Physical Communication 1 (2008) 163182

Contents lists available at ScienceDirect

Physical Communication
journal homepage: www.elsevier.com/locate/phycom

Overview of channel models for underwater wireless communication

networks
Mari Carmen Domingo
Broadband Wireless Networking Laboratory, School of Electrical and Computer Engineering, Georgia Institute of Technology, Atlanta, GA 30332, United States

article info a b s t r a c t
Keywords: Acoustic communication in Underwater Wireless Communication Networks (UWCNs) has
Acoustic communications
several challenges due to the presence of fading, multipath and refractive properties of the
Channel models
sound channel which necessitate the development of precise underwater channel models.
Deep water
Shallow water Some existing channel models are simplified and do not consider multipath or multipath
Underwater networking fading. In this paper, a detailed survey on ray-theory-based multipath Rayleigh underwater
channel models for underwater wireless communication is presented and the research
challenges for an efficient communication in this environment are outlined. These channel
models are valid for shallow or deep water. They are based on acoustic propagation physics
which captures different propagation paths of sound in the underwater and consider
all the effects of shadow zones, multipath fading, operating frequency, depth and water
temperature. The propagation characteristics are shown through mathematical analysis.
Transmission losses between transceivers are investigated through simulations. Further
simulations are carried out to study the bit error rate effects and the maximum internode
distances for different networks and depths considering a 16-QAM modulation scheme
with OFDM as the multicarrier transmission technique. The effect of weather season
and the variability of ocean environmental factors such as water temperature on the
communication performance are also shown. The mathematical analysis and simulations
highlight important considerations for the deployment and operation of UWCNs.
2008 Elsevier B.V. All rights reserved.

1. Introduction
Underwater Wireless Communications Networks
(UWCNs) are constituted by devices such as sensors and
Underwater Autonomous Vehicles (UAVs) that interact to-
gether to perform specific underwater applications such as
collaborative monitoring over a given volume of water [1].
Underwater acoustic communication in these networks
is an important challenge due to the presence of fading,
multipath and refractive properties of the sound channel.
Therefore, for an efficient deployment of these networks,
the underwater channel characterization is absolutely nec-
essary. Some existing simple channel models only distin-
guish between shallow (depths lower than 100 m) and
Tel.: +1 404 894 6616; fax: +1 404 894 7883.
E-mail address: mdomingo3@ece.gatech.edu.
1874-4907/$ see front matter 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.phycom.2008.09.001
deep waters (depths more than 100 m) and do not con-
sider multipath and fading effects. However, the underwa-
ter communication is severely affected by different propa-
gation phenomena as well as by the time-dependent vari-
ations of the channel.
In this paper we provide a comprehensive character-
ization of the underwater wireless channel and discuss
the research challenges for an efficient communication
in this environment. We present a complete survey on
ray-theory-based multipath Rayleigh underwater channel
models for shallow and deep waters. These models are
based on acoustic propagation physics. They capture not
only different propagation paths of sound in the under-
water but also other effects such as multipath fading, op-
erating frequency, depth, water temperature as well as
the existence and effects of the shadow zones. The ef-
fect of seasons and the variability of ocean environmen-
tal factors, e.g., water temperature, on the communication
164 M.C. Domingo / Physical Communication 1 (2008) 163182
performance are also investigated. The bit error rates and
maximum internode distances are studied by considering
a 16-QAM modulation scheme with OFDM as the multi-
carrier transmission technique. The mathematical analysis
and simulations carried out in this paper reveal that under-
water communication is severely affected by physical and
chemical properties of the water, especially by the depth of
transceivers and by water temperature. Our results high-
light important considerations for the deployment and op-
eration of UWCNs.
The paper is structured as follows: In Section 2 we dis-
cuss the related work about underwater channel mod-
elling. In Section 3 we introduce the sound propagation
in the underwater and focus on transmission loss as the
quantitative measure in the reduction of sound intensity
between two points. The characteristics of the underwater
channels are described in Section 4. In Section 5, the effects
of variations in water temperature are analyzed. Finally,
Section 6 discusses the research challenges for UWCNs and
concludes this paper.
2. Related work
Ray theory and the theory of normal mode are very
effective methods for channel modelling. Ray-theoretical
models calculate the transmission loss on the basis of ray
tracing, whereas the normal-mode solutions are derived
from an integral representation of the wave equation.
The normal-mode approaches tend to be limited to
acoustic frequencies below 500 Hz due to computational
limitations (and not due to the underlying physics).
Specifically the number of modes required to generate a
reliable prediction of the transmission loss increases in
proportion to the acoustic frequency [2]. Ray theory is very
useful for the description of sound propagation at high
frequencies in the underwater.
Some simple underwater channel models already exist
where the transmission loss is related to the distance
as in Thorps equation [3] or use a propagation loss
formula combined with a stochastic fading component
calculated from two Gaussian variables [4]. However,
these models do not describe the underwater channel
accurately because they do not consider several effects
such as multipath, fading, shadow zones, which exist in
underwater environments.
Considering the received signal as the sum of a large
number of multipath arrivals, each of which is modelled
as a complex Gaussian stochastic process, the resulting
model is the Rayleigh fading channel. Some researchers
have modelled the deep water channel as a Rayleigh
fading channel, although there is a limited number of
available measurements. The shallow water has also been
modelled as a Rayleigh fading channel [5] but there
is still no consensus among the research community
on a model which is applicable for shallow waters [6].
In [7], a simple stochastic channel model is proposed
and some experiments are provided for its validation in
the shallow water environment in Southampton, United
Kingdom. An underwater acoustic channel is introduced
in [8] where there can be several propagation paths named
eigenpaths over which a signal can propagate from a
source to a receiver. Each eigenpath signal contains a
dominant stable component and many smaller randomly
scattered components named sub-eigenpaths. The number
of eigenpaths reaching a receiver has a Poisson distribution
and the envelope of the eigenpath signal is described
by using a Rice fading model. However, no experimental
results are presented in [7].
Moreover, the model does not include acoustic propa-
gation physics, e.g., spreading and absorption.
Recently, a ray-based model is proposed for medium-
range very shallow waters in [9] where the signal strength
along each ray exhibits independent Rayleigh fading. The
model results have been validated through measurements
in Singapore. This model includes acoustic propagation
physics such as spreading and absorption and captures
transmission losses due to sea surface and sea bottom.
However, none of these channel models capture
transmission losses due to other different propagation
phenomena in the underwater such as surface duct,
convergence zones, deep sound channel and reliable
acoustic paths as well as the existence and effects of
shadow zones.
This paper addresses the observations that have been
made in the field of physics and presents a comprehensive
overview of the existing underwater channel models.
The introduced underwater channels are ray-theory-based
models for multipath propagation; the envelope of each
ray is modelled as a Rayleigh fading model. We focus
on important issues such as multipath and multipath
fading and analyze the effects of all existing different
propagation phenomena such as surface reflection, surface
duct, bottom bounce, convergence zone, deep sound
channel, reliable acoustic paths, and other effects such
as operating frequency, transmission range, depth, water
temperature and the shadow zones. The bit error rates
and maximum internode distances are also studied by
considering a 16-QAM modulation scheme with OFDM
as the multicarrier transmission technique. The research
challenges for UWCNs are also pointed out based on our
results.
3. Underwater signal propagation
3.1. Sound propagation in the sea
The study of sound propagation in the sea is funda-
mental to the understanding of underwater acoustic phe-
nomena and to the development of underwater acoustic
applications. Sound propagation in the sea is affected by
the physical and chemical properties of the sea water [10].
If a sound source radiates a signal with a source level SL
decibels at a unit distance (1 yd = 0.9144 m) on its axis,
the sound intensity reaching the receiver becomes SL TL.
This means that the sound intensity is reduced by the
transmission loss TL as the sound travels from the source to
the receiver. Transmission loss analysis is essential for the
study of underwater signal propagation. Therefore in the
following sections we carefully analyze TL and introduce
simple models for the characterization of shallow water
and deep water channels.
 M.C. Domingo / Physical Communication 1 (2008) 163182 165

