This is a very complex issue and the question raises many further questions, for example:   What water depths?
What ranges? Interferometric systems have a greater swath width to depth ratio than beam-forming systems, which means that the sound rays have a shallow angle to the water layers, resulting in more refraction.
What is the overall error budget needed?
Is target location important? Refraction affects the apparent position of targets on the seabed as well as their depth.
As a rough estimate, the depth error with significant variations of temperature and salinity in the water column will probably be in the range of 3% to 5% of depth. You will also get distortions of the seabed profile (known in the business as “smileys” or “frowneys”, according to whether the profiles curve up or down at the end). This will cause significant along-track features on seabed 3D images and “dragging” of contours, particularly on flat seabeds. So even though depth accuracy may not be critical, un-corrected SVP refraction can significantly degrade the usability of the survey data.
There are various papers and standards touching on this, including:
http://www.imca-int.com/media/73578/imcam200.pdf
http://www.hydro-international.com/issues/articles/id1290-Solving_the_Uncertainty_Management_Puzzle.html
http://ccom.unh.edu/sites/default/files/publications/beaudoin_paper.pdf

Extracts:
“Temperature
The temperature at the sea surface varies with the geographic position on the earth, with the season of the year and the time of the day [Pickard and Emery, 1990]. The temperature field distribution is a complex one and can not be predicted with enough accuracy for hydrographic surveys; through the water column the behaviour of the temperature is also very complex. Such unpredictability necessitates a comprehensive distribution of sound velocity profile casts, both temporally and spatially, to maintain a representative currency of the sound velocity profiles for the survey area.

The depth measurement is quite sensitive to variations of the sound velocity profile; a variation of one degree Celsius in temperature translates to approximately 4.5 m/s in sound velocity variation.
The temperature variation is the dominant factor for sound velocity variation between the surface and the lower limit of the thermocline36, thereafter pressure becomes the principal influence.

Salinity
The salinity is a measure of the quantity of dissolved salts and other minerals in sea water. It is normally defined as the total amount of dissolved solids in sea water in parts per thousand (ppt or ‰) by weight.
In practice, salinity is not determined directly but is computed from chlorinity, electrical conductivity, refractive index or some other property whose relationship to salinity is well established. As a result of the Law of Constancy of Proportions, the level of chlorinity in a sea water sample is used to establish the sample’s salinity37.
The average salinity of sea water is around 35 ‰. The rate of variation of sound velocity is
approximately 1.3 m/s for a 1 ‰ alteration in salinity. Typically, the salinity is measured with a CTD cast (Conductivity, Temperature and Depth) using the observable electrical conductivity, see 2.2.1.2.

Pressure
The pressure also impacts significantly on the sound velocity variation. Pressure is a function of depth and the rate of change of sound velocity is approximately 1.6 m/s for every alteration of 10 atmospheres, i.e. approximately 100 metres of water depth38.
The pressure has a major influence on the sound velocity in deep water. “

“Sound speed profiling (for MTES and SBES systems; mandatory for MBES)
Acquisition
When determining the speed of sound in the working area using a sound velocity profiler (a velocimeter or CTD), the probe is lowered to the maximum depth of the site or at a depth at which the acceptably uniform speed of sound is obtained. The resultant cast from a velocimeter or CTD will provide a table listing the speed of sound at recorded depth intervals. The results obtained from the sound speed cast should be entered directly into a multibeam echo sounder. Data acquired with a MTES and SBES may have the sound speed corrections applied during post processing.
Unless specified otherwise sound speed casts should be done on a daily basis or when deemed necessary.
The results of these casts should be checked against a second probe or another system in order to verify that the sound speed is correctly being measured. These checks should be performed at least once a week or as the conditions permit.
A proper record or log should be kept to identify the SVP casts acquired during the project including their time, date and their geographical locations. Sound speed profilers should be capable of measuring sound speed to a precision of at least 0.1m / s.

Data Processing
Profile location and time acquired should be recorded in any processing logs and referred to when applying SVP corrections in post processing.
Profiles should be edited for anomalies or spikes in the data. Whenever possible, profiles collected with different probes in the same survey area should be compared. When applying SVP corrections in post processing, ensure that the data within the profile contains information for depths up to or greater than the soundings being corrected. If required, an SVP editor may extend the profile to depths greater that the soundings being corrected.”

A standard sound velocity profiler, such as Valeport’s miniSVP (www.valeport.co.uk) is the standard way to do this; it records depth and sound velocity internally, and the data can be uploaded on recovery. That works well when a separate member of the survey team is available to do SVP “dips” in parallel with other survey work.

Another way to take SVP dips is to use a continuous-reading instrument, such as Valeport’s miniSVS, with the optional pressure meter and cable on a reel added (www.valeport.co.uk). That way, the profile can be recorded directly to the Bathyswath software through a serial port on the computer. The Bathyswath software includes “Here” and “Now” buttons, which attach the current position and time to the profile that has just been taken, and stores the profile for later use in post-processing.

