6 th Annual CMAS Conference October 3 rd

6 th Annual CMAS Conference, October 3 rd , 2007 Off-line Air Quality Modeling Paradigms: Study of WRF-CMAQ simulations for the Texas. AQS 2000 Episode Fong (Fantine) Ngan, Daewon W. Byun and Soontae Kim IMAQS, University of Houston, Tx

Contents v Introduction v Application of WRF in CMAQ modeling v Result of WRF simulations v Result of CMAQ simulations Full CMAQ simulations for 8/23 – 8/31 2000 (9 days) CMAQ-tracer simulations for 8/27 – 8/28 2000 (2 days) v Summary and future works

Introduction Ø WRF-ARW (Advanced Research WRF) (Skamarock et al. , 2005) new generation operational & research weather forecasting model Providing met. input for air quality modeling Off-line modeling for air quality assessment with WRF-CMAQ WRF-ARW was demonstrated to have accurate numerics and high quality mass conservation characteristics. The governing set of equations, coordinate system, numerical algorithms, and computational framework of WRF-ARW are closer to CMAQ than MM 5. In WRF-ARW, the continuity equation is one of the prognostic equation and dry air density is one of the state variables. v The formulations of WRF can be derived from CMAQ’s FCGSEs by just replacing the Jacobian and defining the vertical coordinate transformation. (Kim and Byun, 2002) v The mass coordinate dynamics of WRF inherits the benefit of the hydrostatic formulations. (Byun and Kim, 2003)

Application of WRF in Air Quality The weather parameters (wind, T and PBL Height) are controlling the development of ozone production and the transportation of other pollutants. A successful meteorological simulation is one of the required steps to predict air quality phenomena realistically under such complex conditions over HGA. The evaluation of WRF performance is necessary before using the output for air quality modeling. Through the comparisons of WRF simulated met. conditions and extensive observations during the Tex. AQS-2000 episode, we are able to assess how WRF simulates the weather condition on HGA, especially on the parameters affecting air quality modeling. v Does WRF generate comparable met. features like MM 5? v What science settings can provide the best meteorological for WRF to model air quality on HGA?

Application of WRF in Air Quality (Cont. ) How to transfer the meteorological information effectively into the chemical model is very important to establish a reliable air quality modeling system for use in applications studies to relate emissions sources and air quality problems. Mass inconsistency problem is a possibility during the transition of the data or the computation inside air quality model because of the dynamic and numeric discrepancies of two models. In air quality modeling, the mass-consistent simulation of trace gas species is a necessary property. (Byun, 1999 a, b) Mass conservation: a principle concept in physics well treated in numerical algorithms Mass consistency: special quality of met. data determined by how the continuity equation is satisfied v What are the advantages for using WRF instead of MM 5 to drive air quality model? v How does WRF maintain mass consistency in off-line coupling paradigms?

6 th Annual CMAS Conference, October 3 rd , 2007 WRF simulations 8/22 – 9/1 (11 days) 1 st layer wind, 1. 5 temperature, PBL height

WRF simulation with grid FDDA – 1 st layer wind vectors grid FDDA W 22 FD In W 22 simulation (no FDDA): Lack of northerly component in the morning & delay in tuning of wind direction from E to SE in the afternoon

Effects of LULC data in WRF modeling WRF v 2. 2 WRF v 2. 1 603 10 WRFv 2. 2 more accurately represents LULC around the coast (but not the best still). Diurnal variation of temperature can be generated well. With the nudging process, the underestimation of maximum temperature, shown in simulation W 21 & W 22 (especially on Aug 30 & 31), was reduced in W 22 FD.

WRF simulation of PBL height W 22 Low PBL height bias was reduced. High PBL bias was from 8/30, 31 (sunny, high temp) W 22 FD MRF & YSU scheme show the same feature. MYJ scheme underestimated for whole period. During the stable condition, PBL top was reported to be the height of the first layer, around 17 m (sigma=0. 996) effect the NOx conc. at nighttime

6 th Annual CMAS Conference, October 3 rd , 2007 Full CMAQ simulations 8/23 – 8/31 (9 days) Ozone, NOx

Full CMAQ simulations with WRF met. data – O 3 conc. downtown NW, loop 610 E, ship channel CBfd & CBw 21 used NEI 99 emissions. O 3 peak couldn’t be generated (eg. 8/25) even though WRF met. data were improved by the grid-FDDA in CBfd. On 8/30 & 8/31, O 3 peaks were affected by met. conditions. Better prediction of O 3 max can be seen in CBfd after improvement of met. data. CBn 1 uses “imputed” 2000 Texas Emission Inventory (2000 TEI) improving ozone simulations. (All emission data were generated by Dr. Soontae Kim)

Spatial plots of O 3 @ 8/25 21 UTC NEI 99 level emission CBfd, wrf with FDDA Imputed level emission CBn 1, imputed emiss With imputed level emission, CBn 1 simulated O 3 max much better when similar met. data from WRF was used.

Full CMAQ simulations with WRF met. data – NOx WRF reports PBL height as the height of 1 st layer (~17 m) at nighttime stable condition. Very high level of NOx was simulated. In CBn 1 simulation, min PBL was set to 50 m over the ocean and 300 m at urban cell under stable condition. Extreme NOx value can be eliminated.

CMAQ tracer simulations with WRF met. data Modeling of a tracer species with CMAQ v Tuning off sources, removal processes, chemical reactions and turbulent diffusion. v IC/BC values of the trace species were set at a uniform distribution of 1 ppm. v Tuning off mass correction process in CMAQ. v Simulation period: 8/27 – 28, 2000 (wind was S to SE). Layer average weighted by sigma The 24 th hour 1 st layer The 24 th hour B A Ideally, the tracer concentration should remain 1 ppm (orange color). By averaging all layers, mixing ratio varies between 0. 99 to 1. 01 ppm after 24 hour simulation. At 1 st layer, much more variations of IC 1_BC 1 happened over the land.

Vertical cross section of IC 1_BC 1 & WWIND (Animation) Right below topopause A PBL over the ocean during nighttime B A Development of PBL during daytime B

1 hourly met. data Time series of mixing ratio. AVE: whole domain average Lay 01: 1 st layer average Lay 43: the top layer average Near SFC layers, mixing ratio change according to the heating & cooling of the surface. Lowest 5 layers Top layer got the largest variation of IC 1_BC 1. 00

15 mins met. data Tracer simulation with WRF output in 15 minutes. Time series of mixing ratio. AVE: whole domain average Lay 01: 1 st layer average Lay 43: the top layer average Pattern of mixing ratio change with the 15 mininterval WRF input is very similar to the same with the hourly WRF input. Higher frequency met data does not necessarily reduce mass error in the CMAQ simulations. Lowest 5 layers

Summary & Future works Ø With the updated physics, FDDA function & nesting ability in the latest version of WRF (v 2. 2), WRF model can be used to provide met. data for AQ study. Accurate LULC data has to be implemented into WRF Grid & OBS FDDA has to be used to model met. condition Ø By using imputed emission & updated met. data from WRF, O 3 peaks during Tex. AQS 2000 period can be generated. Ø Variation of mixing ratio is similar when the hourly & 15 mininterval WRF input is used in the tracer simulations. Total air density Tracer simulation with hourly initialization

~ The End ~