3.2. Transmission loss A1 is the Boric Acid Component in sea water

8.68
The sonar parameter transmission loss (TL) is defined A1 = 10(0.78 pH5) (5)
as the accumulated decrease in acoustic intensity when c
an acoustic pressure wave propagates outwards from a where pH is the pH of water and c is the sound speed in m/s:
source. This magnitude can be estimated by adding the
c = 1412 + 3.21T + 1.19S + 0.0167z (6)
effects of geometrical spreading, absorption and scattering.
Spreading loss refers to the energy distributed over an with z being the depth (in m).
increasingly larger area due to the regular weakening of a P1 is the depth pressure for the boric acid term in sea
sound signal as it spreads outwards from the source [3]. water and is given by:
Absorption is a process that involves the conversion of
P1 = 1 (7)
acoustic energy into heat due to the internal friction at
a molecular scale within the fluid. At certain frequencies, A2 is the Magnesium Sulfate Component in sea water
absorption is increased due to ionic relaxation of certain
S
dissolved salts. Scattering occurs when sound waves are A2 = 21.44 (1 + 0.025T ) (8)
redirected when they interact with a body. The scattering c
of energy by bodies and bubbles in the underwater P2 is the depth pressure for the magnesium sulfate term in the
medium causes propagation loss at sea [11]. sea water:

P2 = 1 1.37 104 z + 6.2 109 z 2 (9)

3.2.1. Transmission loss in shallow water
Acoustic signals in shallow water propagate within a Furthermore, A3 is the Pure Water (Viscosities) Compo-
cylinder bounded by the surface and the sea floor; as a nent in the sea water:
result, cylindrical spreading appears [2].
4.937 104 2.59 105 T + 9.11 107 T 2

The transmission loss caused by cylindrical spreading
+ 9.11 107 T 2 1.50 108 T 3


and absorption can be expressed as follows [12]:


for T 20 C

A3 = (10)
TL = 10 log r + r 103 (1) 3.964 104 1.146 105 T
+ 1.45 10 T 6.5 10 T

7 2 10 3
where represents the absorption coefficient with the unit

for T > 20 C

dB/km and r is the range expressed in meters.
The ocean sound is attenuated by two main mecha- where P3 is the depth pressure for the pure water (viscosities)
nisms named viscous absorption (viscosity can be described term:
as the resistance of a fluid to flow) and ionic relaxation ef-
fects due to the presence of minute concentrations of boric P3 = 1 3.83 105 z + 4.9 1010 z 2 (11)
acid (B(OH)3 ) and magnesium sulfate salts (MgSO4 ) in sea Fig. 1 illustrates the effect of depth and frequency
water [13]. The effect of viscous absorption is significant at on the absorption coefficient in the sea water for
high frequency (above 100 kHz), whereas the ionic relax- T = 15 C, pH = 8 and S = 35 ppt.
ation effects due to boric acid affect at low frequency (up The absorption coefficient increases significantly with
to a few kHz), and due to magnesium sulfate affect at in- increasing operating frequency; at the zero depth for f =
termediate frequencies (up to a few 100 kHz). 1 kHz, the absorption coefficient is a few hundredths
Several different formulas for the absorption coefficient of dB/km, for f = 10 kHz is around 1 dB/km, for f =
have been derived in [1418]. 100 kHz, it is several tens dB/km and for f = 1 MHz
The total absorption coefficient in dB/km is derived several hundreds of dB/km. The depth has a large impact
in [16,17] for viscous and chemical absorption in the sea on the underwater data transmission. We observe that
water depending on acoustic frequency, pressure, acidity, the absorption coefficient diminishes when the depth
temperature and salinity (valid for 100 Hz < f < 1 MHz): increases particularly for frequencies above a few 100 kHz.
At frequencies above a few hundred Hz the absorption
A1 P1 f1 f 2 A2 P2 f2 f 2 coefficient can be calculated using Thorps expression
= + + A3 P3 f 2 (2)
f2 + f12 f 2 + f22 as [3]:
where f1 (in kHz) stands for the relaxation frequency for 0.1f 2 40f 2
boric acid and is given by: = + + 2.75 104 f 2 + 0.003 (12)
1+ f2 4100 + f 2
 0.5
S For lower frequencies, the following formula can be
f1 = 2.8 10[41245/(273+T )] (3)
used [19]:
35
with S being the salinity (in parts/1000) and T the f2
temperature (in C). = 0.002 + 0.11 + 0.011f 2 (13)
1+f2
Furthermore, f2 (in kHz) stands for the relaxation
frequency for magnesium sulfate and is obtained from: Fig. 2 shows the total transmission loss in shallow
water at a depth of 15 m. As the transmission range and
8.17 10[81990/(273+T )] the operating frequency increase, the absorption losses
f2 = (4) assume greater importance and affect mostly the TL values.
1 + 0.0018(S 35)
166 M.C. Domingo / Physical Communication 1 (2008) 163182

Fig. 1. Absorption coefficient as a function of operating frequency and depth.

Fig. 2. Total transmission loss in shallow water as a function of frequency and distance range.

For frequencies around 1 kHz or less there is no restriction
for distance ranges caused by excessive attenuation; at
10 kHz the distance range is limited to several tens of
kilometers; at 100 kHz the distance range is 1 km and for
higher frequencies in the MHz range the distance range
is drastically reduced to less than 100 m. Therefore we
can conclude that transmission to large distances requires
low frequencies; at low frequencies cylindrical spreading
is the most important factor affecting TL because it does
not depend on the operating frequency.
3.2.2. Transmission loss in deep water
Considering deep water as a homogeneous unbounded
medium, the transmission loss caused by spherical spread-
ing and absorption would be:
TL = 20 log r + r 103 (14)
where represents the absorption coefficient and has the
units dB/km and r is the range expressed in meters.
Fig. 3 shows the total transmission loss TL in deep water
at a sea depth of 10 000 m. As the transmission range and
the operating frequency increase, the absorption losses
assume greater importance. Thus, the transmission to large
distances requires low frequencies. Spherical spreading
is the most important factor affecting transmission loss
at low frequencies because it does not depend on the
operating frequency. The TL values in deep water (Fig. 3)
are larger than in shallow water (Fig. 2) for low frequencies
or low transmission ranges because the spreading term
dominates. But when the distance ranges as well as the
frequencies increase, the absorption losses are the major
cause of TL and they diminish with depth; as a result TL at
z = 10 000 m (deep water) will be lower than TL at a depth
of 15 m (shallow water).
 M.C. Domingo / Physical Communication 1 (2008) 163182 167

Fig. 3. Total transmission loss in deep water as a function of frequency and distance range.

