Hurricane Prediction using High-Resolution Numerical Models

1
Hurricane Prediction using High-Resolution
Numerical Models
James D. Doyle Naval Research Laboratory,
Monterey, CA james.doyle_at_nrlmry.navy.mil
C. Amerault1, S. Chen1, H. Jin1, R. Hodur2, Y.
Jin1, C.-S. Liou1, C. Reynolds, C. Sampson1, K.
Sashegyi1, J. Schmidt1 1Naval Research
Laboratory, Monterey, CA 2SAIC, Monterey, CA
DoD HPCMP IBM P5, P6, Cray XT5 (NAVO), SGI
Altix 4700 (AFRL) Cores 3000,
4800, 12,900,
9216

• Outline
• Background motivation
• COAMPS-TC overview
• Tests during T-PARC/TCS08
• Predictability

Typhoon Saomai (08W) and Tropical Storm Bopha
(10W) 02Z 8 Aug 2006 (NASA MODIS)
2
Tropical Cyclone Forecast Skill 1990-2008

• Forecast skill of TC track has been steadily
  improving (better models, new satellite
  observations, dropsondes, improved forecasting
  techniques, .).
• TC intensity forecast skill has improved very
  little (poor understanding of the processes
  controlling intensity, inadequate models,
  inadequate observations).
• Economic impact of hurricanes is huge 137B in
  insured losses (1987-2006)
• DoD impact is significant (e.g., Adm Halsey,
  Fleet sortie forecasts, etc.)

Jon Moskaitis (MIT, NRC)
3
Tropical Cyclone Forecast Skill 1990-2008
Objective Develop, evaluate and transition to
DoD operations a coupled tropical cyclone
prediction system to analyze, initialize, and
predict TC position, structure, and intensity,
using a high-resolution mesoscale air/ocean/wave
coupled modeling system.
Jon Moskaitis (MIT, NRC)
4
COAMPS-TC Overview Model Framework Components
ESMF Modeling Superstructure
Atmospheric Model
Ocean Model
Wave Model
Dynamics
Physics Interface to Physics Suites
Flux Coupler
Share Community Models/Share Community Physics
Relatively easy to add new modeling components
5
COAMPS-TC Background

• COAMPS-TC New version of COAMPS? developed for
  tropical cyclone analysis and prediction (track,
  intensity, structure)
• New TC analysis and physical process
  representations
• Moving nested grids track the TC
• Development of air-sea-wave coupling for TCs
• Adjoint of COAMPS including microphysics
• Allows for the mathematically rigorous
  calculation
• of sensitivity to changes in the initial state

• Real-time forecasts in support of THORPEX Pacific
  Asian Regional Campaign / Tropical Cyclone
  Structure 08
• COAMPS-TC 45/15/5 km, moving nests (15/5 km),
  track, structure intensity
• COAMPS Adjoint Targeted observations

6
THORPEX Pacific Asian Regional Campaign (T-PARC)
Tropical Cyclone Structure 08 (TCS08)

• Observe TCs and Environment Genesis to
  extratropical transition.
• Targeted Observing Take observations where they
  will improve forecasts.
• Scope 9 nations 4 aircraft (lidar, radar,
  dropsondes), driftsondes, rapid-scan satellite
  observations, off-time radiosondes, AXBTs.
  (Aug-Oct 2008)

DOTSTAR
7
COAMPS-TC Track Forecasts for T-PARC/TCS08 Track
Forecast Verification
Black line Warning positions Colored lines
COAMPS forecasts
8
COAMPS-TC Real-Time Modeling for T-PARC/TCS08
9
COAMPS-TC Intensity Forecasts Prediction of Super
Typhoon Jangmi
0000 UTC 26 September 2008 (72-h forecast)
Animation of COAMPS predicted radar reflectivity
every 30 minutes on 5 km moving grid

• COAMPS-TC forecasted rapid intensification of
  Jangmi, however TC was stronger than observed at
  72 h since it never made landfall.
• Convection was spotty and disorganized early in
  forecasts (cold starts).
• Overall, intensity forecasts were not as skillful
  as some other TC models/tools.

10
COAMPS-TC W. Pacific Forecasts (2008) Improvements
Following T-PARC/TCS08
Intensity Error (homogeneous comparison)
New Version of COAMPS-TC shows much improved TC
intensity forecasts
Intensity Error (kts)
Forecast Time (h) (Number of Cases)
COAMPS-TC (new) intensity forecasts in W. Pacific
for T-PARC/TCS08 are now similar to
other skillful models.
11
COAMPS-TC Prototype Coupled Model Tests
COAMPS-TC Air-Ocean Coupled Prediction of
Hurricane Gustav Initial Time 1200 UTC 30 August
2008
12
Edward Lorenz (1917-2008) Father of Chaos Theory
one flap of a sea-gulls wing may forever
change the future course of the weather (Lorenz,
1963)
Kyoto Prize in 1991 for "profoundly influencing
a wide range of basic sciences and brought about
one of the most dramatic changes in mankinds
view of nature since Sir Isaac Newton
13
Predictability of Tropical Cyclones COAMPS
Ensembles
Ensemble 48-h forecasts from 00 UTC 09 July 2005
(TC Dennis)
Perturbed initial state, boundaries, physics
results in intensity variability (987 to 1002
hPa) as well as track
14
Predictability of Tropical Cyclogenesis T-PARC/TCS
08 Typhoon Sinlaku and Jangmi
Sinlaku Intensification

• Adjoint allows for the mathematically rigorous
  calculation of sensitivity to changes in the
  initial state
• Real-time COAMPS adjoint for mesoscale targeting
  guidance.
• First time targeting of TC formation has been
  attempted
• Adaptive response function box.

