The Ocean has started to generate lot of interest in the researchers. There are numerous situations under the sea that can be explored and the information can be collected. This information can be used for various useful purposes like tactical surveillance, disaster prevention, civilian and military applications, etc. Underwater sensor networks (UWSNs) is the enabling technology for underwater communications. Radio, optical, and acoustic waves are the available physical layer technologies for UWSNs. Acoustic waves will be staying as the major carrier of UWSNs due to the relatively low attenuation of sound in sea water. Propagation of acoustic waves in the sea water depends on speed, absorption loss, spreading loss, scattering loss, ambient noise, Doppler Effect and multipath. Three types of underwater acoustic sensor networks architectures exist: Static Two-dimensional UWSNs, Static Three-dimensional UWSNs and Three-dimensional Networks of AUVs. They can be delay-tolerant or real-time depending on their application. Underwater acoustic sensor networks have their own complex set of issues that cannot be addressed by traditional terrestrial sensor networks. Thus, quite a few companies and research groups have developed underwater acoustic modems for various undersea applications. But due to these issues, they are very limited in their functionality. This poses a great challenge for researchers and a wealth of new problems are at their hands. This paper discusses the fundamentals challenges in the design of underwater acoustic sensor networks.
Key words: Underwater sensor network communication technologies, acoustic waves, underwater acoustic sensor networks
Underwater sensor networks can be deployed for a variety of applications. These applications include underwater data collection, monitoring, exploration, navigation and surveillance. To make these applications possible, there is a need to enable underwater communications among the underwater devices. Underwater devices consist of a number of sensors and the vehicles equipped with sensors, the Autonomous Underwater Vehicles (AUVs) and Unmanned Underwater Vehicles (UUVs). To enable underwater communication among these underwater sensor nodes and vehicles, they must be able to coordinate their operation by exchanging configuration, location and movement information, and to relay monitored data to an onshore station. This can be obtained by connecting underwater devices by means of wireless links and forming underwater sensor networks.
2. Applications of underwater sensor networks
The potential of underwater sensor networks is huge. They may enable a wide variety of new applications that significantly surpass the existing underwater applications. In this section, we will explore their potential .
2.1 Ocean Monitoring
UWSNs can be used to monitor the ocean environment. An example may be Odyssey-class AUVs . It is a network of sensors and autonomous underwater vehicles (AUVs) that can perform three dimensional cooperative sampling of the oceanic environment . The ability to observe and predict the characteristics of the oceanic environment has been greatly improved by the arrival of sophisticated new robotic vehicles that incorporate advanced ocean models as suggested by the recent underwater experiments [4, 5].
2.2 Environmental Monitoring
UWSNs can perform pollution monitoring of every sort: chemical, biological, and nuclear. For example, it may be possible to detail the chemical slurry of antibiotics, estrogen-type hormones, and insecticides to monitor streams, rivers, lakes, and ocean bays . Monitoring of ocean currents and winds, improvement in weather forecast, detection of climate changes, understanding and prediction of the effect of human activities on marine ecosystems, biological monitoring such as tracking of fishes and other microorganisms, are other possible applications of underwater sensor networks. They can also be used for more specific purposes such as to detect thermoclines, which are considered to be a breeding ground for certain marine microorganisms .
2.3 Undersea Reservoir, Discovery and Management
UWSNs can help detect underwater oilfields or reservoirs. They can also assist in the exploration for valuable minerals. They can facilitate the determination of routes for laying undersea cables. Later, they can be used to monitor and assess the performance of underwater oilfields or reservoirs. UWSNs would also enable the monitoring of expensive equipment after the deployment.
2.4 Seismic Monitoring
UWSNs can measure seismic activity from remote locations and then can provide earthquake warnings . Later, the recorded seismic activity can be used for further research i.e., to study and analyze the effects of seaquakes. Sensor networks that measure seismic activity from remote locations can provide tsunami warnings to coastal areas , or study the effects of submarine earthquakes, the seaquakes.
UWSNs can assist in sea navigation by identifying hazards on the seabed. Potential hazards can be like dangerous rocks, shoals in shallow waters, submerged wrecks. They can then be used to avoid these hazards, by locating dangerous rocks or shoals in shallow waters, mooring positions, and submerged wrecks. They can also be used to perform bathymetry profiling .
