Responsibilities of Harvard Atmospheric Chemistry Modeling Group Loretta Mickley, Lu Shen, Daniel Jacob, and Rachel Silvern 2.1 Objective 1: Compile comprehensive air pollution, weather, emissions, and GIS datasets for the entire continental US for the period spanning 2000 2015. 2.1.3 GEOS Chem Simulations. 16 years (2000 2015) 0.5 o x 0.625 o horizontal resolution. NOx VOC PM chemistry informed by our involvement in recent SEAC4RS aircraft campaign. Updated isoprene chemistry. Updated halogen chemistry. Simulations will be started this summer.
Two examples from recent research using GEOS Chem to understand air quality over the Southeast. NEI emissions GEOS Chem with 60% cut in NOx emissions from mobile and industrial sources better matches observations. Travis et al., 2016 Improved emissions 1990 2013 trend in isoprene SOA Improved SOA mechanism We find SOA decreases over 1990 2013 due to decreasing sulfate, which drives trends in aerosol volume and acidity. Old SOA mechanism Marais et al., 2017
2.4 Objective 4. Forecast weather and air quality changes for each region for the period of 2015 2040 using archived results from an ensemble of climate models. Method. 1. Characterize statistically significant relationships between observed meteorological parameters and observed ozone and PM 2.5 for present day. 2. Apply the relationships from #1 to an ensemble of climate model projections for the 2040s atmosphere. IPCC models Start with the present day. Observed relationships between meteorology and air pollutants + Future meteorology Future air quality
1990 2016 mean JJA MDA8 ozone Regardless of emission trends, meteorology can strongly perturb air quality. EPA actions have led to much cleaner air for millions of people Eastern US mean There is large interannual variability in ozone due to the effects of meteorology 2012 drought + heat Site in southern Illinois From Lu Shen s interactive online tool for tracking US summer ozone: https://lshen2009.github.io/
Statistical models have some advantages over dynamical models in projecting future air quality. 1. Chemistry climate models or chemistry transport models Make it easy to investigate the physical processes. May have difficulty in capturing O 3 and PM 2.5 variability. High computational expense. 2. Statistical models Based on observed relationships between meteorological variables and pollutants. Low computational expense. Can take advantage of large ensembles of climate models, leading to more robust results. Can be used to validate dynamical models. Need relatively long records of observations. May fail to include some important physical processes
Recent results. Results 1. Completed project that informs current EPA ACE projects. Results 2 4. New projects, funded by EPA ACE.
1. We find that number of surface ozone episodes per summer increases across the US by 2050s. (a) ozone Change episode in ozone episode days by days 2050s in 2050s Sites where inclusion of Tmax does not improve EVT model. Episode = MDA8 O 3 > 75 ppb days 15 12 9 6 4 3 2 1 0.5 0-1 We build a model using extreme value theory and daily maximum temperatures. We find that ozone episodes increase by 3 9 days per summer in Northeast and California by 2050s. U.S. average change is 2 more episode days per summer. Shen et al., 2016
2. For annual or seasonal PM 2.5, we include synoptic circulation factors into the statistical model. Cross validated coefficients of determination (R 2 ) between observed and predicted 1999 2013 monthly PM 2.5 in US PM 2.5 PM 2.5 mean R 2 mean R 2 Meteorological variables: T, RH, precip, windspeeds Shen et al., 2017a
We find that climate change alone could increase annual mean PM 2.5 uniformly across the East by the 2050s. PM2.5 by 2050s ( ) y μg m 3 1.5 1.0 0.5 We apply the model to large ensemble of climate model output. 0.0 5 0 5 annual mean ( ) -0.5-1.0-1.5 3 We find that annual mean PM 2.5 could increase 1 1.5 μg m 3 in the eastern US, larger than previous estimates. 2 1 0-1 Change is especially large in summer, due to faster oxidation and more greater biogenic emissions. JJA mean -2-3 Shen et al., 2017a