You still need ancillary devices such as motion sensor (IMU, INS, etc.), positioning sensor (GNSS, etc.), Sound Velocity Sensor (SVS), Compass.

Polyurethane (PU) over-moulding on the transducer does have anti-fouling compound added (upto a certain limit so that it does not affect acoustic performance). If the transducer(s) are supposed to stay in the water for a long duration (i.e. Bathyswath-Static) then the end-user can add if needed an ordinary marine anti-fouling paint (There is no prefered or recommended brand). First, clean the surface of the transducer with a soft brush (e.g. a nail brush) and washing-up liquid. Then, paint on the thinnest layer possible.

An SVP is recommended for survey work. Is it always needed?
You need an SVP (sound velocity profiler) for three reasons:

• To measure the sound velocity at the transducers; the software needs to know that to calculate the angle of the sound echoes from the bottom
• To calculate the range to the bottom, knowing the time between transmitted pulse and echo
• To compensate for refraction effects, when the sound ray is bent as it passes between layers of water at different depths Sound velocity in water changes with two things: salinity (the amount of dissolved salt) and temperature.

In lakes and rivers, we don’t have to worry about the first one. Temperature is easy and cheap to measure, and you can work out the sound velocity using a simple formula. So, if the temperature is the same at all water depths, then all you need is a thermometer.
However, it is very common for the top surface layer to be warmer than the rest of the water, particularly in lakes. So, we need to compensate for that. For example, on Lake Annecy, where we do most of our tests, if we don’t compensate for this warm surface layer in the summer then there is a definite bend to the depth profiles.

Therefore, what I suggest is to get a simple temperature sensor on the end of a wire (these are easy to get for a few tens of euros), and mark the wire at 1-metre intervals. Then, tie a weight to the temperature sensor, lower it into the water, and read off the temperature at each metre. The top 10 metres is probably enough, as the temperature will be fairly constant after that. A simple Excel spreadsheet will give you the sound velocity profile for a fraction of the cost of an SVP.

Bathyswath needs a motion sensor to work with it, such device is also called attitude sensor, motion reference units (MRU) or inertial navigation sensors (INS).

It provides the roll, pitch, heading and heave information that allows the range and angle information from the sonar to be properly positioned in space. If the motion measurement is not accurate, then you see “motion artefacts” in the measurements of the depths, where the depth map has waves in it that match the movement of the boat.
Which one is best for my application?

Some Motion sensors are quite expensive so choosing the right one is therefore important.

• The first thing to check is the roll measurement accuracy. We recommend 10cm max. error in depth so the angular accuracy should be less than 0.05 degrees [arctan (0.01/100)].
• At sea, heave motion (up and down) is also an important error factor, and needs an accurate sensor; but on lakes and rivers, there is almost no heave if the weather is good, so we don’t have to worry about that too much.
• Sensor response with regards to large and fast-changing motions is also important (i.e. in case of bad sea states)
• Sensor data timing is an important consideration. Sensor data can be timed relative to an external time source, usually GPS time, and synchronised with a timing signal, e.g. PPS pulses. Or, it can be time-stamped in the computer at the time the data arrives. Some sensors can receive PPS signals and output time stamps, and some do not. So, this should also be something to look for when choosing a sensor.
• Some sensors input position information from a GNSS (global navigation satellite system, e.g. GPS), and use their motion data to improve the accuracy of the position data, and to keep it updating if satellite signals are lost for a few seconds, for example if the boat goes under a bridge.
• Some motion sensors can also have problems with long-period heave motion. This happens with ocean swell, but not so much in inland waters or smaller sea areas.

In general, the more money you spend, the worse weather you can survey in. All motions sensors have their good and bad points, so we are reluctant to give one definite suggestion. The following sensors have been used successfully with Bathyswath:

• Ellipse, Ekinox and Apogee IMUs from SBG Systems
• POS-MV from Applanix.
• SPAN from Novatel (PwrPak7)
• Dual antenna GNSS systems from Trimble and Hemisphere, providing low cost but low accuracy roll and heading from the antennas
• IMU-108 from Ship Motion Control (SMC)
• F180 from CodaOctopus
• And several others …

Sonar specifications often give numbers for along and across track beam widths. What are these numbers for Bathyswath?
Along and across beam width determine the size of objects that can be detected: smaller beam widths allow you to see smaller objects, and large beam widths cause objects on the bottom to be “smudged out”. Bathyswath, like all swath bathymetry systems, looks sideways from the vessel, using a beam that is wide in the vertical direction and narrow at right angles to the direction that it is looking (in the direction that the vessel is going: “along track”).

