Multibeam and swath echosounders

Brief Description

Multi-beam or interferometric swath echosounders have become increasingly common and provide a powerful seabed-modeling tool. Each transducer produces a fan of acoustic beams to provide sounding information either side of the vessel's track. The high-performance systems have wide-angle swaths that cover an area up to 7 times water depth; more typically, the swath width is twice the water depth. As water depth increases, range increases, but maximum range becomes limited due to acoustic energy depletion of the outer beams. The accuracy of swathe systems is critically dependent on the correction applied for vessel motion, (heave, pitch, roll, yaw etc); consequently, a swath system is integrated into many other specialist sensors within the survey system. The chief advantage of swath bathymetry systems is the high rate of productivity and excellent data sample density, especially in deeper water. Swath systems can be hull mounted in the ship, installed in a towed body (tow-fish) or in other remotely operated platforms. While hull mounted systems are easier to calibrate than towed systems, a towed system offers more portability and can be deployed closer to the seabed. Many swath bathymetry systems also record backscatter (reflected energy) from the seabed, similar to side-scan sonar images.

See also NOAA's summary table Summary view of multibeam sounding technique (541 KB PDF).

NOTE: The content below is derived from Chapter 5 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

Multibeam and Swath Sonar | Principles and data processing of Multibeam Sonar Systems (MBSS) | Seafloor classification procedure

Multibeam and Swath Sonar

Multibeam sonar systems (MBSS) have rapidly evolved over the last few decades, and currently are the most advanced acoustic tool for remote observations and characterisation of the seafloor. Among the existing sonar systems, MBSS's provide maximum information about the bottom properties along with a wide coverage of seafloor mapping, which is essential for rapid assessment of the benthic habitat over large areas of coastal waters. The swath width of modern MBSS is approaching the raster width of acoustic sidescan systems, while the width of individual MBSS beams can be as small as tenths of a degree. The main advantage of a modern MBSS with respect to discrimination and classification of different seafloor types is its ability to provide simultaneously a high-resolution bathymetry map and a backscatter image of the surveyed area. Most of the seafloor classification techniques developed recently are based on a mutual analysis of bathymetry or backscatter data using certain characteristics of the seafloor roughness and backscattering strength. The estimated parameters of different types are clustered to define selection criteria for individual classes of the seafloor.

The capability of multibeam systems in remote acoustic characterization of the seafloor is a subject of current research. A number of organisations and companies are currently working to provide seabed classification capability for multibeam systems. Simrad has produced the Triton system for this purpose, and an evaluation of this product has been carried out as part of the CSIRO deep water program in Australia (Kloser, private communication). The QTC and RoxAnn organisations are developing multibeam classification systems and other organisations are active in the field. The Seabeam multibeam instrument corporation has recently been advertised as offering seafloor classification capability. Several advantages of multibeam over single beam classification procedures can be noted. Firstly, multibeams provide greater area coverage, approaching that provided by sidescan systems. Secondly, the ability to infer cross track bathymetry allows for corrections to be made for seabed slope across the vessel track direction, while along track slopes may be derived from sequential depth assessments. Within the spatial limitations imposed by beam geometries and operating depths, it is thus possible to determine local angles of incidence. This relates to a third advantage, that it is in principle possible to build a model of surface roughness based on the variation of backscatter amplitude with the angle of incidence at the seabed. Multibeam systems are, however, expensive and in general require larger vessels than single beam systems. It seems likely that over time multibeam or possible interferometric sidescan systems will become the acoustic systems of choice where budget provisions allow. As is more the case than with the less effective single beam systems, however, significant research, development and proving needs to be done to realise the potential of the more complex systems for classification.