3. On the large scale, we find that sea surface temperatures can influence US air quality. We develop a statistical model to predict summertime ozone variability using meteorology from the previous spring. Correlation of SSTs in spring with JJA ozone in eastern US Eastern ozone MAM ΔSST The model relies on: Memory of springtime SSTs in the following summer. Dependence of US temperatures on SSTs Dependence of surface ozone on US temperatures. Shen et al., 2017b
We predict 45% of the variability of JJA MDA8 ozone in the eastern US using patterns of SSTs and sea level pressure in the preceding spring. ozone anomaly (ppbv) -6-4 -2 0 2 4 6 Mean JJA MDA8 ozone over eastern US model, HF method observations r(ma) =0.59 r(hf) =0.67 obs (HF) model, prediction MA method (HF) prediction (MA) 1980 1985 1990 1995 2000 2005 2010 Ozone has been detrended via two methods: 7 year moving average (MA) 7 year Henderson filtered trend (HF) Shen et al., 2017b
4. We find that teleconnections involving the Atlantic Multidecadal Oscillation can also influence US air quality. AMO and summertime temps in East US Warm AMO Cold AMO SSTs across the north Atlantic show a decadal long variation known as the AMO. Temperatures averaged across East correlate with the smoothed AMO index. PM 2.5 over one half cycle AMO Both PM 2.5 and ozone increase over 1/2 cycle AMO. These effects are overlooked in dynamical models. Shen et al., 2017c
Next steps. Perform and validate 16 year GEOS Chem simulation. Wrap up work on statistical models of future air quality. Interface with Project 4: Use knowledge gleaned from statistical models to help interpret trends of past air quality. Interface with Project 5: Provide GEOS Chem results for presentday. Published papers. Shen L., L.J. Mickley, and L.T. Murray, Influence of 2000 2050 climate change on particulate matter in the United States: Results from a new statistical model. Atmos. Chem. Phys., 17, 4355 4367, 2017a. Shen, L., and L.J. Mickley, Influence of large scale climate patterns on summertime U.S. ozone: A seasonal predictive model for air quality management, PNAS, 114, 2491 2496, 2017b.
The influence of meteorology on air quality can be broken down into different spatial scales. Local meteorology Local T, RH, wind speed, precipitation, etc. Global background Changing climate change e.g., changes in meridional temperature gradient or water vapor Synoptic circulation Meteorological patterns on larger spatial scales (~ 1000 km) -- e.g., polar jet wind and Bermuda High
Projections of future PM 2.5 under a warmer climate using GCM CTMs are sometimes inconsistent. ΔPM 2.5 by 2050s (July, PCM CMAQ) ΔPM 2.5 by 2050s (July, GRE CAPs ) Avise et al. [2009] Day et al. [2015] Previous studies have not evaluated the long term ( ~10 years) sensitivity of PM 2.5 to the major meteorological variables in the CTMs.
2. On the regional scale, we find that PM 2.5 responds to meteorology across many states. Shen et al., 2015, 2017a Correlations of PM 2.5 in one sample grid box with surrounding meteorology Sample grid box in Georgia Mean May June July, 1999 2013 Anticyclonic circulation affects PM 2.5 in gridbox
The statistical model predicts an especially strong response in PM 2.5 to 2050s climate in summer. 2000 2050 changes in PM 2.5 PM 2.5 in JJA ( ) μg m -3 3 2 1 0-1 -2-3 PM( 2.5 ) in DJF μg m -3 3 2 1 0-1 -2-3 Faster oxidation rate, more biogenic emission, and stagnation Volatilization of ammonium nitrate Shen et al., 2017a
4. We find that teleconnections involving the Atlantic Multidecadal Oscillation can also influence US air quality. Warm AMO Cold AMO SSTs across the north Atlantic show a decadal long variation known as the AMO. Temperatures averaged across East correlate with the smoothed AMO index. Temperatures in the East vary by 0.5 1.0 K over one half cycle of AMO, suggesting an influence on air quality. Shen et al., 2017c