Tropical Cyclones Features Inferred from SAR Images

Abstract

Due to relatively small amount of in situ data available for the open oceans, particularly during extreme events, remote sensing techniques take an important role in the retrieval of geophysical information under such
conditions. This study is based on a dataset of wide swath Synthetic
aperture radar (SAR) images (400 km x 400 km coverage), acquired by the European ENVISAT satellite, to observe tropical. Wide swath SAR images offer the opportunity to observe the whole structure of tropical cyclones and
measure some characteristic parameters like cyclone size, rain band distribution, eye wall size and shape, as well as the radius of maximum wind speed. The objective of this study is to investigate the performance of SAR to improve the existing model for the retrieval of information on wind field under extreme wind and wave conditions using information from a
parametric Holland type model. Performing a 2-dimensional spectral analysis of the SAR images wavelength and direction of boundary layer rolls for information on mixed boundary layer depth and wind direction are determined. The work aims at the improvement of prediction of the
cyclone track, intensity and sea state.

Full text

Page 1

ABSTRACT
Due to relatively small amount of in situ data available
for the open oceans, particularly during extreme events,
remote sensing techniques take an important role in the
retrieval of geophysical information under such
conditions.
This study is based on a dataset of wide swath Synthetic
aperture radar (SAR) images (400 km x 400 km
coverage), acquired by the European ENVISAT
satellite, to observe tropical.
Wide swath SAR images offer the opportunity to
observe the whole structure of tropical cyclones and
measure some characteristic parameters like cyclone
size, rain band distribution, eye wall size and shape, as
well as the radius of maximum wind speed.
The objective of this study is to investigate the
performance of SAR to improve the existing model for
the retrieval of information on wind field under extreme
wind and wave conditions using information from a
parametric Holland type model.
Performing a 2-dimensional spectral analysis of the
SAR images wavelength and direction of boundary
layer rolls for information on mixed boundary layer
depth and wind direction are determined.
The work aims at the improvement of prediction of the
cyclone track, intensity and sea state.

1. INTRODUCTION
Tropical cyclone is a generic term which comprises
hurricanes in the Atlantic Ocean and Northeast Pacific
Ocean, typhoons in the Northwest Pacific Ocean, and
cyclones in the Indian Ocean and Southwest Pacific
ocean.
Tropical cyclones account for a significant fraction of
damage, injury and loss of life. There are still a lot of
aspects in the physics of tropical cyclones, which are
not well understood. Of particular importance are
processes taking place at the air sea interface, which is a
key component in the heat flux driving the cyclone.
Because of their all weather capability space borne
active microwave sensors like the Synthetic Aperture
Radar are of particular interest in this context [1].
Fig. 1 (left) shows three consecutive ENVISAT wide
swath SAR images of hurricane Katrina. The images
Figure1: ENVISAT Wide Swath images of hurricane
Katrina(left) acquired over the Gulf of Mexico on Aug.
28 2005 15:50 UTC. On the rigth is shown the
collocated Meris image.

were acquired on Aug. 28, 2005 at 15:50 UTC over the
Gulf of Mexico when the hurricane was at cat. 5 stage
according to the Saffir-Simpson scale.
The SAR image clearly shows very dark spiral shaped
features, which are likely due to extreme precipitation
events (rain bands). The red and blue circles correspond
to the radius of maximum wind speed (25 Km) and the
radius of hurricane force winds (130 Km) respectively.
On the rigth side of Fig.1 the collocated optical Meris
image is shown. This yelds the relationship between
cloud coverage and the area of destructive wind field.
On the SAR image is superimposed the retrieved wind
field using the SEASAR algorithm [2], based on the
scatterometer geophysical model function (GMF)
*
*
TROPICAL CYCLONES FEATURES INFERRED FROM SAR IMAGES

A. Reppucci, S. Lehner, J. Schulz-Stellenfleth

German Aerospace Center (DLR), Oberpfaffenhofen 82234
Wessling, Germany. Tel. +49 8153282101, email: antonio.reppucci@dlr.de
_____________________________________________________
Proc. ‘Envisat Symposium 2007’, Montreux, Switzerland
23–27 April 2007 (ESA SP-636, July 2007)

Page 2

Figure 2:Raeanalyzed wind field of hurricane Katrina,
processed by NOAA HRD.

