Acoustic sediment classification

Brief description


This data acquisition technique involves the use of the acoustic return signal from a bathymetric echosounder to make qualitative estimates of the seabed composition.

See NOAA's summary table Summary view of acoustic sediment classification technique (451 KB PDF)

NOTE: The content below is derived from Chapters 3.3 - 3.6 of Acoustic Techniques for Seabed Classification (2005) (11 MB PDF) by J D Penrose, P J W Siwabessy, A Gavrilov, I Parnum, L J Hamilton, A Bickers, B Brooke, D A Ryan and P Kennedy.


Back to top of page


Contents

The RoxAnn System | The QTC View System | The ECHOplus System | Comparisons of systems

The RoxAnn System

The RoxAnn system is manufactured by Marine Micro Systems of Aberdeen, Scotland, UK. The system uses the first and second acoustic bottom returns in order to perform bottom sediment classification. The first acoustic bottom return is reflected directly from the seabed and the second acoustic bottom return is reflected twice off of the seabed and once off of the sea surface (and the vessel hull). This method was earlier used by experienced fishers using regular echosounders (Chivers et al., 1990). The double bottom interaction of the second echo causes it to be strongly affected by the acoustic bottom hardness, with roughness effects becoming secondary. Figure 1 below shows the voltage trace of typical echosounder output for a single ping. Normally, this output is not available to the user because the system processes the data and stacks up a number of pings prior to display.

Shown on the left of Figure 1 is the tail end of the transmission, which may be ringing in the transducer, or reflections from close in structure or from entrained air bubbles beneath a hull-mounted transducer. In the middle of Figure 1, the first acoustic bottom return from the seabed is shown. In general, where epibenthos does not provide significant scattering, the first acoustic bottom return from the seabed is composed of at least three components: (1) the initial reflection from the seabed immediately beneath the transducer, (2) backscatter from an area surrounding the point on the seabed below the transducer and (3) possibly reflections from the sub-bottom (sub-bottom reverberation). Shown on the right of Figure 14 is the second acoustic bottom return from the seabed which has undergone an additional trip to and from the sea surface. The arrival time of the second acoustic bottom return is approximately twice that of the first acoustic bottom return. The second acoustic bottom return, however, is not just a scaled and delayed version of the first echo. It indeed carries further information (see below).

Typical echosounder output

Figure 1. Typical echosounder output.

The RoxAnn system uses echo-integration methodology to derive values for an electronically gated tail part of the first return echo (E1) and the whole of the first multiple return echo (E2). While E2 is primarily a function of the gross reflectivity of the sediment and therefore hardness, E1 is influenced by the small to meso-scale backscatter from the seabed and is used to describe the roughness of the bottom. By plotting E1 against E2 various acoustically different seabed types can be discriminated (Chivers et al., 1990; Heald and Pace, 1996). In principle E1 and E2 are related dominantly to acoustic roughness and hardness respectively, although each contains components of both.

The scattering geometries, along with parts of interest of bottom returns in the transducer response, are shown in Figure 2. An echosounder transducer transmits over a broad range of angles. If the seabed were perfectly smooth all the energy transmitted normally to the seabed would return to the transducer and energy at other angles would be reflected away. The seabed is not completely flat and a fraction of the energy transmitted at other angles is returned to the transducer. This backscatter mechanism is illustrated in Figure 2(a). The rougher the seabed, the more energy is scattered back to the transducer and the more energy appears in the tail of the echo, which also lengthens.

To avoid contamination of the backscattered energy with energy that has been directly reflected from below the transducer only the tail of the first echo is used in the analysis as shown by the dark curve in Figure 2(b).

The characteristics of the second echo are not as simple as the first and there are at least two rationales for the underlying physical mechanisms.

In the first rationale by Chivers et al., (1990), the dominant ray paths for the second echo undergo two reflections at the seabed and a single scattering at the sea surface as shown in Figure 2(c). The amount of reflection is related to the difference in acoustic impedance between seawater and the sea bottom. Therefore, the harder the seafloor the more energy is reflected forward to the surface and back to the transducer and the more energy appears in the dark curve in Figure 2(d).

Scattering geometry and parts of interest of 1 st and 2 nd bottom returns

Figure 2. Scattering geometry and parts of interest of 1 st and 2 nd bottom returns.

In the second rationale by Heald and Pace (1996), the important observation is that because of the presence of the sea surface, the configuration is really a bistatic one with the transmitter and the receivers vertically displaced by twice the water depth. The impact of this is that the receiving transducer is in the near field scattering zone of the seabed and the scattered energy is therefore driven by the reflection and hence hardness properties of the seabed.

Both rationales, however, support the concept that the harder the seabed the more energy appears in the dark curve in Figure 2(d). However, a very rough, hard surface can scatter so much energy that it appears acoustically softer than expected. In deep sea applications "Reflection from a very rough rocky bottom may appear to be less than that from a muddy sediment" (Brekhovskikh and Lysanov 1982; section 1.9). Similarly, losses due to roughness effects can cause sand with ripples, sandwaves, holes, and scours to appear to some acoustic measures to have the same properties as mud. Suitable averaging of echoes can overcome much of this variability, however acoustic bottom classification results are sometimes ambiguous, a point which must always be remembered.

The parameters E1 and E2 are plotted against each other, and different pairings of the two are expected to be related to different bottom types. The user must determine which parameter combinations are related to particular bottom types by taking bottom samples. The approach is purely empirical, but works very well for flatter bottoms (Hamilton et al., 1999). Some rationale is given for this approach by noting that smaller scale sediment roughness may be physically related to grainsize (McKinney and Anderson 1964). McKinney and Anderson (1964) expected backscatter to be a function of particle size and bottom relief, and proposed sediment particle size influenced the size of bottom relief. Burns et al. (1989) state this as "harder ground has a greater capability of exhibiting roughness" , effectively the rationale assumed for RoxAnn operation. However these relations are lost over rougher topographies (Hamilton et al., 1999). E2 and E1 are often referred to as "hardness" and "roughness" , implying measures of mechanical hardness and geometrical or physical roughness, but they are simply acoustic indices with some unknown relation to seabed conditions. E1 is a bottom backscatter index, and E2 is related to acoustic reflectivity.

Over rougher bottoms e.g. those with ripples, the energy lost to the second echo by backscatter can lead to lower than expected values of RoxAnn acoustical "hardness" for a particular sediment type, so that careful calibration against sediment samples is needed to obtain inferences of bottom type from the acoustics. See Hamilton et al. (1999) for more details. Depending on beam angle, unreliable E2 values are returned even for small slope values, a problem not widely appreciated. 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".

RoxAnn system configuration

Figure 3. RoxAnn system configuration.

