Sensitivity of Simulated Tropospheric Ozone in the Mid-Atlantic Region in June 2016 Using the WRF-Chem Model

Open Access
Thomas, Andrew M
Graduate Program:
Master of Science
Document Type:
Master Thesis
Date of Defense:
October 11, 2017
Committee Members:
  • Amy Kathleen Huff, Thesis Advisor
  • Fuqing Zhang, Thesis Advisor
  • William Henry Brune, Committee Member
  • ozone
  • air quality
  • numerical modeling
  • air quality model
  • emission
Numerical predictions of tropospheric ozone, a secondary pollutant harmful to human health, contain large sources of uncertainty. Various studies have been conducted on the influence of meteorological initial and boundary conditions, PBL scheme, and emissions inventory, but there is no published comparative study. To address this need, we ran seven simulations of WRF-Chem, a regional online atmospheric chemistry model, with different emissions inventories, PBL parameterizations, and emissions inventories over the eastern United States within the Mid-Atlantic region. The simulations were run for June 2016, which had multiple occurrences of high mixing ratios of observed ozone within the Mid-Atlantic region, particularly along the urbanized Interstate-95 Corridor. Special attention was paid to the 20% of days with the highest observed ozone. Our analysis of the average standard deviation of 8-hour average ozone showed that ground-level ozone was most sensitive to the uncertainty in the initial and boundary conditions, with peaks in average standard deviation of 6 to 7 ppbv, while the emissions inventory uncertainty was of secondary importance with peaks in the average standard deviation of 4 to 5 ppbv. We compared the temporal average of the ensemble members and showed that the emissions inventory has the greatest influence on the average ozone mixing ratio. The ERA-interim generated more ozone than the other initial and boundary conditions ensembles, with ground-level peaks of 70 to 75 ppbv 8-hour averaged ozone. In comparison to the emissions inventory and the meteorological initial and boundary conditions, tropospheric ozone was less sensitive to the PBL scheme. The modeled sensitivity was enhanced aloft over Interstate 95, a source of ozone precursor emissions from mobile sources. Based on our data, we estimated that updating the emissions inventory yields an average decrease of 0.6 ppbv yr-1 in the mean absolute error, and an average decrease of 0.8 ppbv yr-1 in mean daily eight-hour ozone. There are many avenues for further research, including expanding the ensemble size. Our results suggest that the choice of initial and boundary conditions may affect other air quality studies.