Fig. 4. Total transmission loss in deep water as a function of distance for several depths (z ).

Fig. 4 presents the total TL in deep water for different
depths z and f = 150 kHz. As depth increases, the
absorption coefficient decreases because the magnesium
sulfate component of the equation is the dominant term
and it decreases with depth; this means that nodes located
at larger depths will experience less transmission loss in
their communications for the same ranges (for distance
ranges larger than 200 m).
4. Underwater channel characteristics
4.1. Propagation paths in the sea
The sound propagates in the sea through many different
paths. The particular paths travelled depend upon the
sound-speed structure in the water as well as the source
and receiver locations.
The sound speed is a function of temperature, pressure
(or depth) and salinity of sea water, and can be expressed
by [20]:
c = 1448.96 + 4.591T 0.05304T 2 + 0.0002374T 3
+ 1.340 (S 35) + 0.0163z + 1.675 107 z 2
0.01025T (S 35) 7.139 1013 Tz 3 (15)
This equation is valid for 0 T 30 , 30 S 40
and 0 z 8000.
The characteristic sound velocity profile for deep
oceans at mid-latitudes is shown in Fig. 5. The profile is
divided into layers, each one with different characteris-
tics [21]. Below the sea surface is the surface layer where
168 M.C. Domingo / Physical Communication 1 (2008) 163182

Fig. 5. Typical deep-sea velocity profile divided into layers.

the sound velocity varies with local changes (heating, cool-

ing and wind action). The surface layer may contain a
layer of isothermal water mixed by the wind where sound
is trapped or channelled. Below the sea surface lies the
seasonal thermocline layer, a layer where temperature
decreases with depth according to the seasons. It is charac-
terized by a negative thermal and velocity gradient (tem-
perature and velocity decreasing with depth). Below the Fig. 6. Shallow water propagation mode.
seasonal thermocline is the main thermocline, a region
of the water-column where the temperature decreases the principal mode of sound propagation (see Fig. 6). Re-
rapidly with increasing depth. This region is character- flections at the surface have little loss, whereas reflections
ized by a negative sound-speed gradient. Below the main at the bottom generally suffer a higher attenuation de-
thermocline and extending to the sea bottom is the deep pending on the sediment type, incidence angle and fre-
isothermal layer, where the temperature is nearly constant quency.
(near 4 C) and the sound speed increases with depth as a A semi-empirical expression for evaluating the trans-
function of increased sound pressure. This velocity profile mission loss has been developed in [22] based on the
changes with latitude, season, time of day and meteorolog- measurements in the frequency range of 0.110 kHz. A pa-
ical conditions. On the other hand, the velocity profile for rameter F , the skip distance (in km), is defined as the max-
shallow water tends to be irregular and unpredictable due imum range at which rays make contact with either the
to surface heating and cooling, salinity changes and water surface or the bottom and is expressed as:
currents [3].  1/2
1
Depending on the location of transmitter and receiver F = (d + z ) (16)
there will be several different ways of multipath propaga- 3
tion. Therefore, the different types of multipath propaga- d is the mixed-layer depth in meters.
tion for shallow water (direct path, surface reflection and The transmission loss, TL, in shallow water can be
bottom bounce) can be present in deep water if transmit- calculated as:
ter and receiver are located near the surface/bottom. But on
TL = 20 log r + r

the contrary, several specific propagation paths from deep
+ 60 kL for r < F


water (bending of the rays as in deep sound channel) are


r 
TL = 15 log r + r + aT

not present in shallow water. Multipath propagation de- 1 + 5 log F
pends mainly on depth but also on frequency and trans- F (17)
+ 60 kL for F  r 8F 
mission range.

r
TL = 10 log r + r + aT

In the following sections we describe the different
1 + 10 log F
F


+ 64.5 kL for r > 8F

multipath propagation alternatives for shallow and deep
water and derive the related transmission loss expressions.
where r is the transmission range in km, is the
absorption coefficient of sea water in dB/km, kL is the
4.1.1. Shallow water paths and losses near-field anomaly (mean contribution to the field of
In shallow water sound is propagated to a distance by the multiple bottom and surface reflections) in dB and
repeated reflections from the surface and bottom. This is depends on the sea state (wave height), frequency and
 M.C. Domingo / Physical Communication 1 (2008) 163182 169

= 20 log r + r 103 + TLRL_surface 


bottom composition, and aT is a shallow water attenuation
coefficient related to the additional loss due to coupling
of energy from the wind-roughened sea surface to the
bottom and is given in dB per cycle of bottom and surface
reflections.
The resulting transmission loss TL in shallow water
with a bottom composition of sand and a sea state 1
(calm, rippled) is shown in Fig. 7. The values of kL and
aT can be found in [22]. Transmission loss increases with
distance and frequency; when compared with the single-
path model, the multipath model results in a higher
transmission loss because more acoustic energy is lost
due to reflections. The proposed model for shallow water
(which is called Colossus) has been extended for negative
sound-speed profiles in [23]. In [24] another model is
proposed for predicting transmission loss in shallow water,
which gives the same results as the Colossus model when
the same inputs are used.
4.1.2. Deep water paths and losses
The six basic propagation paths between a source
and a receiver in the deep sea are illustrated in Fig. 8:
Surface reflection (A), surface duct (B), bottom bounce (C),
convergence zone (D), deep sound channel (E) and reliable
acoustic path (F).
In the following sections we study each of these
phenomena in detail and derive expressions for the
transmission loss at each different propagation path. We
also tackle the existence of shadow zones.
4.1.2.1. Surface reflection. We use the term surface re-
flection to describe the reflection of sound from the sea
surface. The roughness or smoothness of the sea affects
reflection. The Rayleigh parameter R gives a measure of the
acoustic roughness of the sea surface:
R = kH sin
2
where k = is the acoustic wavenumber, H is the crest-
to-trough rms wave height and is the grazing angle.
When R  1, the surface is considered to be acoustically
smooth and it is primarily a reflector; when R  1, the
surface is considered to be acoustically rough and it is
primarily a scatterer and energy is lost in all directions.
When the sea is rough, the loss on reflection can be
found using the BeckmannSpizzichino surface reflection
model, which allows the calculation of the acoustic
intensity reduction [25]:
1 + (f /f1 )2
 