Beginning in April 2000, the project Marine biological and resource surveys, South East Region was begun by CSIRO Marine Research under an Agreement with the National Oceans Office. The project constitutes Australia’s largest benthic survey program to date. The interim report (Kloser et al., 2001a) concentrates on the comparison of high resolution swath mapping with pre-existing or known habitat types, to provide an initial evaluation of the application of rapid assessment methods based on swath mapping. Other techniques used in the field program carried out in April-May 2000 included normal incidence echo sounding at three frequencies and a variety of conventional sampling systems. The conclusions of this report indicate that swath mapping and associated software for classification from backscatter returns, even at the present limited state of development, produces a powerful surrogate for habitat type. The report authors suggested that Australia should establish a national program of seafloor mapping using the suite of technologies used in the CSIRO project. More recently CSIRO has acquired a dedicated 30 kHz Simrad system which is mounted on the research vessel Southern Surveyor and further seafloor mapping using this multibeam system has begun in Australian waters.

Back to top of page


Principles and data processing of Multibeam Sonar Systems (MBSS)

General design principles of modern MBSSs are illustrated in Figure 1. The sonar transducer emits acoustic pulses propagated inside a wide across-track and narrow along-track angular sector. The receive array directed perpendicularly to the transmit array forms a large number of receive beams that are narrow across track and steered simultaneously at different across-track directions by a beamforming process. Thus the system performs spatial filtration of acoustic signals backscattered from different portions of the seafloor along the swath. Modern shallow-water MBSSs, such as Simrad EM 3000 and Reson SeaBat 8125, operate at hundreds of kHz, transmit short pulses of several tens of microseconds, and form hundreds of beams of about 1 degree width. Because they employ short pulse lengths, narrow-beam MBSSs are capable of resolving small features a few decimetres wide in the seafloor relief in the horizontal plane along with finer bathymetry details.

Typical geometry of the transmit and receive beams of MBSS.

Figure 1. Typical geometry of the transmit and receive beams of MBSS.

The operational principles of MBSS give evident advantages for seafloor mapping. However, they result in much stricter requirements for ship’s navigation than that for single-beam and sidescan systems. If ignored, ship’s attitude affected by roll, pitch, and heave may distort irreparably the bathymetry and backscatter images, especially for the oblique beams of small grazing angles for which ship’s motion induces a large horizontal deviation of the footprint location (see Figure 2, for example). Therefore swath bathymetry mapping must be accompanied with simultaneous tracking of ship’s motion, including roll, pitch, heave, and yaw. The procedure for compensating for a ship’s attitude for swath mapping data is well developed and described in the literature (US Army Corps of Engineers, 2002).

Bathymetry images before Bathymetry images after

Figure 2. Bathymetry images before (left panel) and after (right panel) compensation for ship’s motion (from the results of the Coastal Water Habitat Mapping project of the Coastal CRC).

Another serious problem specific to MBSS seafloor mapping is acoustic refraction in the water column with a depth dependent sound speed, which distorts the acoustic ray trajectories and hence the bathymetry images. Correction of swath data for the sound speed change in presence of ship’s 3-D motion is a complicated problem. Possible artifacts in MBSS bathymetry imaging due to mutual effects of the sound speed variation and ship’s motion are considered by Kammerer, et al. (2000) suggesting a new, more accurate method for removing those artifacts. The problem becomes even more complicated if the sound speed profile is time and range dependent. Hughes Clarke et al. (2000) propose a new operational procedure for swath mapping integrated with underway oceanographic measurements, which improves the results of acoustic mapping in a range-varying environment.

The most prevalent approaches to seafloor characterisation by using MBSSs are briefly discussed below.