CMOD5 [3] and a new interpolation technique for the
determination of the wind directions. Some artefacts in
the estimated wind field are clearly visible and the wind
speed error is less than 10% for measurements up to 20
m/s. In this range in fact the difference between the
SAR retrieved wind speed and two buoys measurement
(see red asterisks Fig. 1) is less than 2 m/s. The
maximum wind speed retrieved, of about 45 m/s, is
much lower than the in situ measured wind speed of
about 70 m/s as given by Hurricane Research Division
shown in Fig 2. Thus the retrieved wind field give a
good estimate of the effective hurricane size which is
not possible to recover from the optical image.
The paper aims at deriving a complete wind speed
estimate for the hurricanes
The paper is organized as follows: the dataset used for
this study is described in section 2. Section 3 presents
an analysis of size and orientation of boundary layer
rolls in hurricanes. The influence of rain on the radar
cross section and limits in the retrieval of wind speed in
typhoon conditions are discussed in section 4.
Furthermore a new technique to estimate the maximum
wind speed and the effective wind field in tropical
cyclones is presented, using an interpolation procedure
based on an analytical model. Finally in section 5 the
conclusions and future steps are summarized.




Figure 3: Simulation of wind speed for hurricane
Katrina using the Holland model.
2. DATA SETS
In this study ENVISAT ASAR images and numerical
model are used.

2.1 ENVISAT Wide Swath Images
Hurricane Katrina image acquired in ScanSAR mode
(400 Km x 400 Km) is used to observe and analyze the
mesoscale features associated with tropical cyclones,
e.g. ocean wave fields, wind rolls, rain bands.
ENVISAT has the ability to acquire data in the so called
ScanSAR mode, in which several beams are combined
to generate an image of 400 Km x 400 Km with 150 m
resolution. In ScanSAR mode a large part of a tropical
cyclone can be imaged synoptically.

2.2 Model Data
In order to improve wind speed measurement in
hurricane conditions the SAR measured wind speed is
compared to the wind speed simulated using the
parametric Holland Model [4]. The Holland Model is an
analytical model for the radial profiles of the near
surface (10 m above sea level) wind speed Vr in a
hurricane:
2
2
1
4
2
2
exp
/
fr
f
r
B
r
c
p
n
pBA
r
V
B
rA
⋅
−⋅+
⋅
−⋅⋅
=














−






ρ
(1)
where:

A =
B
MW
R


cn
M
−
pp
v
eB
=
2
ρ

Here, pc is the central pressure, pn is the environment
pressure (assumed constant at 1015 haP), r is the
distance from the center, f is the Coriolis parameter, and
ρ is the air density (assumed constant at 1.15). The

Page 3

Figure 4: Wind streak of hurricane Katrina. Image
acquired in the Gulf of Mexico on Aug28 2005 15:50
UTC.

model contains two parameters A and B that account for
the location of the maximum in the profile, relative to
the origin, and its shape [4]. The input parameters for
the model are the central pressure pc, the radius of
maximum wind speed RMW and the maximum wind
speed vM. Fig. 3 shows a simulation of Hurricane
Katrina at the time of the acquisition of the SAR image.
The input parameters pc, RMW and vM for this simulation
are taken from the NOAA Hurricane Research Division
(HRD) web site.
The standard Holland Model [4] does not include
information on wind direction. In this simulation we
used 20° spiral wind directions, i.e. VS is given as
follows:


⋅ °−
=
sinβ
sin20





⋅ °+⋅ °−
⋅−
cosβ
cos20
cosβ
sinsinβ
cos20
VV
rS
(2)

where β is the radial wind direction.

In the final step a vector corresponding to the direction
and speed of forward movement of hurricane VA (see
with arrow Fig.1 has been added to the Holland wind
field to obtain a more realistic simulation :

AS
VVV
+=
(3)

3. BOUNDARY LAYER ROLLS
On the SAR images often so-called wind streaks are
visible, ranging in wavelengths from 600 to 2,000
meters [5]. This variation in sea surface roughness is
explained by changes in surface wind speed due to the
formation of boundary layer rolls. The direction of these


Fig.5: Histogram of wavelength of the boundary rolls
measured from the image of hurricane Katrina.

streaks is used to derive the wind direction [6] and the
wavelength is taken to be a measure of roll size and thus
mixed layer depth. To determine wavelength and
direction of boundary layer rolls a spectral analysis on
subscenes of 10 km x 10 km is performed. In Fig.4 the
direction and wavelength of the boundary layer rolls
measured from the center image of hurricane Katrina
(Fig.1) are shown. These were most prominent in the
front semicircle of the storm. Fig 5 shows the histogram
of the wavelength of the boundary rolls measured from
the center image of hurricane Katrina. The mean
retrieved wavelength is about 990 m, the same order of
magnitude as found in literature [7].
Due to many sea surface feature is not possible retrieve
the wind direction from all over the swat. Thus in [2]
the direction are determined localzing the wind rolls in
the image and then appliyng a new interpolation
technique on all the swath.