TLRL_surface = 10 log
1 + (f /f2 )2
(90 w)
  2
1+
60 30
10f2 and f2 = 378w , where w is the 2
where f1 =
windspeed in knots (1 knot = 0.514 m/s). TLRL_surface is
a function of the angle of incidence to the horizontal
measured in degrees.
The final expression of the total transmission loss for
the rays that suffer surface reflection (two-path channel
model) for a near-surface source and receiver can be
computed as follows:
TLTOT (20)
where represents the absorption coefficient (in dB/km)
and r is the range expressed in meters.
We notice how TLRL_surface increases in Fig. 9 with higher
frequencies; besides, the loss curve slides leftwards with
increasing windspeed and downward with increasing an-
gle of incidence to surface indicating increasing attenua-
tion at lower frequencies at the sea.
4.1.2.2. Surface duct. If the surface layer has a positive
gradient (the velocity of sound increases with depth
because of the pressure effect on sound velocity) (see
Fig. 10) and that layer is deep enough, the sound may
be bent back towards the surface then reflected back
into the layer. In many cloudy and windy ocean areas of
the world, the water beneath the surface is isothermal
because it is heated daily by the sun and mixed by the
wind. Sound emanating from a source in an isothermal
layer is prevented from spreading in all directions and is
channelled or confined between the boundaries of the so-
called mixed sound channel or surface duct. In a mixed
layer the sound propagates to long ranges by successive
reflections from the sea surface along ray paths that are
long arcs of circles (see Fig. 10).
The duct expressions for the mixed sound channel
are [3]:
TL = 20 log r + ( + L ) r 103
r < 350H 1/2 (short ranges)
(21)
TL = 10 log r0 + 10 log r + ( + L ) r 103
r > 350H 1/2 (long ranges)
with 10 log r0 = 20.9 + 5 log H, where H stands for the
26.6f (1.4)S
mixed-layer depth in meters and L = [(1452+3.5t )H ]1/2
where S stands for the sea state number and t stands for
the temperature.
(18)
In Fig. 11 we observe that TL (Eq. (21)) is practically
not affected by frequency and basically increases with
transmission range. For distance ranges around 2800 m
and frequencies of 3.5 kHz TL becomes 71 dB, which means
that medium-range systems can communicate properly
with moderate transmission losses.
4.1.2.3. Bottom bounce. We use the term bottom bounce
to describe the reflection of sound from the sea floor.
The reflection loss of sound incident (see Fig. 12) at a
grazing angle 1 to a plain boundary between two fluids
of densities 1 and 2 and of sound velocities c1 and c2
using the Rayleigh formula is given by the intensity of the
reflected wave Ir related to the intensity of the incident
wave Ii [19]:
 
Ir
(19) TLRL_bottom = 10 log
Ii
2 !
m sin 1 n sin 2

= 10 log
m sin 1 + n sin 2
"
1/2 #2
m sin 1 n2 cos2 1
= 10 log 1/2
(22)
m sin 1 + n2 cos2 1


c1
where m = 2 and n = c2
.
1
170 M.C. Domingo / Physical Communication 1 (2008) 163182

Fig. 7. Transmission loss in shallow water as a function of distance for several frequencies with r < F .

Fig. 8. Propagation paths between a source and a receiver in deep water.

Fig. 13 illustrates TL (Eq. (22)) caused by the bottom
for different types of sediments: sandy bottom (m =
1.95, n = 0.86), clayed mud (m = 1.56, n = 1.008),
muddy very fine sand (m = 1.73, n = 0.98) and
sandmudclay (m = 1.41, n = 1.006).
We observe that the curves of TL as function curves of
TL as function of grazing angle 1 are different depending
on the relationships between n and m:
n < 1, n < m: it is the most common condition for
natural bottoms. The total reflection (zero loss) occurs
at grazing angles less than the critical angle 0 . Applying
cos (0 ) = n, 0 = 30.68 for sandy bottom and 0 =
11.48 for muddy very fine sand.
1 < n < m: in many soft mud bottoms, the sound
velocity is less than that in the water above and an angle
  12
m2 n2
of intromission B may exist. From cos B = m2 1
we can get B = 6.08 for clayed mud and B = 6.34
for sandmudclay.
The attenuation coefficient S due to the presence of
sediments at the sea floor is:
S = kf n (23)
where n is an empirical constant (n is typically 1 for many
measurements on sands, silts and clays). k [with the unit
dB/(m* kHz)] depends upon porosity and is approximately
equal to 0.5 over the range of porosity 35%60% [26].
The final expression for total transmission loss for the
rays that suffer bottom bounce (two-path channel model)
for a near-bottom source and receiver can be computed as:
TLTOT = 20 log r + r 103 + S r + |TLRL_bottom | (24)
 M.C. Domingo / Physical Communication 1 (2008) 163182 171

Fig. 9. Transmission loss due to surface reflection as a function of frequency for different grazing angles (ang) and windspeeds (w).

Fig. 10. Surface sound channel, sound-speed profile and ray diagram.

4.1.2.4. Convergence zone. The sound propagation in the
deep ocean from a shallow source causes the formation
of convergence zones. In very deep water, the sound rays
emitted by the source are refracted or bent downward
as a result of decreasing temperature until the increase
in pressure bends the sound rays upward in the deep
isothermal layer [21]. When the refracted sound rays
approach the surface, they often intersect, leading to areas
of higher intensity levels at, and immediately below the
surface [12]. This enhanced region is the convergence
zone and the range at which it occurs depends on the
characteristics of the sound-speed profile. The negative
temperature gradient bends the rays again downward and
if the signal is strong enough, it may be detectable at
distances of two or three convergence zones or more.
In the North Atlantic Ocean convergence zones appear
at intervals of approximately 65 km, with zone widths
of about 4 km. Charts of surface temperature and water
depth can be used to predict the formation of convergence
zones [2].
The convergence zone is formed when the upper ray
of the sound beam becomes horizontal, i.e., at the depth
where the sound speed is equal to the higher of the speeds
at either the surface or of the bottom of the surface layer.
The difference in depth between the ocean bottom and the
depth at which the upper ray becomes horizontal is called
the depth excess (see Fig. 14). The depth excess must be of
the order of a few hundred meters because otherwise the
rays would be cut off by the bottom and they would not
converge.
There exist some requirements for the formation of
convergence zones. The water depth must be greater
than the critical depth (the depth excess is the vertical
separation between the critical depth and the bottom, see
Fig. 14). In insufficiently deep water, rays are reflected by
the bottom and do not converge.
The total expression of transmission loss for the rays
within a convergence zone would be [12]:
TLTOT = 20 log r + r 103 Gconv ergence_zone (25)
where Gconv ergence_zone refers to the gain in the convergence
zone which is a complicated function of the position in the
convergence zone and can be determined numerically [27].
172 M.C. Domingo / Physical Communication 1 (2008) 163182
Fig. 11. Transmission loss as a function of frequency and distance in a surface duct.
Fig. 12. Reflected and transmitted rays at a discontinuity between two
mediums.
Common values for convergence gains can be found in the
range between 5 and 20 dB [3]. TLTOT is significantly lower
than spherical spreading within the zones, but significantly
higher than spherical spreading between zones.
4.1.2.5. Deep sound channel. The deep sound channel
exists due to the sound velocity profile in the deep ocean;
the sound velocity profile has a minimum that varies
about 1000 m at mid-latitudes to near the surface in
polar regions. Above and below the minimum the velocity
gradient continually bends the sound rays towards the
depth of minimum speed. Consequently, a portion of the
power radiated by the source is trapped in the deep
sound channel or SOFAR (Sound Fixing and Ranging)
channel and propagates over internally refracted paths.
The propagation of sound in this channel at very long
ranges is excellent and a significant portion of the acoustic
energy remains in the channel with no acoustic losses
by reflection from the sea surface and sea bottom and
low transmission loss, especially if the source and receiver
are located near the depth of minimum velocity (sound
channel axis). If the source is located at low depths
the deep sound channel contains convergence zones and
shadow zones; some rays can be reflected at the sea surface
and bottom or refracted and reflected from either the sea
bottom or the sea surface.
The transmission loss model for the deep sound channel
and long acoustic signals is given by [3]:
TL = 10 log r0 + 10 log r + r 103 (26)
where r0 is the transition range from cylindrical to
spherical spreading and is calculated by (see Fig. 15) [12]:
 r   D 1/2
s s
r0 = (27)
8 zs
where zs is the depth of the source, DS is the axis depth and
rS is the skip distance.
rs = 2 (Ds + Dr ) (2cmax /c )1/2 (28)
where (Ds + Dr ) is the vertical extent of the channel, cmax
is the largest speed of sound found in the channel and c
is the difference between cmax and the speed of sound on
the axis of the channel cmin .
4.1.2.6. Reliable acoustic path. When the source is located
very deep in the ocean and the receiver is located
in a shallow position, the propagation to moderate
ranges takes place via the so-called reliable acoustic path
(RAP). The sound energy is refracted downward due to
temperature decrease until the deep isothermal layer is
reached; then it is refracted upward to or near the surface
due to pressure when the sound velocity below the axis
of the deep sound channel exceeds the maximum near-
surface sound velocity. The path is reliable because it is
sensitive neither to near-surface effects nor to bottom
reflection. If the rays leave the source with an angle of
elevation greater than 0 in Fig. 15, a RAP is formed from
a deep source to the surface. The rays steep enough can
be considered straight lines and TL can be approximated as
spherical spreading over the slant range r with losses due
to absorption [12]:
TL = 20 log r + r 103 (29)
 M.C. Domingo / Physical Communication 1 (2008) 163182 173