UWSNs can monitor areas for surveillance, reconnaissance, targeting, and intrusion detection systems. AUVs and fixed underwater sensors can be used to collaboratively monitor the areas. As an example, by employing a three dimensional underwater sensor network, a tactical surveillance system can be realized, that is able to detect and classify submarines, small delivery vehicles, and divers based on the sensed data from mechanical, radiation, magnetic, and acoustic micro-sensors. As compared to the traditional radar or sonar systems, underwater sensor networks can reach a higher accuracy, higher coverage, and robustness and they can enable detection and classification of low signature targets as well, by combining measures from different types of sensors .
2.7 Mine Scouting
UWSNs can be used to perform underwater scouting operations. They can assess underwater environment in order to detect underwater objects of interest. These objects of interest may be the mines. They can be detected by the simultaneous operation of multiple AUVs with acoustic and optical sensors. This scheme can provide a lot of flexibility. By employing this scheme, a rapid environmental assessment can be performed quickly and efficiently to detect mine-like objects .
The capabilities of underwater sensor networks have motivated the development of communication techniques for the underwater environment. The vast amount of work already done in the field of terrestrial WSNs may provide a valuable insight into this new area. But many challenges, unique to underwater communications, may exist due the significantly different characteristics of communication medium.
3. Underwater sensor network communication technologies
Present underwater communication systems involve the transmission of information in the form of radio frequency waves, optical waves, or acoustic waves. Each of these techniques has its own advantages and limitations. In the following subsections, we will explore the effect of the underwater environment on RF, optical and acoustic waves and describe the rationale for using acoustics for underwater acoustic sensor networks.
3.1 Radio Waves
Radio frequency waves are electromagnetic waves in the frequency band below 300GHz. An electromagnetic wave is a wave of energy having a frequency within the electromagnetic spectrum. Underwater radio frequency communications have been investigated since the very early days of radio , and had received considerable attention during the 1970s , however few underwater RF systems have been developed due to the highly conducting nature of salt water .
Pure water is an insulator. Water contains dissolved salts and other matter, which makes it a partial conductor . As the propagating waves cycle energy between the electric and magnetic fields, the conduction leads to strong attenuation of electromagnetic propagating waves .
Fig. : Attenuation of radio waves in fresh and sea water versus frequency.
Sea water has a high salt content and thus high conductivity varying from 2 Siemens/meter (S/m) in the cold arctic region to 8 S/m in the Red Sea . Average conductivity of sea water is considered to be 4 S/m while the conductivity of fresh water is typically on the order of a few mS/m . Therefore, average conductivity of sea water is much higher than the conductivity of fresh water .
Attenuation of radio waves in water not only depends on conductivity but also on their frequency. It increases both with increase in conductivity and increase in frequency. It can be calculated from the following formula :
is the attenuation dB/meter, is the frequency in Hertz, and is conductivity in S/m. The attenuation as a function of frequency can be seen in Fig. 1, both for sea water (4 S/m) and fresh water (0.01 S/m). Attenuation in sea water is very high as compared to fresh water for a given frequency. Therefore, to communicate at any reasonable distance, it is necessary to use very low frequencies. However, using low frequencies for underwater communications requires the use of larger antennas to capture the signal of larger wavelength.
The wavelength of a sinusoidal wave is the distance over which the wave's repeats its shape. For underwater radio waves, the wavelength can be calculated from following formula :
is wavelength in meters, is frequency in Hertz, and is conductivity in S/m. A plot of wavelength versus frequency can be seen in Fig. 2, for sea water, with conductivity 4 S/m, and fresh water, with conductivity 0.01 S/m. It can be observed that a signal's wavelength in air is considerably reduced underwater, especially in salt water, leading to differences in antenna engineering for terrestrial and underwater communications.
3.1.3 Refraction Loss on Air and Water Interface
As the radio waves experience considerably higher attenuation in water, higher transmission distances may be achieved by making the signal to leave the water near the transmitter, travel via an air-path, where attenuation loss is low, and re-enter the water near the receiver. However, as RF waves travel from air to water or water to air, there is a refraction loss due to the change in the medium. This loss can be calculated via the following formula :
is the loss in dB, is frequency in Hz, and is conductivity in S/m. A plot that illustrates the refraction loss as a function of frequency can be seen in Fig. 3, for both fresh and sea water.
