1. The Influence of Lateral and Top Boundary Conditions on Regional Air Quality Prediction: a Multi-Scale Study Coupling Regional and Global Chemical Transport Models Youhua Tang 1, Gregory R. Carmichael 1, Narisara Thongboonchoo 1, Tianfeng Chai 1, Larry W. Horowitz 2, R. B. Pierce 3, J. A. Al-Saadi 3, M. A. Avery 3, G. W. Sachse 3, David G. Street 4, Jeffrey M. Vukovich 5, T. B. Ryerson 6, J. Holloway 6 and other ICARTT Measurement Teams 1 Center for Global and Regional Environmental Research, University of Iowa 2 NOAA GFDL Laboratory 3 NASA Langley Research Center 4 Argonne National Laboratory, Illinois, USA 5 Carolina Environmental Program (CEP), University of North Carolina at Chapel Hill 6 NOAA Aeronomy Laboratory

2. Most regional air quality/chemical transport models use typical concentration profiles as their boundary conditions. These predefined boundary conditions determine the model’s background concentration, inflow airmass characteristics and associated ambient reactions and products. This kind of boundary conditions usually lacks temporal and spatial variations. STEM simulations for ICARTT field experiment (July-August, 2004) were tested with the top and lateral boundary conditions from 3 global chemical transport models (MOZART-NCAR, MOZART-GFDL and RAQMS), and these results were compared with ICARTT aircraft measurements. We utilize this experiment to examine the impacts of temporal and spatial variations in the boundary conditions’ on model performance. This sensitivity study was performed in 60km and nested 12km STEM domains from July-August, 2004 over North America.

3. MOZART-NCARMOZART-GFDLRAQMS Top & Lateral BCs (NCAR) Top & Lateral BCs (GFDL) Top & Lateral BCs (RAQMS) STEM 60km Simulation with NCAR BCs STEM 60km Simulation with GFDL BCs STEM 60km Simulation with RAQMS BCs Comparison and Analysis for ICARTT period Temporal Averaging Time-fixed BCsProfile BCs Spatial Averaging STEM 60km Simulation with Time-fixed BCs STEM 60km Simulation with Profile BCs Comparison and Analysis for ICARTT period Time-varied BCs for the 12km domain Nested 12km Simulation with Time-varied Original BCs Time-fixed BCs for the 12km domain Temporal Averaging Profile BCs for the 12km domain Spatial Averaging 12km Simulation with Time-fixed BCs 12km Simulation with Profile BCs Analysis Framework

4. MOZART-GFDL RAQMS MOZART-NCAR Mean O 3 top boundary conditions (ppbv) during the ICARTT period

5. Mean Lateral-Boundary CO (ppbv) from MOZART-GFDL Mean Lateral-Boundary CO (ppbv) from MOZART-NCAR Mean Lateral-Boundary CO (ppbv) from RAQMS

6. the DC-8 flight #8 on 07/15/2004

7. Observed and Simulated O 3 and CO concentrations for the DC-8 flight 8 on 07/15/2004 Simulations with MOZART-GFDL and RAQMS boundary conditions tend to overpredict O 3 near top troposphere in this scenario

8. MOZART-NCARMOZART-GFDLRAQMS O 3 (ppbv) CO (ppbv) MOZART-NCARMOZART-GFDLRAQMS STEM 60km simulated O 3 and CO concentration in the 10km layer with boundary conditions in the 10km layer from the three global models for DC-8 flight 8 on July 15 (the upper-left shows the flight path).

9. MOZART-NCAR MOZART-GFDL RAQMS STEM 60km simulated O3 and CO oncentrations in the 5.6km layer with boundary conditions from the three global models for WP-3 flight 12 on July 28 (the upper-left plot shows the flight path).

10. Observed and simulated O 3, CO, and NO y concentrations for the WP-3 flight 12 on 07/28/2004

11. Observed and simulated CO and O 3 concentrations for the WP-3 flight 13 on 07/31- 08/01, 2004 Sensitivity of the model to temporal and spatial variations of boundary conditions

12. CO (ppbv)O 3 (ppbv) Oringal BCs Time-Fixed BCs Profile BCs 12km simulated CO (left column) and O 3 (right column) concentrations in the 1km layer, at 0 UTC, 08/01/2004, driven by 3 different boundary conditions. The WP-3 flight path is shown in each plot.

