Comparing modeled and observed smoke concentrations for the 2007 Southern and 2008 Northern California wildfires

The California wildfires of 2007 and 2008 had significant smoke impacts on air quality across the state and beyond. Throughout these wildfire events, a large number of inter-agency monitors recorded fine particulate matter (PM2.5) surface concentrations. At the same time, several regional and national models produced smoke concentration predictions. We examine how these model predictions compared with the observations as collected by AirNow. The observation data allow for direct evaluation of overall and site specific model surface concentration predictions. The modeled data were analyzed using observed data from 15 (2007) and 134 (2008) stations. The 2007 analyses include stations that were primarily hit with smoke throughout the duration of the event while the 2008 analyses include stations that may not have been directly impacted by wildfire smoke. The analyses show that overall, the models performed well but tend to mis-time the observed concentration peaks. These statistical analyses identify ways that the model concentrations may be meaningfully and usefully translated, and highlight the need for probabilistic forecast interpretations.