Fig. : Wavelength of radio waves in sea water and fresh water versus frequency.
Fig. : Refraction loss of radio waves in sea water and fresh water versus frequency.
As frequency increases refraction loss decreases. Similar communications could be carried out underground depending on the conductivity of the surrounding rocks [9, 13].
3.1.4 Existing RF Systems
Because the conductivity of sea water poses severe attenuation to radio wave, only a few systems have been designed that employ radio waves as the physical layer technology. Historically, extremely low frequency radio signals (ELF) have been used in military applications. During the World War II, Germany pioneered the use of radio waves for underwater for communication between the submarines. Their "Goliath" antenna was capable of outputting up to 1 to 2 Mega-Watt (MW) of power. It was strong enough to send signals to submarines submerged in the Indian Ocean . Later, a U.S. and Russian based ELF system used 76 Hz and 82 Hz radio frequency signals respectively. It transmitted a one-way `bell ring' to call an individual submarine to the surface .
Until recently, the use of high frequency waves for underwater communication had been considered impractical. However, with new antenna designs, recent experiments indicate that radio waves within the frequency range 1-20 MHz can propagate over distances up to 100 m, at rates beyond 1 Mbps, using dipole radiation with transmission powers of the order 100 W [17, 18]. The antennas are very different from those used for terrestrial communications [15, 17, 18]. They do not have direct contact with seawater. Instead, the are surrounded by waterproof electrically insulating materials [17, 18]. This allows them to transmit an electromagnetic signal into a body of seawater and pick up an electromagnetic signal sent by a distant receiver from the body of seawater.
The first commercial underwater radio-frequency (RF) modem in the world was the SeaText. It was released by Wireless Fibre Systems  in September 2006. It can communicate over several tens of meters at a rate of 100 bps. Wireless Fibre Systems released a second RF modem named SeaTooth. It can communicate within a range of 1 meter at a rate of 1-100 Mbps .
3.2 Optical Waves
Optical waves are electromagnetic waves. Their wavelengths are between 400nm, the blue light, and 700nm, the red light. Light in the ocean travels at a velocity equal to the velocity of light in vacuum, divided by the index of refraction. The index of refraction is typically 1.33 for the ocean . Due to their high speed in ocean, optical waves offer the possibility for very high speed underwater communications. They can potentially exceed 1Gbps. However, optical waves used as wireless communication carriers are generally limited to very short distances because of severe water absorption at the optical frequency band and strong backscattering from suspended particles .
The amount of light attenuation in sea water can be described by Beer's Law :
is intensity of light of wavelength observed at the receiver meters away from the source of intensity . is the attenuation coefficient that can be expressed as :
is absorption and is the turbidity of water, a measure of scattering caused by suspended particles . The absorption of pure sea water is given in Fig. 4. Sea water is composed primarily of H2O and H2O absorbs heavily towards the red spectrum. Along with the H2O, the sea water also contains dissolved salts such as NaCl, MgCl2, Na2SO4, CaCl2, and KCl. These salts absorb specific wavelengths . As can be seen in figure, pure seawater is least absorptive around 400-500nm, the blue-green region of the visible light spectrum [25, 26].
Fig. : Absorption Coefficient of Pure Seawater [25, 26].
Scattering in the ocean is due to both inorganic and organic particles floating within the water column. In coastal waters and continental shelf, inorganic matter contributes to 40-80% of the total scattering where in the open ocean scattering comes mainly from organic particles like phytoplankton . The scattering process of optical waves can be evaluated by the Mie scattering theory . This theory can explain the wavelength dependence of underwater optical channels because it is valid for all possible ratios of particle diameter to wavelength. According to the Mie theory, light interacts with a particle over a cross-sectional area larger than the geometric cross section of the particle when the light wavelength is similar to the particle diameter. The scattering cross section area is defined as the total energy scattered by a particle in all directions :
is the scattered light intensity, is the incident light intensity, and is the radius of the particle. The scattering cross section area is related to turbidity as:
is the particle diameter, is the scattering cross section for particles with diameter and is is the probability distribution function of particle size.