The RoxAnn system consists of a head amplifier, which is connected across an existing echosounder transducer in parallel with the existing echosounder transmitter (Figure 3), and tuned to the transmitter frequency. The parallel receiver accepts the echo train from the head amplifier (Schlagintweit, 1993). The installation requires no extra hull fittings, simply room for the processing equipment. The required processing equipment includes an IBM compatible computer together with a monitor. Software, specifically written to handle RoxAnn data, must then be installed on the computer for processing analysis.

Papers describing the RoxAnn type system include Kloser et al. (2001b), Hamilton et al. (1999) Siwabessy et al. (1999, 2000), Bax et al. (1999), Sorensen et al. (1998), Greenstreet et al. (1997), Davies et al. (1997), Ryan et al. (1997), Magorrian et al. (1995), Schlagintweit (1993) and Voulgaris and Collins (1990).

Kloser et al. (2001b) and Voulgaris and Collins (1990) experienced a depth dependence in their RoxAnn data that could not be explained by differences in bottom type as determined from sediment and photographic samples (Figure 4). When the depth trend prior to data clipping was removed from the E1 and E2 results, the resulting data compared favourably with the data derived from a CSIRO developed processing algorithm (Kloser et al., 2001b). Although Hamilton et al. (1999) did not observe depth dependence in their RoxAnn data, they have warned that RoxAnn data might vary with depth and water column properties because water column absorption and scattering are not allowed for by the RoxAnn system. The depth dependence reported by Kloser has also been experienced in RoxAnn data gained in Antarctica (Pauly Ref 1, personal communication ). In addition, Hamilton et al. (1999) noticed that the RoxAnn system when sampling seabed with great slopes or depth changes could have problems detecting the second echo and part of it could be included in the first echo.

Despite the claim by the manufacturer that the RoxAnn system is not dependent on vessel speed, Hamilton et al. (1999) found that E2 was inversely related to vessel speed. They also found that E1 sometimes experienced change in synchronisation with E2. Similarly, Schlagintweit (1993) observed a consistent seabed classification by the RoxAnn system only at constant speed. He suggested that this might be related to changes in aeration and engine noise. Hamilton et al. (1999) also found that the occasional engine noise as the vessel was held on station or manoeuvred gave an unexpected result i.e. the RoxAnn system did not function well when the vessel was essentially stationary where it was expected to be more reliable. Wilding et al. (2003) have also reported RoxAnn classification results that vary with vessel speed, amongst other factors.

The RoxAnn manufacturer recommends the use of RoxAnn squares, introduced by Burns et al. (1989), to assess seabed classifications and encourages users to adopt it. A typical scatterplot of E2 versus E1 together with the RoxAnn squares is shown in Figure 5. Each of the squares represents one particular seabed type and is determined arbitrarily based upon ground truth. A number of problems in using the RoxAnn squares, however, have been notified by Voulgaris and Collins (1990), Greenstreet et al. (1997) and Hamilton et al. (1999).

While Greenstreet et al. (1997) observed inconsistency in the allocation system of the RoxAnn squares in boundaries between different seabed types, Voulgaris and Collins (1990), Greenstreet et al. (1997) and Hamilton et al. (1999) found that E1 and E2 parameters were not independent but linearly related such that data form an elongated roughly elliptical envelope inclined to E1 and E2 axes (Hamilton et al., 1999). Since E1 and E2 are not orthogonal in RoxAnn space, Hamilton et al. (1999) argued that the RoxAnn squares cut across the data trend. In addition, Schlagintweit (1993) believed that an unsupervised classification method would be the best alternative, i.e., let the system select the natural groupings and then look at ground truthing.

Scatterplot of RoxAnn E1 (roughness) and E2 (hardness) indices with depth collected in the South East 
Fisheries region

(a)

Scatterplot of RoxAnn E1 (roughness) and E2 (hardness) indices with depth collected in the South East 
Fisheries region

(b)

Figure 4. Scatterplot of RoxAnn E1 (roughness) and E2 (hardness) indices with depth collected in the South East Fisheries region (After Kloser et al., 2001b).

Typical plot of E2 versus E1 together with the RoxAnn Squares, each of which represents one particular 
seabed type

Figure 5. Typical plot of E2 versus E1 together with the RoxAnn Squares, each of which represents one particular seabed type (After Chivers et al., 1990).

Kloser et al. (2001b) and Schlagintweit (1993) observed the dependency of seabed classification on acoustic frequency. For the same seabed feature, different roughness indices were observed for two different frequencies they used. Schlagintweit (1993) found that the differences arising from 40 and 208 kHz data were due to the different seabed penetration depths of these frequencies on various sea floor types. That is, the frequency dependent penetration factor into the sea floor depended on the local sea floor itself. Schlagintweit (1993) felt that the frequency should be chosen according to the application.

At low frequencies where acoustic wavelengths are larger than the scale of seabed roughness, the seabed surface will appear acoustically smooth. In this case, seabed reflection will dominate seabed scattering. On the other hand at high frequencies such that acoustic wavelengths are smaller than the scale of seabed roughness scattering can dominate the returning signal and the seabed may be considered to be acoustically rough. In addition, as the seabed absorbs less energy at low frequency than it does at high frequency, layers underneath the seabed surface might be acoustically visible. As such, seabed backscatter and sub bottom reflection at low frequency may arrive at the same time from different angles.

Hamilton et al. (1999) and Kloser et al. (2001b) noticed a bias due to slope or a sudden rise or drop of the seabed in their RoxAnn data. High slopes or sudden rises or drops of the seabed normally produce long tails in the first bottom echo which thus provide large acoustic roughness index estimates. For a sudden rise or drop of the seabed, this bias can be easily noticed in the echograms. Similarly, this bias can be picked up easily in the echograms if the vessel steams normal to the high slopes. If on the other the vessel is transecting parallel to the slope, this bias can only be interpreted once seabed types are plotted on the corresponding bathymetric map. This bias however can be used as a unique indication to identify such seabed types or areas (Greenstreet et al., 1997; Hamilton et al., 1999; Kloser et al., 2001b). In addition, Kloser et al. (2001b) found that a narrow beamwidth was more sensitive to slopes than a wider one.

Results from North West Shelf and Southeast Fisheries Regions

CSIRO Marine Research has developed its own system using the first and second echo technique. Using a SIMRAD EK 500 scientific echosounder operating three frequencies (12, 38 and 120 kHz), acoustic volume reverberation (s V) data are continuously logged using ECHO, a software package developed by CSIRO Marine Research (Waring et al., 1994; Kloser et al., 1998). The quality control and the derivation of E1 and E2 indices are conducted by using the ECHO software as well. The ECHO software provides several algorithms to derive E1 and E2 indices including a constant depth and a constant angular algorithm. This system has been tested in at least two Australian areas, namely the North West Shelf and the Southeast Fishery regions.