The along-track beam width is controlled by the “azimuth beam width” of the transducers. This is 1.1 degrees for the 468kHz and 234kHz transducers, but because these beams are narrow for both the transmit and receive parts, the transmit and receive beams combine to give an effective beam width that is half that: 0.55 degrees. Across-track beam width is a number that applies to beam-forming sonars, which form separate beams that the receive data is detected in. The more separate receive elements there are, the more beams and the narrower each separate beam is. Interferometers like Bathyswath don’t form beams like this; instead, they detect the angle of the incoming sound wave at different times.

So, the resolution (the size of the smallest object that can be detected) is controlled by two things:
– The rate at which angles are measured: using modern electronics and computers, this can be as fast as you like; Bathyswath 2 does this at 100kHz, 100,000 times a second. So this is not a limit to performance.
– The width of the sonar transmit pulse: the sonar sends out a short “pulse” and listens to the echo. If the pulse is long, then the area covered by each one when it hits the bottom is wider, and the resolution is worse. But longer pulses put more energy into the water, so the echo is stronger and the data quality and range are better. The Bathyswath software controls the pulse length automatically during the survey to maintain data quality (although the operator can over-ride this if required).

In a typical survey, we see that the pulse length was 46 cycles, so the length of the pulse was 46 wavelengths at 468kHz. One wavelength is 3.2mm at this frequency, so each pulse is 15cm long. At a water depth of 10m and a range of 30m, this corresponds to an angle of 0.1 degrees.

Bathyswath basically *is* a sidescan system; or, rather, it is four sidescan systems used together to measure depths as well as sidescan. One user in the USGS told us that he considers it “an excellent sidescan system that happens to give depths too”.
The other difference between Bathyswath and most sidescan-only systems is that Bathyswath is usually fixed to the vessel, but sidescan systems are used from a tow-fish, which is set to fly close to the bottom, to give the best variation in angles with the seabed as the shape of the seabed changes. If the water is fairly shallow, that shouldn’t be too much of a problem. In deeper water, do what you can to get closer to the bottom, perhaps choosing a low tide or using a longer pole.

Some interferometric systems have a gap in the swath below the transducers: the area called the “nadir”. Does this affect Bathyswath?
The data density of depth measurements is lowest at nadir for all interferometric systems. This is because they measure angle to the seabed for a set of equally-spaced ranges. Simple geometry means that the horizontal distance between measurements is greatest below the transducers and least further out. This means that, for most survey situations, although depths are measured right across the swath, the density of data below the transducers may not be high enough for some survey specifications. For example, IHO S44 Special Order typically requires a denisty of 9 processed points per square metre. In some cases, Bathyswath meets this requirement, but in challenging sonar conditions, the data density may fall below this requirement.

This issue can be resolved in several ways:
– Run the survey so that the far range of some lines overlaps with the nadir region of other lines. This is called a “sidescan search pattern”.
– Fit a third, forward-looking transducer. See here.
– Use Bathyswath together with a beam-forming multibeam. Several Bathyswath customers do this; using Bathyswath in very shallow water, where the wide swath width lets them finish the job more quickly; using the beam-former in deep water, beyond the range of Bathyswath; and using both systems together in depths in between, combining data from the two systems to give good coverage right across the swath. We hear that this works very well.
Some interferometers offer a single-beam echosounder to give a single data point at nadir. This option is available to Bathyswath users – the Bathyswath software reads data from and displays the results of echosounders. But it isn’t often chosen by our users, as a single data point doesn’t make much difference to the data density.

The depths measured by Bathyswath come from many different measurements and calculations. How can I be sure that they are all working correctly to give me the right answer?
There are several ways that Bathyswath users prove the accuracy of their systems. These include:
– Comparison with previous surveys using other systems. Most sonar software can output data in ASCII XYZ (easting, northing and depth written to a text file). The Bathyswath Grid Processor can input these files and subtract them from a Bathyswath grid to give a comparison between them.
– Surveys in areas where the depth is known. Places where the water can be taken away are good for this, such as dry docks and locks. We helped a customer in Belgium to do this recently, using a large river lock, where the depth of the concrete bottom had been accurately measured using land-survey techniques. By carefully eliminating all sources of error in their combined system, we were able to prove agreement between the sonar data and the land survey to better than 1cm!
– Bar-checking: echosounder users use a metal bar or disk at the end of a rope hung beneath the sonar transducers. This is harder to do with multibeam systems, but it can work. However, it is usually easier to use the rope to measure the distance to the seabed itself.

Many sonar systems advertise accuracy and resolution data. What does this information mean?

Depth Accuracy tells you how sure that you can be that a depth measurement is correct. If the sonar tells you the water depth is 10 metres, is it really 10.1m or 9.9m?
Depth Resolution tells you the smallest depth difference that can be detected. So if the depth resolution is 1cm, you should be able to see a 1cm “step” in the bottom of the sea.

