Comparison of CENRAP 36 km and 12 km MM5 Model Runs for 2002

1
Comparison of CENRAP 36 km and 12 km MM5 Model
Runs for 2002

• Jeremiah Johnson, Yiqin Jia, Chris Emery and
  Ralph Morris
• ENVIRON International Corporation
• Zion Wang and Gail Tonnesen
• University of California at Riverside
• Central Regional Air Planning Association
  (CENRAP)
• Modeling Workgroup
• May 23, 2006

2
Introduction

• CENRAP is performing regional Particulate Matter
  (PM) modeling to address the requirements of the
  Regional Haze Rule (RHR)
• Both the CMAQ and CAMx models are being applied
  for the 2002 annual period on a 36 km continental
  U.S. domain
• 12 km sensitivity modeling was conducted for
  three episodes on a Central States domain
• Spring March 7th to April 23th
• Summer August 27th to September 13th
• Winter November 26th to December 15th
• Model Performance was compared using 36 km and 12
  km grids
• CAMx 12 km flexi-nesting was used to determine
  whether effects of 12 km domains are due to 12 km
  meteorology or emissions

3
12 km Sensitivity Results

• Presented at Feb 6-8, 2006 CENRAP meeting in
  Baton Rouge
• http//pah.cert.ucr.edu/aqm/cenrap/ppt_files/Morri
  s_36vs12km_Feb6-8_2006.ppt
• CMAQ model performance fairly insensitive to 36
  km and 12 km grid
• CAMx exhibited more sensitivity when 12 km was
  used
• Summer SO4 over-prediction bias greatly reduced
  using 12 km grid
• CAMx flexi-nesting (12 km meteorology w/ 36 km
  emissions) found improved SO4 performance due to
  12 km meteorology and not 12 km emissions

4
Recommendations 12 km Modeling

• Key Findings
• Except for CAMx SO4 little difference in model
  performance using 36 and 12 km grid
• Substantial more resources needed for 12 km grid
• Effects of grid on emission controls not nested
• Feb 6-8, 2006 Baton Rouge CENRAP Meeting
  Recommended the Following
• Perform most runs using just 36 km grid
• Process 12 km MM5 for modeling in case needed
• Need to evaluate 12 km MM5 data
• If 12 km runs are needed, results suggest can
  just run with 12 km meteorological inputs

5
Model Setup for CENRAP Runs

• National RPO 36 km Grid, 164 x 128 x 34
• 12 km Grid, 264 x 240 x 34
• Lambert Conformal Projection
• Pleim-Xiu/ACM LSM/PBL
• RRTM Longwave and Dudhia Shortwave

6
Model Setup for CENRAP Runs (continued)

• No shallow convection
• Kain-Fritsch 2 cumulus convection
• Reisner 1 moist physics
• Nudging analysis u/v surface
• u/v/t/q
  aloft

7
Surface Evaluation Method

• Use Metstat statistical package to compare
  model surface
• output to surface station data
• Metstat calculates statistics which describe
  how well the model
• output fields agree with observations
• Metstat analysis divides U.S. into subdomains
  to show
• how MM5 performs region by region
• Metstat statistics summarized in scatter plots
  which contain a
• model performance benchmark box as well as
  results of other
• MM5/RAMS runs used for air quality modeling

8
METSTAT Evaluation Package

• Average observed and predicted
• Absolute Bias and Error
• RMSE
• Index of Agreement (IOA)
• Daily and, where appropriate, Hourly Evaluation
• Statistical Performance Benchmarks
• Based on an analysis of gt 30 MM5 and RAMS runs
• Not meant as a pass/fail test, but to put
  modeling results in the proper perspective

9
Subdomains for Model Evaluation
1 Pacific NW 2 SW 3 North 4 Desert SW 5
CenrapN 6 CenrapS 7 Great Lakes 8 Ohio
Valley 9 SE 10 NE 11 MidAtlantic
10
Datasets for Met Evaluation

• NCAR dataset ds472 airport surface met
  observations

11
MM5 36 km vs. 12 km Model Performance

• Previously evaluated CENRAP 36 km MM5 simulation
  across US domains and compared results with
  VISTAS and WRAP 36 km MM5 runs
• http//pah.cert.ucr.edu/aqm/cenrap/ppt_files/CENRA
  P_VISTAS_WRAP_2002_36km_MM5_eval.ppt
• Now evaluate CENRAP 12 km MM5 and compare with 36
  km MM5 results
• CenrapN and CenrapS domains
• Wind Speed and Direction, Temperature and
  Moisture
• Compare against performance Benchmarks
• Follows are examples for Jan, Mar, Jul and Oct
  2002

12

• Highest temperature error for CenrapN subdomain
  in January

13

• Comparable humidity performance for both runs

14

• Comparable wind performance for both runs

15

• CenrapN subdomain outside goal, but better than
  January

16

• Comparable humidity performance for both runs

17

• CenrapS subdomain falls just within goal

18

• Greatest temperature performance improvement
  using 12km grid
• CenrapS outside goal for 36km run

19

• Highest humidity error in July, still within goal

20

• Comparable wind performance for both runs

21

• Temperature performance comparable for 36 and
  12km runs

22

• Comparable humidity performance for both runs

23

• Comparable wind performance for both runs

24
Overview of CENRAP 36/12 km Surface Performance

• Generally, performance was within the benchmarks
  over the two CENRAP regions, and little
  improvement was gained with the addition of the
  12km grid (exception of 12 km temperature for
  July)
• Both 36 and 12km runs had very similar wind and
  humidity performance
• Temperature performance for the CenrapN subdomain
  fell outside the benchmark in January and March
  for both the 36 and 12km runs

25
Overview of CENRAP 36/12 km Surface Performance

• Wind performance slightly better in the summer
  than in winter for both runs
• Both runs display a cold bias in the winter and
  warm bias in the summer over the CenrapN
  subdomain
• Humidity performance is worse in summer than in
  the winter for the 36 and 12 km runs

26
Overview of CENRAP 36/12 km Surface Performance

• Adding the 12 km CENRAP grid resulted in only
  modest performance increases for both CENRAP
  subdomains
• 12 km grid requires significantly more computer
  time than for the 36 km coarse grid
• It may not be worth the extra processing time to
  include the 12km CENRAP grid, especially for
  annual modeling purposes.

27
CENRAP 36/12 km Surface Performance

• Evaluation of 36 and 12 km MM5 runs found
• Little difference in MM5 model performance
• No new problems found with 12 km MM5 fields
• Conclusions from CMAQ/CAMx 36/12 km modeling
  still hold
• For regional haze modeling of Class I areas the
  36 km grid appears to be sufficient
• Conclusions may not be true for 8-hour ozone and
  PM2.5 modeling of urban areas
• Midwest and Southeast U.S. modeling indicates
  that higher resolution than 36 km (e.g., 12 km
  and/or 4 km) is needed for 8-hour ozone and PM2.5
  issues