MARCH 2007AMERICAN METEOROLOGICAL SOCIETY
|
311
AFFILIATIONS: CHEN, ZHAO, AND DONELAN—Rosenstiel School
of Marine and Atmospheric Science, University of Miami, Miami,
Florida; PRICE—Woods Hole Oceanographic Institution, Woods
Hole, Massachusetts; WALSH—NASA/Goddard Space Flight
Center, Wallops Island, Virginia
CORRESPONDING AUTHOR: Dr. Shuyi S. Chen, RSMAS/
University of Miami, Miami, FL 33149
E-mail: schen@rsmas.miami.edu
DOI:10.1175/BAMS-88-3-311
©2007 American Meteorological Society
The CBLAST-Hurricane Program and the Next-
Generation Fully Coupled Atmosphere–Wave–Ocean
Models for Hurricane Research and Prediction
BY SHUYI S. CHEN, JAMES F. PRICE, WEI ZHAO, MARK A. DONELAN, AND EDWARD J. WALSH
sphere–wave–ocean modeling system that is capable
of resolving the eye and eyewall at ~1-km grid resolu-
tion, which is consistent with a key recommendation
for the next-generation hurricane-prediction models
by the NOAA Science Advisor Board Hurricane
Intensity Research Working Group. It is also the Na-
tional Centers for Environmental Prediction (NCEP)
plan for the new Hurricane Weather Research and
Forecasting (HWRF) model to be implemented op-
erationally in 2007–08.
AIR–SEA INTERACTION AND HURRI-
CANES. Hurricanes rarely reach their maximum
potential intensity (MPI, as de ned by Kerry Emanuel
and Greg Holland). Many factors can prevent a given
storm from reaching MPI, including environmental
vertical wind shear, distribution of troposphere water
vapor, hurricane internal dynamics, and air–sea in-
teractions. e e ect of air–sea interactions on hurri-
cane structure and intensity change is the main focus
of the CBLAST-Hurricane program. Intensi cation of
a hurricane depends upon two competing processes
at the air–sea interface—the heat and moisture uxes
that fuel the storm and the dissipation of kinetic en-
ergy associated with wind stress on the ocean surface.
Air–sea interaction is especially important within
the extremely high winds (up to 75 m s
−1
) and strong
gradient zones of temperature and pressure located
in the inner core (eye and eyewall) of a hurricane.
e enthalpy and momentum exchange coe cients
under the extreme high-wind conditions are, of
course, very di cult to determine in precisely the
regions where they are most important. e stress is
supported mainly by waves in the wavelength range
of 0.1–10 m, which are an unresolved “spectral tail”
in present wave models.
In the November 1995 Journal of the Atmospheric
Sciences, Emanuel proposed that storm intensity is
largely controlled by the ratio of the air–sea enthalpy
T
he record-setting 2005 hurricane season has
highlighted the urgent need for a better under-
standing of the factors that contribute to hurri-
cane intensity, and for the development of correspond-
ing advanced hurricane prediction models to improve
intensity forecasts. e lack of skill in present forecasts
of hurricane intensity may be attributed, in part, to de-
ciencies in the current prediction models—insu cient
grid resolution, inadequate surface and boundary-layer
formulations, and the lack of full coupling to a dynamic
ocean. e extreme high winds, intense rainfall, large
ocean waves, and copious sea spray in hurricanes push
the surface-exchange parameters for temperature, water
vapor, and momentum into untested regimes.
e Coupled Boundary Layer Air–Sea Transfer
(CBLAST)-Hurricane program is aimed at develop-
ing improved parameterizations using observations
from the CBLAST-Hurricane field program (de-
scribed by Peter Black and colleagues elsewhere in
this issue) that will be suitable for the next generation
of hurricane-prediction models. e most innovative
aspect of the CBLAST-Hurricane modeling e ort is
the development and testing of a fully coupled
1
atmo-
1
e so-called fully coupled model here refers to two-way
coupling with simultaneous communication between the
two models.
MARCH 2007
|
312
and momentum ux exchange coe cients, C
k
/C
D
.
Using a simple axisymmetric model with idealized
environmental conditions, Emanuel showed that
this ratio needs to be equal to or greater than 1 for
hurricanes to intensify. As shown in many studies,
C
D
is sea-state dependent while C
k
has relatively little
sensitivity to sea state. Other research shows that the
e ects of sea spray on the air–sea exchange may also
be important. Recent laboratory experiments con-
ducted at hurricane wind speeds by Donelan and col-
leagues have shown that C
D
reaches a saturation point
at high wind speeds greater than about 33 m s
−1
, when
ow separation begins to occur, and that C
k
remains
relatively constant. The airborne turbulence flux
measurements from CBLAST-Hurricane reported in
the Journal of the Atmospheric Sciences by Drennan
and colleagues also support these laboratory results,
indicating that C
k
/C
D
is less than 1 for intensifying
storms (e.g., Hurricane Fabian in 2003). e rapid
increase in computer power and recent advances in
observation technology have made it possible for us
to develop a strategy for the next generation of high-
resolution hurricane prediction models.
