You are on page 1of 5

Notice of Retraction

After careful and considered review of the content of this paper by a duly
constituted expert committee, this paper has been found to be in violation of
IEEE's Publication Principles.
We hereby retract the content of this paper. Reasonable effort should be made to
remove all past references to this paper.
The presenting author of this paper has the option to appeal this decision by
contacting TPII@ieee.org.

2010 Sixth International Conference on Natural Computation (ICNC 2010)

An Inverse Model for Coastal Acoustic Tomography


and its Application to Chinese Coastal Sea
ZHU Xiao-Hua
State Key Laboratory of Satellite Ocean Environment
Dynamics, Second Institute of Oceanography, State Oceanic
Administration,
Hangzhou, China
Abstract An inverse model for coastal acoustic tomography
(CAT) to reconstruct the tidal current distributions was built and
applied to the Zhitouyang Bay near the Zhoushan Island, China.
The range-averaged current (post-inversion result) obtained
along the transmission lines by the inverse model showed a
prominent semidiurnal oscillation and was in good agreement
with that (pre-inversion result) obtained from the travel time
difference data, producing a root-mean-square (rms) difference
of 0.036 m/s. Also both the range-averaged currents agreed with
the ADCP results, and their rms differences were 0.097 m/s for
the post-inversion result and 0.042 m/s for the pre-inversion
result. The horizontal distributions of tidal current obtained from
the inverse model were in rough agreement with that of the
shipboard acoustic Doppler current profiler velocity. The rms
differences for the eastward and northward currents were 11.70
cm/s and 16.50 cm/s, respectively. The above results indicate that
the inverse model worked well, and could be used for mapping
tidal current fields in Chinese coastal region.

Wu Qingsong, Zhang Chuangzheng


State Key Laboratory of Satellite Ocean Environment
Dynamics, Second Institute of Oceanography, State Oceanic
Administration,
Hangzhou, China
II.

THE INVERSE MODEL

A. Ray travel time

Keywords: Inverse model; coastal acoustic tomography; tidal


current; application; field experiment; Zhoushan Islend.

I.

INTRODUCTION

The ocean acoustic tomography (OAT) proposed by Munk


and Wunsch in 1979 is one of the most powerful acoustic
techniques for ocean observation [1-2]. The coastal acoustic
tomography (CAT) is an application of OAT to the coastal sea
[3]. The CAT is proposed to map the current fields in coastal
sea, such as ports, bays and semi-enclosed and inland seas
without disturbing shipping traffic, and fishing activities.
The CAT system using high frequency sound has been
developed by the acoustic tomography group of Hiroshima
University and applied to the current measurement of coastal
seas since 1995 [4].
A CAT experiment using five sets of self-contained type
system was successfully carried out in 1999 in a Channel of the
Seto Inland Sea of Japan. A strong tidal vortex pair was well
captured and imaged [5]. The similar CAT experiments were
also operated to map the tidal current and circulation in
Kanmon Strait, Tokyo Bay and Hiroshima Bay in Japan [6-8].
However, to date, there is no such experiment to be carried out
in the region outside of Japan, especially in China [9-11].
In this paper, we build an inverse model for CAT to
reconstruct the tidal current distributions, and applied this
model to the Zhitouyang Bay near the Zhoushan Island, where
the first CAT field experiment in China was successfully
carried out during July 12-13, 2009.

978-1-4244-5961-2/10/$26.00 2010 IEEE

Figure 1. Sketch of inverse model domain and tomography domain. The


acoustic tomography stations are indicated with solid circles. The solids line
connecting the acoustic stations are the sound transmission lines.

Figure1 shows the inverse domain together with the


tomography domain, where the CAT experiment will be
performed. The inverse domain is LxLy, which is twice the
tomography domain. The reciprocal sound travel time between
the station pair A and B can be expressed by (1) and (2) [2].
t1 =

L
C0 + U 0

(1 )

t2 =

L
C0 U 0

( 2)

The ray path averaged sound speed (C0) and along-sectional


average velocity (U0) can be obtained separately after simple
calculation of the coupled (1) and (2) as follows:
L1 1 L
+
2 t1 t2 t m

(3)

2
L 1 1 C0

t
2 t1 t 2 2 L

(4)

