Stephen J' Lord

1
Advanced Hurricane Modeling at EMC The HWRF

Stephen J. Lord and the HWRF Team
NCEP/Environmental Modeling Center
WHERE AMERICAS CLIMATE AND WEATHER SERVICES
BEGIN
2
HWRF TEAM Naomi S., Qingfu L., Vijay T., Young
K., Bob T., Zhan Z. and Janna OC, H.-S. Kim
Also, thanks to Isaac G., J.-W. Bao, And HYCOM
team at MMAB/EMC
3
Overview

• HWRF 08 implementation
• Current Operational HWRF configuration
• 2008 HWRF end-of-the-season performance
• TE for 2009 HWRF upgrades
• The Advanced HWRF

4
HWRF 08 implementation

• HWRF Analysis Changes
• Improved initialization of weak storms
• Dynamical Balance surface pressure adjustment
  based on gradient wind stream function (more
  accurate)
• Changes to POM initialization
• Feature based assimilation of mesoscale oceanic
  features
• Assimilate more than one warm core rings as well
  as cold core rings
• Rings can be assimilated close to the loop
  current
• HWRF Physics changes / Bug fixes
• Initialize TKE to zero (HWRF does not use TKE
  based PBL)
• Adjustment of temperature and pressure fields
  after nest motion to remove noise at the lateral
  boundary of nest domain
• HWRF Script Changes
• Improved functionality of all HWRF scripts
• HWRF scripts for EMC parallels are consistent
  with operational configuration
• Test results follow

5
Intensity Error Dennis, Katrina, Rita, Wilma,
Dennis 2007 HWRF vs 2008 HWRF GFDL
H041 Operational HWRF implemented FY 2008
2007 HWRF
2008 HWRF
Reduced intensity forecast errors (mostly due to
reduced negative bias)
6
(No Transcript)
7
Atlantic Track Error Dennis, Katrina, Rita,
Wilma, Dean 2007 HWRF vs 2008 HWRF
H041 Operational HWRF implemented FY 2008
2007 HWRF
Improved Track Forecast Errors
8
2008 HURRICANE SEASON REVIEW ATLANTIC
9
Atlantic Track Error 2008 HWRF 2008 GFDL
HWRF track forecast errors are comparable to
those from GFDL for the Atlantic Multi-model
ensemble Approach
2008 HWRF
2008 GFDL
10
Atlantic Track Error 2008 HWRF 2008 GFDL GFS
HWRF track forecast errors are comparable to
those from GFDL and GFS for the Atlantic (MME)
11
Atlantic Intensity Error 2008 HWRF GFDL
HWRF Performed better than GFDL (about 1-2 kts)
up to 84-hr forecast, some degradation (about 2
kts) at 96- and 120-hr forecasts.
12
BERTHA (02L)
GFDL
HWRF
13
HWRF taking Fay to Gulf
14
Gustav (07L)
HWRF
GFDL
HWRF tracks are more consistent (less forecast to
forecast variation)
15
Ike (09L)
HWRF
GFDL
Ike One of the difficult 2008 storms for HWRF
16
Ike northward turn into Fl during early stages
of forecast Sept. 05, 00Z
24 hrs later. Sept. 06, 00Z
17
48 hrs later. Sep. 10 00Z
48 hrs later. Sep. 08 00Z
18
2008 HURRICANE SEASON EASTERN PACIFIC
19
EPAC Track Error 2008 HWRF 2008 GFDL GFS
HWRF, GFS still have some problems in the Eastern
Pacific track forecasts
20
EPAC Intensity Error 2008 HWRF 2008 GFDL
Positive intensity bias in the early stages of
HWRF forecasts (possibly due to lack of ocean
feedback?)
21
2009 HWRF IMPLEMENTATION

• Refine Hurricane Initialization (Q. Liu)
• Gravity Wave Drag Parameterization (Kwon)
• Sea Spray Parameterization (Fairall, Bao)
• Bug Fix for Cooler Land Surface Temperatures in
  the inner domain (Tuleya, Zhang)
• HWRF-HYCOM Coupled System will run in parallel to
  the operational HWRF-POM System (H-S Kim, MMAB)
• Testing Tallapragada, OConnor, et al.