A COUPLED MODELING SYSTEM. is paper
describes the strategy and current activity in develop-
ing and testing a new, high-resolution, coupled atmo-
sphere–wave–ocean model for hurricane research and
prediction. e fully coupled atmosphere–wave–ocean
modeling system includes the following three compo-
nents: the atmospheric model, a surface wave model,
and an ocean circulation model. e basic coupling
parameters—that is, the data passed between the mod-
els—are noted in a schematic
in Fig. 1. A speci c issue we
emphasize here is the determi-
nation and parameterization
of the air–sea momentum and
enthalpy uxes in conditions
of extremely high and time-
varying hurricane winds.
The atmospheric model. The
atmospheric component of the
coupled modeling system will
be either the nonhydrostatic
fifth-generation Pennsylvania
State University (PSU)–Na-
tional Center for Atmospheric
Research (NCAR) Mesoscale
Model (MM5) or the Weather
Research and Forecasting (WRF) model. A special prop-
erty of the atmospheric model is that it must support
very high spatial resolution. The radius of maximum
wind (RMW)—that is, the distance from the center
of the eye to the ring of the maximum wind speed—is
typically 15–30 km, although occasionally it is as small
as 5–10 km, as observed in Hurricanes Lili (2002) and
Wilma (2005). To resolve the eye and eyewall structures
in a hurricane, the horizontal resolution (or grid spac-
ing) in numerical models needs to be on the order of
~1 km. Figure 2 shows an example of model-simulated
Hurricane Floyd (1999) with various grid resolutions
and a comparison with the observations (this figure
and much of the work described here can be found in
more detail in a paper submitted by Chen and Tenerelli
to Monthly Weather Review). Only the 1.67-km simula-
tion can reproduce the observed inner-core structure,
whereas the simulations with 5- and 15-km resolution
clearly do not (Fig. 2).
To capture the long life cycle of hurricanes and
resolve the inner-core structure, we developed a vor-
tex-following nested grid that allows the model to be
integrated for 5 days or longer at a very high resolu-
tion (~1 km) in the innermost domain. e high-reso-
lution elevation and land-use data are refreshed on the
ne meshes each time they are initialized or moved.
We use four nests with 45-, 15-, 5-, and 1.67-km grid
spacing. e three inner domains move automatically
with the storm. e same vortex-following moving-
nest capability has been adapted in WRF. e model
has been used to simulate Hurricanes Bonnie (1998),
Floyd (1999), and Lili (2002). e inner core of hur-
ricanes is simulated explicitly in the cloud-resolving
FIG. 1. Schematics of a coupled atmosphere–wave–ocean modeling system with
the component atmosphere, surface wave, and ocean circulation models, as well
as the coupling parameter exchanges between each of the component models.
MARCH 2007AMERICAN METEOROLOGICAL SOCIETY
|
313
mode. e microphysics scheme is based on the work
of Wei-Kuo Tao and Joanne Simpson. e Blackadar
PBL scheme is used on all grids, but over water we
include a modi cation (based upon Garratt’s 1992
book, e Atmospheric Boundary Layer) in which we
introduce di erent roughness scales for temperature
z
t
and moisture z
q
. In the uncoupled MM5 and WRF
applications, the momentum roughness length z
o
over
the open ocean is calculated from the relationship
described in Charnock’s 1955 article in the Quarterly
Journal of the Royal Meteorological Society. e NCEP
global analysis elds (6 hourly and 1°
× 1°) and the
high-resolution (~9 km) Advanced Very High Resolu-
tion Radiometer (AVHRR) Path nder SST analysis
as well as the Tropical Rainfall Measuring Mission
(TRMM) Microwave Imager (TMI) Advance Micro-
wave Sounding Radiometer (AMSR) SST (~25 km) are
used to initialize the uncoupled MM5 and provide
continuous lateral and lower boundary conditions.
Ocean model. Hurricanes draw energy from the ocean
surface, and they cool the ocean by wind-induced
FIG. 2. (a) The NOAA/Atlantic Oceanographic and Meteorological Laboratory (AOML)/HRD airborne radar-
observed reflectivity (dBZ, over an area of 360 km x 360 km) and the MM5-simulated rain rate (mm h
−1
) using
(b) 1.67-, (c) 5-, and (d) 15-km grid resolution in Hurricane Floyd at 0000 UTC 14 Sep 1999.
MARCH 2007
|
314
surface fluxes, which are crucial to the hurricane, and
by vertical mixing, which is dominant with the ocean
surface layer. This vertical mixing occurs as a response
to the large-amplitude near-inertial currents gener-
ated within the oceanic mixed layer (OML). Upward
vertical motion generated by the wind-driven currents
(upwelling) may also enhance SST cooling by lifting
the base of the OML and bringing cooler water closer
to the sea surface. The amplitude of SST cooling result-
ing from a given hurricane thus depends in part upon
the thermal stratification of the upper ocean, which is
sometimes represented by an integrated heat content.