C0 =

U0 =

Where t m t1 t 2 t m = 12 (t1 + t 2 ) , and t is the reciprocal


travel time between A and B, which also can be written as:

3792

Un
ds
2
A C0

t = t1 t 2 = 2
x2

(5)

2
2
C0

x2

x2
udx
tan

vdx
+

x1
x1

v ( x, y ) =

Nl

(12)

k =0 l =0

(6)

B. Inverse Problem
The horizontal velocity u and v at (x, y) can be expressed
using the stream function (x, y) as follows:

N k Nl

u (k , l ) c Ck ,l + u (k , l ) d Dk ,l +

0 a0 + 0 b0 +

k =0 l =0 tan (v( k , l ) c C k ,l + v ( k , l ) d Dk ,l )

Nk

Where U n = u cos + v sin .

u ( x, y ) =

2
2
C0

= a0 + b0 + ( A(k , l ) Ck ,l + B (k , l ) Dk ,l )

2
U
1
1
dx = 2 2 (u + v tan )dx
t = 2 n2
cos
x1 C 0
x1 C 0

t =

(7 )

(8)

Eq.(12) then reduces to:


N k Nl
A(k , l ) Ck ,l
2

t = 2 a0 + b0 +

C0
k =0 l =0 + B ( k , l ) Dk ,l

(13)

Eq. (13) represents the relationship between the ray travel time
and current velocity field for only one ray. Eq. (13) can be
written for i rays as:
t1
t 1
2 2
=
#
i
t i

(x, y) can be expanded into a truncated Fourier series:

1
2
#

a0
b
0
A(0,0)1 B (0,0)1 A(1,0)1 B (1,0)1 " A( N k , , N l )1 B ( N k , , N l )1 C 00

A(0,0) 2 B (0,0) 2 A(1,0) 2 B (1,0) 2 " A( N k , , N l ) 2 B ( N k , , N l ) 2 D00

C10
#
#
#
#
#
# %
#
A(0,0) i B (0,0) i A(1,0) i B (1,0) i " A( N k , , N l ) i B ( N k , , N l ) i

C N k Nl
DN N
k l

Finally the above equation reduces to the simplified form

( x, y )

2kx 2ly

+
Ck ,l cos
Nk Nl
Ly

Lx
= a0 x + b0 y +

2kx 2ly
k =0 l =0

+ Dk ,l cos L + L
y
x

y = Ex + n

(9)

Substituting (9) into (7) and (8), we obtain

2kx 2ly
2l

cos
+
Ck ,l
L
Nk Nl
Ly
Ly

x
u = b0 +

2kx 2ly
k =0 l =0
2l

Dk , l L cos L + L
y
y
x

(10)

(14)

Where y is the travel time difference data, x the solution


vector and E the operator matrix. The noise vector n is
introduced to represent both the measurement error and any
other inaccuracies in the model.
C. Sulution
Eq. (14) is a well-known linear inverse system, where the
number of unknown variables N is not equal to the rank of
operation matrix E, the uniqueness of solution is not assured.
The least squares method seeks the solution vector x to
minimize the objective function J that is the summation of the
squares of the error vector n. The objective function for the
damped least squares is given as [5]:

J = nT n + 2 xT x = (y Ex ) (y Ex ) + 2 xT x
T

2kx 2ly
2k

+
cos
C k , l
L

Nk Nl
Lx
L
x
y

v = a0 +

2kx 2ly
k =0 l =0
2k

+ Dk ,l L cos L + L
x
y
x

(11)

Where (Lx, Ly), (2kx/Lx, 2kx/Ly) are the size of inverse


domain and wavenumber in (x, y). a0, b0, Ck,l, Dk,l are the
Fourier coefficients. Substituting (10) and (11) into (6), we
obtain:

(15)

Where is the constant value called the damping factor and


provides a constrain to the vector norm of solution x . Then
the estimated solution is determined by minimizing the above
objective function J. The solution reduces to:

x = ET E + I ET y
The expected error is also determined

n = y Ex = I E E T E + 2 I

(16)

ET y

(17)

The L-curve method [5] is used to specify the optimum value


of the damping factor . The optimum value of is given as a

3793

point where the curvature of the concave L-curve become a


maximum, making both the sizes of solution ( ) and
estimated error ( ) as small as possible in a balance (Fig.2).