22
2009 HWRF Testing Protocol

• Test 2008 HWRF with GSI upgrades - DONE
• Test each upgrade candidate on relevant cases for
  2008 ops GFS- DONE
• Test each upgrade candidate with GSI upgrades
  IN PROGRESS
• Combine candidates into single experimental
  system for final testing, including
• GSI upgrade
• Any GFS changes
• IN PROGRESS

23
QL08 New HWRF Initialization Early results
indicate improved track and intensity forecasts
for Hurricane Dean
24
No GWD
GWD
50nm improvement at t120hr
25
H41A Bench Mark 2008 HWRF H044 w/Gravity Wave
Drag. Dennis/Katrina/Rita/Wilma (2005) and Dean
2007
Negligible impact on Atlantic Hurricanes
(neutral) for strong benchmark cases (testing in
progress for weaker storms)
GWD
No GWD
26
Sea Spray Negligible impact on Atlantic
Hurricane track forecasts (slight degradation at
120 hrs)
Sea Spray
No Sea Spray
27
Sea Spray Increased intensity errors due to
increased positive intensity bias
Sea Spray
No Sea Spray
28
Cold Land Surface Temperatures in HWRF moving grid
Cooler (by about 6-10 deg) land surface
temperatures due to irregular computation of
radiation in the inner domain (due to nest motion)
Problem solved by calling radiation for the nest
at regular intervals
Radiation called at nest motion threshold
interval (9 min)
Radiation consistent with parent domain (54 min.)
29
HWRF-HYCOM Experiments

• Configuration of HWRF-HYCOM coupled system is
  complete.
• HYCOM better represents the Ocean physics than
  POM. However, there is a cold bias in SST,
  possible due to heat budget and wind stress
• Preliminary results show weaker storms in the
  Atlantic.
• Work in progress.
• Improved version of HWRF-HYCOM slated to go in
  parallel FY 2009.

30
Hurricane Forecast Improvement Project (HFIP)

• Accelerate skill of numerical guidance and
  associated products
• Intensity
• Track
• Major foci
• Ensemble systems
• Data assimilation
• Impact of high resolution
• Physics improvements
• Established Teams
• Regional data assimilation
• Ensemble-based guidance
• Global and regional models and physics
• Ocean model and coupling
• Diagnostics
• Product generation

31
Hurricane Diagnostics

• Ongoing and continuous efforts to develop a
  system for comprehensive model diagnostics for
  hurricane forecasts
• Primary tasks include
• Evaluation of initial storm structure (analyzed),
  
• Vortex evolution in the forecasts,
• Representation of large-scale flow in HWRF and
  GFDL compared to the GFS
• Impact of boundary conditions, domain
  configurations
• Impact of physics, ocean feedback, horizontal and
  vertical resolution
• Evaluation of derived diagnostic products
  including energy, angular momentum and PV budgets
• Collaborative effort with Mark DeMaria

• HPLOT capabilities
• Model side-by-side
• comparison
• Standard diagnostics
• vert. shear
• x-sections, etc.
• Based on generic software

32
NOAAs Aircraft in Hurricanes
Working on flight strategies for GIV and
P-3s Working to obtain resources for flight
hrs/addl crews Operational requirement for core
data at 0Z, 12Z to initialize HWRF core
circulation (06,18Z ?) Operational requirement
for AXBTs Two mission profiles
Environment CORE Observations
GPS, Radar, AXBTs New operational status of
P-3s
33
Advancing the HURRICANE WRF System
09 10 11
12 13
Mesoscale Data Assimilation for Hurricane Core
Doppler Radial velocities Reflectivity
Experimental advanced assimilation
techniques EnsDA 4D-var Hybrid

Atm. Model physics and resolution upgrades
(continuous)
Atm/ocean boundary layer, enthalpy fluxes
Microphysics,
radiation



Incr. Res 4-6km/100L?
HWRF Ensembles
.

