17.4 Use of Generalized Additive Models to Understand the Meteorological Influences on O3 and Identify Possible Exceptional Events

Thursday, 26 January 2017: 2:15 PM
4C-3 (Washington State Convention Center )
Dan Jaffe, University of Washington, Bothell, WA; and X. Gong and P. D. Sampson

Ozone is a complex secondary pollutant that is formed from photochemical reactions of nitrogen oxides and hydrocarbons. In most of the US, anthropogenic emissions of these precursors have decreased, resulting in significant reductions in ozone mixing ratios. At the same time the ozone standard has become tighter and emissions of these precursors in the western US from wildfires may be increasing and causing exceedances of the air quality standard.  EPA rules allow for exclusion of such “exceptional events” if it can be shown that an exceedance is due to a natural or uncontrollable source. Thus attribution of the sources of ozone is becoming increasingly important. Because wildfires are highly variable and emissions are significantly different from typical industrial emissions, we wanted to find a method that could directly model the observed ozone as a function of key meteorological variables.  Generalized Additive Models (GAMs) appear to be an ideal tool for this work. We have used GAMs with key meteorological variables to model the Maximum Daily 8-hour average (MDA8) ozone for a number of cities in the Western US. The models have the ability to predict summer ozone with R2 values between 0.5-0.8. The residuals from the GAM can then suggest dates with unusual sources of ozone. We have applied this method to large fires that burned in Washington State in summer of 2015. The smoke from these fires was transported widely across the western US. We used a GAM for each individual city that was impacted. The GAM residuals indicate that wildfire ozone increased with distance from the fire, due to secondary production, and that the fires contributed to high MDA8 values in urban areas up to 1000 km away.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner