WRF Real-Time Modeling in Antarctica
Under AMPS— Efforts and Issues
Jordan G. Powers and Kevin W. Manning
Mesoscale and Microscale Meteorology Division
NCAR Earth System Laboratory
National Center for Atmospheric Research
Boulder, CO
Workshop on Polar Simulations with the WRF Model
Columbus, OH
November 2–3, 2011
AMPS— The Antarctic Mesoscale Prediction System
• Real-time WRF to support US Antarctic Program
Antarctic weather forecasting and science
• AMPS effort includes:
(i) WRF polar performance analysis
(ii) Physics improvement
http://www.mmm.ucar.edu/rt/amps
Motivations for AMPS: Antarctic Forecasting Needs
– Weaknesses seen in the NWP available to the
McMurdo forecasters (circa 2000)
The Antarctic Weather Forecasting Workshop (May 2000, BPRC)
1) Inadequate resolution
 To capture mesoscale wx features
 To capture local topography
2) Need to improve high-latitude model performance
 Better physics/parameterizations needed
for the Antarctic PBL / troposphere
– Proposed: A real-time mesoscale modeling system
tailored for the USAP forecasters
AMPS WRF
(Nov. 2011)
15 km
Version 3.2.1
Western Ross Sea
5 km
McMurdo
NB: The official, released WRF has been
“Polar” since WRF V3.1, April ‘09
South Pole
5 km
6 Grids: 45-, 15-, 5-, 1.67-km
45 km
●●
5 km
Christchurch (CHC)
Antarctic
Peninsula
Primary air route: CHC–MCM
15 km
1.67 km
McMurdo Station
●
(MCM)
●
●
Topography
shaded
Ross Is.
●
●
South Pole
Palmer Station
McMurdo
AMPS WRF Configuration
Fcsts
Top: 10 mb
45 km— 120 h
15 km— 120 h
5 km— 36 h
1.67 km— 36 h
Levels (η): 44
Lowest ½-η: 12 m AGL
Physics
LSM: Noah (w/ latest polar mods)
PBL: MYJ scheme
Microphys: WSM 5-class
Sfc layer: Eta scheme
Cu: Kain-Fritsch: 45-km,15-km
Fully-explicit: 5-km, 1.67-km
LW rad: RRTMG
SW rad: Goddard
Sea ice: Fractional representation
Sea ice analyses:
Nat’l Snow & Ice Data Center
AMPS Plans for WRF Improvement
Goal: Improve WRF for Antarctica through
polar modifications and model evaluation
1) Base AMPS Effort
a) Investigation of forecast clouds
– Evaluation of WRF microphysics to identify
best performance for Antarctic
conditions
b) Stable boundary layer performance
– Comparison of MYNN PBL scheme with
MYJ scheme in stable conditions in
Antarctica
AMPS Plans for WRF Improvement (cont’d)
2) NSF OCI (Ofc of Cyberinfrastructure)– Supplemental Effort
a) Review, revise the organization of polar mods
within WRF
– Shift ice-related mods from Noah to
separate module
b) Prepare the revised code for next WRF release
– Spring 2012
c) Investigate new polar capabilities
▪ Variable sea ice thickness
▪ Variable snow depth on sea ice
▪ Variable sea ice albedo
Real-Time WRF over Antarctica: Problems Encountered
1) Stability
• WRF blow-ups
– Most often seen over Antarctic Peninsula in winter
– Assoc’d with: Strong winds over steep terrain
– WRF config adjustments
 Shortened Δt
 Increased w damping and depth of damping
layer
• WRFDA Analysis
– Problems with Var failing or producing bad analyses
Real-Time WRF over Antarctica: Problems Encountered (cont’d)
– Variational bias correction (VBC) bug:
Use of AMSU-A radiances w/ VBC
70 ms-1
WRFDA Input
Min= 0 Max= 85
Wind speed (ms-1)
250 ms-1
WRFDA Output
Min= 0 Max= 270
Level: η1/2= 5 (approx. 25 mb)
Real-Time WRF over Antarctica: Problems Encountered (cont’d)
2) Performance
– Wind speed low bias in high-wind events
McMurdo
(E.g., May ‘04, Nov. ‘07, Apr. ‘09, Nov. ‘09)
– Warm sfc T biases in AMPS over Plateau
▪ Coastal sites, Ross Ice Shelf,
West Antarctica show varying
warm / cold biases
▪ Original, larger warm bias reduced in AMPS
with cycling of subsurface temps from
WRF (cf. GFS input)
Real-Time WRF over Antarctica: Problems Encountered (cont’d)
– Warm sfc T biases in AMPS over Plateau
Obs WRF
Obs WRF
Bias RMSE (°C)
Fcst hr
Winter (Apr.–Jul. ‘11)
Bias RMSE (°C)
Fcst hr
Summer (Nov. ‘10-Jan. ‘11)
Real-Time WRF over Antarctica: Problems Encountered (cont’d)
– Forecast cloud bias
▪ Lower-than-obsv’d cover over the Southern Ocean
(Nicolas and Bromwich 2011)
▪ Q: What are impacts on incoming/outgoing radiation?
Monthly Mean
Cloud Fraction
Feb. 2011
(from
Nicolas and
Bromwich 2011)
AMPS
CloudSat / CALIPSO
satellites
Re: Road to Improvement of WRF in AMPS
• Latest versions of WRF do not always immediately
improve things
– Verification for warm and cold seasons in AMPS
done before implementation
• Polar modification gains: Incremental, but steady
• Forecast improvements from assimilation of
new data types individually small
– Large error reductions not seen as individual new
obs types added: Small positive impacts
– Note: GFS first-guess in AMPS reflects prior DA
Advancing WRF for Polar Simulations: Possibilities
1) Restructuring of Current Polar Mods (Noah LSM)
– Revision effort in progress under AMPS
2) Subsurface Initialization (esp. for Antarctica)
– What different approaches are taken?
– Is there a best approach?
(Approach may depend on the first guess)
3) Sea-ice Model Coupling
– Goal: Improved sea-ice treatment and evolution
Advancing WRF for Polar Simulations: Possibilities (cont’d)
4) Ocean Model Coupling
– How much need for high-latitude WRF?
5) Improved High-latitude Cloud Representation
Ex: WRF low bias over Southern Ocean
6) Radiation scheme improvements
Ex: Modified RRTM longwave scheme
– What are the current deficiencies?
Advancing WRF for Polar Simulations: Possibilities (cont’d)
7) Blowing Snow Parameterization
– What are the possibilities? Coupled snow model?
8) Diamond Dust / Clear-sky Precip
– Significant source of pcp over Antarctic plateau
– Development of parameterization or forecast
algorithm possible?
On Needs for WRF for High-Latitude Applications
• What are the need areas for WRF polar improvement?
– Can comprehensive list be compiled?
– Identification of 2–3 most pressing areas
– Better simulation of certain meteorological
conditions?
(e.g., fog, stable PBL, polar clouds)
– Reduction of specific forecast errors?
• What WRF physics areas easiest to improve for low-cost
model advancement (i.e., low-hanging fruit)?
• Are there WRF deficiencies seen at one pole and not the
other?