These two figures are very different. A sonar might be able to resolve a small object (resolution), but it might not be able to tell exactly how deep it is (accuracy).
Accuracy and resolution can also be stated in range (distance from the sonar transducers) and horizontal distance. Survey specifications such as IHO S44 define acceptable limits for the vertical and horizontal accuracy and resolution of the data from a survey. It uses the word “uncertainty” instead of “accuracy”, which can be related the statistical spread of measurements made. Horizontal resolution is important here, and it is related to the density of measurements on the seabed: the number of measurements per square metre. The accuracy and resolution are specified for the complete survey, not just the sonar system. The quality of the position and attitude measurement systems are as important as the sonar system for this, and the way that the system is used is also very important. A survey made when the sea is very rough will not be as accurate as one made in calm weather! Few survey organisations give “official approval” of sonar systems. All that can be said is that a sonar system has been used to generate data that has met the approval of hydrographic authorities. Bathyswath and SWATHplus have met this requirement many times in different countries around the world.

Depth accuracy and horizontal resolution are related to each other, according to the filtering that is applied to the data. Raw Bathyswath data has very high resolution: the resolution of the un-filtered bathymetry data is the same as that as the sidescan image data, which has to be very good in order to see small objects on the seabed. This necessarily means that there is a wide statistical spread of the raw depth measurements. The data has to be filtered to reduce the statistical spread of depth measurement, and so improve the accuracy. That filtering process reduces the horizontal resolution. Detailed analysis shows that, when Bathyswath data is filtered to the limit of the data density requirement of IHO S44, the accuracy (uncertainty) is better than the limit required by the specification. The “Accuracy and Resolution” section of the Bathyswath Technical Information document explains this in detail.
Some sonar systems that are similar to Bathyswath provide additional resolution information, and the Bathyswath brochure provides this same information for comparison.

This information includes:
– Measurement resolution limit: this is the smallest difference in distances that the system can measure. Bathyswath measures range as a count of sonar cycles, and so the measurement resolution limit is the same as the wavelength of the sonar. This is 3mm for the 468kHz version, 6mm for 234kHz, and 12mm for 117kHz.
– Resolution detection limit: this is the smallest object that can theoretically be detected by the sonar. Sonar theory says that this is half the sonar wavelength, and so it is half the “measurement resolution limit” for Bathyswath.
– Across track resolution: this is the smallest difference in range that can practically be measured, and it is related to the length of the transmit pulse: the burst of sound that the sonar sends out. It is 1cm for 468kHz, 6cm for 234kHz, and 13cm for 117kHz. So although the sonar electronics can measure differences in distances equal to the measurement resolution limit, the length of the transmit pulse spreads out the sound echoes to the size of the across track resolution.

All three measurements are theoretical limits, and conditions in the real sea and seabed will restrict the actual resolution. Most similar systems (interferometers) work the same way, and so they have similar theoretical performance limits. The actual limits to performance are related to the quality of the sonar transducers, electronics, and software, and are best judged by examining real sonar data obtained from the sonar system.

Some multibeam systems offer beam steering (beamsteering) in roll and pitch to make sure that the beam footprint is always regular and stable when the ship moves. Does Bathyswath have that capability?
Interferometers don’t need roll beam steering.

A problem with some beam-steering multibeams is that the beams are formed in fixed directions relative to the ship, so when the ship rolls from side to side, then the area of seabed that is covered by the beams also moves from side to side. That means that survey lines have to be run closer together to make sure that the edges join up. Some beam-forming systems have the ability to adjust the angles at which the beams are formed, by reading in roll from a motion sensor in real time, so that the beams always point in the same direction relative to vertical. This is called roll beam steering.

Interferometric systems, like Bathyswath, measure the angle of the incoming wavefronts of the back-scattered sound signal directly: they don’t form beams relative to the ship. The edge of the swath is determined by the range at which not enough sound signal is back-scattered to the transducers, not by any angle relative to the ship. So, the swath stays parallel to the ship’s track, and there is no need to steer the beam.

The footprint of each ping on the seabed also moves forwards and backwards when the ship pitches. So, some beam-forming systems also adjust the fore-and-aft direction the transmitted beam to keep the pings at a regular interval along the seabed. This effect is most noticable in deeper water, below about 200m water depth, where the fore-aft shift of the ping footprint with a typical pitch angle becomes large compared to the distance between pings, which is determined by the speed of the ship over the ground and the ping interval. Bathyswath is typically used in shallower depths, where the advantages of interferometers (wide swath angle and sidescan imagery) are more useful. Pitch steering is not worth the extra cost and complexity in these depths.

Although Bathyswath does not require roll and pitch beam steering to keep the swath parallel to the ship’s track and to keep the ping footprints at a regular spacing along-track, it is still essential that the depth measurements are corrected for ship motion with a good-quality motion sensor.