Making use the system developed by CSIRO just mentioned, Siwabessy et al. (1999, 2000) developed a procedure for seabed classification using a multi-frequency technique. The procedure of seabed classification involves multivariate analysis, in particular Principal Component Analysis and Cluster Analysis (the iterative relocation (k-means) technique). This procedure has been tested in the two regions studied.

It is assumed that the linearly increasing trend of E1 with depth shown in much of Figure 6 is likely to be an artefact of beam geometry and choice, and reliability, of TVG. Despite the fact that there is no law of nature which requires that must equal to 0, as his working hypothesis, Siwabessy (2001) assumed that on average where E1 is the acoustic roughness index and R 0 is the depth. To implement the above assumption, he used the constant angular integration interval algorithm within ECHO software developed by CSIRO Marine Research (Kloser, et al., 1998). This algorithm ensures that the proportion of the tail sector being integrated is similar regardless of depth.

In his study, Siwabessy developed an alternative approach to the use of echosounder returns for bottom classification. The approach used, while similar to that used in the commercial RoxAnn system, involves several further developments. In grouping bottom types, multivariate analysis (Principal Component Analysis and Cluster Analysis) was used instead of the RoxAnn squares mentioned previously. In addition, the approach adopted allowed for quality control over acoustic data before further analysis was undertaken, an issue of considerable importance in handling many real data sets. Three different frequencies, i.e. 12, 38 and 120 kHz, were operated. Principal Component Analysis was used in his study to reduce the dimensionality of the roughness indices from 3 to 1. The same was done for the hardness indices. The k-means technique was applied to cluster the resulting E1-E2 pairs formed for each set of six parameters to see if this would produce separable seabed types. This produced four separable seabed types, namely soft-smooth, soft-rough, hard-smooth and hard-rough seabeds (Figure 6). Principal Component Analysis was also used to reduce the dimensionality of the area backscattering coefficient s A, widely accepted in fisheries acoustics as a relative measure of biomass of benthic biota, here assumed to be mobile.

There are some possible drawbacks to this technique, in that what is being done may merely be density clustering of a continuous data cloud (see Figure 9) without physical meaning. A seabed type which is sampled less often than another type may be lost (absorbed into another type), even if it is dramatically different from other types, unless it appears as extreme outliers. This seems to be an unrealised problem in acoustic bottom classification. Greenstreet et al. (1997) noted this problem for their bottom grab samples, these being evenly spaced throughout their study area, rather than there being the same number of samples for each bottom type. A similar problem exists with acoustic data (Hamilton 2001).

The bottom classification described above appeared to be robust in that, where independent ground truthing was found, acoustic classification was consistent (Figure 7 and Figure 8). When investigating the relationship between the derived bottom type and acoustically assessed total biomass of benthic mobile biota, no trend linking the two parameters, however, appears. Figure 9 and Figure 10, however, reveal some patterns between seabed types and derived fish communities formed from key species according to Sainsbury (1991) for the NWS region and to Bax et al. (1999) for the South East Fisheries (SEF) region. In addition, using the hierarchical agglomerative technique applied to a set of variables containing the:

  • average and centroids of first principal component of roughness and hardness indices associated with the four seabed types;
  • average first principal component of the area backscattering coefficient (s A);
  • species composition of fish community;

from which two main groups of acoustic population were observed in the North West Shelf (NWS) study area and three groups were observed in the SEF study area.

The two main groups of acoustic population in the NWS study area and the three main groups of acoustic population in the SEF study area were associated with the derived seabed types and fish communities of the key species.

Soft-Smooth Soft-Rough Soft-Smooth Soft-Rough

Soft-Smooth Soft-Rough

Hard-Smooth Hard-Rough Hard-Smooth Hard-Rough

Hard-Smooth Hard-Rough

Soft-Smooth Soft-Rough Soft-Smooth Soft-Rough

Soft-Smooth Soft-Rough

Hard-Smooth Hard-Rough Hard-Smooth Hard-Rough

Hard-Smooth Hard-Rough

Figure 6. Representative examples of seabed images taken by a 35 mm Photosea 1000 camera system in the SEF area - upper 4 images (after Kloser et al., 2001b) and the NWS study area – lower 4 images (after Siwabessy et al., 1999).

Map of acoustically derived seabed types along the track, bathymetry and coastline for the SEF study 
area.

Figure 7. Map of acoustically derived seabed types along the track, bathymetry and coastline for the SEF study area. · = soft-smooth (SoSm); · = soft-rough (SoRg); · = hard-smooth (HdSm); · = hard-rough (HdRg).

Map of acoustically derived seabed types along the track, bathymetry, coastline and benthic habitat types (pie charts)

Figure 8. Map of acoustically derived seabed types along the track, bathymetry, coastline and benthic habitat types (pie charts) from Althaus et al. (in prep) for the NWS study area. · = soft-smooth (SoSm) = habitat 4 (H4) ; · = soft-rough (SoRg) = habitat 5 (H5) ; · = hard-smooth (HdSm) = habitat 3 (H3); · = hard-rough (HdRg) = habitats 1 & 2 (H12).

Pictorial plot of four derived fish communities given as pie charts overlaid into four seabed types 
given as a Cartesian plot of PC1_E1 versus PC1_E2 for the SEF study area.

Figure 9. Pictorial plot of four derived fish communities given as pie charts overlaid into four seabed types given as a Cartesian plot of PC1_E1 versus PC1_E2 for the SEF study area. Colour definition for seabed types remains the same with that given previously. n , n , n and n are communities 1, 2, 3 and 4, respectively.

Pictorial plot of four derived fish communities given as pie charts overlaid into four seabed types 
given as a Cartesian plot of PC1_E2 versus PC1_E1 for the NWS study area.

Figure 10. Pictorial plot of four derived fish communities given as pie charts overlaid into four seabed types given as a Cartesian plot of PC1_E2 versus PC1_E1 for the NWS study area. Colour definition for seabed types remains the same with that given previously. n , n , n and n are communities 1, 2, 3 and 4, respectively.

