A Comparison of Observed and Simulated Long-Term Ozone

A Comparison of Observed and Simulated Long-Term Ozone Fluctuations and Trends Over the Northeastern United States C. Hogrefe 1, 2, W. Hao 2, E. E. Zalewsky 2, J. -Y. Ku 2, B. Lynn 3, C. Rosenzweig 4, M. Schultz 5, S. Rast 6, M. Newchurch 7, L. Wang 7, P. L. Kinney 8, and G. Sistla 2 1 Atmospheric Sciences Research Center, University at Albany, NY, USA 2 New York State Department of Environmental Conservation, Albany, NY, USA 3 Weather It Is, LTD, Efrat, Israel 4 NASA-Goddard Institute for Space Studies, New York, NY, USA 5 Forschungszentrum Jülich, Germany 6 Max Planck Institute for Meteorology, Hamburg, Germany 7 University at Alabama, Huntsville, AL, USA 8 Mailman School of Public Health, Columbia University, New York, NY, USA CMAS Conference, Chapel Hill, NC, October 19 -21, 2009

Modeling System Setup n n n n Simulation period: 1988 – 2005 Domain: Northeastern U. S. Meteorology: MM 5 v 3. 7. 2 Emissions: NEI 1990, 1996 -2001, OTC 2002, OTC 2009, processed by SMOKE Air Quality: CMAQv 4. 5. 1, CB 4, aero 3 Grid resolution 36 km / 12 km Boundary Conditions for 36 km Grid: n n Time-invariant default climatological vertical profiles (CMAQ/STATIC) Derived from ECHAM 5 -MOZART (CMAQ/ECHAM 5 -MOZART) 2

Domain-Total Anthropogenic NOx and VOC Emissions For the 1988 – 2005 CMAQ Simulations • Substantial reductions in NOx emissions from power plants starting in 1995 • Continuous reductions of NOx and VOC emissions, to a large extent driven by mobile 3 source reductions

Observed and CMAQ/STATIC Simulated Ozone Variability and Trends 4

Overall Model Performance Mean Bias RMSE Normalized Correlation Observed CMAQ (ppb) Bias (%) Error (%) Coefficient (ppb) All Days 1 -hr DM May-Sep 8 -hr DM 59. 6 61. 5 1. 9 15. 2 3. 2 19. 5 0. 72 51. 4 55. 9 4. 5 14. 5 8. 9 22. 0 0. 70 1 -hr DM 94. 6 90. 2 -4. 4 9. 8 -4. 7 8. 5 0. 76 8 -hr DM 82. 4 81. 1 -1. 3 7. 8 -1. 6 7. 7 0. 75 1 -hr DM 28. 5 38. 5 10. 0 11. 4 35. 8 37. 1 0. 19 8 -hr DM 22. 7 34. 9 12. 3 13. 4 55. 8 56. 8 0. 16 95 th %ile

Spectra Calculated from 18 years of Hourly Observed and CMAQ Ozone Time Series CMAQ/STATIC appears to capture the variability in the diurnal and synoptic bands CMAQ/STATIC underestimates variability in the high -frequency (intraday) and lowfrequency (seasonal and longterm) bands of the spectrum 6

Inter-Annual Variability (IAV) of Observed and CMAQ/STATIC 8 -hr Daily Maximum Ozone Observations CMAQ/STATIC • IAV is defined as (standard deviation) / mean • IAV is calculated separately at each site for 5 th, 25 th, 50 th, 75 th, and 95 th percentiles of May – September 8 -hr DM O 3 • The box plots show the distribution of IAV for a given percentile across all sites Percentiles of May – September 8 -hr DM O 3 • CMAQ/STATIC IAV is lower than observed IAV for all percentiles 7

Ratio of CMAQ/Observed Inter-Annual Variability (IAV) of 8 -hr Daily Maximum Ozone The Median IAV Ratio Across All Sites is Shown for Each Percentile CMAQ/STATIC IAV is lower than observed IAV for all percentiles, the underestimation is more pronounced for lower percentiles 8

Time Series of 5 th, 50 th, and 95 th Summertime Percentiles Estimated From May – Sep 8 -hr Daily Maximum Ozone, 1988 - 2005 Calculated for Domain-Wide Ozone Averaged Over All Monitors CMAQ/STATIC appears to capture the trend in the upper range of the ozone distribution rather well, but less so the mid and lower range 9

Observed (top) and CMAQ/STATIC (bottom) Least-Squares Trends in the 95 th (left) and 5 th (right) Percentiles of May – September 8 -hr Daily Maximum Ozone, 1988 - 2005 Observations CMAQ/STATIC Good agreement between linear trends estimated for the 95 th Percentile of observed and CMAQ summertime 8 -hr DM ozone concentrations Observations CMAQ/STATIC While the observed trends at almost all stations are upward for the 5 th percentiles of summertime 8 -hr DM ozone concentrations, CMAQ shows a mixture of upward and downward 10 trends

