SUMMARY OF HFIP REGIONAL MODEL DEVELOPMENT
Contributors to Talk:
Ji -Wen Bao (ESRL)
Morris Bender (GFDL)
Ligia Bernardet (DTC)
Chris Davis (NCAR)
James Doyle (NRL)
Chris Fairall (ESRL)
Isaac Ginis (URI)
S.G. Gopalakrishnan (HRD)
Alexandr Khain (Hebrew U.)
Young Kwon (EMC)
Vijay Tallapragada (EMC)
HFIP HFIP Annual Review Meeting
Tuesday, November 9th,, Miami, FL
EMC FY10 Status
1. Performed PBL sensitivity tests using different parameters,
such as critical Richardson number and mixing length. Also
test has been done using a local scheme.
2. HWRF with Noah LSM option is running in parallel for the
2010 season and plan to evaluate the performance.
3. Sensitivity test on horizontal diffusion will be conducted
using the final bug-fixed version of HWRFV3.2
4. Successful transition of HWRF version2.0 to version 3.2
with several bug fixes, such as stripes and triple nest
(collaboration with DTC and HRD)
5. Upgrade of the HWRF’s vortex initialization is in progress
in order to get more balanced initial state.
Noah LSM test
HWRF with slab model
HWRF with Noah LSM
P: 939.7hPa
W: 58.4 m/s P: 943.5hPa
W: 74.1 m/s
P: 928.4hPa
W: 77.2 m/s
HWRF/original diffusion
HWRF/reduced diffusion
GFDL model
Track forecast: similar to before Intensity forecast: error reduced.
1. Microphysics:
Calibrate the operational scheme using the output of a bin
microphysics
2. Convection scheme:
Test the new SAS (simplified Arakawa Schubert) and new
shallow convection schemes implemented in 2010 GFS model
(To be tested in both GFDL and HWRF )
3. Triple nest:
Conduct test simulations for 27/9/3km triple nest HWRF with
help of HRD
4. Community Repository:
Continue supporting the community repository version of
HWRF with collaborating with DTC
Future NCEP Work Plan
1. Released Beta version in March.
2. Expanded community code to match 2010 baseline
configuration (H050) and later 2010 operational
configuration
3. Exposed and/or fixed bugs in various components
4. Upcoming:
1. Expand code to match 2011
2. Expand code to third nest
3. Expand code to idealized simulations
4. New HWRF release
High-Resolution HWRFV3.2 SystemHigh Resolution Research version of the HWRF system geared to complement operations
• Improved resolution and improved understanding of forecast at about 1-3 km resolution
• Incorporate observations on vortex scale for improved model initialization
• Incorporate appropriate representation of physical processes in tropics for
1-3 km resolution
Instantaneous snap shot of the “thermal plumes” and the warm core structure
3 km 3 km
Idealized Framework
Advanced Real-Time Products to
support Operations (parallels)
Multiple Moving Domains to Support Operations at 27:09:03
Contributions from HRD: New nest motion algorithm Advanced diagnostics capability (Diapost) Third nest capability Idealized simulations Improved vortex assimilation procedure for operations
Baseline version testing (without coupling) - HRH cases, 2009 and 2010 Atlantic cases
Use results from the 2010 Stream 2 demo (9:3 from HWRFX and HWRFV3.2) as the basis for evaluation at 3 km resolution
Development of coupler for the third nest (EMC) Improved vortex initialization for the third nest (EMC) Retrospective testing of 27:09:03 for Stream 1.5 demo in 2011 and
future operational implementation Test new physics suitable for higher resolutions Integrate the modeling component of research and development
activities at HRD with operational HWRF Synchronize the developmental efforts with DTC repository for
community purpose
1. In general their performance was no better then the
operational regional models for either track or intensity
2. The operational regional models did not do as well as
the best global models for track.
3. As in 2009, high resolution (6 km to 1 km) – alone - did
not produce desired Improvement
4. Continues to suggest the need for development of
physics that can properly resolve physical processes
simulated by the higher resolution
11
WHAT IS THE VERDICT ON HOW WELL WE DID OVERALL?
