7.3
Development of an Automated Statistical Air Quality Forecasting System
PAPER WITHDRAWN
Timothy S. Dye, Sonoma Technology, Inc., Petaluma, CA; and D. S. Miller and C. P. MacDonald
Providing advanced warning of unhealthy air quality to the public is becoming increasingly important to the U.S. Environmental Protection Agency (EPA) and state and local air quality agencies. These warnings need to be accurate, consistent, and timely to gain the public’s confidence in using air quality predictions to change personal or organizational behavior. Because the science of ozone forecasting is still quite young, many of the techniques currently used by air quality agencies to forecast ozone do not provide these needed qualities. In this paper, a technique is discussed that may help elevate this forecast quality issue. One statistical tool that improved weather forecasting is Model Output Statistics (MOS). MOS is based on multi-linear regression that relates surface weather variables to numerical weather prediction (NWP) model output. MOS’s strength is in developing the statistical equations using NWP model forecast fields that contain hundreds of weather variables, many of which are not available from traditional observations.
This paper presents the development, operation, and initial results of an air quality forecasting system developed using a similar approach to MOS. Five years of historical NWP forecast fields from the 40-km Eta model were used to develop statistical relationships between forecasted weather parameters and air quality data (ozone and particulate matter). The air quality data were correlated with forecasted weather variables (900 mb temperature, 700 mb wind speed, etc.) to identify key predictor variables. Regression equations were developed using the key variables in the high spatial/temporal data set provided by the NWP model. The regression equations were then tested for several air pollution episodes during 2001.
Session 7, integrated modeling/measurement systems for emissions and air quality predictions (Parallel with Joint Sessions J3 & J4)
Thursday, 23 May 2002, 9:00 AM-12:57 PM
Previous paper Next paper