated as it identifies strong and systematic patterns of CMAQ model bias across a myriad of spatial and temporal scales that are neither constrained to geopolitical boundaries nor monthly/seasonal time periods (a limitation of many current studies). The technique also identifies locations (station–grid cell pairs) that are used as indicators for a more thorough diagnostic evaluation thereby hastening and facilitating understanding of the probable mechanisms responsible for the unique behavior among bias regimes. A sampling of results indicates that biases are still prevalent in both SO42− and NH4+ simulations that can be attributed to either: 1) cloud processes in the meteorological model utilized by CMAQ, which are found to overestimated convective clouds and precipitation, while underestimating larger-scale resolved clouds that are less likely to precipitate, and 2) biases associated with Midwest NH3 emissions which may be partially ameliorated using the bi-directional NH3 exchange option in CMAQ.