Fig. 13. Transmission loss as a function of the grazing angle.

Fig. 14. Depth excess for the convergence sound path.

Fig. 15. Sound transmission from a source in the deep sound channel.

4.1.2.7. Shadow zones. Shadow zones are areas where
there is a little propagation signal energy. Thus, it could
be difficult (sometimes impossible) to communicate with
a receiver located in a shadow zone. When the sound
speed has a negative gradient just beneath the surface (see
Fig. 16), a shadow zone is formed because the acoustic
rays are refracted downward. Refraction produces shadow
zones that sound waves do not penetrate because of their
curvature. The sea bottom can produce a shadow zone as
well when the rays are refracted upward. Shadow zones
174 M.C. Domingo / Physical Communication 1 (2008) 163182
Fig. 16. Shadow zone formation when sound velocity monotonically decreases with depth.
can also appear beneath the mixed layer for a source
located near the ocean surface because the acoustic energy
is trapped in the surface duct. The shadow zone is usually
bounded by the lower boundary of the surface duct and
the limiting ray. Shadow zones can also appear between
convergence zones. If the source is located at the same
depth of the underwater sound channel axis, the shadow
zone will disappear [19].
Theoretically, no sound should enter the shadow
zone, but some sound may enter due to scattering and
diffraction. However, some experiments show that for high
frequencies signal levels are typically at least 40 dB less
than those at the edges of the shadow zone [12]. For low
frequencies the signal loss in the shadow zone is less severe
and for sufficiently low frequencies the shadow zone may
cease due to the strong diffraction.
The transmission loss model for shadow zone would be:
TLTOT = TLprev ious + |TLshadow_zone |
where TLprev ious represents the transmission loss due to
the propagation phenomena studied in Section IV and
TLshadow_zone refers to the transmission loss inside the
shadow zone.
4.2. Multipath fading and bit error rate
The multipath channel model described in Section 4
for shallow water and the two-path/multipath models de-
scribed in Section 4 for deep water simulate the main
propagation characteristics of sound waves underwater. So
far we have considered a time-invariant acoustic channel
often called macro-multipath because it models determin-
istic propagation paths [28]. However, the multipath is
usually time-varying due to multiple causes. The rough-
ness of the sea surface and bottom causes not only
reflection, but also refraction. In the shallow water chan-
nel surface scattering due to waves contributes mostly
to the overall time variability [28]. In the deep water
channel internal waves are the most important contribu-
tors to the time variation of the signal propagating along
the deterministic paths in addition to surface scattering.
These random signal fluctuations, which account for the
time variability of the channel response, are named micro-
multipaths [29]. As a result multipath fading should also
be considered in addition to the basic two-path/multipath
channel models.
We consider that each path in the underwater channel
is Rayleigh distributed such that the envelope of the signal
from each path in the two-path model is modelled as an
independent Rayleigh distributed random variable, i , i
{1, 2}. Consequently, for the one-path model, the received
energy per bit per noise power spectral density is given
by = 2 Nb , which has a distribution as f ( ) =
E
  0
exp , where 0 = E 2 Nb and Eb /N0 can be found
1 E
 
0 0 0
from the signal-to-noise ratio (SNR) of the channel.
Similarly, we model the randomness in the two-path
channel using the Rayleigh fading channel model, where
the envelope of the signal (absorption coefficient) is
modelled as a Rayleigh distributed random variable .
Since the two-path Rayleigh model does not have a closed-
form expression for SNR distribution, we have computed
one pair of Rayleigh distributed variables and have added
(30) them. We have repeated the same for 5000 trials and
have done the average [30]. So we can obtain and this
value is used to find the expected bit error rate (BER) for
the multipath shallow water model and the deep water
model with a deep sound channel (SOFAR). The BER will
depend on the SNR. The SNR of an emitted underwater
signal at the receiver can be expressed by the passive sonar
equation [3]:
SNR = SL TL NL + DI DT (31)
where DT has been defined as the detection threshold, SL is
the source level, TL is the transmission loss, NL is the noise
level and DI is the directivity index.
Four basic sources model the ambient noise NL in the
ocean: turbulence, shipping, waves and thermal noise. The
overall power spectral density on the ambient noise is
given by [31]:
N (f ) = Nt (f ) + Ns (f ) + Nw (f ) + Nth (f ) (32)
where the ambient noise due to turbulence, shipping,
waves and thermal noise is described by the following
equations.
10 log Nt (f ) = 17 30 log f (33)
10 log Ns (f ) = 40 + 20 (s 0.5) + 26 log f
60 log (f + 0.03) (34)
 M.C. Domingo / Physical Communication 1 (2008) 163182 175

Fig. 17. The maximum internode communication distance of one-path channel using different modulation schemes for shallow water.