HFIP REGIONAL MODEL VERIFICATIONS
VS.
CURRENT OPERATIONAL MODELS
(HWRF, GFDL, GFS)
(Through Nov. 7th, 12z)
NORMALIZED TRACK ERROR NORMALIZED INTENSITY ERROR
Operational Regional Models Still not
as good as the statistical models for
Intensity.
HWRF had significant negative bias
GFS global model showed more skill
for Track at Every Time Level(HWRF very poor performance in Tomas
significantly affected its overall seasonal
performance)
Participating Regional models:
– HWRF (Operational) 9km
– GFDL (Operational) 7.5km
– HWRF-x (AOML) 3km
– WRF/ARW/NCAR 1.3km
– WRF/ARW/FSU 4km
– COAMPS-TC 5km
14
NORMALIZED TRACK ERROR NORMALIZED INTENSITY ERROR
HWRFx Neutral or Inferior for Track HWRFx Slightly Improved Intensity
NORMALIZED TRACK ERROR NORMALIZED INTENSITY ERROR
Most Skillful of the Regional Models for
Intensity through 2 Days
NORMALIZED TRACK ERROR NORMALIZED INTENSITY ERROR
COAMPS-TC Regional Model Summary
FY10 Major Accomplishments I
•Diagnosis of 2009 HFIP Real-Time Demo Results
•Nest and diagnostic tracker problems (particularly for weak storms)
• Initialization / spin-up problems early in forecast
•Over-intensification in the latter part of forecast
•Track statistics show a slow and right of track bias
•Spotty convection for weak storms, which is a challenge for the tracker
•New Initialization Development
•Tropical Cyclone Dynamic Initialization (TCDI) development and evaluation
•Experiments with a digital filter option
•Sensitivity tests to synthetic observation specification (size, depth, structure)
•New synthetic observation method developed – Discussions with EMC / HWRF
•New Data Assimilation Capabilities
•3D-Var experiments with SSM/I Total Precipitable Water (TPW) retrievals
• Inclusion of additional satellite wind observations
• Improved super-observation specification (scatterometer obs)
•Development of a COAMPS-TC EnKF capability within DART
COAMPS-TC Regional Model Summary
FY10 Major Accomplishments II
•Improvements to TC Physical Parameterization
•New version of the NRL microphysics that mitigates over-production of ice
• Implementation and testing of WRF (Thompson) microphysics
• Implementation and testing of new version of Kain-Fritsch (outer 2 meshes)
•Sensitivity tests to formulation of mixing length (Bougeault, M-Y, Klemp-Wil.)
•Dynamical Core Improvements
•Tests done with a semi-Lagrangian selective monotonic advection (SMA)
•Lateral Boundary Conditions
• Improvements and bug fixes to the two-way nesting
•Sensitivity tests for LBCs using NOGAPS and GFS
•Air-Sea Coupled Capability
•Additional coupling infrastructure and evaluation of COAMPS-NCOM system
• Improved ocean data assimilation capability using 3D-Var
COAMPS-TC Regional Model Summary
FY11 Plans
•Diagnosis of Real-Time 2010 Results
•Analyze statistical performance, diagnose systematic errors
•Improvements to COAMPS-TC System
• Improved initialization (TCDI or new synthetics), possibility of digital filter
•Data assimilation (3D-Var) of more observations, improvements to 3D-Var
• Improvements to air-sea coupling (better integration of NCOM, ocean 3D-Var).
