Bottom classification - practical notes

The quoted content (below) are observations about bottom classification that may useful for your equipment set up, recording or analysis. A change in text color indicates particularly useful information within general commentary.

List of data-conditioning steps often used before the extraction of seabed echo features
Frequency and the seabed
Absorption coefficient values
Depth and TVG
Maximum depth
Transmit pulse length, sample rate and Nyquist criterion
Single beam sonar
Single beam echosounder - classification qualification
Slope and beam width (Anderson)
Slope and beam width (Hamilton)
Slope and vessel noise/speed
Operational design of acoustic instrumentation (for Acoustic Seabed Classification (ASC))
Further observations (Hamilton)

Overview of generic data collection issues

p51-52 Anderson et al (2007).

"The following collection procedures need to be considered to ensure that relative or absolute backscatter data are obtained from measured seabed incidence angles.

Geo-referencing of signal returns and signal timing issues;
Angle of incidence to the seabed (e.g. transect direction, across or along slope);
Acoustic noise, e.g. vessel (engine, propeller, and other instruments), background (biological, environmental);
Electrical noise – instruments/machinery;
Propagation medium characteristics – sound speed and absorption coefficient for the region; sound speed at the transducer face for beamforming systems such as MBES;
Environmental gradients;
Aeration close to transducer – monitor when caused by rough weather, at high speed or poorly located transducer.
Calibration – precision/accuracy:
- Beam pattern characteristics (transmitters/receivers);
- Instrument stability (transmitter, receiver, transducer);
- Equivalent area insonified and absorption time variable gain (TVG) corrections;
- Instrument integration/timing issues;
- Timing issues for bathymetry MBES;
- Between-beam backscatter gains for MBES.
Instrument information:
- Frequency band;
- Pulse length/shape/amplitude stability;
- Transducer transmit and receiver characteristics and beamforming method;
- Transducer depth;
- Pitch, roll, heave, and yaw;
- Calibration settings;
- Software and hardware version recording;
- Preprocessing algorithms and pre-applied TVG gain.
Environmental conditions (wind, swell direction, tide level, etc.).

Many of these parameters can be monitored if the digitized echo from pulse transmission to end of seabed echo (second echo in some cases) is recorded along with instrument configuration and peripherals with good time recording. When using multibeam sounders, integration of the motion reference units with the geo-referencing and echosounder signals is important for accurate geo-location of measurements and for calculation of incidence angles. In addition to these generic issues for range-independent measurements, special attention is required to account for correct compensation of the equivalent insonified area and for exclusion of unwanted targets such as pelagic and demersal fish."

List of data-conditioning steps often used before the extraction of seabed echo features

p65 Anderson et al (2007). This list identifies particular data-conditioning steps (which may also be listed as a generic data collection issue) and often cites a referenced article.

"Table 5.2. List of data-conditioning steps often used before extraction of seabed echo features.

Single-beam echosounders are referred to as SBES. Multibeam echosounders are refereed to as MBES."

