Update on 8Hour Ozone Early Action Compact Modeling for Oklahoma

1
Update on 8-Hour Ozone Early Action Compact
Modeling for Oklahoma
October 2, 2003
2
Update on 8-Hour Ozone Early ActionCompact
Modeling for Oklahoma

• Ralph Morris
• Yiqin Jia
• Edward Tai
• ENVIRON International Corporation
• 101 Rowland Way
• Novato, CA 94945
• [phone]
• rmorris_at_environcorp.com
• Presented at
• Oklahoma Ozone Modeling/SIP Protocol Meeting
• October 2, 2003

3
OK 8-Hour Ozone Modeling Update Contents

• MM5 Meteorological Modeling and Evaluation for
  full August 13 through September 1, 1999 episode
• Jia and Morris 2003
• Emissions Modeling for August 1999 Episode
• CAMx Ozone Modeling for First Two Base Cases
• OTAG Clean Boundary Conditions (BCs)
• Updated BCs (DFW, E-TX, SOS)
• Corrected Base Case with Updated BCs

4
OK 8-Hour Ozone Study Participants

• ENVIRON International Corporation
• Ralph Morris, Project Manager
• Yiqin Jia, Edward Tai, Steve Lau
• Advisory Committee
• INCOG, Tulsa
• ACOG, Oklahoma City
• Oklahoma DEQ
• EPA Region VI
• Technical Advisory Committee
• INCOG, Tulsa
• ACOG, Oklahoma City
• Oklahoma DEQ
• EPA Region VI

5
Summary of Oklahoma 8-Hour EAC Modeling Protocol

• December 2002 Modeling Protocol
• August 15-25, 1999 Episode Selected Initially
• MM5 Meteorological, EPS2x Emissions, and CAMx
  Photochemical Models
• Concern that Episode Doesnt have Sufficient
  Stagnant Periods
• Oklahoma DEQ Analyzes 10 Years of Meteorological
  Data
• Episode Extended to September 1, 1999
• August 13, 1999 September 1, 1999 Modeling
  Period

6
Oklahoma Ozone Monitoring Network (1999-2001)
7
Oklahoma 8-Hour Ozone Exceedance Days During Aug
15 - Sep 1, 1999 Episode
8
CAMx 36/12/4-km Modeling Domain
9
Meteorological Modeling

• MM5 Prognostic Meteorological Model
• 36/12/4-km Two-Way Nesting
• August 13 September 1, 1999
• Four Dimensional Data Assimilation
• EGAS Analysis Fields
• NCAR Observations
• Evaluation of MM5 using METSTAT Program
• NCAR Observations
• MESONET Observations

10
MM5 108/36/12/4 km Modeling Domain
11
MM5 (blue) and CAMx (red) Modeling Domain
12
NCAR Meteorological Observations Database
13
MESONET Met Observations Database
14
Summary MM5 Performance OK 4-km Domain
15
MM5 Model Performance in Tulsa And OK City
16
MM5 Modeling Summary

• MM5 model performance mostly meets performance
  benchmarks for winds, temperatures and relative
  humidity
• Worst performing days are August 17-18 and August
  26-27
• Details in September 15, 2003 Report (Jia and
  Morris, 2003)

17
Emissions Modeling

• EPS2x Emissions Model GLOBEIS Biogenic
• EPA NEI99 Emissions Database
• MOBILE6 NONROAD Mobile Source Models
• Local Travel Demand Model (TDM) Data for Tulsa
  and Oklahoma City
• Major Point Sources Day-Specific CEM Data from
  Acid Rain Database
• Local Data as Available

18
Emissions Modeling

• Several Streams of Emissions Processing

19
Photochemical Modeling

• CAMx Photochemical Grid Model
• 36/12/4-km Two-Way Nesting
• CB-IV Chemistry
• PPM Advection Solver
• Clean Boundary Conditions (BCs) on Outer
  36-km Grid Domain
• MM5CAMx to Process MM5 Data
• Evaluation Following EPA Guidance

20
EPA Ozone Model Performance Goals

• EPA 1991 1-Hr Ozone Modeling Guidance
• Unpaired Peak lt /-20
• Normalized Bias lt/-15
• Normalized Gross Error lt35
• EPA 1999 Draft 8-Hr Ozone Guidance
• Bias 8-Hr Daily Max near each monitor within
  20 at most monitors
• Graphical Figures of Model Performance
• Spatial Maps
• Time Series
• Scatterplots
• Q-Q plots

21
Oklahoma Run1 and Run2 Base Case Simulations

• Run1 OTAG Clean Boundary Conditions
• Ozone underprediction bias
• Boundary conditions too clean for eastern,
  northern and western boundaries
• Run2 Updated Boundary Conditions (SOS) that
  vary spatially
• Underprediction bias still present but not as
  severe
• Updated BCs more technically justifiable

22
Run 1 OTAG Clean (left) and Run2 Updated
Spatially Varying (right) Boundary Conditions
23
Definition of Boundary Segments for Spatially
Varying Boundary Conditions
24
Oklahoma Run3 Base Case Simulation

• Run3 Corrected Error in Emissions
• Link-based emissions incorrectly allocated to
  wrong location
• Affected Tulsa and Oklahoma on-road mobile
  sources
• Use updated spatially varying boundary conditions
  (more technically justifiable)
• Still have overall ozone underprediction bias
  except for some overprediction bias in unpaired
  peaks

