1. Importance of chemistry-climate interactions in projections of future air quality Loretta J. Mickley Lu Shen, Daniel H. Cusworth, Xu Yue Earth system models

2. How will surface air quality change in changing atmosphere? Number of people living in U.S. counties with air quality concentrations above the NAAQS thresholds in 2014 One or more pollutants Ozone (8-hour) PM 2.5 (annual or 24-hour) PM 10 (24- hour) increasing temperatures? trends in stagnation? increasing wildfires in the West? ….. http://www3.epa.gov/airtrends/aqtrends.html How will these bars change in response to:

3. St. Louis Number of days MDA8 ozone > 75 ppb 1980 2014 Anthropogenic emissions Even as emissions decline, meteorology drives variability in air quality. GDP

4. 1. Chemistry-chemistry models Two ways to quantify air pollution in future climate. 2. Statistical models Construct relationships between meteorological variables and pollutants. Statistical relationships are then applied to ensemble of climate projections. winds Winds carry pollutants. Emissions + chemistry calculated within box T, RH, etc. Site in Maryland mean One  MDA8 ozone, JJA 2003-2012 Shen et al., in review

5. Statistical method to project future air quality can take advantage of large ensemble of climate projections. Yue et al., 2013  Temp JJA 2000-2050 oCoC 2000-2050 change in different ecoregions  Temp Temp (K) Close analysis reveals large spread among model projections. We can thus calculate statistical significance of results.

6. 1. Build statistical model of area burned using meteorological variables and fire indices. 2. Apply statistical model to CMIP3 climate projections. 1. Projections of wildfire activity and smoke concentrations over the western United States by 2050s Yue et al., 2013, 2014, 2015 Area burned (10 5 ha) R 2 ~ 0.5 Area burned over Southwest model observed Area burned (10 5 ha) Timeseries of area burned observed model 3. Apply median areas burned for present-day and future into GEOS-Chem. Area burned in Southwest doubles by 2050.

7. Much of California sees a doubling of “smoke waves” – episodes of enhanced smoke at the 98 th % level. Many populous counties in the West experience 40-150% increases in smoke PM 2.5 by 2050s. Percent increase in smoke PM 2.5 by midcentury relative to 2000s, by county. Large cities affected by fires With collaborators from Yale, we are examining the health impacts using Medicare data. Liu et al., in review Yue et al., 2013, 2014, 2015

8. 2. Projection of U.S. ozone episodes in 2050s atmosphere. We build a model using extreme value theory. Probability of surface ozone above a specified threshold. Only meteorological variable considered is Tmax. location factor scale factor shape factor Model also needs to take into account the phenomenon known as “ozone suppression” at high maximum temperatures. Site in Maryland mean One  Observed MDA8 ozone, JJA 2003-2012 Shen et al., in review

9. Ozone suppression at high temperatures is a widespread phenomenon across the United States. Sensitivity studies with GEOS-Chem suggest the cause is due to the nonlinearity of meteorological effects on ozone during heatwaves. We include suppression in statistical ozone model and apply to 19 CMIP5 models. Sites demonstrating significant decrease in O 3 -max T slope at high temperatures. AQS Shen et al., in review Steiner et al. (2010) found similar phenomenon in California.

10. Mean change in ozone episode days due to climate change in by 2050s in RCP4.5 scenario. Episode = MDA8 O 3 > 75 ppb Statistical model finds that number of surface ozone episodes per summer increases by 2050s. Ozone episodes increase by 3-9 days yr -1 in Northeast and California. U.S. average change is ~2 days yr -1.  episode days 2000-2009 vs 2050-2059 Sites where inclusion of Tmax does not improve EVT model. Shen et al., in review

11. New results suggest that annual mean PM 2.5 could increase 1-1.5 μg m -3 in the eastern US, larger than previous estimates. μg m -3 New study, Lu Shen Predictors: Local meteorology and regional scale circulation patterns Tai et al. (2012) Predictors: local meteorological variables μg m -3 3. Improved projections of PM 2.5 concentrations in 2050s. Here focus is on climate penalty on anthropogenic and biogenic PM 2.5. Change in annual mean PM 2.5 by 2050s Note different color bars!

12. μg m -3 ΔPM 2.5 in JJAΔPM 2.5 in DJF Faster oxidation rate, more biogenic emission, and stagnation Volatilization of ammonium nitrate Statistical model predicts different responses in different seasons for change in PM 2.5 in 2050s atmosphere. Lu Shen, work in progress

13. Dynamical models may have difficulty capturing observed sensitivity of monthly mean PM 2.5 to temperature. Slopes of monthly mean JJA PM 2.5 vs surface temperatures μg m -3 K -1 slope > 1 μg m -3 K -1 within contour Lu Shen, work in progress observations ACCMIP model another ACCMIP model We are investigating the mismatch with observed slopes in GEOS-Chem. Preliminary results point to problems in capturing the sensitivity of sulfate to temperature.

14. Take home messages. Many populous counties in the West could experience doubling of smoke PM 2.5 by mid-century. Ozone episode days could increase by as much as 3-9 days per summer in the 2050s due to climate change. Summertime PM 2.5 concentrations could increase 3  g m -3 by 2050s over much of the eastern United States. We are working with the CDC to put together an interactive website with these results.