Description	Note
Echosounder calibration	Foote et al., 1987 Cochrane et al., 2003
TVG adjusted to actual sound speed and absorption profiles	Hutin et al., 2005
Normalization of maximum echo amplitude to 1 to remove the effect of propagation loss with range caused by spreading and absorption	Preston et al., 2004b
Correction for refraction of beam paths in stratified water columns	Mitchell, 1996
Alignment of the bottom echo envelope through the ping series relative to a given threshold in the bottom echo rise	Sternlicht and de Moustier, 2003a, 2003b
Consideration of the pitch, roll, and heave fluctuations of the transducer platform (see Section 3)	Mitchell, 1996
Removal of range-dependent effects in computing E1 features by mean of a linear regression	Siwabessy et al., 1999
Incident angle alignment by compensation for the range-dependent duration of the bottom echo leading edge by resampling (see Section 3)	Preston, 2003 Pouliquen, 2004
Rejection of side lobe bottom echoes in tilted SBES using time of flight and split-beam phase angle; rejection of data where ray bending due to sound speed profile was not negligible	Jackson et al., 1986a; Jackson and Briggs, 1992
Taking into account the seabed local slope in estimating the incident angle	Stewart et al., 1994
Alignment of bottom echoes, normalization of the amplitude to a maximum of 1, and summation by stacks of five pings	Preston et al., 2004b
Average of backscattering strength over groups of 20 and 100 successive pings or time periods, rejection of low SNR and significant angular oscillations of the platform	SBES Jackson and Briggs, 1992; Sternlicht and de Moustier 2003a, 2003b; Pouliquen and Lurton, 1994; Lurton and Pouliquen, 1994 MBES Hughes-Clarke et al., 1997
Average of backscattering strength over 0.05 nautical miles or 20 pings at 10 knots in estimating E1 and E2 features	SBES Siwabessy et al., 1999; Kloser et al., 2002
Backscattering strength moving average on five successive pings along survey track, removal of depth trends in the backscattering strength features using a regression against bottom depth	SBES Hutin et al., 2005
Multibeam backscattering strength correction, subtraction of the local trend estimated from a moving average along the survey track	MBES Kostylev et al., 2001, 2003
Removal of automatic modifiers applied by MBES equipment manufacturer to bottom backscattering data, such as beam-pointing tramlines, Lambert’s law, and center beam smoothing	MBES Kloser et al., 2002
Standardization of the features to the same numerical range and/or normalization

Frequency and the seabed

p10-11 Anderson et al (2007).

"... It has been determined that the directional character of the scattered sound depends on the acoustic impedance of the seabed material and the surficial topography at scales that are comparable with the wavelength of the sound. It also depends on any anisotropy in the near-surface seabed structure, particularly the constantly varying current and wind wave-generated sand ripples and waves. At high acoustic frequencies, the roughness and surface structures dominate the scattering responses. At low frequencies, the sound tends to penetrate into the seabed, and volume scattering from the heterogeneity tends to be more important. Of course, any bottom-associated flora or fauna also creates heterogeneity on the sediment’s surface and within the volume, and this usually increases variability in the observed scattering. ..."

Acoustic absorption values

p32 Hamilton (2001).

"... An echo sounder ping experiences both spherical spreading loss and absorption losses. The amount of absorption is dependent on the distance traveled through the water, and the temperature, salinity, and pressure of the water. For a temperature of 20°C, salinity of 35, and depth of 10m, the absorption at 10, 30, 50, and 100 kHz is 0.761, 5.19, 13.0, and 38.0 dB/km respectively (see Francois and Garrison 1982). ..."

Depth and TVG

p33-34 Anderson et al (2007).

"... Gain is the ratio of output-to-input voltage; linearity requires that the output is proportional to the input; and dynamic range is defined by the minimum and maximum input levels between which amplification is linear. The dynamic range must include the entire variation in possible echo amplitudes from all seabed types at a given depth as well as TVG that compensates for beam spreading and absorption. An exceptional dynamic range, such as 160 dB (Simrad, 1993), is required if echoes from a large depth interval (e.g. 5– 500 m) are to be recorded with the same receiver gain settings. Insufficient dynamic range leads to saturation, also called clipping, a severe distortion that flattens echo peaks that exceed a certain level.

The purpose of TVG is to maintain a constant sensitivity for the observation of a given target at any range. For a circular piston transducer with beam width θ, and a relatively flat, smooth, and horizontal seabed, the transmit pulse will insonify a circular area or pulse footprint. Relatively uniform insonification of the seabed occurs when the pulse footprint is smaller than the beam footprint. This is the case when:

Where c, τ and R give sound speed, pulse duration and range, respectively. The size of the pulse footprint is given by the area covered within a pulse length near the beam axis. As the range increases, the beam footprint will exceed the pulse footprint, insonification of the seabed will start from the acoustic axis, then travel outward as an annulus towards the beam boundary. Assuming uniform scattering from the seabed, the TVG will change from 20 to 30 logR + 2αR as the pulse footprint changes from greater to smaller than the beam footprint. Here α is the absorption coefficient for seawater. Thus, if range is the only variable, the TVG will change from 20 to 30 logR as range increases.