In 1976, N. G Jerlov published the book titled "Marine Optics". This book proposed a system for classifying the clarity of the water . This system divided the globe's sea water into two categories: oceanic waters, the blue water, and coastal waters, the littoral zone. The oceanic group is further subdivided Type I-III and the coastal group is subdivided into Types 1 through 9. The clarity of oceanic water is much greater than the clarity of coastal waters. Oceanic waters are so clear that 10% of the light transmitted below the sea surface can reach a depth of 90m. On the other hand, coastal waters, such as the Baltic Sea, have a 10% transmittance depth of only 15 m due to large quantities of chlorophyll, causing highly turbid water .
Therefore, the use of optical waves as wireless communication carriers is generally limited to very short distances because of severe water absorption at the optical frequency band, strong backscattering by the inorganic and organic particles floating within the water column, and the turbidity .
3.2.4 Existing Optical Systems
Water clarity plays such a significant role in determining whether optical waves can be used for underwater communication or not. No commercial underwater optical modems have been developed due to this reason. However, there has been a recent interest in short-range high-rate optical underwater communications . For example, a dual mode transceiver has been reported which uses both acoustic and optical technologies for underwater communications . The optical transceiver can achieve a data rate of 320 Kbps over a few meters. Another optical modem that achieves a range of 2 meters for a rate of 57 Kbps has also been reported . An optical modem prototype has also been designed for deep sea floor observatories that can operate up to 10 Mbps over 100 meters . An LED-based communication system has also been demonstrated that can reach 1 Mbps in a 12 feet, 1200 gallon tank .
Furthermore, a recent study using Monte Carlo simulations indicates that optical communication data rates greater than 1 Gbps can be supported over seawater paths of several tens of meters . Recent experiments have also involved a mixed acousto-optical approach. This approach uses laser to generate sound underwater. In 2009, the U.S. announced the use of blue/green lasers to produce bubbles of steam that pop to create explosions of up to 220 decibels . The laser beam incident at the air-water boundary is exponentially attenuated by the medium. This creates an array of thermo-acoustic sources related to the heat
Fig. : Vertical profile of sound speed (m/s) in seawater as the lump-sum function of depth (m) [21, 36].
energy and physical dimensions of the laser beam in water. The fluctuations in local temperature give rise to volume expansion and contraction. The volume fluctuations in turn produce a propagating pressure wave which has the characteristics of an acoustic signal. Controlling the rate of temperature fluctuations could provide a means for transmitting the information .
3.3 Acoustic Waves
Acoustic waves are caused from variations of pressure in a medium. Acoustic waves have been widely used in underwater communication systems. This has been due to the relatively low attenuation of sound in water . However, there are many factors that can affect the propagation characteristics of sound waves in water and in the subsequent subsection, we will discuss them one by one .
The extremely slow propagation speed of sound through water is an important factor that separates them from radio waves. The speed of sound in water depends on the water properties of temperature, salinity and pressure. Typical speed of sound in water near the ocean surface is about 1520 m/s, which is more than 4 times faster than the speed of sound in air, but five orders of magnitude smaller than the speed of light.
The speed of sound in water increases with the increase of water temperature, salinity and depth. In the surface ocean, the changes in sound speed can be majorly attributed to the changes in temperature. Salinity has little effect on sound speed because the salinity changes in the open ocean are small. Near shore and in estuaries, the salinity varies greatly. There it can have a more significant effect on the speed of sound in water. As depth increases, the pressure of water has the largest effect on the speed of sound. Sound will travel faster in warmer water and slower in colder water. Approximately, the speed of sound increases 4.0 m/s for a 1Â±C change in water temperature, 1.4 m/s when salinity increases 1 practical salinity unit (PSU), and 17 m/s as the depth of water, and therefore the pressure, increases 1 km. But in general, the variations in sound speed for a given property are not linear. The overall effect of sound speed in seawater is illustrated in Fig. 5 [21, 36].