•New microphysics (new NRL microphysics or another scheme)
• Improvements to TKE parameterization (esp. mixing within clouds)
•Retrospective Forecasts with 2011 COAMPS-TC
•Establish skill prior to the 2011 Demo using 500 retro cases for stream 1.5
•2011 Real Time forecasts
•45/15/5 km (or 36/12/4 km) every 6 h for WATL, EPAC, CPAC, WPAC
•COAMPS-TC for the HFIP Multi-Model Ensemble
• Integrate better with HFIP partners; COAMPS-TC code repository for HFIP
•COAMPS-TC Validation Test Report for Navy
1. Delivery time• Forecasts were not finished until t+10• Considered two cycles old at NHC
2. Multiple storms• Could only run one storm per cycle
3. Behavior on 12-km domain• Spurious genesis• Overintensification
4. Size of outer domain• Trouble with recurving storms farther north
5. Physics• Cases of overintensification (Earl)
6. Consistency with EnKF• Currently differing domains and physics could
lead to spurious behavior in forecast
1. Delivery time• Remove 1.33-km nest; use smaller 12-km domain• Estimated 2.5 h completion
2. Multiple storms• Add more nodes for more storms; all forecasts finish at same
time
3. Behavior on 12-km domain• 12-km domain now a buffer between 36 and 4-km domains• Modified Kain-Fritsch trigger function reduces spurious
genesis
4. Size of outer domain• Use larger 36-km domain as outer grid
5. Physics• Overdevelopment decreased on 4-km inner domain• Re-examine use of double-moment scheme
6. Consistency with EnKF• Make domain sizes and physics as close as possible (no 4-km
domain or ocean coupling in EnKF)• Select member nearest ensemble mean for high-res. forecast
REPORT ON RECENT
PHYSICS DEVELOPMENT
24
Wave Model
• Includes effects of wind-wave-current interaction and sea spray
• Allows for different algorithms of sea-state and sea-spray parameterization
Air-sea Interface for
Coupled Hurricane-Wave-Ocean Models
(Possible 1.5 Candidate for GFDL in 2011)
Hurricane Model
Sea state,
SprayWind
Momentum
Flux
Current
SST,
Current
Heat
Fluxes
Atmosphere
Ocean
Ocean Model
25
GFDL Hurricane-Wave-Ocean Model
Hurricane Earl, Initial time: 12Z Aug 30, 2010
Red – GFDE (waves)
Blue – GFDL (operational)
26
Black – no sea spray
Red – with ESRL sea spray
Effect of Sea Spray on Drag Coefficient
in GFDL Hurricane-Wave-Ocean Model
Magenta Triangles: ARW SAS
Brown Triangles : ARW BMJ
Black Triangles : ARW None
Red Circles: HWRF SAS
Yellow Circles: HWRF BMJ
Gray Circles: HWRF None
Max. Surface Wind Speed (m/s) Min. Sea-level Pressure (mb)
HWRF spins up faster than ARW!
The Hebrew University Cloud Model (HUCM)
with imbedded SBM already used
successfully for optimizing the two-moment
bulk microphysics scheme in the Relocatable
Regional Weather Forecast model in Germany
(Seifert et al. 2006)28
29 Seifert et al, 2006)
CONCLUDING THOUGHTS FOR
FUTURE DIRECTION IN REGIONAL
MODELING DEVELOPMENT
With transition of HWRFx to HWRFV3.2 System HFIP Regional
Model Development is at a Crucial Crosswords to achieve its
goals within 5 years to significantly improve intensity skill.
There must be a careful and focused effort over next 3-4 years
to develop physics that will correctly represented hurricane
inner core physics processes that can be adequately resolved
at high resolutions.
Still need to devote resources to find out what is the minimum
horizontal resolution needed to adequately resolve the
hurricane inner core (1km, 2km, 3km ??)
Collaboration between the academic community and NOAA will
be essential to help achieve needed physics improvements.
Closer collaboration is also needed between observational and
model development teams to improve physics (e.g., surface,
micro-physics)
Improvement in physics must be a key HFIP priority
both for Stream 1 and Stream 2 development.
Better coordination and closer collaboration between
all parts of the HFIP community (e.g., NOAA, Navy,
NCAR) remains essential to achieve our goal .
There are a lot of pieces to the puzzle that still
need to be understood, but we can succeed if
we work together !!!!