Processing bathymetry details

Modern high frequency, narrow-beam MBSSs used for shallow-water surveys produce high-resolution bathymetry maps with a cell size of the horizontal resolution of several decimetres, which makes it possible to perform a small-scale textural analysis of seafloor relief. The purpose of the textural analysis is ideally to determine a 2-D, generally anisotropic spectrum of the bottom roughness. In practice, the capability of such an analysis is limited by the spatial resolution of MBSS, which becomes coarser with the increase of sea depth. The other problem is a large number of parameters to be estimated to define a mask for classification (spectrum width / roughness correlation length in different directions, roughness rms height, etc.). Therefore simplified criteria are commonly introduced in the mapping procedures to identify local topography features and roughness of the bottom surface. After the procedure of cleaning and gridding, the bathymetry data are represented on a regular topographic grid of which the mesh size depends on the MBSS angular resolution and typical sea depth in the surveyed region. In a gridded form, the bathymetry data are much more suitable for raster-based processing, including 2-D filtering and a spectral analysis. The most common approach is to use a rectangular window of several grid cells width (variable in general) moving along the grid to determine local characteristics for further terrain analysis. The most prevalent characteristics are the average elevation, spatial derivatives, a topographic variability index (TVI), and a topographic amplitude index (TAI). The first-order derivatives give the slope and aspect of the bottom surface. The second-order derivatives express the curvature of the surface. The TVI and TAI indexes were introduced by Chavez et al. (1995) as a measure of topographic variability. Both TVI and TAI are generated from the high-pass filtered (HPF) surface elevation, which is usually obtained by subtracting the low-pass filtered (i.e. averaged within a 2-D integration window) depth values from the original bathymetry grid. The TVI is determined by sorting the pixels into two categories such that the pixel values lie either within or outside a predefined divergence from the average value. Every pixel is designated as 0 or 1 in accordance with its category, which corresponds to either negligibly small or noticeably large variations respectively. Finally, the TVI is calculated by averaging the indexes over the window. Figure 3 illustrates an example of terrain mapping using TVI analysis .

The TAI is defined as maximum absolute deviation of the high-pass filtered elevation of the surface relative to the average value within the integration window.

MBSS bathymetry maps are generally gridded with a variable mesh size, because the horizontal resolution of MBSS is a nearly linear function of depth. For variable grid spacing, the results of seafloor mapping by terrain attributes become somewhat indefinite, if the estimates of the terrain characteristics are markedly dependent on the mesh size of the grid. It is shown by Diaz (1999) that the TVI index is nearly invariant to changes in grid spacing. The estimates of the TAI index and spatial derivatives (surface slope and curvature) are much more dependent on the mesh size, which follows directly from the ambiguity of numerical differentiation with a variable sampling interval.

Slope draped over a 3-D bathymetry over the Morinda Shoal region Bowling Green Bay TVI draped over a 3-D bathymetry over the Morinda Shoal region Bowling Green Bay

Figure 3. Slope (left panel) and TVI (right panel) draped over a 3-D bathymetry over the Morinda Shoal region in the Bowling Green Bay, Queensland, Australia (from the results of the Coastal CRC CWHM project).

Fractal analysis is nowadays another method that is widely used for seafloor classification using high-resolution MBSS bathymetry data. The fractal concept applied to modeling topographic relief is, in a certain sense, a modification of the 2-D spatial spectrum analysis that confines the variety of modeling spectra within a single class of fractal spectra. The shape of a fractal spectrum is defined by only two parameters, which are a fractal dimension, and a cut-off wavenumber that determines the roughness correlation length. In the general case of an anisotropic surface, the cut-off wavenumber is different along X and Y directions. The fractal power spectrum has the following form:

Sand-ripple roughness of seafloor surface modelled by an anisotropic fractal spectrum with a fractal dimension of 2.5.,

where u c and v c are the cut-off wavenumbers, K is a coefficient dependent on the rms height of roughness, and n is related to fractal dimension D by n = 6.5 - 2D. Figure 4 demonstrates a sand-ripple seafloor surface modelled by an anisotropic fractal spectrum with D = 2.5.

Sand-ripple roughness of seafloor surface modelled by an anisotropic fractal spectrum with a fractal dimension of 2.5.

Figure 4. Sand-ripple roughness of seafloor surface modelled by an anisotropic fractal spectrum with a fractal dimension of 2.5.