Back to top of page


The QTC View System

The QTC View system is manufactured and distributed by Quester Tangent Corporation of Sidney, BC, Canada. Like the RoxAnn system, the QTC View system uses the existing echosounder transducer; however, the QTC View system uses only the first echo. The QTC View system operates in the following manner:

First, both the transmitted echosounder signal and return signals are captured and digitised by the QTC View system as a continuous time series. Second, the seabed echo is located (bottom pick), and an averaged echo from several consecutive returns is computed. This reduces the effects of acoustic and environmental variability. Next, the effects of the water column and beam spreading are removed such that the remaining waveform represents the seabed and the immediate subsurface (Collins et al., 1996). Quester Tangent's echo shape analysis works on the principle that different seabeds result in characteristic waveforms. Through principal component analysis, complex echo shapes are reduced into common characteristics. Each waveform is processed by a series of algorithms which characterise it by a cited 166 shape parameters (Collins et al., 1996). A covariance matrix of dimension 166 x 166 is produced and the eigenvectors and eigenvalues are calculated. In general, three of the 166 eigenvectors account for more than 95 per cent of the covariance found in all the waveforms. The 166 (full-feature) elements of the original eigenvector are reduced to three elements ("Q values" ). These reduced feature elements will cluster around locations in reduced feature space, which may correspond to a seabed type.

The QTC View system was designed to operate in either the supervised or unsupervised classification mode. In the unsupervised classification mode, class assignment is based solely on the statistical nature of the data (the distribution of data) without knowledge of the spread of data of any seabed types. This can only be done in post processing. In the supervised classification mode, however, the classification is based on the knowledge of the spread of data of each seabed type revealed from ground truthing encountered in a particular survey. Classification then becomes a function of bottom sampling.

QTC View system configuration.

Figure 11. QTC View system configuration.

The QTC View system is comprised of a head amplifier and PC (see Figure 11). The head amplifier is connected in parallel across the existing transducer and to the PC via an RS232 cable. The PC also accepts GPS data in NMEA-0183 standard GGA or GGL format for georeferencing of data (Collins et al., 1996). The PC displays three windows: one for the reduced vector space, one for the track plot and classification and the third for seabed profile and classification.

Papers on the performance of the QTC View system include Hamilton et al. (1999), Bornhold et al. (1999), Collins and McConnaughey (1998), Galloway and Collins (1998), Collins and Rhynas (1998), Tsemahman et al. (1997), Collins and Lacroix (1997). The first paper is the only one without authors from Quester Tangent Corporation.

Tsemahman et al. (1997) used the QTC View system in selected areas near Vancouver Island. They found that the system was able to discriminate between four different acoustic classes. After a calibration, QTC View was found to agree with each ground truthed area and showed good transition from seabed type to seabed type (Tsemahman et al., 1997). Preston et al. (2000) recently have described the use of a modified QTC system in very shallow waters, with depths as small as one metre.

Unlike the RoxAnn system, the depth dependency of the QTC View system has not been reported in any papers on the performance of the system. Hamilton et al. (1999), nonetheless, have warned that QTC data might vary with depth and water column properties because water column absorption and scattering are not allowed for by the system. In addition, they observed that the QTC data were not obviously dependent on vessel speed. They also found that classification was consistent and did not change regardless of speed or even when the vessel was stationary or manoeuvred. They, however, noticed a bias due to slope or a sudden rise or drop of the seabed in their QTC data. This bias however can be used as a unique indication to identify such seabed types or areas (Hamilton et al., 1999).

Results from Wallis Lake , NSW

This section summaries results of acoustic mapping of estuarine benthic habitats in Wallis Lake, NSW, fully reported in Ryan et al. (2004). A collaborative field trial of the Quester-Tangent View Series 5 single beam acoustic benthic mapping system was conducted in Wallis Lake by Geoscience Australia and Quester Tangent Corporation, with the assistance of NSW Department of Infrastructure, Planning and Natural Resources, Great Lakes Council and Curtin University. The survey, in June 2002, involved acquisition of the acoustic backscatter data from the northern channels and basins of Wallis Lake. Quester-Tangent software (IMPACT v3) was used to classify acoustic signals (echograms) that returned from the lake bottom into statistically different acoustic classes. The classification was based on characteristics of the echograms, which were further reduced using a principal component analysis (PCA) algorithm. An algorithm then divided the data into logical clusters, each cluster representing a unique acoustic class. Six acoustically different substrate types were identified in the Wallis Lake survey area (Figure 12 and Figure 13).

In conjunction with the acoustic survey, ground-truthing was undertaken to identify the sedimentological and biological features of the lake floor that influenced the shape of the return echograms. Samples of lake-bottom sediment were collected from over 100 sites (located by DGPS) by snorkel divers, and pole-coring. For each sample, laboratory measurements were made of grain size, wet bulk density, total organic carbon, CaCO 3 content. Mass of coarse fraction (mainly shell) material was recorded for samples taken in December. At most sampling sites, diver observations provided information about the morphology of the lake bed, such as density of animal burrows, and seagrass species and coverage; and at several sites photographs of the substrate were taken. A scheme of simple ranked indices was developed to reflect the relative density of each of these site features.

Clusters identified in the Wallis Lake acoustic dataset, based upon a Principal Components Analysis

Figure 12. Clusters identified in the Wallis Lake acoustic dataset, based upon a Principal Components Analysis (using QTC Impact software). The ellipsoidal shapes (representing 95% confidence limits) include data points that make up the six acoustic classes (Class 1 = Red, Class 2 = Green, Class 3 = Blue, Class 4 = Cyan, Class 5 = Yellow, Class 6 = Pink).

Spatial representation of the acoustic data, coded for acoustic class, and plotted on a digital 
aerial photograph.

Figure13. Spatial representation of the acoustic data, coded for acoustic class, and plotted on a digital aerial photograph. Aerial photographs courtesy of Land and Property Information, NSW.

 Cluster analysis dendrogram of the sediment data. Ground-truthing sample site numbers are shown 
on the x-axis.

Figure 14. Cluster analysis dendrogram of the sediment data. Ground-truthing sample site numbers are shown on the x-axis. The four main cluster subgroups have been labelled Groups A-D.

Multi-dimensional scaling plot (MDS) of the sediment data.

Figure 15. Multi-dimensional scaling plot (MDS) of the sediment data. The four main cluster subgroups, identified in Figure 17, have been labelled Groups A-D.

Statistical cluster analysis and multi-dimensional scaling were utilised to identify any physical similarities between groups of ground-truthing sites (Figure 14 and Figure 15). The analysis revealed four distinct and mappable substrate types in the study area. These comprised dense, shelly sands, low density mud without shells, low density shelly mud, and poorly sorted muddy sand. Very little relationship between seagrass distribution and sediment type was detected. Cluster analyses and multi-dimensional scaling were also used to indicate the degree of association between acoustic classes and both sediment parameters and observed biophysical features (Figure 16 and Figure 17). Groundtruthing sample sites were coded for acoustic class, based on their proximity to the acoustic survey track lines. The analysis revealed that, based on the parameters measured, not all of the six acoustic classes were uniquely linked to distinct sedimentological facies, indicating that factors other than the sediment parameters influence the return acoustic signal. However, the well recognised estuarine sedimentary environments, such as marine sands and muddy basin sediments, are clearly delineated.