4. ANALYSIS
HURRICANE SAR IMAGES
OF RAIN IMPACT ON
Strong rain can cause damping of the microwave radar
signals in the atmosphere and additional roughening or
damping of the sea surface, depending on the rain rate
and the radar wavelength and incidence angle [8]. There
are also strong indications that the special wind
conditions within a rain band can lead to a reduction of
the ocean surface roughness and thus the respective
NRCS [9]. Finally there is a potential impact of ocean
waves on the surface stress. In the complicated
hurricane sea state conditions this effect is not very well
understood by now.

4.1 Forward simulation of NRCS
In fig. 6 a cut through the center of the hurricane's eye
in the SAR image of Fig. 1 along the azimuth direction
is shown in red. The blue curve represents a simulation
Wavelength [m]

Page 4

Figure 6: Along track cut of Fig 1 ((red) and simulated NRCS (blue) using the Holland model. Brown parts
correspond to land acquisition.




of the expected NRCS using wind speed from the
Holland model of Fig. 3 and the geophysical model
function CMOD-5 [3]. From the cut it is possible to
estimate the radius of maximum wind speed RMW as
approximately 25 Km. This is consistent with the in situ
measurement reported by NOAA.
Furthermore one can see a strong damping of two dB or
more in the centre of the rain band. The measured
NRCS is lower than the expected one within a distance
of 300 Km from the center of the hurricane. This
damping can be in part explained by the atmospheric
attenuation due to rain and by the change on the surface
due to the impinging rain drops. In Fig.7 the attenuation
due to different rain rates is plotted considering a rain
layer of height H equal to 5 Km as a function of the
Figure 7: NRCS attenuation due to the rain vs.
undisturbed surface NRCS for different rain rates.
undisturbed NRCS according to the radiative transfer
model [10]:

1
~
)sec(2
00
ησσ
a
k

where ka is the attenuation coefficient and η isthe
volume backscattering coefficient, θ is the radar
~σ is the attenuated NRCS and
)1 (
2
)sec(2
θθ
kHkH
ee
−−
−+=
(4)
incidence angle,
00
σ is
the undisturbed NRCS.
As can be seen for a rain rate of 50 mm/h, which can be
easily reached in hurricane conditions, the attenuation
can be close to -1 dB. Moreover must be considered that
the CMOD5 GMF has been tuned for high wind speed
using the measurement acquired by a scatterometer, on
board an airplane operating at C-band in hurricane
conditions [2], [11] excluding rain cases. Some recent
studies using the same scatterometer [12] have shown
that for wind speed above 30 m/s and rain rate of 15
mm/h the error in the wind speed estimation can be
more than 10 m/s, increasing with wind speed. Since the
airplane used for that experiment operated at heights
between 1.5 Km and 2 Km, the effect of atmospheric
attenuation are expected to be relatively weak. This
suggests that in high wind speed conditions the rain can
strongly modify the sea surface causing an additional
damping of the measured NRCS that must be added to
the one due to atmospheric attenuation (Fig.7).
Thus it does not seem possible to apply a rain correction
without dedicated measurements.

4.2 New Method to Estimate the Maximum
Hurricane Wind Speed.
Fig.8 shows the wind speed retrieved from the NRCS
profile of Fig.6 (in red) and the wind speed for a new fit

20 mm/h
30 mm/h
40 mm/h
50 mm/h
Holland simulated
NRCS
Measured NRCS
Radius Max
Wind Speed

Page 5

of the Holland model (in blue). The blue profile has
been obtained fitting an appropriate Holland model, as
given by Eq. 1, that matches the retrieved wind speed in
the radius over 300 Km, where the wind speed is below
17 m/s (tropical storm force winds). In the fitting
procedure the radius of maximum wind speed has been
fixed at 25 Km, as measured in the SAR image, while
the central pressure and the maximum wind speed are
free variables to be optimized. Then the retrieved
maximum wind speed is the one that minimizes the cost
function Eq. (5).

