12th Conference on Atmospheric Chemistry
16th Conference on Air Pollution Meteorology

J11.5

AQMOS: Air Quality Model Output Statistics

Dianne S. Miller, Sonoma Technology, Inc., Petaluma, CA; and C. P. MacDonald, T. S. Dye, and K. Craig

Air quality forecasters routinely use numerical air quality models for forecast guidance, such as the National Oceanic and Atmospheric Administration (NOAA) ozone and the BlueSky Gateway PM2.5 models to issue local forecasts.  These models are still rapidly evolving and continue to contain biases in their predictions.  Air Quality Model Output Statistics (AQMOS) was developed in an attempt to remove these biases.  AQMOS is a tool that dynamically computes regressions between recent historical air quality model predictions and observations.  These regressions are model run, site specific, and applied to the current model forecast for over 300 forecast cities in the AIRNow Program (www.airnow.gov).  Because the models are evolving, only data from the last 365 days are considered for the regression.  In addition, only historic warm-season data are used during the warm season months (April through October) and vice versa for the cold season.  Furthermore, the tool filters data based on current predicted concentrations.  If it is within the highest 15% of historical forecast concentrations for the same model run and site, only days with historical forecasts in the highest 15% range are used in the regression; otherwise, all days are included.  In this presentation, we will describe how AQMOS works, evaluate the performance of AQMOS, and demonstrate the website on which forecasters can access predictions.

Recorded presentation

Joint Session 11, Air Quality Forecasting II
Thursday, 21 January 2010, 3:30 PM-5:00 PM, B316

Previous paper  Next paper

Browse or search entire meeting

AMS Home Page