Operational use of numerical air quality model forecast guidance: Current practice and benchmark skill

Numerical air quality model forecast guidance has recently (since approximately 2004) become routinely available to local air quality forecasters. Operational air quality forecast model guidance is available from both private and public sources, and a growing number of local area models (see, e.g., Pacific Northwest, http://lar.wsu.edu/airpact-3/) are also available. While fine particulate (PM2.5) forecast models are still primarily in the developmental phase, photochemical ozone (O3) models are reasonably mature and stable enough to form part of the standard operational air quality forecast rote. Because coupled meteorological-chemistry models are more complex, and contain many more sources of uncertainty, than the more familiar operational weather models, and this uncertainty can vary significantly across relatively small spatial scales, the adoption of numerical air quality model forecast guidance by members of the far-flung air quality forecasting community has been spotty. This paper presents a variety of methods, easily implemented in the operational forecast environment, to interpret and analyze model guidance in a manner that makes it more useful to the forecaster. As an example, these methods are applied to O3 forecasts in the Philadelphia metropolitan area for the period 2007 to 2009 using the NOAA-EPA National Air Quality Forecast System (NAQFS) output. The results are compared to traditional O3 forecast methods and form a benchmark for the quantification of future improvements in the system.