The oceanic part of a coupled modeling system then
should include a realistic thermal stratification, an
appropriate parameterization of vertical mixing, and
the possibility of upwelling to achieve an accurate
representation of the SST cooling.
e ocean model in the coupled system will be a
three-dimensional, primitive equation, hydrostatic
upper-ocean model (the 3DPWP of Price et al.’s 1994
Journal of Physical Oceanography paper) or the Hy-
brid Coordinate Ocean Model (HYCOM; developed
by Chassignet et al.). It is used to simulate the upper-
ocean current and temperature elds underneath the
hurricane. e 3DPWP model domain is the same
as the outer domain of the atmospheric model with
15-km grid spacing. It has 30 vertical levels with grid
spacings of 5–10 m for the top 20 levels and 20 m for
the remaining levels. It is initialized using observed
and climatological temperature and salinity pro les.
The temperature profile is blended from selected
prestorm airborne expendable bathythermograph
(AXBT) observations from Hurricane Research
Division (HRD) aircra missions and LEVITUS94
climatological temperature data for depths greater
than those sampled by AXBT observation. At each
time step of the ocean model (10 min), the ocean
model takes surface stress and heat and moisture
uxes from the atmospheric and wave models and
steps ahead the ocean dynamics. At the same time in-
terval, the ocean model passes back the SST anomaly
(the di erence between the initial and current SST)
to the atmospheric model, and passes back the ocean
surface current to the wave model.
Surface wave model. The coupling of the atmosphere
through waves to the ocean is best served by a direct
calculation of the evolution of the wave field and
the concomitant energy and momentum transfer
from wind to waves to upper-oceanic layers. A
third-generation wave model, the WAVEWATCH
III (WW3), is used to simulate ocean surface waves
in the atmospheric–ocean–wave coupling system. It
was developed by Tolman in Weather and Forecasting
(2002) for wind waves in slowly varying, unsteady,
and inhomogeneous ocean depths and currents.
WW3 is extensively evaluated and validated with
observations. The wind waves are described by the
action density wave spectrum N (k, θ, x, y, t). We use
25 frequency bands, logarithmically spaced from
0.0418 to 0.41 Hz at intervals of Δƒ/ƒ = 0.1, and 48
directional bands at 7.5° intervals. The water-depth
data used in the wave model are the 5-min gridded
elevation data from the National Geophysical Data
Center. In the coupled system, the wave model inputs
the surface-wind and ocean-surface current fields
from the atmospheric and ocean models and outputs
surface stress integrated from a new wind–wave cou-
pling parameterization developed by the CBLAST
modeling team.
Most of the wind stress is supported by surface
waves with a wavelength that is less than the cuto ,
typically 10–20 m, of existing third-generation wave
prediction models. In order to correct this shortcom-
ing, a new wave and wind-stress prediction model has
been developed and is being tested against both eld
data, with respect to its prediction skill in rapidly
changing wind conditions, and laboratory data of
FIG. 3 (opposing page). Coupled model-simulated (a) rain rate (color, mm h
−1
) and surface wind speed (black
contour with 10 m s
−1
interval); (b) enthalpy (sensible + latent) flux (color, W m
−2
) and surface wind (vector);
(c) significant wave height (color, m) and wave propagation direction (white vectors); and (d) mean wavelength
(color, m) and surface wind (black vectors) in Hurricane Frances at 1200 UTC 31 Aug 2004. The black “+”
indicates the storm center of Hurricane Frances. The arrow in the lower-left corner indicates the direction
of the storm motion. (e) Observed and (f) model-simulated SST (color, °C) and ocean surface current (vec-
tors) in Hurricane Frances at 1200 UTC 4 Sep 2004. Storm tracks are overlaid on each of the SST maps. The
red stars indicate the locations of the Electromagnetic-Autonomous Profiling Explorers (EM)-APEX floats:
1636 (on the storm track), 1633, and 1634 (50 and 100 km away from the storm center, respectively). The
observed vertical temperature and salinity profiles from the EM-APEX floats are used in the ocean model
initialization and evaluation.
MARCH 2007AMERICAN METEOROLOGICAL SOCIETY
|
315
(e)
direct measurements of wave spectra and Reynolds
stress. We developed a new wind–wave parameteriza-
tion that includes e ects of the wave spectral tail on
the drag coe cient. It calculates directional stress
using surface-wave directional spectra by parameter-
(f)
izing “spectral tails” (frequency > cuto frequency)
unresolved by the wave models. The wind–wave
parameterization incorporates an important feature
shown in the 2004 Donelan et al. article in Geophysi-
cal Research Letters: that the drag coe cient becomes