This simple model can only provide a guide as backscatter cross section changes with incident angle and as seabeds are seldom flat, horizontal surface scatters. In addition, the seabed echo is dominated by volume backscatter at sonar frequencies below about 10 kHz and by surface backscatter above about 100 kHz. A similar trend exists when changing from soft to hard seabeds. In most practical cases, equation 20 logR + 2αR provides a good approximation. In exceptional cases of short pulse length, long range, and smooth bottom, 30 logR + 2αR is used; 40 logR + 2αR is never appropriate for seabed backscatter. A detailed and insightful discussion of the subject is given by Lurton (2002). ..."

Maximum depth

p54 Hamilton (2001).

"...The sounder must be set to twice the maximum depth expected, so as to receive the second echo. ..."

Transmit pulse length, sample rate and Nyquist criterion

p36 Anderson et al (2007).

"... 3.3.1 Typical echoes and digital sample rate requirements

Figure 3.5 shows a typical echo envelope from a point target (a target that is small compared with the transmit pulse length). Transmit pulse length is 1 ms or approximately 75 cm. The sample rate is 7.5 kHz or 7.5 points per transmit pulse length, satisfying the Nyquist criterion with a comfortable margin. A sufficient number of data points are used to describe the smooth pulse shape. Little would be gained by using a higher sample rate because the natural bandwidth of the envelope is being sampled fully. Minimum sample rates of four samples per transmit pulse length (twice Nyquist) are now used in scientific echosounders. Echoes recorded with this, or a higher sample rate, are suitable for seabed classification. ..."

Single-beam sonar

p52-53 Anderson et al (2007).

"... 4.3.1 Single-beam sonar (SBES)

For good data quality, the instrument’s frequency and transducer beam pattern shape need to be carefully chosen to suit the depth and resolution of the study (see Section 3). For example, a narrow beam-width, low side lobe SBES operated around normal incidence is not as useful in seabed mapping as an instrument with higher side lobes. This is a result of the greater separation of seabed types at higher incident angles off normal incidence (Figure 4.4). Once an instrument configuration is chosen, it is important to understand the noise characteristics of the whole survey system. This could include monitoring the noise on the system caused by vessel speed and working out the optimal speed for low-noise recordings while also ensuring timely completion of a survey.

First echo interference problems:

Platform motion and aeration
Bottom detection stability and consistency
Fish schools and their multiple scatter that extend to the seabed
Seabed slope
Noise interference masking echoes originating off the acoustic axis (e.g. notably from side lobes)

Second echo interference issues:

More susceptible to acoustic and electrical noise
Vessel hull size, shape, and depth
Water surface roughness
Vessel speed

Figure 4.5 highlights some specific noise/interference problems that need to be removed or quantified before undertaking processing of the seabed signal. The effect of aeration under the transducer, caused by bubble sweep down from the vessel hull or by surface bubbles, significantly degrades backscatter intensity. Note the reduced and occasionally non-existent first seabed echo. The TVG compensation is TVG =2αwR +20logR. This TVG will not be suitable for all components of the seabed echo return for varying ranges and incidence angles (Equations (4.5) and (4.6); Lurton, 2002). Vessel noise changes with speed and, depending on the magnitude, can dramatically affect the intensity and shape of both the first and second seabed echo. As shown in Figure 4.2c, the seabed echo can change shape as a result of slope and not necessarily because of a change in the seabed properties. Fish schools in the water column may also cause interference; they may cause excessive attenuation of the seabed echo, the echosounder’s bottom detection may use the top of the fish school as the start of the seabed signal, or the multiple scattering within the school may influence the seabed echo. Figure 4.5 also shows the effect of interference from other echosounders. The second seabed echo is used in many commercial and visual seabed classification systems, and this echo is also susceptible to changes in sea surface roughness and vessel hull shape; because of the lower signal strength, it is more sensitive to background noise (Figure 4.5). ..."

Single-beam echosounder - classification qualification

p96 Anderson et al (2007).