( )
=x x=
( + )
Nr

i =1

Nr

= nT n =
i =1

( ) =

i 2 u iT y

2 2

(18)

( )
( + )
4 ui T y
2

2 2

(19)

' ' ' ' ' '

{( ')

+ ( ')

2 3/ 2

(20)

Figure 2. left panel: L-curve and right panel: K-curve. The X indicate the
maximum point of K, where =0.54.

III.

obtained from the reciprocal sound transmission between the


paired CAT systems. The inverse domain is 23 23 km,
which is twice the tomography domain. The y, x, E in the
equation system (14) become a column vector order 21 (the
total numbers of station pair), an unknown 20-element vector
and a 21 20 matrix, respectively.

Figure 3. Location maps of the experimental site. The positions of the CAT
stations C1-C7 are shown with solid circles. The thick solid lines connecting
the CAT stations indicate the sound transmission lines. Bathymetric contours
are in meters.

IV.

THE COASTAL ACOUSTIC TOMOGRAPHY EXPERIMENT

A. Site and method


The experiment using seven CAT systems was carried out
during July 12-13, 2009 in the region near the Zhoushan Island,
around the mouth of Hangzhou Bay, China (Figure 3). This
region is connected with four channels, the Loutou Channel,
Fodu Channel, Qingzimen Channel and Laoshumen Channel.
The area of the tomography domain is about 11.4 km11.5 km.
The water depths are deeper than 40 m in the southwestern part
of the bay and shallower than 20 m in the northeastern part of
the bay.

Seven CAT systems numbered from C1 to C7 were set up


around the bay using fishing boats anchored at the each station.
A pseudo random signal, called the 10th order M sequence, was
transmitted every 3 minutes from the broadband transmitter of
central frequency 5.0 kHz. The transmission signals were
received by the transmitter equipped at other six stations,
except the transmitted station.
The shipboard acoustic Doppler current profiler (ADCP,
RDI 600-kHz) surveys were also performed during the CAT
experiment inside the tomography domain to collect the data
for comparing with the CAT data.
B. Application of the inverse model
The inverse model is applied to reconstruct the tidal
current distributions on the basis of travel-time difference data,

RESULTS

A. Results

Figure 4. Horizontal current distributions at 18:00 of July 12 (left panel) and


01:00 of July 13 (right panel) obtained by the inverse model. The tidal level is
shown in the top-right corner in each panel.

The inverse results provide a continuous mapping of


horizontal tidal current distributions at 3-minute interval. The
typical horizontal tidal current structures are shown in Fig. 4.
The strong eastward tidal current with a maximum velocity of
2.34 m/s entered into the tomography site from the Loutou
Channel at the western part of the bay, and separated into two
branches which flowed into the Laoshumen Channel,
Qingzimeng Channel and Fodu Channel during ebb tide, while
the strong westward tidal current jointed by the currents from
Laoshumen Channel, Qingzimeng Channel and Fodu Channel,
went back to the Loutou Channel during the flood tide.

3794

B. Evaluation of the inverse results


The range-averaged current velocity (post-inversion result)
along the horizontal transmission lines of C2 to C5 obtained by
the inverse model is shown in Fig.5, together with that obtained
by substituting the travel time difference into (4) (pre-inversion
result). The post-inverse result showed a prominent semidiurnal
oscillation and was in good agreement with the pre-inversion
result, producing an rms difference of 0.056 m/s. Also the both
range-averaged current velocities agreed with the ADCP
results, and their rms differences were 0.11 m/s for the postinversion result and 0.10 m/s for the pre-inversion result.

current with a maximum velocity of 2.34 m/s entered into the


tomography site, and separated into two branches during ebb
tide, while the strong westward tidal current went back from
east to west.
The range-averaged current velocity was in good agreement
with that obtained from the travel time difference data. Also
both the range-averaged currents agreed with the ADCP results.
The horizontal distributions of tidal current obtained from the
inverse model were in rough agreement with that of the
shipboard acoustic Doppler current profiler velocity. These
results indicate that the inverse model worked well, and could
be used for mapping tidal current fields in Chinese coastal
region.
ACKNOWLEDGMENT

Figure 5. Time plots of the range-averaged current velocity calculated


along the C2-C5 horizontal transmission line, before (solid lines) and after
(dots) the inverse analysis. The sectional mean velocity obtained by the
shipboard ADCP is indicated with the horizontal line.