3.3.2 Absorption loss
The absorption of acoustic waves in sea water depends on the temperature, salinity, acidity of the sea water, and their frequency. The absorption loss for acoustic waves can be expressed as , where is the propagation distance and is the absorption coefficient of frequency . For seawater, the absorption coefficient at frequency in kHz can be written as the sum of chemical relaxation processes and absorption from pure water :
The first term is the contribution from Boric Acid and is its relaxation frequency; the second term is the contribution of Magnesium Sulphate and is its relaxation frequency; the third term is the contribution of pure water. The pressure dependencies are given by . are constants. The variation in total absorption versus the frequency can be observed in Fig. 6 for different oceans of different temperature, pressure, and pH . Since increases with frequency, high frequency waves will be considerably attenuated than the low frequency waves for a given distance. Therefore, the low frequency waves can travel farther.
3.3.3 Spreading Loss
The energy radiated from an omni-directional source spreads spherically through water. Not all of the energy is directed in a single direction. Therefore, much of the energy is lost. This is called spreading loss. The spreading loss is frequency independent .
Fig. : Acoustic Absorption as a function of temperature, pressure, and pH [11, 41].
In the deep sea water, the spreading loss is proportional to the square of the distance from the source.
Due to the restriction of sound waves by the surface and the sea floor in shallow sea water, instead of spherical spreading, cylindrical spreading occurs. In this case, the spreading loss increases linearly with the distance from the source.
For a practical underwater setting, the spreading loss falls somewhere between spherical and cylindrical spreading, with power loss proportional to where is between 1 for cylindrical spreading and 2 for spherical spreading . In logarithmic terms, the classical equation for spreading loss in dB is:
A plot of this equation can bee seen in the Figure 7. Spreading loss not only depends on the distance but also on the spreading coefficient . Spreading loss is also higher for the spherical spreading then the cylindrical spreading for a given distance.
3.3.4 Ambient Noise
Ambient noise is the noise associated with the background. It is generated by a variety of natural and man-made sources. Natural sources of ambient noise include breaking waves, rain, marine life, bubbles. Man made sources include, surface-ships, and military sonar. As a result, this noise is attributed to multiple sources and no one source dominates the received field. Therefore, the individual sources could not be identified generally [42, 43].
Ambient noise is frequency dependent. The primary source of ambient noise can be categorized by the frequency of sound waves. In the frequency range of 20-500 Hz, ambient noise is primarily generated by distant shipping. In the range of 500-100KHz ambient noise is mostly due to spray and bubbles associated with breaking waves. At frequencies above 100KHz, the noise generated by the Brownian motion of water molecules, also called the Thermal Noise, dominates. Thus, when selecting a suitable frequency band for communication, besides path loss, noise should be also considered [39, 44]. The ambient noise in underwater environment can be divided into 4 major categories: turbulence noise, shipping noise , wind noise and thermal noise . The total noise can be calculating using equation :
A summary of background noises in the ocean in the form of a graph that displays typical noise levels at different frequencies is available from Wenz 
Often one or more localized non-uniformities in the water, such as particles and bubbles, force the sound waves to deviate from a straight trajectory, resulting in the scattering of acoustic waves. Scattering also causes the reflected sound waves to deviate from the angle predicted by the law of reflection. When the wind speed increases, it causes the surface to roughen and the effect of surface scattering becomes even more evident. Surface scattering introduces not only power loss, but also spreading in delay of each surface bounce path [21, 39].
In multipath propagation, waves travel over multiple different paths before reaching the receiving end. Multipath propagation exists in the acoustic underwater communications. It can be attributed to the two fundamental mechanisms that underlie the formation of multiple paths: reflection and the ray bending. Reflection of sound waves is caused by the sea bottom, sea surface or any objects in the water. As for the ray bending, rays of sound always bend towards regions of lower propagation speed because the speed of sound is function of temperature, salinity, and depth . Multipath due to reflections from the surface and bottom is common in shallow waters whereas multipath due to ray bending is common in deep waters.
Fig. : Spreading loss for cylindrical and spherical spreading
Multipath can adversely affect communications because it introduces time dispersion, which in turn can cause severe inter-symbol interference. Typical underwater channels may have a delay spread around 10ms, but occasionally, the delay spread can be as large as 50 to 100 or as small as 3 .
