IMPACTS OF MODELING CHOICES ON RELATIVE RESPONSE FACTORS IN ATLANTA, GA

1
IMPACTS OF MODELING CHOICES ON RELATIVE RESPONSE
FACTORS IN ATLANTA, GA
Byeong-Uk Kim, Maudood Khan, Amit Marmur, and
James Boylan6th Annual CMAS ConferenceChapel
Hill, NCOctober 2, 2007

2
Objective

• Investigate the effects of modeling choices on
  Relative Response Factors (RRFs) in Atlanta, GA
• Horizontal grid resolution 4 km and 12 km
• Chemical Transport Model CMAQ and CAMx

3
Approach

• Exercising typical SIP modeling
• Model Performance Evaluation (MPE)
• Measures and methods following the EPAs guidance
  document (EPA, 2007)
• Modeled Attainment Test
• Relative Response Factors
• Additional analyses
• MPE with graphical measures
• Partial implementation of PROMPT (Kim and
  Jeffries, 2006)
• Investigation of day-by-day and site-by-site
  variation of model predictions

4
Modeled Attainment Test

• Future Attainment Status is determined by Future
  Design Value (DVf)
• DVf should be less than 0.85 ppm.
• DVf RRF x DVb
• Where,
• DVb is Baseline Design Value and
• RRF is Relative Response Factor defined as

5
8-Hour Ozone Attainment Status in GA
6
Modeling System Setup

• Base case modeling period
• May 21, 2002 Sep 13, 2002 UTC (3 spin-up days )
• MM5 (v 3.x)
• Pleim-Xiu model for Land-Surface interaction
• Asymmetric Convective Mixing
• SMOKE (v 2.x)
• VISTAS Base G version 2 inventory
• CMAQ and CAMx
• Inputs made to be close to each model for a same
  grid configuration.

Georgia
7
4 km
7x7 array for 4-km runs
12 km
8
MPE with statistical metrics
9
Time series
10
Time series
11
Time series
O3
Mon Tue Wed
Thur Fri Sat
Sun
O3
12
Time series
NO2
Mon Tue Wed
Thur Fri Sat
Sun
ETH
13
Time series
O3
Mon Tue Wed
Thur Fri Sat
Sun
O3
14
NO2
Mon Tue Wed
Thur Fri Sat
Sun
ETH
15
Spatial distribution (12km)Daily Max 8-hr O3
2002-06-12
2002-07-23
2002-07-24
CAMx
ppb
CAMx-CMAQ
ppb
16
Relative Response Factors
RRFs from max O3 nearby grid cell arrays

• Two possible methods to calculate RRFs
• Max value in nearby grid cell arrays
• Value at each monitoring site grid cell
• Spatially averaged RRFs vary from 0.891 to 0.897
  by modeling choices
• If DVb 100 ppb, 0.001 difference in RRF will
  result in 0.1 ppb in DVf.

17
Conclusion (1)

• Reasonable performance with respect to
  statistical metrics by all four models, CMAQ and
  CAMx with 4-km and 12-km grids
• 4-km emissions had 11 lower NOx in
  non-attainment areas
• 4-km MM5 runs showed poor nighttime performance.
  
• Higher biases during nighttime by CMAQ and during
  daytime by CAMx
• Gross overestimation of ozone by CAMx for several
  days
• Lower biases from 4-km simulations
• Probably due to emission discrepancies in 4-km
  inputs compared with 12-km emissions.
• No significant daytime NOx biases

18
Conclusion (2)

• Stable or insensitive RRFs
• Due to higher absolute concentrations predicted
  by CAMx, CAMx might show quite lower RRFs than
  CMAQ.
• Max-Value based RRFs fell within 0.863 0.914
  for all simulations.
• Effect of RRF calculation methods
• Despite of noticeable differences between 4-km
  and 12-km modeling inputs, Max-Value based RRFs
  does not reflect this fact significantly.
• Cell-Value based RRF distinguished grid
  configuration differences.
• For all 11 monitoring sites, maximum RRF
  difference due to model choices were 0.036 and
  0.033 by Max-Value based and Cell-Value based RRF
  calculation.

19
Future Work

• Process Analysis to explain large variation of
  predicted ozone concentrations with similar
  modeling inputs
• Detail study on the relationship between model
  performance including day-by-day and site-by-site
  meteorological model performance and RRFs

20
Acknowledgement

• ENVIRON International Corporation
• Ralph Morris for CMAQ-to-CAMx utilities

21
Byeong-Uk Kim, Ph.D.Georgia Environmental
Protection Division4244 International Parkway,
Suite 120Atlanta, GA 30354Byeong_Kim_at_dnr.state.
ga.us 404-362-2526
Contact Information