This study was supported by the scientific research fund of


Second Institute of Oceanography, SOA (JT0903), the
National High Technology Research and Development
Program (2006AA09Z102), the Zhejiang Provincial Natural
Science Foundation (2007R10G3050004) and National
Natural Science Foundation of China (40776021 and
40706017). We thank Mrs. Wen Y. Z., Zeng D.Y., Lin Y., Xu
L.J., Ding Y. B. and Zhou Z. J. for their supports in the
shipboard work during the experiment.
REFERENCES
[1]

Figure 6. Vector plots of the CAT (Solid) and ADCP (dotted) data.

The horizontal distributions of tidal current obtained from


the inverse analysis were in rough agreement with that of the
shipboard ADCP velocity smoothing though a 10-minute
running mean (Fig.6). The rms differences for the eastward
and northward currents were 0.084 cm/s and 0.184 cm/s,
respectively.
V.

SUMMARY

The inverse model for CAT to reconstruct the tidal current


distributions was built, and was applied to the Zhitouyang Bay
near the Zhoushan Island, where the first CAT field experiment
in China was successfully carried out during July 12-13, 2009.
The inverse results showed that the strong eastward tidal

The ocean tomography group. A demonstration of ocean acoustic


tomography, Nature, 1982, 299, 121-125.
[2] Munk W, Worceser P F, Wunsch C. Ocean Acoustic Tomography,
Cambridge: Cambridge University Press, 1995, 1-433.
[3] Elisseeff, P., H. Schmidt, M. Johnson Acoustic tomography of a coastal
front in Haro Strait, British Columbia, J. Acoust. Soc. Am., 1999, 169184.
[4] Zheng, H., N. Gohda, H. Noguchi, T. Ito, H. Yamaoka, T. Tamura, Y.
Takasugi and A, Kaneko, Reciprocal Sound Transmission Experiment
for Current Measurement in the Seto Inland Sea. Japan, J. Oceanogr.,
1997, 53, 117-127.
[5] Yamaoka, H., A. Kaneko, J.- H. Park, H. Zheng, N. Gohda, T. Takano,
X-H. Zhu and Y. Takasugi, Coastal acoustic tomography system and its
field application, IEEE J. Oceanic Eng., 2002, 27(2), 283-295.
[6] A. Kaneko, K. Yamaguichi, T.Yamamoto, N. Gohda, H. Zheng, S. Fadli,
J. Lin, H.Q. Nguyen, Coastal acoustic tomography experiment in the
Tokyo Bay, Acta Oceanologica Sinica, 2005, Vol.24, 86-94.
[7] Yamaguchi K, J. Lin, A. Kaneko, T. Yamamoto, N. Gohda, H.-Q.
Nguyen and H. Zheng, A continuous mapping of tidal current structures
in the Kanmon Strait, J. Oceanogr., 2005, 61, 283-294.
[8] Nguyen, H., A. Kaneko, J. Lin, K. Yamaguchi, N. Gohda and Y.
Takasugi, Acoustic Measurement of Multi Sub-Tidal Internal Modes
Generated in Hiroshima Bay, Japan, IEEE Journal of Oceanic
Engineering, 2009,34,102-112.
[9] Zhu X.H., H. Zheng, Q. Wu and C.Zhang, Development and Application
Experiment of an Acoustic Current Velocity Meter, Chinese Journal of
Scientific Instrument, 2009, 30, 20-23 (in Chinese with English abstract).
[10] Zhu, X.-H., H. Zheng G.H.Liao and C. J. Liang, Design and
implementation of coastal acoustic tomography, Journal of Harbin
Engineering University, 2010, 31(1), 64-68 (in Chinese with English
abstract).
[11] Zhu, X. H., and Zheng H., Design and Field Experiment of the Coastal
Acoustic Tomography, in Proceedings of the 2nd International
Conference on Image and Signal Processing, edited by Peihua Qiu,
IEEE Conference Publishing, 2009, Volume 9, 4566-4570.

3795

You might also like