()
[
i
Mc
vpJ
σ
where
()
Mc
vp ,
σ
= Holland simulated NRCS
Figure 8: Wind Speed retrieved from the NRCS Profile of Fig.6 (red) and the simulated wind speed using the
Holland model (blue), eq.1. Brown corresponds to acquisitions over the land
()
]
o
i
o
iMc
o
i
vp
σσ
2
,
,
∑
−
=
(5)



The results of the fitting yield a new maximum wind
speed of 72 m/s and a wind profile as shown in Fig. 8.
Using this technique the max wind speed estimated
agree with the HRD analysis (Fig.2) of 70 m/s.
5. CONCLUSION
The analysis showed that ENVISAT ASAR tropical
cyclones images can be used to infer information on
structure, eye size, radius of maximum wind speed and
rain band distribution from the image.
The available geophysical models for the retrieval of
wind field do not account for the rain contamination that
for high wind speed and high rain rate can produce a
strong damping of NRCS. This turn in a strong
underestimation of wind speed in hurricane conditions.
A new fitting technique to derive the maximum wind
speed and so the hurricane strength is shown.
6. REFERENCES
1
J. Horstmann, D.R. Thompson, F. Monaldo, H.C.
Graber, and S. Iris,(2005). Can Synthetic
Aperture Radars be used to Estimate Hurricane
Force Winds?. Geophys. Res. Let., vol. 32.
J. Schulz-Stellenfleth, S. Lehner, A. Reppucci, S.
Brusch, and T. König. On the divergence and
Vorticity of SAR Derived Wind Fields. This
issue
H. Hersbach, (2003) CMOD5 an improved
geophysical model function for ERS C-band
scatterometry Internal report, European Centre
for Medium- Range Weather Forecast.
G. J. Holland, (1980). An Analytical Model of the
Wind and Pressure Profiles in Hurricanes.
Monthly weather review, vol.108, pp. 1212-1218.
W., Alpers, and B. Bruemmer, (1994). Atmospheric
boundary layer rolls observed by the synthetic
aperture radar aboard the ERS-1 satellite.
J. Geophys. Res., vol. 99, no. C6,pp. 12,613–
12,622, 1994.
S. Lehner; J. Horstmann, W. Koch; W. Rosenthal
(1998).Mesoscale wind measurements using
recalibrated ERS SAR images. Journal of
Geophysical Research, Vol. 103, no. c4, pp.
7847-7856.
I. Morrison, S. Businger, F. Marks P. Dodge, J. A.
Businger, “An Observational Case for the
Prevalence of Roll Vortices in the Hurricane
Boundary Layer”, Journal of the Atmospheric.
Sciences, vol.62, 2005, pp.2662-2673.
C. Melsheimer, W. Alpers and M. Gade (1998).
Investigation of
multipolarization radar signatures of rain cells
over the ocean using SIR-C/X-SAR data. J.
Geophys. Res, vol. 103,no. c9, pp. 18867-18884
2
3
4
5
6
7
8
Multifrequency-

Holland wind speed
Retrieved wind speed
tropical storm force winds
o
o
i
σ = Measured NRCS.
Wind Speed [m/s]

Page 6

9
M. Powel, (1990). Boundary Layer Structure and
Dynamics in Outer Hurricane Rainbands. Part I:
Mesoscale Rainfall and kinematic Structure.
Monthly Weather Review, Vol. 118, no. 4, pp.
891–917 .
10 F.T. Ulaby, R.K. More, and A.K. Fung, (1981).
Microwave Remote Sensing: Active and Passive.
MA: Addison-Wesley, vol. 1.
11 W.J. Donnely, J.R. Carswell, R. E. Mc Intosh, P.S.
Chang, J. Wilkeson, F. Marks, P. G. Black,
(1999). Revised ocean backscatter models at C
and Ku band under high-wind conditions. J.
Geophys. Res, vol. 104,no. c5, pp. 11485-11498.
12 J. Yang, J. A. Zhang, X. Chen, Y. Ke, D. Esteban,
J. R. Carswell, S. Frasier, D. J. Mc Laughlin, P.
Chang, P. G. Black, F. Marks, (2004). Effect of
precipitation on ocean wind scatterometry. In
Proc. 2004 IEEE IGARSS.