Land surface Coupling Waves multi-grid/surf-zon
e physics Ocean 4km. - continuous upgrades
in Ocean Data Assimilation


34
THANKS Questions?
35
Additional Slides

• ATLANTIC HURRICANE TRACKS

36
Hurricane-Wave-Ocean-Surge-Inundation Coupled
Models
NCEP/Environmental Modeling Center Atmosphere-
Ocean-Wave-Land
NOS land and coastal waters
HWRF SYSTEM

NMM hurricane atmosphere
NOAH LSM
runoff
High resolution Coastal, Bay Estuarine
hydrodynamic model
surge inundation
fluxes
Atmosphere/oceanic Boundary Layer
radiative fluxes
winds air temp.
other fluxes
elevations currents 3D salinities
temperatures
SST currents
HYCOM 3D ocean circulation model
wave spectra
WAVEWATCH III Spectral wave model
wave fluxes
37

• 2009-2012 HURRICANE UPGRADE PLAN
• Data assimilation
• Advanced initialization for hurricane core -
  assimilate airborne doppler radar obs (new radar
  on GIV, P-3s) to define storm strength and
  storm structure in HWRF analyses. (will run in
  parallel in 07)
• Continuous upgrades to HWRF hurricane core
  initialization through advanced 4-D data
  assimilation for winds and reflectivity
  (requires nearly continuous obs for hurricane
  structure from storm top to surface beginning in
  depression stages through evolution of storm
  lifecycle). Augment doppler data with GPS
  sondes. Also, explore use of satellite microwave
  data.
• Model resolution upgrades
• Increase in resolution Horizontal 3-6km.
  /Vertical resolution 100levels (dependent on
  results of current studies).
• Hurricane Ensembles Hi-res hurricane model
  ensembles, e.g. 4km? Multi model ensembles?
  (e.g. GFDL, GFDN, HWRF) Work on development of
  HWRF ensembles in progress.

38

• Model Physics Continuous upgrades to
  atm/ocean boundary layer (fluxes), microphysics,
  deep convection (cloud resolving scales?),
  radiation
• Coupling to land surface model w/ adv. Sfc.
  Physics for improved rainfall forecasts at
  landfall. Important input to hydrology and
  streamflow models to address inland flooding.
• Advance Wave Model (WAVEWATCH III) to forecast
  waves up to the beach, ie improve non-linear
  interactions, surf zone shallow water physics,
  wave interactions with currents
• ALL hurricane model physics upgrades dependent
  on allowable resolution, e.g. affordable
  complexity of microphysics, sea spray, explicit
  representation of all clouds.

39
ARTHUR (01L)
GFDL
HWRF
40
Cristobal 03L
HWRF
GFDL
41
Dolly (04L)
GFDL
HWRF
42
Edouard (05L)
GFDL
HWRF
43
Josephine (10L)
HWRF
GFDL
44
Kyle (11L)
HWRF
GFDL
45
Laura (12L)
HWRF
GFDL
46
Marco (13L)
HWRF
GFDL
47
Nana (14L)
GFDL
HWRF
48
Omar (15L)
HWRF
GFDL
49
Paloma (17L)
HWRF
GFDL
50
Additional Slides

• Eastern Pacific HURRICANE TRACKS

51
Alma (01E)
HWRF
GFDL
52
Douglas (04E)
HWRF
GFDL
53
Elida (06E)
HWRF
GFDL
54
Fausto (07E)
HWRF
GFDL
55
Genevieve (08E)
HWRF
GFDL
56
Hernan (09E)
HWRF
GFDL
57
Iselle (10E)
HWRF
GFDL
58
Julio (11E)
HWRF
GFDL
59
Karina (12E)
HWRF
GFDL
60
Norbert (15E)
HWRF
GFDL
61
Odille (16E)
HWRF
GFDL
62
Polo (18E)
HWRF
GFDL
63
Additional Slides