Fig. 18. The maximum internode communication distance of one-path channel using different modulation schemes for deep water.

where s is the shipping activity factor whose value ranges
between 0 and 1 for low and high activity, respectively.
10 log Nw (f ) = 50 + 7.5w 1/2 + 20 log f
40 log (f + 0.4)
where w is the windspeed in m/s and
10 log Nth (f ) = 15 + 20 log f
The dominant noises according to the operation fre-
quencies are turbulence noise for f < 10 Hz, distant
shipping for f = 10100 Hz, surface motion caused by
wind-driven waves for f = 100 Hz100 kHz (operating
region used by the majority of the acoustic systems) and
thermal noise for f > 100 kHz. The noise level generally
decreases with increasing frequency.
The selected modulation method does also affect the
BER. We have used three different types of basic modu-
(35) lation methods (FSK, PSK and QAM) to illustrate the ef-
fects of the underwater channel on BER. The relationship
between the maximum internode distance of the single-
(36) path model and the water depth is plotted in Fig. 17 for
shallow water and in Fig. 18 for deep water. The max-
imum internode distance is found subject to a BER tar-
get of 103 for different modulation methods. Although
the FSK modulation method provides the largest range,
QAM achieves larger bandwidth efficiency and provides
the second longest range. Orthogonal frequency division
176 M.C. Domingo / Physical Communication 1 (2008) 163182

Fig. 19. BER as a function of the internode distance with different transmission power for shallow water.

multiplexing (OFDM) is suggested for underwater envi- power increases, the BER decreases. The distance can
ronments due to the low complexity of the receivers re- be extended to 1 km with the limitation that BER is
quired to deal with highly dispersive channels [32]. OFDM below 103 . In Fig. 20 the relationship between BER and
is a multicarrier transmission technique, which divides the frequency is shown for the shallow water case. Again BER
available spectrum into many carriers; the required ampli- decreases with increasing transmission power. There is an
tude and phase of the carrier are then calculated based on optimal operating frequency (4 kHz in Fig. 20) for which
the modulation scheme (typically differential BPSK, QPSK the minimum BER is obtained and the internode distance
or QAM). The spectrum is efficiently used by spacing the is increased.
channels closer together and making them orthogonal to In Fig. 21 the relationship between BER and distance
avoid interference between them. The scalable OFDM with is shown for the deep water case. The BER decreases
16-QAM modulation has been successfully tested in the with increasing transmission power. The distance can
settings with bandwidths 12 kHz, 25 kHz and 50 kHz, lead- be extended to 800 m with the limitation that BER is
ing to data rates about 12 kbps, 25 kbps and 50 kbps respec- below 103 . In Fig. 22, the relationship between BER
tively [33]. Consequently, we consider OFDM transmission and frequency is shown for the deep water case. Again
with QAM modulation in our analysis. BER decreases with increasing transmission power. As in
The BER for the modulation scheme 16-QAM is given shallow water, there is an optimal operating frequency
by: (4 kHz in Fig. 22) for which the minimum BER is obtained
s ! and the internode distance is increased; the BER values are
3 4 Eb BN much lower than in the shallow water case, which means
16QAM
pb = erfc , Eb /N0 = SNR (37) SOFAR channels improve underwater communication.
8 10 N0 R
The effective communication range is another impor-
where SNR = 10SNRdB (d,f )/10 , BN is the noise bandwidth and tant parameter for the design of effective networking
R is the data rate. protocols in UWCNs. The maximum internode distance
The signal level SL is related to the intensity It and hence for the multipath model in shallow water is shown in
to the transmit power of the transceiver for shallow water Fig. 23 according to a BER of 103 and different frequencies.
as follows: As frequency increases, the maximum internode distance
  decreases. As depth increases, the communication range
Pt It can be improved for f = 1 kHz and f = 4 kHz. This
It = , SL = 10 log (38)
2 1mz 0.67 1018 suggests that for low operating frequencies it is better
to locate nodes deeper in shallow water but for higher
where Pt is the transmission power in watts and z is the
frequencies the depth will not affect the maximum intern-
depth in meters.
ode distance. Fig. 24 shows the maximum internode dis-
In deep water Eq. (38) becomes:
tance for the two-path model with SOFAR in deep water. As
frequency increases, the maximum internode distance de-
 
Pt It
It = , SL = 10 log (39) creases. As depth increases, the communication range can
4 1m 0.67 1018
be improved specially for high frequencies.
In Fig. 19 the relationship between BER and distance This suggests that if the operating frequency is high
is shown for the shallow water case. As the transmission it is better to locate nodes deeper in deep water but for
 M.C. Domingo / Physical Communication 1 (2008) 163182 177

Fig. 20. BER as a function of the frequency with different transmission power for shallow water.

Fig. 21. BER as a function of the internode distance with different transmission power for deep water.

lower frequencies the depth will not affect the maximum
internode distance. The maximum internode distances are
larger in deep water in comparison with shallow water.
5. Effects of temperature variations in underwater
The results obtained from our study of channel models
clearly show that ocean environmental data helps us
to do acoustic predictions and their variations affect
communication performance. In this section we will study
how the water temperature (T ) variability affects our
acoustic prediction.
From Section 4 we know that T changes with depth;
we use the following formula to represent the ocean
temperature profile (see Fig. 25):
T (z ) = TB + T0 ez /H (40)
where TB is the temperature in C of the deep isothermal
layer, T0 is a shifted extrapolated surface temperature in
C, z is the depth in meters and H is a thickness scale of the
main thermocline in meters. A mixed layer of depth zM is
introduced so that T (0) = T (zM ).
The water temperature data shown in this paper
were obtained from the Bermuda Institute of Ocean Sci-
ences [34] at the Sargasso Sea (location 31 400 N 64 100 W,
sea depth of 4300 m) for the months of February, May and
August of 2006. At this location H = 720 m [35]. T0 is
the surface temperature and will vary with each season
178 M.C. Domingo / Physical Communication 1 (2008) 163182

Fig. 22. BER as a function of the operating frequency with different transmission power for deep water.

Fig. 23. Maximum internode distance as a function of depth for shallow water.

