CRGAQS: Initial CAMx Results

1
CRGAQSInitial CAMx Results

• Presentation to the
• Gorge Study Technical Team
• By
• ENVIRON International Corporation
• October 31, 2006 (boo!)

2
Todays Presentation

• Describe initial CAMx simulations
• Model configuration
• Performance evaluation metrics
• Performance for PM and light extinction
• Next Steps

3
Model Configuration
4
Model Configuration

• CAMx version 4.40
• Run 1 configuration
• Maximize model speed
• Mechanism 4 CF (static 2-mode PM chemistry)
• Bott advection solver
• No PiG
• OMP parallel processing on Linux quad-CPU
• OBrien Kv profile with 0.1 m2/s minimum
• 10-day model spin-up period
• 36-km grid only first 8 days
• 36/12-km grid last 2 days

5
Model Configuration

• Episodes
• August 10-22, 2004
• Meteorology from MM5 Run 6, 36/12-km grids
• Flexi-nesting to the 4-km grid
• November 4-18, 2004
• Meteorology from MM5 Run 3, all grids
• Identified issues
• Small temporal profile problem for fires
• No on-road vehicle ammonia emissions in 4-km grid

6
Performance Evaluation

• Review spatial plots
• Review statistical performance
• Species
• Individual PM species
• Total PM2.5 and PM10
• Light scattering/extinction
• Monitors
• IMPROVE, Gorge sites, FRM/STN
• Focus on sites along Gorge

7
Performance Evaluation

• Metrics
• Fractional bias and gross error
• Regression and correlation
• Need to develop time series of scattering and
  extinction

8
Performance Evaluation
9
Performance Evaluation
10
Performance Evaluation
11
Performance Evaluation
12
Statistics IMPROVE
August NO3
August SO4
13
Statistics IMPROVE
August NH4
August OC
14
Statistics IMPROVE
August EC
August Primary Fine
15
Statistics IMPROVE
August Primary Coarse
August Total PM2.5
16
Statistics IMPROVE
August Total PM10
August Bext
17
Statistics Gorge
August NO3
August SO4
18
Statistics Gorge
August NH4
August OC
19
Statistics Gorge
August EC
August Bscat
20
Statistics STN
August NO3
August SO4
21
Statistics Gorge
August NH4
August OC
22
Statistics Gorge
August EC
August FRM PM2.5
23
Initial August Modeling

• NO3, SO4, and NH4 were mostly under predicted
• CAMx predictions higher at Gorge sites later in
  episode, but data were not available
• More NH3 may increase NO3 and NH4
• OC was mostly over predicted at IMPROVE and Gorge
  sites
• Driven by fires

24
Initial August Modeling

• EC was over predicted at IMPROVE but under
  predicted at Gorge sites
• Reasons not clear is Gorge EC backed out from
  aetholometer readings?
• Total PM2.5 looks good
• Balance of component over and under predictions
• CM and PM10 were mostly under predicted at
  IMPROVE site

25
Initial August Modeling

• Reconstructed scattering and extinction look good
• Due to good PM2.5 predictions and dry conditions
• Contribution from under predicted hygroscopic
  salts just about balance contribution from over
  predicted carbon
• CM doesnt play a large role in visibility

26
Performance Evaluation
27
Performance Evaluation
28
Performance Evaluation
29
Statistics IMPROVE
November NO3
November SO4
30
Statistics IMPROVE
November NH4
November OC
31
Statistics IMPROVE
November EC
November Primary Fine
32
Statistics IMPROVE
November Primary Coarse
November Total PM2.5
33
Statistics IMPROVE
November Total PM10
November Bext
34
Statistics Gorge
November NO3
November SO4
35
Statistics Gorge
November NH4
November OC
36
Statistics Gorge
November EC
November Bscat
37
Statistics STN
November NO3
November SO4
38
Statistics Gorge
November NH4
November OC
39
Statistics Gorge
November EC
November FRM PM2.5
40
Initial November Modeling

• More NO3 and SO4 was observed and predicted than
  in the August episode.
• OC and PM2.5 were much lower over domain than in
  August since wildfire season was over

41
Initial November Modeling

• SO4 performed well at IMPROVE sites, but was
  under predicted at the GORGE and STN sites
• NO3 performance was scattered
• NH4 was mostly under predicted
• OC and EC performance OK, but there were some
  over prediction outliers

42
Initial November Modeling

• Primary fine and coarse (soil) was over predicted
  at all IMPROVE sites on all dates
• Fine PM emissions may not reflect squelching
  effect of recent rains
• Reconstructed total PM2.5 and PM10 slightly over
  predicted
• Driven by carbon and primary over predictions

43
Initial November Modeling

• Extinction at IMPROVE sites was generally too
  low, while scattering at Gorge sites exhibits
  little skill
• Under predicted salts are not contributing
  sufficient scattering
• More NH3 might not help
• Lack of modeled fog probably is not generating
  enough sulfates and nitrates

44
Moving Forward

• Sensitivity/Diagnostic runs
• Revised emissions (fix identified problems)
• Kv sensitivity
• CMAQ approach
• Alternative minimum Kz
• Met sensitivity for August? (use Run 3)
• Increase NH3 emissions
• Reduce primary fine/coarse emissions in November