• Details of Upgrades to Initialization

64
2009 Upgrades - HWRF Initialization

• Change in the minimum surface pressure
  specification make the surface pressure-wind
  relationship more consistent with the model
  instead of the observation
• Redefine the filter domain size based on
  observation.
• Filter domain is defined too large in some cases,
  particularly near frontzone or waves.
• Define filter domain in GFS analysis data as 1.1
  times the radius of outmost isobar.
• Set the limit for small size storm
• HWRF has problems to forecast intensity for small
  size storms, we set
• Radius of maximum wind 19 km
• Radius of the outmost isobar 334 km
• Search maximum wind position at 10m height
  instead of the first level height
• Convert the model level 1 wind to 10m, then
  search the position of the maximum wind
• Increase the search area for the maximum wind in
  high latitude storms
• set the maximum search radius as 4.5 degrees for
  storms stronger than 30m/s and in north of 300N.

65
Additional Slides

• Details of GWD Parameterization

66
Gravity wave vertical propagating and breaking
Mountain blocking
67
Influence of orography on the atmosphere
Create obstacles and additional turbulence
Gravity wave drag
Change the large scale flows
Generation of vertically propagating gravity waves
Change the track of hurricanes
68
Why do we need GWD parameterization?
NWP models use grid-averaged (smoothed) terrain
data
Coarse resolution models ( gt 4km) cannot resolve
the GWD caused by subgrid scale topography
69
Simulation results of various resolutions show
that a coarse model does not produce enough
gravity wave drag caused by terrain
?X33km
?X10km
Z
X
?X3.3km
70
Motivations

• Track forecast skills of HWRF on Eastern Pacific
  storms are not as good as those on Atlantic
  storms
• Diagnotics of HWRF indicates the anomalous flows
  developed over Mexican Plateau seems to cause the
  less skillful track forecast of HWRF
• Proper GWD representation might improve the track
  forecast of HWRF

Results NEXT PAGE
71
Additional Slides

• Details of Sea Spray Parameterization in HWRF

72

• The ESRL Sea-Spray Scheme
• Wind-dependent source parameterization
• Thermal feedback warm-up of surface layer at
  high winds
• Momentum feedback surface drag reduction
• Purpose To take into account sea-spray effects
  on the surface momentum and heat fluxes

• Implementation in the HWRF model
• The ESRL sea-spray scheme is called by (and
  appended to the end of) the subroutine
  module_bl_gfs.F.
• A function routine (function qsat_spray) is added
  to the function subroutine module_gfs_funcphys.F.
• Evalution runs have been conducted at EMC for
  Katrina (2005), Rita (2005), Dennis (2005) ,
  Wilma (2005) and Dean (2007).

73
Results of Evaluation Testing

• For strong storms (such as Katrina and Rita), the
  scheme tends to produce a greater positive bias
  of intensity during the first 48-72 hours than
  the control runs, while the impact on track is
  negligible.
• For weak storms (such as Dennis), the scheme
  tends to produce an intensity bias that varies
  around that of the control runs, while the impact
  on track is slightly degraded after 72 hours.
• The storm structure is affected by the sea-spray
  mediated momentum and heat fluxes, suggesting a
  strong connection between the surface fluxes and
  the vortex dynamics through the convection in the
  eyewall.
• The performance of the scheme can be improved by
  tuning the source function and the degree of
  feedback effects.