Adjoint z Sensitivity
C130 Flight Track
15
Next Generation High-Resolution Modeling New
Dynamical Cores for COAMPS

• Spectral Element Dynamical Core
• High order accuracy
• Appropriate for the sub-mesoscale
• Scales to many processors
• Bridge to unstructured methods

• Advancements to 2D SE Model
• Semi-implicit solver
• Addition of stretching, BCs, tracers
• Microphysics and moist tests

In Collaboration with Frank Giraldo (NPS)
16
Challenges and Future Directions

• New generation of models are needed to achieve
  skillful tropical cyclone intensity and structure
  forecasts
• New physics algorithms (sea spray, turbulence,
  cloud physics)
• Sophisticated numerical methods (nonhydrostatic,
  embedded meshes)
• Fully coupled models (atmosphere, ocean
  circulation, waves, surge)
• High resolution needed to resolve details of TC
  interaction w/ environment
• - convective permitting (Dx1-4 km) (250x
  computing 5 to 1 km)
• - turbulence resolving (Dx10-100 m)
• -Estimate the uncertainty (ensembles)
• -New data assimilation methods for the cloud
  scale (hybrid)
• -Community efforts Hurricane Forecast
  Improvement Project (NOAA/Navy)
• Computational attributes
• Faster cores and greater number of cores
  (resolution, physics, ensembles)
• Fast I/O (ensembles, post processing)
• New paradigms may be needed for scaling across
  1000 cores
• Adaptive load balancing for embedded meshes
• Challenges for coupled codes
• Grid / cloud computing challenges for ensembles
  (post processing)
• Visualization / data management challenges

17
COAMPS Community Modeling

• COAMPS Community Model
• Within DoD, academia, other agencies
• gt350 users
• Physical Parameterizations
• Sharing suites and parameterizations with other
  models (e.g., WRF)
• Joint Ensemble Forecast System (JEFS) with AFWA
• Multi-model ensemble as part of a new
  probabilistic approach
• Coupling
• Couple with other models to achieve
  interoperability
• Ocean circulation, waves, hydrology, ice etc.
• Framework
• Community framework, Earth System Modeling
  Framework (ESMF)
• Tropical Cyclone Community
• Hurricane Forecast Improvement Project (HFIP)
• DoD Users and Forecasters
• DoD Information Assurance
• Open source may not be possible for entire code

18
COAMPS-TC Improvements Based on Analysis from
T-PARC/TCS08
Azimuthally average tangential (shaded) and
radial (contour) winds Hurricane Katrina (72 h
valid 00Z Aug 29 2005, Dx3km)
New Version
TCS08

• New Version of COAMPS-TC
• TCS08 Version
• Basic TC 3D-VAR
• No sea spray
• New Version
• Additional synthetic observations
• Improved 3D-VAR data assimilation
• Bougeault type of mixing (PBL above)
• New sfc moisture transfer coefficient
• New ice nucleation
• New dissipative heating formulation
• New sea spray parameterization

• New COAMPS-TC
• Improves initial forecast intensity
• Improves the convective structure
• Good agreement with Doppler obs.

R. Rogers (HRD)
19
COAMPS-TC Overview Current and Future System
Components
Ocean Analysis
Atmospheric Analysis

• Navy Coupled Ocean Data Assimilation (NCODA)
  System
• 2D OI SST
• 3D MVOI T, S, SSH, Sea Ice, Currents
• Complex Data Quality Control
• Initialization Stability check

• Complex Data Quality Control
• Relocation of TC in background
• NAVDAS 3DVAR u, v, T, q, TC option
• Initialization Hydrostatic Constraint on
  Analysis Increments, and/or Digital Filter
• TC Balance Step (underway)

Ocean Models
Atmospheric Model

• Numerics Nonhydrostatic, Scheme C, Moving Nests,
  Sigma-z, Flexible Lateral BCs
• Physics PBL, Convection, Explicit Moist Physics,
  Radiation, Surface Layer
• TC Tools Moving nests, dissipative heating,
  spray parameterization, shallow convection

• NRL Coastal Ocean Model (NCOM)
• Numerics Hydrostatic, Scheme C, Nested Grids,
  Hybrid Sigma/z
• Physics Mellor-Yamada 2.5
• Wave Models (WWIII and SWAN)
• Generalized Flux Coupler (ESMF)

20
COAMPS Scaling
SGI Origin
linux scali
linux scali2

COAMPS has been ported to many platforms SGI,
IBM, Linux, Cray, HP, etc. Efficiency and scaling
addressed in a PET Project