Cluster analysis dendrogram of the sediment data.

Figure 16. Cluster analysis dendrogram of the sediment data. Each sample is coded for acoustic class (Figure 18) on the x-axis. The four main cluster subgroups have been labelled Groups A-D.

Multi-dimensional scaling plot (MDS) of the sediment data, also coded for acoustic class.

Figure 17. Multi-dimensional scaling plot (MDS) of the sediment data, also coded for acoustic class. The four main cluster subgroups, identified in Figure 20, have been circled (Groups A-D).

Much of the seagrass-dominated environments are limited to the shallow-water margins of the basin and channel areas. The wide central basin areas, such as Pipers Bay and the area west of Wallis Island, were dominated by very fine sediments with varying concentrations of shell material, and densely populated by various infaunal organisms. The channel areas, which form intricate networks in northern Wallis Lake, were the deepest areas in the lake, and comprised relatively dense sandy muds and muddy sands, and also featured a high concentration of shell material. The marine tidal delta was also a distinct and mappable habitat, with clean well sorted sand, and dense seagrass beds.

The Quester-Tangent View Series 5 system demonstrated the ability to rapidly survey a relatively large and diverse area, to produce a coherent map of spatially homogenous acoustic classes. Although not all of the acoustic substrate classes could be identified in sedimentological terms, useful linkages were made between the acoustic classes and known estuarine sedimentary environments, illustrating that the Quester-Tangent acoustic mapping system is a useful tool for coastal management and research. The spatial interpretation of the Wallis Lake Quester-Tangent data represents the first quantification of non-seagrass habitats in the deeper areas of the lake, and provides a useful indication of benthic habitat diversity and abundance. For future studies, a more quantitative measure of faunal burrow size and density, and also other sedimentary bedforms, is recommended. Other unmeasured features of the lake bed may also have influenced the echograms, for example gas bubbles produced by plants or evolved from within surficial organic-rich muds.

Back to top of page


The ECHOplus System

The ECHOplus (Bates and Whitehead, 2001a, 2001b) system appears to utilise a similar approach to that of the RoxAnn system, but is digital not analogue. It is noteworthy that the authors report a similar approach to coping with the apparent depth dependence exhibited by data from a number of RoxAnn implementations as had been adopted by CSIRO and Siwabessy (2001).

Principles of operation- The ECHOplus System

A detailed description of the ECHOplus system is given by SEA personnel (Bates and Whitehead, 2001a; Bates and Whitehead, 2001b), below is a summary of this work.

ECHOplus is a digital version of RoxAnn (described in Section 3) produced by SEA (Advanced Products) Ltd. (Aberdeen, Scotland). Like RoxAnn, ECHOplus does not record the complete waveform, but instead measures the parameters E1 and E2 to characterise the acoustic roughness and hardness of the seafloor (see Section 3 for definitions). An advantage of ECHOplus over other acoustic ground discrimination systems (AGDS) available (e.g. RoxAnn, QTC View) is its capability of undertaking simultaneous dual frequency analysis. Moreover, the ECHOplus system attempts to compensate for common problems in single-beam surveys (identified in earlier sections), namely:

  • Depth;
  • Frequency;
  • Power level; and,
  • Pulse length.

Depth

If no compensating action were taken the ECHOplus outputs would depend on depth as both the echo return levels and the total absorption losses are depth dependent. In order to compensate for the former a time varying gain is applied to the digitised voltages within the system. Because of the underlying geometry of the system, in particular the constant beam width of the transducer, this gain factor is a linear function of depth.

Frequency

Unlike other systems (e.g. QTC), ECHOplus automatically detects and operates with any echosounder within its operating range. In addition, ECHOplus has the facility to input and process two frequencies simultaneously. There is no requirement for user intervention to tune the system to a particular frequency. This is achieved by using wideband low loss front-end analogue hardware together with frequency estimation software. The centre frequency of the transmitter is estimated directly from the pulse waveform and used to control the analogue data acquisition and the digital base-banding when listening to receive echoes.

Power level

To compensate for power level, the pulse amplitude is measured by the system on every transmission and the estimated amplitude is used to scale the outputs accordingly. ECHOplus measures the electrical power at the output of the echosounder transmitter. This does not automatically relate to the acoustic power delivered by the transducer into the water column, as there are a number of sources of variation that cannot be compensated for including losses within the transducer cable, efficiency changes between different transducers and efficiency changes for the same transducer as a function of age. Apart from the final one, however, these will be constant for a particular installation.

Pulse length

The pulse length is measured and used to adjust the outputs accordingly. There are separate and independent corrections for roughness (E1) and hardness (E2) values. The measured pulse length is also used to tune the filtering process so that the signal to noise ratio and hence overall performance and reliability are optimised.

A potential drawback for scientific use is that the various compensations made by the ECHOplus system may unwittingly affect the parameters being measured. There is also no justification for assuming linearity between acoustic parameters measured with different frequencies or echosounder characteristics (Hamilton, 2001). In addition, there are other factors that may influence the output levels:

  • Beam pattern and direction,
  • Feed losses and impedance,
  • Transducer transmit and receive efficiencies,
  • Surface scatter.

ECHOplus, therefore, provides a reference calibration facility which can be set by the user post installation. These factors can be compensated for either by using this reference calibration facility on ECHOplus or via a similar function in the analysis software, if it has this capability. However, there is no work available that assesses this calibration facility.

Examples of using ECHOplus

ECHOplus was first trialled by SEA personnel (Bates and Whitehead 2001a; Bates and Whitehead, 2001b) in the sediment dominated Hopavågen Bay , Norway , from which it was concluded: "The results exhibit excellent correlation between acoustic bottom classes and ground truth data" . Although the hardness and roughness values obtained from the surveyed area indicated that ECHOplus can possibly distinguish between different grades of sediment, the study lacked a rigorous accuracy assessment. A better appraisal of the discrimination ability of ECHOplus can be seen in Riegl et al. (in press), which compares macroalgae surveys in the Indian River Lagoon in Florida using both QTC View and ECHOplus. Both systems were able to distinguish between seagrass-algae-bare substratum. A three-class confusion matrix based on QTC View surveys, suggested a total accuracy of around 60%. However, both systems showed high confusion when trying to discriminate two algae classes (sparse versus dense). When comparing the two systems, Riegl et al. (in press) conclude that: "ECHOplus appeared to give a somewhat clearer distinction, which could, however, largely be due to the way classes were assigned as binned intervals of the digital numbers provided by the ECHOplus , while in the QTC View system classes are assigned based on the position of signals in pseudo- 3-dimensional space after PCA" . To support this observation, the study would have benefited from producing a confusion matrix for the ECHOplus results as well.