13. 60km Simulated with MOZART-NCAR BCs 60km Simulated with MOZART-GFDL BCs 60km Simulated with RAQMS BCs Species Observed Mean Simulated Mean SlopeR Simulated Mean SlopeR Simulated Mean SlopeR O 3 (ppbv) 67.061.40.430.5768.20.900.5980.21.360.63 CO (ppbv) 112.8105.21.120.67 122.10.950.67100.61.020.64 HNO 3 (ppbv) 0.550.691.180.620.681.170.610.521.150.63 OH (pptv) 0.220.360.920.430.330.770.380.350.810.40 HO 2 (pptv) 13.115.01.040.54 15.51.050.5215.31.000.51 The statistic result of 60km simulations with the three global-model boundary conditions compared with the observations in all ICARTT DC-8 and WP-3 research flights. 60km Simulated with MOZART-NCAR BCs 60km Simulated with MOZART-GFDL BCs 60km Simulated with RAQMS BCs Species Observed Mean Simulated Mean SlopeR Simulated Mean SlopeR Simulated Mean SlopeR O 3 (ppbv) 59.657.41.070.6157.31.120.6059.81.010.62 CO (ppbv) 144.5148.70.980.55162.00.920.53136.60.900.58 NO z (ppbv) 2.873.201.260.433.141.260.433.041.270.45 SO 2 (ppbv) 1.512.060.750.612.070.750.612.070.760.61 Fine Sulfate (μg/std m 3 ) 4.175.530.800.575.550.800.575.770.800.57 DC-8 WP-3

14. Observed and 60km-simulated CO and O 3 mean profiles and standard deviations for all DC-8 flights (upper) and WP-3 flights (lower)

15. 60km Simulated with MOZART-NCAR BCs 60km Simulated with Time-Mean BCs 60km Simulated with Profile BCs Species Observed Mean Simulated Mean SlopeR Simulated Mean SlopeR Simulated Mean SlopeR O 3 (ppbv) 67.061.40.430.5763.40.370.5162.90.350.49 CO (ppbv) 112.8105.21.120.67106.11.060.66105.51.010.66 HNO 3 (ppbv) 0.550.691.180.620.561.090.610.561.100.61 OH (pptv) 0.220.360.920.430.300.780.500.300.770.49 HO 2 (pptv) 13.115.01.040.5415.50.930.5515.50.940.55 The statistic results of 60km and 12km simulations with the original, time-mean and profile boundary conditions compared with the observations in ICARTT DC-8 and WP-3 flights. 12km Simulated with Original BCs 12km Simulated with Time-Mean BCs 12km Simulated with Profile BCs Species Observed Mean Simulated Mean SlopeR Simulated Mean SlopeR Simulated Mean SlopeR O 3 (ppbv) 59.661.81.130.6563.30.570.5163.70.550.48 CO (ppbv) 144.5153.61.070.61151.20.760.57151.10.760.56 PAN (ppbv) 0.420.780.710.470.730.480.410.720.500.44 SO 2 (ppbv) 1.512.640.780.632.620.760.622.620.770.62 Propane (ppbv) 0.380.310.670.640.300.610.630.300.600.63 Acetone (ppbv) 1.911.030.600.691.060.440.601.060.440.58 60km simulations for DC-8 flights 12km simulations for WP-3 flights

16. 60km Simulated CO near the west boundary 60km Simulated CO near the south boundary 60km Simulated CO near the north boundary 60km Simulated CO near the east boundary Mean CO difference and profiles of standard deviation near boundaries of the 60km domain Comparison among the original MOZART-NCAR, time-mean and profile boundary Conditions

17. 60km Simulated CO near the west boundary 60km Simulated CO near the north boundary 60km Simulated CO near the south boundary 60km Simulated CO near the east boundary Mean CO difference and profiles of standard deviation near boundaries of the 12km domain Comparison among the original, time-mean and profile boundary Conditions in the 12km domain

18. Summary Using global model as boundary conditions for regional models is an efficient way to introduce external forcing and improve regional prediction. The regional model’s performance somehow depends on the predicting quality of global models that could differ significantly. The model’s sensitivity to boundary conditions depends on model resolution, domain scale & locations, and relative strength of the inner pollutant emissions and reactions compared to external forcing. The 12km simulation covering Northeastern USA shows stronger sensitivity to lateral boundary conditions than the 60km simulation in altitudes < 5km. Both 12km and 60km simulations show relatively high sensitivity in high altitudes, especially for O 3, as it has strong stratospheric sources.