Simulating diurnal changes of speciated particulate matter in Atlanta, Georgia using CMAQ

1
Simulating diurnal changes of speciated
particulate matter in Atlanta, Georgia using CMAQ
Yongtao Hu, Jaemeen Baek, Bo Yan, Rodney Weber,
Sangil Lee, Evan Cobb, Amy Sullivan, Armistead G.
Russell
School of Civil and Environmental Engineering and
School of Earth and Atmospheric Sciences, Georgia
Institute of Technology
CMAS conference, October 18th, 2006
Acknowledgements Eric S. Edgerton and John
Jansen ARA and Southern Company
2
Background
3
Speciated particulate matter monitored at two
sites in Georgia Tech's campus, 500m away from
each other
4
Measurements at neighboring sites HYW and ROF

• Frequency twice per day of 12-hr average
  compositions of PM2.5 for daytime (10am10pm) and
  nighttime (10pm10am).
• Items ions, EC/OC, organic compounds and metals.
• Periods Jun. 1518, 2006 and Jan. 1926, 2006.

• Findings
• Compare two sites ROF is cleaner SO4 and
  NH4 no significant difference NO3 ROF is
  higher, but both very low EC and OC HYW is
  significantly higher.
• Compare day and night Higher percentage of
  OC at night Higher percentage of SO4 during day.

5
Other PM2.5 composition monitors in Atlanta Met
6
Sampling frequency

• SEARCH stations JST and YRK, hourly composition
  of PM2.5, as well as daily 24-hr averages
• ASACA stations FTM, TUC, SDK, YGP, daily 24-hr
  average composition of PM2.5
• STN site South De Kalb (same location as SDK),
  every third day 24-hr average composition of
  PM2.5

7
Can CMAQ capture the observed gradient of the
EC/OC concentration at the two closely
neighboring sites?Can CMAQ capture the
observed diurnal changes of PM2.5 and its
components?
Questions
8
Objectives of this work

• Simulating PM2.5 speciation using CMAQ at very
  fine scale.
• Characterize emissions from freeway.
• Compare fine scale CMAQ results to observations
  using detailed speciation of organics and metals
  (just have EC/OC and ions for now). Next to
  freeway, nearby (500m), 2km away, within the
  region.
• Mutual calibration with receptor modeling
  results.
• Reconcile differences Improve emission
  characterization, emissions distributions,
  dispersion, etc.

9
CMAQ v4.5 simulation

• Four nesting domains down to 1.3-km resolution.
• Thirteen vertical layers, first layer 18 meters.
  
• Simulating summer episode currently June 12-20,
  2005.
• SAPRC99 mechanism plus aero4 module.
• MM5 and SMOKE provide meteorology and emission
  rate fields.
• OSU land surface model plus 4DDA (only for 36-km
  and 12-km grids) used in MM5.
• VISTAS 2002 emissions inventory projected to
  2005, CEM data used for EGU sources.

10
Brute force sensitivity simulations
20 sensitivity runs

• sensitivity fields air quality fields basecase
  - air quality fields reduced case

11
Modeling Domains
12
Basecase 1.3-km Grid Emissions
NOx
CO
POA
PEC
13
Simulated Spatial Distributions on 1.3-km Grid
(basecase)
O3
NH4
SO4
EC
OC
NO3
14
First Concern Is 1.3-km grid performance worse
than coarse grid?
15
MM5 Performance 1.3-km grid vs. other
resolutions
Compare with TDL hourly surface observations
16
CMAQ Performance 1.3-km grid vs. other
resolution
Compare with Network measurements from AIRNOW,
STN, CASTNet (O3 only), IMPROVE, SEARCH and ASACA
17
Further Concern Is PM2.5 performance becoming
worse when compared to measurements in higher
temporal resolution?
18
1.3-km grid PM2.5 performanceCompare with 24-,
12- and 1-hr measurements, respectively
19
Limited EC/OC gradient was captured between HYW
and ROF
HIGHWAY
ROOF
20
EC Sensitivity results show a higher contribution
from traffic emissions at HIWAY
HIGHWAY
ROOF
21
Diurnal Changes captured OK for SO4, NH4 and EC,
not OK for OC
Jefferson Street (urban)
Yorkville (rural)
22
OC performance diurnal change
Jefferson Street (urban)
Yorkville (rural)
23
OC Sensitivity does it make sense?
Jefferson Street (urban)
Yorkville (rural)
24
Estimate Secondary OC from OC measurements
When EC was well reproducedAssume Pri OCobs
Pri OCsim,then, we have SOAobs OCobs -
PriOCsim
ROOF
Yorkville
Secondary OC was not captured by CMAQ, both
mechanism and precursor emissions need
improvements.
25
Summary

• Performance of 1.3-km grid is as good as other
  resolutions. This is encouraging.
• Limited EC/OC gradient was captured at
  neighboring sites. Link-base VMT is necessary to
  allocate the mobile emissions more accurately.
• Utilize modeled primary OC to split SOA from
  observed OC. With uncertainty.
• OC diurnal change was not captured. SOA
  prediction needs to be improved. Problems are
  from both mechanism and precursor emissions.

26
4-km grid PM2.5 performanceCompare with 24-, 12-
and 1-hr measurements, respectively
27
EC performance
28
SO4 performance