Observed and CMAQ/STATIC Least-Squares Trends (y-axis) vs. Percentiles of May – September 8 -hr Daily Maximum Ozone, 1988 – 2005 Trends Were Calculated At Each Site, the Median Across All Sites is Shown Here The agreement between the linear trends estimated from observations and CMAQ/STATIC is better for upper than lower percentiles While typical observed trends at stations in the modeling domain tend to be upward for percentiles < 40 and downward for higher percentiles, CMAQ/STATIC trends tend to be downward at all percentiles 11

Impact of Chemical Lateral Boundary Conditions 12

Comparison of Boundary Conditions for CMAQ/STATIC vs. CMAQ/ECHAM 5 -MOZART for Selected Species and Layers O 3 (ppb) NO (ppt) NO 2 (ppt) HNO 3 (ppt) PAN (ppt) Layer Midpoint Height (m) Static ECH Static ECH 1 18 32 49 44 153 89 1, 907 148 904 68 664 8 560 38 55 38 45 76 513 148 875 62 543 10 1, 403 45 57 22 17 44 177 148 802 48 434 12 3, 855 56 63 4 7 8 40 114 259 25 320 13 6, 139 62 69 0 8 0 31 96 180 14 308 14 9, 480 69 69 0 8 0 31 120 180 11 308 15 13, 004 70 69 0 8 0 31 125 180 11 308 n n “Static”: Used for CMAQ/STATIC simulations “ECH”: Used for CMAQ/ECHAM 5 -MOZART simulations, based on monthly average concentrations from archived 1960 – 2006 global ECHAM 5 MOZART simulations performed as part of the RETRO project 13

Impact of Boundary Conditions On Average Daily Maximum 8 -hr Ozone Concentrations As Function of Day -of-Year, Layer 1, 1988 – 2005 Average Observed and Simulated Average Daily Maximum 8 -hr Ozone Concentrations, Averaged Over All Sites Average Daily Maximum 8 -hr Ozone Concentrations, CMAQ/ECHAM 5 -MOZART Minus CMAQ/STATIC, Averaged Over All Sites CMAQ/ECHAM 5 -MOZART generally yields higher concentrations, the differences can be as large as 12 ppb averaged over all sites 14

Differences in Monthly Average Daily Maximum Ozone Concentrations, CMAQ/ECHAM 5 -MOZART Minus CMAQ/STATIC, Layer 1, 1988 – 2005 Average The impact of different boundary conditions on monthly average daily maximum ozone decreases towards the interior of the domain, but still reaches 3 -9 ppb in July for the regions typically exhibiting the highest observed ozone concentrations 15

Comparison of Observed and Simulated Vertical Profiles at Ozonesonde Sites Location of Ozonesonde Sites 1988 -2005 Data Coverage at Ozonesonde Sites 16

Observed and Simulated Vertical Profiles at Huntsville (left) and Wallops Island (right), Concentrations (top) and Standard Deviation (bottom)

Impact of Boundary Conditions on CMAQ Interannual Variability Observed and Simulated IAV for 1988 -2005, defined as (standard deviation) / mean Observations CMAQ/STATIC CMAQ/ECHAM 5 -MOZ • The CMAQ/ECHAM 5 MOZART simulations capture more interannual variability than the CMAQ/STATIC simulations 18

Impact of Boundary Conditions on CMAQ Trend Estimates and Interannual Variability Ratio of CMAQ/Observed IAV by Percentiles; Median Across All O 3 sites in domain, 8 -hr DM Both simulations underestimate observed IAV (ratios < 1), but deriving boundary conditions from ECHAM 5 -MOZART significantly improves the representation of IAV for mid and low percentiles Linear Trends in 1988 -2005 Summertime 8 -hr DM Ozone by Percentiles; Median Across All O 3 sites in domain • The choice of boundary conditions strongly influences the CMAQ trend estimates • CMAQ/ECHAM 5 -MOZART shows a stronger downward trend than either CMAQ/STATIC or observations 19

Summary The 1988 – 2005 CMAQ simulations with boundary conditions derived from the time-invariant default climatological vertical profile exhibit the following features: An underestimation of interannual variability by 30% - 50% depending on the percentiles of the distribution A tendency to capture the trends at the high end of the ozone distribution but not for the central and lower portions The 1988 – 2005 CMAQ simulations with boundary conditions derived from ECHAM 5 -MOZART show: Significant differences in mean ozone concentrations in all seasons throughout the domain, both at the surface and aloft Improved representation of interannual variability Possible propagation of model biases from global to regional scale, especially for lower percentiles of the ozone distribution A strong impact of boundary conditions on trend estimates 20