WRF-CMAQ 2-way coupled
system: Part I
David Wong, Jonathan Pleim, Rohit Mathur,
Robert Gilliam, Tanya Otte, Jeffery Young
NERL/AMD
Frank Binkowski and Aijun Xiu
Institute for the Environment, UNC
CMAS 08
October 6-8, 2008
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory
Outline
 Current operation scheme
 Motivation
 Coupling issues
 Design and features
 Overall coupled structure
 Preliminary results
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory
Current operation scheme
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory
Motivation
 Shortcomings of the current way of running
met model- MCIP-CMAQ
• 32-bit data storage in files
• excessive data interpolation
• lack of feedback (direct and indirect
effect) from CMAQ to the met model
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory
Coupling Issues
 Compatibility of met model and CMAQ
• model physical aspects
• computational aspects
 Coupling tool
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory
Coupling Issues: compatibility
 Model physical aspects
• map projection
• coordinate system and grid format
• layer structure
• time step size
• etc
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory
Coupling Issues: compatibility (cont’d)
 Computational aspects
• domain decomposition
• processor and subdomain mapping
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory
Coupling Issues: compatibility (cont’d)
 Runtime System Library, RSL
 RSL (MM5 and WRF), RSL-Lite (WRF)
 RSL-Lite is slightly faster than RSL
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory
Coupling Issues: Tools
 Earth System Modeling Framework (ESMF)
 Other choices: OpenMI (UK), Model
Coupling Toolkit (MCT) (ANL)
 M3IO IOAPI Buffered file
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory
Design Strategy
Maintain integrity of WRF and CMAQ models
Minimal code changes:
• Keeping the same IOAPI calls
• Easy to incorporate new version of WRF
Efficiency: Low coupling overhead
• buffered file for data transfer (forward and backward)
• buffered file always holds two time steps of data
allowing interpolation
Flexibility: easy to add feedback parameters and outputs
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory
Features
Flexible time stepping
• CMAQ can be called every WRF
timestep or at any user defined multiple
Simple run time switch of feedback
Subdomain tracking
Switch back to disk I/O file for uncoupled
mode
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory
WRF-CMAQ system overview
WRF
CMAQ
Solve.F
vdiff
biogenic emis
plume rise
surface flux
physics drivers
radiation
microphysics
dynamics
aq_prep
met data
CMAQ (time, Dt)
feedback
aerosol data
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory
hadv
zadv
hdiff
cldproc
chem
aero
emission
data
initial
conditions
boundary
conditions
Example of a call frequency ratio of 4
time line
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory
Preliminary results
Offline – 4:1
O3
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory
PM25
Preliminary results (cont’d)
4:1 – 1:1
O3
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory
PM25
Execution time
Run
Met Only
MCIP
Offline CMAQ
Exec. Time
2:15:23
0:04:21
Inline Model 4:1
2:25:18
4:52:07
Inline Model 1:1
9:02:52
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory
4:45:02
Preliminary results (cont’d)
4:1 w/o – w/ feedback
O3
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory
PM25
Preliminary results (cont’d)
w/ feedback 4:1 – 1:1
O3
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory
PM25
Preliminary results (cont’d)
w/ feedback 4x8 – 2x4
O3
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory
PM25
Execution time (cont’d)
case
2x4 1:1
4:1
1:1
4x4 4:1
1:1
4x8
4:1
Exec. time
w/o feedback w/ feedback
16:36:04
8:18:00
9:26:11
9:34:28
5:13:50
5:20:56
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory
5:55:10
3:34:47
To be continued …
Office of Research and Development
Atmospheric Modeling Division, National Exposure Research Laboratory