Based on trials in 2000 (Anonymous, 2001a), the Archaeological Diving Unit based at the University of St. Andrews has purchased ECHOplus as a means of monitoring underwater sites to complement other methods including sidescan. An initial study by Lawrence and Bates (2001) showed statistically distinct clusters of E1-E2 data that were associated with the exposed archaeological material, which distinguished the wreck site from the surrounding seabed. However, it might be expected that a ground discrimination system can differentiate between artifacts and seabed, since their acoustic properties are likely to be very different.

Results from Lough Hyne, Ireland

This section summarises results of mapping sublittoral habitats of an Irish sea lough using ECHOplus, fully reported in Parnum (2003). The aim of this summary is to indicate the ability ECHOplus to distinguish between visually different habitats and the comparison of using subjective box classification with image processing techniques to classify the E1/E2 data.

Lough Hyne

Located on the south coast of Ireland, Lough Hyne Marine Reserve (Figure 18) is of great conservation, scientific and historic importance. It has been extensively studied and is known to have a very high species diversity and species richness for such a small area (Kitching, 1990). As part of a study to investigate the role acoustic techniques can play in mapping and monitoring the sublittoral habitats of Lough Hyne, an ECHOplus survey was performed in August 2003.

Lough Hyne Marine Reserve located on the south coast of Ireland

Lough Hyne Marine Reserve located on the south coast of Ireland

Figure 18. Lough Hyne Marine Reserve located on the south coast of Ireland (courtesy of John Rowlands, University of Wales, Bangor).

Lough Hyne is a deep landlocked bay or ‘marine lake’ joined by a narrow channel (Barloge Creek) to the sea. The actual embayment area (shown in Figure 18) is around 50 ha; Castle Island separates the two shallower (North and South) basins (20m), with the deepest area in the Western trough (50m). Lough Hyne's sublittoral habitats were first classified by Kitching et al. (1976) through a series of diver observations and grab sampling throughout the Lough. There were a total of 6 biotopes identified by Kitching et al. (1976), their descriptions, depth ranges, and maps of the point observations and subsequent contoured map of the biotope distribution can be found in Figure 22. From the work done by Kitching et al. (1976), it is evident that Lough Hyne is dominated by soft sediments. According to the authors, these soft sediment biotopes are depth-dependant, starting at just below the surface (3m) with a zone of the red algae Rhodochorton floridulum covering soft sediment down to 17m. Below the photoreceptive region is a characteristic mud-burrow zone between 17 and 25m. Below 25m the soft sediment becomes distinctly finer (clay/silt), and also contains large numbers of spionid tubes projecting from the surface. Soft sediment was only found to be replaced by coarse or hard substrate in areas of fast moving current such as the Rapids and Whirlpool area, or on underwater cliffs or in very shallow water. Although Lough Hyne is relatively small and despite the authors’ extensive experience surveying the lough over a number of years, it is hard to give great confidence to the actual position of the point observations as there was no form of geo-referencing in the data. Furthermore, although depth is an important factor on the distribution of biotopes, in reality patchiness of the ground and occurrence of intermediate biotopes due to overlapping, means the contoured map generated by Kitching et al. (1976) is unlikely to be precise. Therefore, the results from Kitching et al. (1976) have to be taken as a broad distribution of biotopes present at the time of study.

Biotope distribution

Biotope Description Depth range
Shallow inshore Boulder slopes, either bare or covered in green algae such as Enteromorpha spp. And Stilophora rhizodes. < 5m
Rhodochorton A blanket of the filamentous red algae Rhodochorton floridulum covering soft sediment (mud). 3-17m
Mud burrow A zone of bare mud with many burrows, believed to be the work of various organisms, for example: the decapods Nephrops norvegicus and Calocaris macandreae. 17-25m
Spionid Large numbers of spionid tubes protruded from the very fine sediment (clay/silt) found in this region. >25m
Whirlpool complex Course sediment found near Whirlpool cliff resulting from transport from the nearby rapids. <25m
Rock boulder Solid bedrock <20m

Figure 19. Biotope distribution of Lough Hyne as (a) observed, and as (b) continuous coverage generated by Kitching et al. (1976). See table for biotope descriptions and depth ranges observed by Kitching et al. (1976) in the lough.

Other ground truth information available includes diver transects performed across the South Basin by Thrush and Townsend (1986), which confirms the distinctly different habitats found in the middle of the South Basin (soft sediment) compared to near to the entrance to the Rapids and Whirlpool cliff (coarse gravel-like sediment). Also, an alternative substratum map of Lough Hyne was presented by Wilkins and Myers (1990) as part of their study on the distribution of gobies in the lough. Again this was constructed through diver observations, however, there was no raw point data to assess the accuracy and confidence that can be given to the map. Another source of information, for this study, on the distribution of benthic habitats in Lough Hyne came from diver observations by Turner Ref 2 (personal communication).

Classification of sublittoral habitats

Marine biotopes for the northeast Atlantic have been described and classified as part of the Joint Nature Conservation Committee (JNCC) Marine Nature Conservation Review (MNCR) by Connor et al. (1997). This classification system has been adopted by both the British and Irish conservation agencies. Therefore, it was appropriate to adopt this system to classify the different sublittoral habitats found in Lough Hyne. From the diver observations and grab sample records in previous studies (as mentioned above), Lough Hyne's sublittoral habitats were separated into 5 'Habitat Complexes' described by Conner et al. (1997). These habitat complexes were fine sediment (mud), mixed sediment (predominately shell and mud), macrophyte dominated sediment (e.g. Rhodochorton floridulum on mud), coarse sediment (i.e. gravel) and rock. In addition, a further habitat complex was added to accommodate the observations of Kitching et al. (1976) and Turner (personal communication) that in the centre of the Western Trough below 25m there is distinct transition from fine cohesive sediment (mud) to much finer fluid-like sediment. This fluidised fine sediment, which is commonly found in some ports, was given the code SFluid. Table 1 details these 6 habitat complexes with their biotope code and depth ranges observed by Kitching et al. (1976).

Table 1. Habitat complexes observed in Lough Hyne as described by the JNCC Biotope Marine Classification System (Conner et al., 1997), with their assigned habitat/biotope code and the depth ranges in which they are generally found according to Kitching et al. (1976). *Fluidised fine sediment is an additional habitat complex not found in the JNCC classification system.