(it also depends from the weather, time of the day, envi-
ronmental factors, etc). The temperature profile for differ-
ent seasons of year 2006 is shown in Fig. 26.
The differences in water temperature are larger for
depths lower than 200 m.
The sound-speed profile as well as the absorption
coefficient will be affected by the temperature variability,
which in effect changes the transmission loss and BER.
The sound-speed profile will be useful to determine the
SOFAR axis depth and SOFAR axis speed of the underwater
channel. Substituting Eq. (40) in the sound speed equation
(15) and using the parameter values S = 0.2 ppt, TB =
2.31 C [36], we obtain the sound-speed profiles for the
three different seasons (see Fig. 27). We notice that for
higher values of T0 the SOFAR axis depths and speeds are
higher (in August SOFAR axis depth = zAXIS = 1696 m/s
and SOFAR axis speed = 1498.9 m/s). The differences
in speed resulting from changes in temperature are quite
large, especially near the sea surface.
From the sound-speed profile we can determine the
different propagation phenomena that will affect an
underwater communication depending on the source
and receiver locations and so the transmission loss
and underwater communication performance can be
estimated accordingly; in this way a complete profile of the
underwater channel model is presented.
Now we can estimate the maximum internode com-
munication distance for a BER = 103 using the study
 M.C. Domingo / Physical Communication 1 (2008) 163182
Fig. 24. Maximum internode distance as a function of depth for deep water.
Fig. 25. Temperature profile.
of the channel models introduced in Section 4. Based on
the location of source zS (the receiver will be at the same
depth) we distinguish between several propagation phe-
nomena: Surface reflection for 0 < zS < zM , convergence
zones for zM < zS < zAXIS 500, deep sound channel
for zAXIS 500 < zS < zAXIS + 500, spherical spreading
for zAXIS + 500 < zS < zSEA 300 and bottom bounce
for zSEA 300 < zS < zSEA , where H = 720 m and the
mixed-layer depth zM = 160 m for February, zM = 40 m
for May and zM = 20 m for August (we estimate zM using
the criteria of a temperature change from the ocean sur-
face of 0.5 C), zAXIS is the sonar axis depth and zSEA is the sea
depth. These will be the most important phenomena expe-
rienced by most sound rays at each depth but some other
rays can experience other phenomena which will coexist
179
with the explained ones. Fig. 28 shows the maximum in-
ternode distance when T changes according to seasons and
depths and a case where T is constant (9 C) for all depths.
The results show that the maximum internode distance
will vary substantially depending on the different propaga-
tion paths in the sea. The use of a single measurement of T
does not accurately predict the maximum communication
range, in fact the prediction is most of the time too opti-
mistic; consequently we conclude that T variability can-
not be ignored. The maximum internode distance is very
low for the different seasons if the transceivers are placed
in the mixed layer due to surface reflection. Afterwards,
the communication range increases abruptly (the highest
peak is in August and the lowest in May) (convergence
zones), decreases softly and then abruptly, increases again
progressively (deep sound channel, spherical spreading)
and then decreases abruptly (bottom bounce) with no sig-
nificant variations between seasons. The communication
range is improved mainly due to the existence of conver-
gence zones. The communication range for the same depth
can change drastically between seasons especially if at a
certain depth a surface reflection or a convergence zone is
observed; for example at a depth of 20 m the communica-
tion range is 115 400 m for August (convergence zone) and
45 440 m for February (surface reflection). Consequently,
environmental adaptive protocols, which can adjust the
operating parameters according to the seasons, will be nec-
essary at these depths for a robust operation in UWCNs.
These protocols need to be aware of the depth which is the
best to locate transceivers according to the seasons.
6. Conclusions
In this paper we have described the propagation
characteristics of sound in the underwater. Based on this
study we have presented a detailed survey on ray-theory-
based multipath Rayleigh underwater channel models for
shallow water and deep water.
180 M.C. Domingo / Physical Communication 1 (2008) 163182

Fig. 26. Temperature profile for different seasons of year 2006.

Fig. 27. Sound-speed profile for different seasons of year 2006.