"... A major disadvantage of the echosounder for ASC is the limited coverage of its narrow-beam footprint (see Section 6). Fishery acoustic surveys are typically conducted along a systematic grid design with widely spaced transects; therefore, bottom coverage is often less than 5%. ASC and seabed mapping surveys relying solely on echosounder operations generally conduct systematic grid surveys, and sometimes incorporate an adaptive star-like cruise track in selected shoal regions. Therefore, ASC results from SBES surveys must rely on contour interpolation to produce seabed classification maps because of the limited seabed coverage. ..."

Slope and beam width (Anderson)

p39 Anderson et al (2007).

"...Significant bottom slope has marked effects on echo amplitude and shape. Echoes from normal incidence can be rich in sediment information because the backscatter amplitude depends on near-nadir angles differently for different sediments. Away from nadir this is not true: amplitudes differ among sediments, but all dependencies on angle are similar. There are two situations: either there is a normal incidence reflection within the beam width or the slope is so steep that there is not. The largest backscatter amplitudes, by far, are at normal incidence. Thus, within the range of slopes small enough that there is a normal incidence reflection, the amplitudes do not vary much, and the echo shape remains rich in sediment information. With steeper slopes, amplitudes are much reduced, and echo shape loses its information content. Sediments in areas with slopes more than about a half beam width cannot be classified acoustically in the same manner as flatter regions (von Szalay and McConnaughey, 2002). It is often possible to separate them into a “slope” class, which may be adequate if it is known that only particular sediments, bedrock perhaps, are stable with these slopes. These observations apply also to ship roll and, less commonly, pitch. ..."

Slope and beam width (Hamilton)

p23 Hamilton (2001).

"... Voulgaris and Collins (1990) quote Jagodzinski (1960) as follows: “the second echo cannot be received unless the inclination of the bottom is smaller than the half beam width of the receiving oscillator. As a result the second echo may in some cases not be recorded, especially in the case of rocky bottoms or features such as sandwaves where the inclination changes rapidly on either side of the sand wave”. (Phil Chapple of DSTO notes that this may be quarter beam-width rather than half, depending on the definition of beamwidth). Outcrops will generally also have smaller scale roughness, which could further diminish the second echo (Burns et al 1989). This indicates E2 on the average will be lower over rough terrain than expected, causing the data envelope to bend in the positive E1 direction for high E1. Examples can be seen in Voulgaris and Collins (1990), causing class overlaps for sand ripples, sandwaves, sand/rocks/ripples, and rocks. E2 appears an unreliable classifier over rough topography. Sometimes however, this lowering of E2 may allow a particular bottom type to be distinguished from other bottom types, providing it does not overlap with other classes. ..."

Slope and vessel noise/speed

p54 Anderson et al (2007).

"... Von Szalay and McConnaughey (2002) investigated the effect of seabed slope and vessel speed on the QTC View single-beam classification system. For the system and vessel used, there was no measurable effect from vessel noise between 3 knots and 12 knots, but bottom slopes of more than 5° to 8° significantly changed the seabed classification being derived. Changes in seabed slope are similar to changes in transducer orientation, and their effect would be beam-pattern-specific as well as platform-motion-specific, highlighting that these commercial systems do not resolve seabed incidence angles. ..."

Operational design of acoustic instrumentation (for Acoustic Seabed Classification (ASC))

p95 Anderson et al (2007).

The system design, operational deployment, and its beam pattern determine the acoustic properties of transmitting and receiving sound waves. There is always a trade-off between resolution and operational range when selecting an acoustic system for ASC, especially its frequency. Transducers with smaller apertures tend to produce wider beam patterns at higher frequencies. Smaller apertures have improved shorter pulse length and wider bandwidth capability, resulting in better range resolution, which is particularly important for target detection and angular resolution. For example, a 12-kHz system (with pulse lengths of 0.4– 0.8 ms and bandwidths of 1–2 kHz) may have range resolutions of 30–60 cm, while a 300- kHz system (with pulse lengths of 0.02–0.03 ms and bandwidths of 40–60 kHz) would have improved resolution of 1–3 cm. Longer apertures produce a narrower beam pattern and its longer pulse duration transmits more power through the water column, increasing the operational range. Hence, the maximum operational range of a 12-kHz system (12 000 m) would be much greater than that of the 300-kHz system (200 m).