Substrate Sublittoral (S) Habitat/Biotope code Depth Range (m)
Rock (including boulders) SR <25
Coarse sediment (gravel) SCs <25
Mixed sediment (predominantly a shell-mud complex) SMx <17
Macrophyte dominated sediment (predominantly Rhodochorton floridulum on mud) SMp <17
Soft sediment (cohesive mud) SMu 17-25
Fluidised fine sediment* (non-cohesive clay)* SFluid

>25

ECHOplus data acquisition & processing

The ECHOplus system utilised a Raytheon DE-719B echosounder operating at 210kHz, with a sampling rate of 9Hz and 8° beam width. The ECHOplus survey took place on the 3 rd August 2003. As the area was well known (by the navigator) the track spacing and navigation employed aimed to record the extent of the different biotopes. A track plot of the ECHOplus survey can be found in Figure 20 (only data used in classification shown). There were a limited number of useable acoustic returns obtained from deeper areas (>25m), in particular, in and around the Western Trough. This was due to a combination of the high frequency of the sounder, the limited sounder power available, the presence of steep slopes leading into deeper areas, and the low reflection coefficient of the very soft bottom sediments found in the deeper areas. Also, strong water currents coming in from the rapids generating bubbles by Whirlpool Cliff prevented the ECHOplus system working effectively. Similar problems were encountered from a different sounder in a separate single-beam bathymetric survey performed the previous day.

A track plot of the ECHOplus survey (minus the anomalies removed) performed at Lough Hyne during 
August 2003.

Figure 20. A track plot of the ECHOplus survey (minus the anomalies removed) performed at Lough Hyne during August 2003.

After the survey, ECHOplus data was imported into a spreadsheet, where all erroneous data possibly due to: hardware failure, bad bottom-pick, steep slope or strong water current were removed. These were usually indicated by incorrect depth values (e.g. 500m), or where large deviations (>10%) occurred outside the general run of observations. Also, where the seabed had not been detected a default readings for the depth, E1 or E2 were given, and these were removed. The "clean" data was corrected to standard datum using a tidal correction over the same time period from a tidal gauge deployed in the lough. Then data was converted into Easting and Northing (UTM WGS84 northern region 29). The clean and edited E1 and E2 data were independently interpolated using the Kriging algorithm in Surfer, both with a 2m grid size, and using a blanking file added to distinguish Lough Hyne's coastline, shown in Figure 24.

Classification of ECHOplus data

The objective of the classification of sublittoral habitats of Lough Hyne was to provide a broad scale distribution of habitat complexes found in the lough. The lack of geo-reference ground-truth points limited any attempt to carry out supervised classification of ECHOplus data. Therefore, unsupervised classification was used. The individual E1 and E2 distribution maps (Figure 21) indicated that neither ‘hardness’ nor 'roughness' independently would be suitable for classifying the seabed. Thus, it was optimal to use both E1 and E2 together to classify Lough Hyne's sublittoral habitats. The two methods used were:

  1. Subjective box classification
  2. Cluster analysis of False Colour Imagery (FCI)

Results of the ECHOplus survey of Lough Hyne in August 2003 (a) E1-Roughness and (b) E2-Hardness. 
Values interpolated using Kriging with a 2m grid and the coastline blanked (black).

Figure 21. Results of the ECHOplus survey of Lough Hyne in August 2003 (a) E1-Roughness and (b) E2-Hardness. Values interpolated using Kriging with a 2m grid and the coastline blanked (black).

Unsupervised classification of sublittoral habitats of Lough Hyne

Figure 22. Unsupervised classification of sublittoral habitats of Lough Hyne using (a) subjective box clusters placed on the scatterplot of E2 against E1 obtained from an ECHOplus survey in August 2003; giving (b) the resulting distribution of "acoustic classes" interpolated using the nearest neighbour algorithm in Surfer™. The 5 acoustic classes produced are assigned habitat complexes that they cover (the principle ones in bold) using the codes listed in Table 1.

Subjective box classification of ECHOplus data

The unsupervised classification of Lough Hyne's sublittoral habitats using subjective box classification, is presented in Figure 22, which shows the scatter plot of E1 and E2 obtained from the ECHOplus survey of Lough Hyne with the subjective boxes placed over 'clusters' of points, and the classified points interpolated using nearest neighbour to produce the classified map. There are an infinite number of combinations of boxes, and the final version represents a compromise between the number of habitats present and the number of 'clusters' that can be successfully represented by using boxes. Thus, the classified map produced in Figure 25, is better thought of as the distribution of acoustic classes rather than habitat complexes. Nevertheless, acoustic classes 1, 4 and 5 appear to represent the soft sediment, rock and mixed sediment habitats reasonably well. However, acoustic classes 2 and 3 cover most habitat complexes present in the lough. Although through an iterative process these classes could be refined, the inappropriate nature of using boxes to cluster points would eventually limit any discrimination of habitats present.

Cluster analysis of False Colour Imagery (FCI)

Cluster analysis of False Colour Imagery based on the methods used by Sotheran et al. (1997) (given in more detail in Foster-Smith et al. (1999) and Greenstreet et al. (1997)), was also used to classify the ECHOplus data. Using image processing software (ERDAS™) the interpolated E1 and E2 raster images (Figure 24) were stacked into one False Colour Image (retaining the 2m grid size). Then unsupervised classification using the ISODATA algorithm was performed, data were first clustered into 6 acoustic classes in an attempt to match the number of sublittoral habitat complexes that are known to be present (Figure 23). As for the box classification above, the fine sediment, rock, and mixed sediment appear to be well represented, here, in acoustic classes 1, 4 and 6 respectively. In addition, acoustic class 3 appears to represent the macrophyte dominated regions found from sample points by Kitching et al. (1976), however, there are a few misclassified regions in deeper areas where there will be an absence of algae. Despite acoustic class 5 representing some areas of known coarse sediment, in particular, in the Rapids/Whirlpool area and adjacent to cliffs sections, it also indicates areas of mixed sediment, e.g. around Castle Island and in the shallower regions of the North Basin. Furthermore, there is no distinction between cohesive soft sediment (mud) and the fluid fine sediment found in the Western Trough, with class 2 representing the same habitats as class 1. It is worth noting, that the inclusion of depth as a third layer in the FCI was tried in the classification procedure, however, this produced a classified image that related more to depth contours than to biotope distributions.

Overall, the classification of habitats using cluster analysis of FCI appears to be more representative and plausible than the subjective box classification method. Consequently, Figure 23 was chosen to be further enhanced, first by combining class 1 with 2, and class 5 with 6 to produce a new map shown in Figure 24. Despite this improvement to the accuracy by combining classes, contextual editing was required. In particular, known areas of soft sediment found in deeper water were misclassified as mixed sediment, macrophyte dominated sediment or rock. Misclassification in deeper water has occurred in other surveys using AGDSs (Hamilton, 2001; Kenny, 2000). The nature of sound scattering from steep slopes and the low acoustic impedance of the fluidised fine sediment found in the deeper water of Lough Hyne are the most likely source of misleading E1 and E2 values. After contextual editing, the final classified image of the distribution of sublittoral habitat complexes in Lough Hyne and can be seen in Figure 24.