Understanding the mechanisms of underwater acoustic multipath propagation is based on the Ray Theory and the Theory of Normal Modes. Bellhop is a highly efficient ray tracing model and is a commonly used one. AcTUP is an underwater acoustic propagation modeling software that can perform two-dimensional Bellhop acoustic ray tracing . Output options include ray coordinates, travel time, amplitude, acoustic pressure or transmission loss.
3.3.7 Multipath acoustic channel model
An underwater acoustic communication channel can be regarded as a multipath acoustic channel, where the propagation occurs over multiple paths. Then its impulse response can be written as the sum of impulse responses of propagation paths :
is the impulse response of each propagation path, and is the propagation delay associated with the -th path. Each path has an impulse response of a different shape due to the different attenuation and delay experienced by it. Overall channel response frequency domain is :
is channel transfer function, is the frequency response of the -th path. can be written as :
is the cumulative reflection coefficient along the -th path and is the propagation loss associated with -th path. Now for the :
is the signal frequency, and is the transmission distance of -th path, models the spreading loss, and is the scattering loss associated with -th path.
3.3.8 Path Loss
As can be seen form the last equation, the propagation loss, or the Path Loss, for the -th path depends on distance, spreading loss, absorption loss, and the scattering loss. The overall path loss is given by :
It is taken with reference to some reference path , usually the first multipath. Now using this equation for the over all path loss, one can one calculate the Signal-to-Noise Ration (SNR) which is given by :
is the transmission power, is the overall path loss that contains scattering, spreading and absorption loss, and is the ambient noise. This equation can be used to calculate the maximum transmission distance achievable for a desired signal to noise ratio at the receiver and is also sometimes called the "Passive Radar Equation" .
3.3.9 Doppler Effect
Movements of receiver or transmitter or mismatch between their sampling rates cause frequency shifting and frequency spreading. This is called the Doppler Effect. Velocity of sound in underwater communications can be very low and there is always some motion present in the form of coasting with waves, currents, tides, which may occur at varying speeds. Therefore, the Doppler Effect on the acoustic waves can be drastic, causing the power spectrum of
Table : Comparison of the underwater acoustic, optical and radio wave communication technologies .
received acoustic signals to spread. This spread is called the Doppler spread. The reciprocal of Doppler spread is called the Coherent Time. Slow fading channels have large coherent time and the fast fading channels have a small coherent time. Time selective fading has a large impact on bit error rate of digital signal, so in order to reduce its impact, symbol rate has to be much larger than the rate of fading beat . Table 1 gives a comparison of the underwater acoustic, optical, and radio wave communication technologies .
3.3.10 Existing Underwater Acoustic Modems
Commercial underwater acoustic modems are currently available for purchase from LinkQuest, Teledyne Benthos, TriTech International, Aquatec Group, EvoLogics, and DSPComm, and the Woods Hole Oceanographic Institute (WHOI).
LinkQuest produce a number of modems  ranging from the UWM1000 modem that can communicate up to 350 meters with a data rate of 9600 to 19200 bps to the UWM10000 that can communicate up to 10 km with a data rate of 2500 to 5000 bps.
Teledyne Benthos modems have been used in sub-sea networks including the US Navy Seaweb program , and the Front Resolving Observation Network (FRONT) . Their modems are primarily for point-to-point deep water vertical communication, e.g. from ocean floor directly to the ocean surface . They have demonstrated communication in over 1 kilometer of water at 10240 bps without any errors in ideal situations; however the practical data rates are less than 2400 bps. The modems are also relatively power-hungry consuming 28-84 W transmit power and 0.7 W receive power.
TriTech International developed the Micron Data Modem  for small remotely operated vehicle (ROV) communications. The Micron Modem is the smallest and lightest acoustic modem in the market. Thus it is suitable for on board small AUVs or ROVs. It can transmit for a maximum range of 1km with a data rate of 40 bps. Micron Modem has moderate power consumption: 7.92 W for transmitting and 720 mW for receiving.
The Aquatec Group modems are tailored to suit the application of interest . As part of their AquaModem long-range series, they can offer modem configurations that provide a bit rate of 300-2000 bps and a range of up to 20 km. As part of their AquaModem500 short-range series, they can offer modem modem configurations that provide a bit rate of 25-100 bps, a range of up to 250 meters.