25
EPA Ozone Model Performance Goals

• EPA 1991 1-Hr Ozone Modeling Guidance
• Unpaired Peak lt 20
• Normalized Bias lt15
• Normalized Gross Error lt35
• EPA 1999 Draft 8-Hr Ozone Guidance
• Bias 8-Hr Daily Max near each monitor within
  20 at most monitors
• Graphical Figures of Model Performance
• Spatial Maps
• Scatterplots
• Q-Q plots
• Time Series

26
Daily Max 8-Hour Ozone August 15, 1999
27
Daily Max 8-Hour Ozone August 16, 1999
28
Daily Max 8-Hour Ozone August 17, 1999
29
Daily Max 8-Hour Ozone August 18, 1999
30
Daily Max 8-Hour Ozone August 19, 1999
31
Daily Max 8-Hour Ozone August 20, 1999
32
Daily Max 8-Hour Ozone August 21, 1999
33
Daily Max 8-Hour Ozone August 22, 1999
34
Daily Max 8-Hour Ozone August 23, 1999
35
Daily Max 8-Hour Ozone August 24, 1999
36
Daily Max 8-Hour Ozone August 25, 1999
37
Daily Max 8-Hour Ozone August 26, 1999
38
Daily Max 8-Hour Ozone August 27, 1999
39
Daily Max 8-Hour Ozone August 28, 1999
40
Daily Max 8-Hour Ozone August 29, 1999
41
Daily Max 8-Hour Ozone August 30, 1999
42
Daily Max 8-Hour Ozone August 31, 1999
43
Daily Max 8-Hour Ozone September 1, 1999
44
EPA 8-Hr Ozone Performance Goals

• bias pred/obs mean 8-hr ( 1-hr) daily maxima
  near each monitor (EPA, 1999)
• 20 most monitors (8-hr comparisons only)
  (EPA, 1999)
• How to define near? Use same NX by NY array
  of grid cells centered on monitor as used in the
  attainment test, e.g., 9 by 9 for 4 km grid
• What predicted ozone to select for comparison
  with observed maxima? Analyze three ways (1)
  maximum ozone near monitor (2) best fit
  (closest) near monitor and (3) at monitor
  location (spatially paired).

45
Daily Max 8-Hr Ozone OK Maximum Near Monitor
(EPA Goal Most lt 20) Scatter and Q-Q Plots
46
Daily Max 8-Hr Ozone OK Best Fit Near Monitor
(EPA Goal Most lt 20) Scatter and Q-Q Plots
47
Daily Max 8-Hr Ozone Tulsa Max Near Monitor
(EPA Goal Most lt 20) Scatter and Q-Q Plots
48
Daily Max 8-Hr Ozone Tulsa Best Fit Near
Monitor (EPA Goal Most lt 20) Scatter and Q-Q
Plots
49
Daily Max 8-Hr Ozone OKC Max Near Monitor(EPA
Goal Most lt 20) Scatter and Q-Q Plots
50
Daily Max 8-Hr Ozone OKC Best Fit Near Monitor
(EPA Goal Most lt 20) Scatter and Q-Q Plots
51
Time Series Pred/Obs Running 8-Hr Ozone Skiatook
and Tulsa (Tulsa)
52
Time Series Pred/Obs Running 8-Hr Ozone Glenpool
(Tulsa) and OSDH (OKC)
53
Time Series Pred/Obs Running 8-Hr Ozone Moore
and Goldsby (OKC)
54
EPA 8-Hr Ozone Performance Goals

• Both 1-hour and 8-hour ozone (EPA, 1999)
• Bias lt 15
• Error lt35
• Disaggregate model performance by subregion
• Oklahoma 4-km Domain (12)
• Tulsa (3)
• OK City (3)
• Ponca City (1)
• NE-OK (Tahlequah)
• Lawton (1)
• S-OK (3 Red River Sites)

55
Daily 8-Hour Performance All Oklahoma
56
Daily 8-Hour Performance Tulsa
57
Daily 8-Hour Performance OK City
58
Episode 8-Hour Performance Statistics
59
Episode 1-Hour Performance Statistics
60
Summary of Run3 Model Performance

• Underestimation bias at ozone monitors
• Steep concentrations gradients resulting in
  nearby peaks higher than observed
• Underestimation of ozone transport from south and
  east
• Too low boundary conditions?
• Overstated mixing?
• Overstated wind speeds?
• Understated precursors?
• Overstated dry deposition rate (drought stress)?

61
Summary of Run3 Model Performance

• Spatial displacement of ozone plume downwind of
  Tulsa and OKC ozone monitors
• Too high wind speeds?
• Insufficient VOC emissions?
• Too much NOx emissions?
• Understated photolysis rates?
• Time series indicates model reproduces morning
  rise but peaks and falls too soon
• Too high wind speeds?
• Overstated mixing
• Understated photolysis rates?
• Spatial displacement of plume?

62
OK 8-Hour O3 Modeling Next Steps

• Continue developing future-year base case
  emissions (2002 and 2007) as performing
  diagnostic sensitivity tests to improve model
  performance
• When future-year emissions ready perform dry run
  of attainment demonstration with current best
  base case configuration

63
OK 8-Hour O3 Modeling Next Steps

• Potential Diagnostic Sensitivity Tests
• Drought Stress Sensitivity
• Rerun GLOBEIS for biogenic emissions
• Turn on drought stress in CAMX dry deposition
  module
• Eliminate cloud attenuation of photolysis rates
• Eliminate wet deposition
• Reduce afternoon mixing
• Rerun MM5 with more aggressive Four Dimensional
  Data Assimilation (FDDA) including MESONET
  surface data
• Others???