74
HWRF 08 implementation

• HWRF Analysis Changes
• Weak storm treatment The weak storm is a
  weighted blend of the guess and the composite
  storm. Weights are based on the observed storm
  intensity and the vertical structure of the guess
  storm.
• Storm Balance Instead of assuming gradient wind
  balance, surface pressure is treated as a
  function of non-linear gradient wind stream
  function. Once surface pressure is adjusted,
  vertical temperature fields are adjusted based on
  hydrostatic balance.
• Changes to POM initialization
• Feature based assimilation of mesoscale oceanic
  features
• Assimilate more than one warm core rings as well
  as cold core rings
• Rings can be assimilated close to the loop
  current
• HWRF Physics changes / Bug fixes
• Initialize TKE to zero (HWRF does not use TKE
  based PBL)
• Adjustment of temperature and pressure fields
  after nest motion to remove noise at the lateral
  boundary of nest domain
• HWRF Script Changes
• Improved functionality of all HWRF scripts
• HWRF scripts for EMC parallels are consistent
  with operational configuration.

75
Intensity Forecast Skill Comparison NAOMIex
HWRF-HYCOM Coupled System
76
Sea Spray Parameterization in HWRF

• Collaborative Effort with C. W. Fairall and J.-W.
  Bao of NOAA ESRL
• Experiments for Benchmark Cases conducted on
  Vapor by Bao at ESRL
• Preliminary results indicated increased positive
  bias in the HWRF intensity forecasts (and thus
  increased errors)
• Further tuning should be possible to reduced the
  bias and increase the forecast skill.
• So far negligible impact on track forecasts.

77
Parameterization of Orographic Gravity Wave Drag
(GWD)

• Diagnostics of HWRF forecasts for the Eastern
  Pacific Hurricanes indicates that the anomalous
  flows developed over Mexican Plateau seems to
  cause the less skillful track forecast of HWRF
• Proper GWD representation might improve the track
  forecast of HWRF
• Probable candidate for 2009 implementation

78
2009 Upgrades - HWRF Initialization

• Change the minimum surface pressure specification
  based on observations
• Redefine the filter domain size based on
  observations
• Set the limit for small size storm
• Search maximum wind position at 10m height
  instead of the model first level
• Increase the search area for the maximum wind in
  high latitude storms

79
Marie (14E)
HWRF
Another difficult storm
GFDL
80
Cristina (03E)
HWRF
More Consistent
Increased Spread
GFDL
81
Lowell (13E)
GFDL
HWRF
Difficult storm for HWRF
82
Boris (02E)
HWRF
GFDL
83
48 hrs later. Sep. 12 00Z
Landfall . Sep. 13 06Z
84
6 hours later all others follow suite
Ike Forecast starting at 2008090906
HWRF
85
Hanna (08L)
HWRF
GFDL
86
Fay (06L)
HWRF
GFDL
Tracks stayed south of the gulf coast
87
HWRF vs. GFDL
HWRF still has some problems in the Eastern
Pacific track forecasts.
88
Done
Addl Tuning
Done
2009 Para
Done
89
Hurricane Diagnostics and Plotting Tool HPLOT

• Enhanced capabilities to view model output from
  two different models for a side-by-side
  comparison.
• Ability to plot the differences between two model
  forecasts interpolated to a common resolution
• Plotting tracks and intensities on any displayed
  graphics
• Computation of some standard hurricane-related
  diagnostics including zonal and meridional
  components of vertical shear, arbitrary and
  single point cross sections, mean flow between
  different vertical layers and plotting skew-T
  plots.
• HPLOT is continuously being upgraded to add
  further capabilities of statistical evaluation of
  model forecasts (RMS errors, bias, anomaly
  correlations etc.)
• This is a GUI based software with easy to
  navigate menus and features.

90
HPLOT Visualization Software
Comparison of HWRF/GFDL forecasts for 850-200 hPa
mean wind. Anomalous blocking pattern along the
west coast in HWRF.
91
HWRF Pre-implementation TE FY2008

• Upgrades to HWRF were tested for a selected set
  of Atlantic Hurricanes (Benchmark)
• DENNIS, KATRINA, RITA, WILMA (2005) and DEAN
  (2007)
• Results showed significant improvements to track
  and intensity forecasts
• Major improvements in reducing the negative bias
  of HWRF intensity forecasts