The effect of recent regional emissions reductions on air quality forecasting in the mid-Atlantic

Significant reductions in emissions of oxides of nitrogen, a precursor of both ozone (O₃) and fine particles (PM_2.5), from large power generation facilities have occurred in the eastern United States beginning in 2003 as a result of the so-called “NO_x SIP Rule”.  This reduction has lowered regional O₃ concentrations and may also have lowered PM_2.5 concentrations although the limited observation network for the latter prior to 2003 limits any conclusions in this regard.  Emissions reductions can also occur on transient, season-to-season, scales as evidenced by lower emissions during the summer of 2009 due to the economic recession.    These secular changes in emissions can have a significant impact on air quality forecast skill.  While forecast guidance from operational coupled weather and chemistry models have been available since the mid-2000's, the skill of these models vary by location.   The complexity of the PM_2.5 formation process further limits the skill of numerical model guidance.  As a result, many forecasters continue to rely on empirical and statistical methods that typically require training over years of historical data.  By definition, these models are expected to lose skill as the emission base changes over time.  In this paper, we investigate changes in summer season O₃ and, to a lesser extent, PM_2.5 concentrations since 2002 in the mid-Atlantic and their relationship to changes in forecast skill in the Philadelphia metropolitan area.  In particular, we focus on two recent seasons, 2009 and 2010.  The summer of 2009 was cooler than normal and precursor emissions were also impacted by the economic recession.  The summer of 2010, on the other hand, was extremely warm, particularly during the heart of the summer O₃ season (June 1-July 31).  We show that coupled meteorology-chemistry numerical models, in this case the NOAA-EPA National Air Quality Model (NAQ), are better able to adjust to changing emissions environments than stand-alone statistical methods and show greater forecast skill.  In addition, we show that reductions in regional scale precursor emissions have reduced the frequency of severe O₃ cases (Code Red, 8-hour O₃ ≥ 96 ppbv) and may also have re-introduced day of week variations in observed O₃.