EvoLogics developed six underwater acoustic modems that make use of their patented sweep spread carrier (S2C) underwater technology . Their modems range from high speed, shallow water, medium range modems that can transmit at the rate of 28 kpbs up to 1000 meters to low speed, deep water, long range modems that can transmit at a rate of 6.5 kbps up to 8 km.
DSPComm produces the AquaComm. AquaComm is an underwater wireless modem that provides the user with the ability to set the receive and transmit power thus offering a low-power solution for short range applications . The AquaComm offers data rates of 100 or 480 bps up to 3 km, depending on the model.
The Woods Hole Oceanographic Institute (WHOI) Micro-Modem  is a popular open architecture alternative to the commercial solutions. It is user programmable and can support multiple instruments. It is currently being used by autonomous underwater vehicles, autonomous surface vehicles, and deep-water ocean observatories. The Micro-Modem employs frequency-hopping frequency shift keying for low data rate communications, up to 80 bps, and phase shift keying high data rate communications, up to 5400 bps.
4. Communication architectures of underwater sensor networks
Now the reference architectures for two-dimensional, three-dimensional underwater networks, and three-dimensional networks of AUVs, are introduced. They can enhance the capabilities of underwater sensor networks. The communication architectures introduced here are used as a basis for discussing the challenges associated with underwater sensor networks [1, 63, 64]:
4.1 Static Two-dimensional UWSNs
Static Two-dimensional UWSNs consist of sensor nodes that are fixed to the bottom of the ocean. Because the sensor nodes are fixed, their depth is constant. Therefore, they are also called static UWSNs. They can be used for two types of communications: horizontal and vertical. The horizontal communication occurs between the nodes in a cluster and their cluster-head. The vertical communication occurs between the cluster head and the base station floating in the water. In this architecture, the base station is equipped with acoustic links, or the radio link, or both. A satellite transmitter can also be an option. Data is collected by the floating base station, which then sends it to the satellite, ships, or to the base stations on shore .
4.2 Static Three-dimensional UWSNs
Static Three-dimensional UWSNs consist of fixed portions and mobile portions. Fixed portions include anchored sensors while mobile portions are composed of autonomous vehicles. Depth of each sensor can be adjusted which gives it a third dimension. One way to adjust the depth of sensor is to use the floating buoys . The principal difference between the Static Two-dimensional UWSNs and this architecture is that Static Three-dimensional UWSNs do not build clusters and therefore only vertical communication is possible. Data can be sent directly to the floating base station, and then onwards sent to satellite, ships, or the on shore base stations. They may be used for surveillance applications or monitoring of ocean phenomena.
4.3 Three-dimensional Networks of AUVs
Three-dimensional Networks of AUVs contain fixed portions of anchored sensors and mobile portions of autonomous vehicles. Fixed portions consist of sensor nodes anchored to the bottom of the sea. These nodes can be at different depths in the sea to collect different sorts of information. The sensor nodes then store this information and when the floating AUVs come across these sensor nodes while floating, they collect this information. But this data exchange between sensor nodes and the AUVs requires network coordination. Therefore, in this architecture, it is necessary to have algorithms that guarantee the coordination between the AUV and the sensors, and their correct functionality in the scenario.
5. Design Challenges
While many aspects of networking under water have similarities with the terrestrial sensor networks, there exist many differences that require communication protocols to be tailored for underwater sensor networks. In this section, we discuss the differences between terrestrial and underwater sensor networks . They can be outlined as follows [45, 63, 67]:
5.1 Communication medium
The fundamental difference between terrestrial and underwater sensor networks is the communication medium. The underwater environment results in different propagation, attenuation, and fading characteristics than in air. Despite the fact that underwater networking is a rather unexplored area, underwater communications have been experimented on since World War II, when, in 1945, an underwater telephone was developed in the USA to communicate with submarines . Underwater communication channel is highly irregular due to absorption, scattering, spreading, noise, and multipath. The path loss not only depends on the distance but also on frequency. The available bandwidth is limited and depends on the transmission distance therefore, high bit error rates can occur. Low propagation speed of sound waves under water contributes to large propagation delays in underwater acoustic communications.