We have extracted some conclusions from our overview
of channel model research that will be very helpful to face
the challenges in the design of UWCNs.
Topology design: Before deciding the location of the
transceivers in the underwater, a complete characteriza-
tion of the underwater physical channel must be carried
out, because our theoretical analysis and simulations show
that the environmental factors (salinity, ph, temperature,
etc.) will affect severely the communication performance.
We have demonstrated that the sound-speed profile es-
timation at different locations is essential for a complete
analysis of sound propagation in the ocean because it helps
in determining the propagation phenomena and conse-
quently the multipath types that sound can encounter,
which will affect tremendously the communication qual-
ity. From this study, the optimal location for transceivers
will be deduced. Our simulations show that the conver-
gence zones and deep sound (SOFAR) channels will favor
communication improvement in the underwater, whereas
shadow zones, surface reflections and bottom bounces
should be avoided. The near-surface shadow zones can
be avoided for locating sources and receivers deep in the
ocean. The analysis carried out also illustrates that the op-
timal position for the transceivers will be time-varying be-
cause the chemical and physical properties of water change
with time and there is multipath fading.
The communication performance is also affected by
the depth. Our survey on channel models reveals that
 M.C. Domingo / Physical Communication 1 (2008) 163182
Fig. 28. Maximum internode distance as a function of depth for different seasons.
for low frequencies or low distance ranges, the TL values
are lower in the shallow water; but when the distance
ranges as well as the frequencies increase, the TL values
will be lower in deep water because they diminish with
depth; therefore both parameters (distance and frequency)
must also be carefully studied before deciding where to
locate the transceivers. The maximum internode distance
for a particular BER is decreased with increasing frequency
and is larger in deep water than in shallow water.
Besides, it varies substantially depending on the different
propagation paths in the sea.
Operating frequency: Our study on channel models shows
that attenuation increases with frequency; our theoretical
analysis and simulations reveal that there is a trade-off
between the available bandwidth and the transmission
range: very long-range systems operating at 1000 km
have a bandwidth of less than 1 kHz whereas very short-
range systems operating around 100 m have a bandwidth
larger than 100 kHz. The desired communication range
will depend on the distance between transceivers and on
the requirements of specific applications; therefore, a good
solution to attain the desired communication range any
time would be to use cognitive radio techniques to adapt
the operating frequency for an optimal communication.
The ambient noise will also depend on the selected
frequency. There will be an optimal operating frequency
for which the minimum BER is obtained and the maximum
distance between transceivers is increased.
Environmental-aware protocol design: The variability of
environmental factors such as water temperature with
seasons will affect the communication performance.
We strongly recommend the design of environmental-
aware protocols that can adjust their parameters locally
according to time and space variations.
Cross-layer design: The oceanic environmental changes
affect all communication-related aspects; therefore we
181
believe that a reliable cross-layer solution should be
developed specifically tailored for this dynamic medium
that integrates different communication functionalities to
achieve high performance channel access, routing, event
transport reliability, and data flow control.
The results obtained are important for the founda-
tions of UWCNs because they establish the principles for
achieving good underwater communication performance
between transceivers, which is essential for the design
of communication systems; they will also help in under-
standing the basic needs concerning the development of
environment-aware protocols and will be useful in the de-
sign of specific underwater applications for the deploy-
ment of UWCNs.
Acknowledgment
The author would like to thank Dr. Ian F. Akyildiz for his
invaluable advice and suggestions.
References
[1] I.F. Akyildiz, D. Pompili, T. Melodia, Underwater acoustic sensor
networks: Research challenges, Ad Hoc Networks Journal, (Elsevier)
3 (3) (2005).
[2] P.C. Etter, Underwater Acoustic Modeling and Simulation, 3rd ed.,
Spon Press, New York, 2003.
[3] R.J. Urick, Principles of Underwater Sound, third ed., McGraw-Hill,
1983.
[4] E. Sozer, M. Stojanovic, J. Proakis, Design and simulation of an
underwater acoustic local area network, in: Proc. Opnetwork99,
Washington, DC, August 1999.
[5] J.A. Catipovic, A.B. Baggeroer, K. Von Der Heydt, D. Koelsch, Design
and performance analysis of a digital telemetry system for short
range underwater channel, IEEE Journal of Oceanic Engineering OE-9
(4) (1984) 242252.
[6] M. Chitre, S. Shahabodeen, M. Stojanovic, Underwater acoustic
communications and networking: Recent advances and future
challenges, Marine Technology Society Journal 42 (1) (2008)
103116.
182 M.C. Domingo / Physical Communication 1 (2008) 163182
[7] R. Galvin, R.F.W Coates, Analysis of the performance of an underwa-
ter acoustic communication system and comparison with stochastic
model, IEEE Oceans94, Brest, France, 1994, pp. III/478III/482.
[8] X. Geng, A. Zielinski, An eigenpath underwater acoustic communi-
cation channel model, in: Proc. OCEANS, MTS/IEEE, Challenges Our
Changing Global Environment Conf., 2, Oct. 1995, pp. 11891196.
[9] M. Chitre, A high-frequency warm shallow water acoustic communi-
cations channel model and measurements, Journal of the Acoustical
Society America 122 (5) (2007) 25802586.
[10] J.W. Caruthers, Fundamentals of Marine Acoustics, in: Elseviers
Oceanographic Series, vol. 18, Elsevier Scientific Publishing Co,
Amsterdam, The Netherlands, 1977.
[11] Applied Physics Laboratory (APL), APL-UW High-Frequency Ocean
Environmental Acoustic Models Handbook, TR 9407, APL, University
of Washington, October 1994.
[12] L.E. Kinsler, A.R. Frey, A.B. Coppens, J.V. Sanders, Fundamentals of
Acoustics, 4th ed., John Wiley & Sons cop, New York, 2000.
[13] H. Medwin, C.S. Clay, Fundamentals of Acoustical Oceanography,
Academic Press, San Diego, 1997.
[14] M. Schulkin, H.W. Marsh, Sound absorption in sea water, Journal of
the Acoustical Society of America 34 (1962) 864865.
[15] F.H. Fisher, V.P. Simmons, Sound absorption in sea water, Journal of
the Acoustical Society of America 62 (1977) 558564.
[16] R.E. Francois, G.R. Garrison, Sound absorption based on ocean mea-
surements: Part I: Pure water and magnesium sulfate contributions,
Journal of the Acoustical Society of America 72 (3) (1982) 896907.
[17] R.E. Francois, G.R. Garrison, Sound absorption based on ocean
measurements: Part II: Boric acid contribution and equation for total
absorption, Journal of the Acoustical Society of America 72 (6) (1982)
18791890.
[18] M.A. Ainslie, J.G. McColm, A simplified formula for viscous and
chemical absorption in sea water, Journal of the Acoustical Society
of America 103 (3) (1998) 16711672.
[19] L.M. Brekhovskikh, Y.P. Lysanov, Fundamentals of Ocean Acoustics,
3rd ed., Springer, New York, 2003.
[20] K.V. Mackenzie, Nine-term equation for sound speed in the oceans,
Journal of the Acoustical Society of America 70 (3) (1981) 807812.
[21] A.W. Cox, Sonar and Underwater Sound, Lexington Books, Lexington,
MA, 1974.
[22] H.W. Marsh, M. Schulkin, Shallow water transmission, Journal of the
Acoustical Society of America 34 (1962) 863864.
[23] M. Schulkin, J.A. Mercer, Colossus revisited: A review and extension
of the MarshSchulkin shallow water transmission loss model
(1962), Appl. Phys. Lab., Univ. Washington, APL-UW 8508, Dec.
1985.
[24] P.H. Rogers, Onboard prediction of propagation loss in shallow
water, NRL Rep. 8500, Naval Res. Lab., Washington, DC, 1981.
[25] R. Coates, An empirical formula for computing the Beckhmann
Spizzichino surface reflection loss coefficient, IEEE Transactions on
Ultrasonics, Ferroelectrics, and Frequency Control 35 (4) (1988)
522523.
[26] E.L. Hamilton, Geoacoustic modeling of the sea floor, Journal of the
Acoustical Society of America 68 (1980) 13131340.
[27] A. Beilis, Convergence zone positions via ray-mode theory, Journal
of the Acoustical Society of America 74 (1) (1983) 171180.
[28] M. Stojanovic, Underwater acoustic communication, in: John
G. Proakis (Ed.), Entry in Encyclopedia of Telecommunications, John
Wiley & Sons, 2003.
[29] D. Brady, J.C. Preisig, Underwater acoustic communications,
in: H. Vincent Poor, Gregory W. Wornell (Eds.), Wireless Communi-
cations, Prentice-Hall PTR, 1998.
[30] M.C. Vuran, I.F. Akyildiz, Cross-layer packet size optimization for
wireless terrestrial, underwater, and underground sensor networks,
in: Proc. IEEE INFOCOM 08, Phoeniz, AZ, April 1318, 2008.
[31] M. Stojanovic, On the relationship between capacity and distance in
an underwater acoustic channel, in: Proc. First ACM International
Workshop on Underwater Networks (WUWNeT06)/MobiCom 2006,
Los Angeles, CA, September 2006.
[32] B. Li, S. Zhou, M. Stojanovic, L. Freitag, P. Willett, Multicarrier com-
munication over underwater acoustic channels with nonuniform
Doppler shifts, IEEE Journal of Oceanic Engineering 33 (2) (2008).
[33] B. Li, S. Zhou, J. Huang, P. Willett, Scalable OFDM design for
underwater acoustic communications, in: Proc. of Intl. Conf. On
ASSP, Las Vegas, NV, Mar. 30Apr. 4, 2008.
[34] Bermuda Institute of Ocean Sciences http://www.bios.edu/.
[35] S. Levitus, Climatological Atlas of the World Ocean, Professional
Paper 13, NOAA, ERL/Geophysical Fluid Dynamics Laboratory, 1982.
[36] K.P. Bongiovanni, W.L. Siegmann, D.S. Ko, Convergence zone
feature dependence on ocean temperature structure, Journal of the
Acoustical Society of America 100 (1996) 30333041.
Mari Carmen Domingo received her Lic. de-
gree in Telecommunications Engineering and
her Ph.D. in Telematics Engineering from the
Technical University of Catalonia, Barcelona,
Spain in 1999 and 2005, respectively. She cur-
rently works as Lecturer at the Department of
Telematics Engineering, doing research in wire-
less networks. She has published several papers
in international journals and a book chapter.
Dr. Domingo has worked on a number of na-
tional and European R&D projects. Her current
research interests are in the area of mobile ad hoc networks, wireless sen-
sor networks, heterogeneous networks and distributed algorithms. She is
an IEICE and IEEE member. She received the ALCATEL Best Ph. D. the-
sis in wired-wireless convergence: applications and services award from
the Spanish Telecommunication Engineers Official Association (COIT) in
2006. From February 2008, she is a postdoctoral researcher at the Broad-
band Wireless Networking Laboratory, Georgia Institute of Technology,
under the supervision of Prof. Ian F. Akyildiz.