Unsupervised classification of sublittoral habitats of Lough Hyne using cluster analysis of a False 
Colour Image composed of E1 and E2 values obtained from an ECHOplus survey performed in August 2003.

Figure 23. Unsupervised classification of sublittoral habitats of Lough Hyne using cluster analysis of a False Colour Image composed of E1 and E2 values obtained from an ECHOplus survey performed in August 2003. The 6 acoustic classes produced are assigned habitat complexes that they cover (the principle ones in bold) using the codes listed in Table 1.

The 6 acoustic classes from Figure 26 are reduced to 4 'habitat' complexes that they cover using the 
codes listed in Table 1

Figure 24. The 6 acoustic classes from Figure 26 are reduced to 4 'habitat' complexes that they cover using the codes listed in Table 1.

The final classified image of the sublittoral habitat complexes found at Lough Hyne (Figure 24) agrees well with previous studies (Kitching et al., 1976; Thrush and Townsend, 1986; Wilkins and Myers, 1990; and Turner, personal communication), which is maybe to be expected as these studies were used to classify and help in contextual editing. However, as the point data in this study was interpolated using more sophisticated geo-statistics than subjective contouring round diver observations (as in previous maps), the map's distribution of habitats has a more natural look and is less contrived than previous maps produced. Nevertheless, without a formal accuracy assessment (which was not possible) the habitat map produced in this study (Figure 24) cannot be shown or concluded to be more accurate than any previous efforts. Moreover, as the habitat maps have been produced from unsupervised classification of interpolated remotely sensed data, they are not definitive like a road map, but are predictive distributions of habitats known to occur there. Therefore, there will always be a large degree of uncertainty associated with these types of habitat maps.

Unsupervised classification with contextual editing of sublittoral habitats of Lough Hyne.

Figure 25. Unsupervised classification with contextual editing of sublittoral habitats of Lough Hyne.

Despite these concerns over accuracy and misclassification of the classified image produced, the study does conclude that using cluster analysis within image processing software is a more robust and repeatable method than using subjective box clusters to classify AGDS data. A conclusion that is shared by other authors (Foster-Smith et al., 1999; Fox et al., 1998; Greenstreet et al., 1997). However, a drawback of this technique is that the original data is not preserved, furthermore, spatial interpolation between tracks (as done for both methods here) implies a knowledge of bottom type there which does not exist, while along track smoothing to achieve regular pixel sizes may remove real changes (Hamilton, 2001). Finally, as this study had to employ unsupervised classification, future work that obtains geo-referenced ground truth data from Lough Hyne, could develop signatures of E1 and E2 (and possibly depth) to perform supervised classification on the ECHOplus data to produce a more reliable habitat map and a better appraisal of ECHOplus.

Conclusions

The limited examples of the use of ECHOplus hinder a thorough assessment of the system's performance in seabed discrimination. In particular, more comparisons of ECHOplus with other systems (e.g. RoxAnn and QTC) would be beneficial. Also of value would be studies that utilise its dual frequency capability, which is an advantage of the system. Siwabessy (2001) showed that having multiple frequencies increases the discrimination ability, as different scales of seafloor roughness and levels of seafloor volume contributions are being examined simultaneously, thus, exploitation of this feature could optimise the use of ECHOplus. Nevertheless, from the limited examples available, ECHOplus does seem to provide good discrimination between noticeably different seafloor types (e.g. rock, algae and sand). However, more subtle changes in seafloor habitat (e.g. sparse versus dense algae cover) are perhaps not picked up. Furthermore, misclassification can occur in deeper water (as with RoxAnn), this is can be mainly attributed to the nature of the scattering of sound from rougher bottoms and its effect on the E1/E2 parameters (Hamilton, 2001). The biggest shortcoming of ECHOplus (and other systems) is that the raw waveform is not recorded. As useful as E1 and E2 parameters are, they are no substitute for the complete echogram, which can be used to identify bad bottom-picks (e.g. due to mid-water reflectors) and perhaps to derive other acoustic parameters from either the seafloor (e.g. rise time) or water column (e.g. fish).

Back to top of page


Comparison of Systems

Hamilton et al. (1999) compared performance of the RoxAnn and QTC View systems in the Cairns area of the Great Barrier Reef lagoon. QTC outperformed RoxAnn, but one outcome of the resulting analysis was that RoxAnn results were consistent only when obtained at constant speed. This was contrary to the manufacturer’s prescriptions which specify that any speed can be used. However any system may be subject to ship noise and aeration. QTC results did not appear to be speed affected, but insufficient data was obtained for RoxAnn within any particular speed band to draw any real conclusions. One interesting result was that Hamilton et al. (1999) were able to obtain a direct mapping of QTC classes to RoxAnn space.

Smith et al. (2001) did not find much difference between QTC View and RoxAnn when searching for oyster beds. Both systems performed well, although tests were not comprehensive.

In 2001, a report describing a comparison of the. RoxAnn and QTC View systems appeared (Foster-Smith et al., 2001). This work emerged from an extensive multi-year British program funded by the UK Department for Environment, Food and Rural Affairs and the Crown Estate. The report, and its companion document Brown et al. (2001) describe acoustic biotope assessment in four areas off the southeast coast of England. A significant component of the motivation for the work was to facilitate mapping of gravel sites. Three study regions were of 48 sq km area and one had area of 336 sq km. The two lengthy reports arising from this work constitute a valuable resource for the conduct of echosounder based benthic classification, both for location of gravels and related substrate types and for more general classification requirements.

Foster-Smith et al. (2001) concluded that there was little difference between the performance of RoxAnn and QTC View, although the detailed conclusions throughout the report favoured RoxAnn. It is noteworthy that the authors do not report the depth dependent parameter variation discussed in Section 3.3 above. It appears that the depth ranges involved always enabled a second return to be gained, as no reference exists in the reports to the loss of the second echo. The availability of a second echo is central to the RoxAnn technique and not relevant to QTC View. In this regard, QTC View does not offer an advantage over RoxAnn in relatively shallow water operation.

  1. Dr. T. Pauly is a Director for Business Development for SonarData and the Verdant Group, Hobart, Tasmania. He spent ten years as a Research Scientist and Hydroacoustician with the Antarctic Marine Living Resources Group of the Australian Antarctic Division.
  2. Dr. J.R. Turner is a Senior Lecturer in Marine Ecology at the School of Ocean Sciences in the University of Wales, Bangor and has dived Lough Hyne extensively over a 20 year period.

Back to top of page