Terrestrial sensor nodes are expected to become smaller in size. They are also becoming increasingly inexpensive. On the other hand, underwater sensors are expensive devices. This is mainly attributed to three reasons. Underwater transceivers are more complex. They require extra hardware protection for the extreme underwater conditions. They also have a relatively small number of suppliers.
Terrestrial sensor networks are densely deployed. The deployment in case of underwater sensor networks is sparser. This can be mainly due to the high costs involved in their deployment under water and also due to the unusual challenges associated with the deployment itself.
Underwater sensors need higher power for their underwater communications. This is due to three reasons: firstly because of use of acoustic waves rather than radio waves as the physical layer technology, secondly due to the higher distances, and thirdly due to the use of more complex signal processing techniques implemented at the receivers to compensate for the channel impairments. Hence, more power is consumed and higher battery power is required. The battery power is limited and the batteries cannot be charged by solar energy. This makes the situation even more difficult.
5.5 Memory and Processing power
Terrestrial sensor nodes have very limited storage capacity. But due to the highly irregular nature of underwater acoustic channel, underwater sensors may need to be able to do some data caching. To remove the distortions introduced by the underwater channel and to recover the original information, underwater sensors will require more processing power at their disposal.
Underwater sensors are prone to problems that are uncommon for terrestrial networks like fouling and corrosion of sensors. This can cause failures and may lead to higher failure rates. Underwater sensors need more protection against corrosion and fouling, which may impact their performance.
Underwater sensors need to be more robust to a high range of temperature variations. They also should be stable against underwater drifts. Drifts can lead to higher Doppler spreads in the signals. High bit error rates, temporary losses of connectivity, shadow zone, can also be experienced due to harsh underwater environmental conditions.
5.8 Network topology
The network topology is a crucial factor in determining the cost, reliability, capacity, and energy consumption of a network. Because the cost of underwater devices is high, the network topology needs to be carefully optimized. The deployed underwater network should be highly reliable to avoid the failure of single or multiple devices that can compromise the functionality of entire network. The capacity of the underwater channel is severely limited. Network topology should be engineered to avoid any communication bottlenecks. Finally, due to the limited underwater battery power and due to the absence of the possibility of a solar recharge, the topology should be energy efficient.
5.9 Real-Time Networking vs. Delay-Tolerant Networking
It is important to develop the protocols that can deal with the typical characteristics of the underwater acoustic channel so that they take into account these issues and can respond to them in timely manner by taking appropriate actions depending upon their application. The application can be real-time or delay tolerant. For example, sensor networks that record seismic activity have a very low duty cycle and upon activation, they generate large amounts of data. This data is then relayed to a monitoring station where it can be analyzed to predict future activity. These sensor networks are delay tolerant. On the other hand, sensor networks intended for disaster prevention require immediate delivery of information and hence they require real-time protocols. Therefore, underwater acoustic sensor networking protocols should always be aware of the characteristics of the underwater environment, and the requirements of their application: real-time or delay-tolerant. The Delay-Tolerant Networking Research Group (DTNRG) [69, 70] has developed mechanisms to resolve the intermittent connectivity, long or variable delays, asymmetric data rates, and high error rates by using a store and forward mechanism. This mechanism is based on middleware between the application layer and the lower layers. Similar methodologies can be developed for other applications .
Underwater communication has generated a lot of flurry in academia in recent years. Its applications are countless, ranging from civil and military purposes to exciting new unexplored areas for researchers. But the challenge is still there. A wealth of new problems has to be met before it can be realized in full potential. Currently, there are three major technologies for underwater communication: acoustic waves, radio waves and optical waves. Acoustic waves remain the most robust and feasible carrier to date. They will be staying as the major carrier of underwater communication in UWSNs due to the high attenuation of radio waves and severe absorption of optical waves in sea water. But, the use of acoustic waves for underwater communication also has its problems and faces quite a number of issues at physical layer. These include larger delays, limited bandwidth, channel impairments, higher error rates, shadow zones, limited energy resources, higher failure rates, the topology issues, networking issues, are just to name a few. Consequently, many networking paradigms need to be revisited to address these major challenges in the design of underwater sensor networks as discussed in this paper.