The fractal approach to topographical classification of the seafloor has two obvious advantages:

  • The seafloor topography can be well modelled by a fractal structure for a broad class of seafloor relief (Hastings and Sugihara , 1994) and
  • The number of estimated parameters is small, which simplifies the classification procedure.

A model of acoustic backscattering from a rough seafloor surface of fractal structure is developed by Lubniewski et al. (2000), who also show that the parameters of fractal spectra of the seafloor surface can be derived from acoustic sonar data.

Processing of backscatter data

Processing of MBSS backscatter data is a more complicated procedure than the analysis of bathymetry data. There are basically two different approaches to utilisation and interpretation of backscatter data, which are commonly used in acoustic classification of the seafloor. These approaches are a textural analysis of backscatter images and an analysis of the angular dependence (AD) of backscattering strength. The textural analysis is most common, since the methods of statistical analysis of MBSS backscatter intensities are similar to those utilized for processing of side-scan sonar images. A large number of statistical characteristics are determined from backscatter data for discrimination of seafloor classes. Before performing the analysis, the backscatter intensity data are usually corrected for the angular dependence of acoustic backscattering and for inequality of the MBSS sensitivity along different beams. An improper correction of the backscatter angular dependence may produce large errors in the estimates of the basic statistical characteristics, such as mean intensity, standard deviation, and higher-order moments, derived from backscatter images. The simplest model of well-known Lambert’s law is frequently applied to backscatter data for removing angular dependence. However, this model is not accurate enough for many classes of seafloor cover, especially at near-nadir (steep incidence) angles. Hellequin et al. (2003) employed a simple composite model that treated the angular dependence of backscattering using the tangent plane (Kirchhoff) high-frequency approximation for near-nadir incidence angles and a Lambert-like term dominating at off-specular angles:

Backscatter intensity image of the seafloor build from five overlapping swath lines

where BS is the backscattering strength and q is the angle of incidence. The coefficients A, B, a , and b are estimated by least-mean-square fitting of the model function to the average angular dependence of backscatter intensity observed across representative (or training) areas. At higher frequencies of hundreds of kHz, the roughness height of the seafloor surface becomes much larger than the acoustic wavelength, so that both Kirchhoff and Lambert approximations are not accurate enough for numerical prediction of the angular dependence. For such conditions, an empirical approach to angular correction of backscatter intensity data has been developed within the CWHM project of the Coastal CRC. The developed algorithm involves calculation of an average angular response for backscatter intensity level within a spatial window of a programmed length that slides along the swath line with a 50 per cent overlap. The average angular dependence is then subtracted from the backscatter intensity level within each section of the swath line that spans the central half of the averaging window. Then the absolute level of backscatter is reconstructed by adding the average level measured within the interval of 30 ± 2 degrees. As a result, the algorithm removes artifacts of the angular dependence from backscatter images, minimizes the boundary effects due to angular correction, and, at the same time, tracks gradual variations of backscattering strength over the surveyed area and allows for preserving information on the backscatter angular dependence at a relatively high spatial resolution. The last two features of the algorithm bring certain advantages to swath backscatter mapping of the seafloor when comparing to the empirical method recently suggested by Beaudoin et al. (2002), in which the backscatter intensity data are corrected for the angular dependence averaged over each whole swath line. The efficiency of the newer algorithm is demonstrated in Figure 5.

To calculate the basic statistical characteristics, the backscatter intensity image of the entire surveyed area (backscatter mosaic) is divided into small rectangular patches for which the mean intensity, standard deviation, and higher-order moments are determined. The patch size is selected taking into consideration the spatial resolution of MBSS and data quality. In contrast to low-frequency deep-water MBSSs that have a footprint size much larger than the correlation length of seafloor roughness, modern high-frequency shallow-water swath systems operate with such narrow beams and such short pulses that the seafloor area insonified instantly is small enough to resolve individual small-scale features of seafloor relief (small rocks, sand ripples, shellfish patches, etc.). Under such conditions, the statistics of backscatter intensities is another important measure that can be used for identification of morphological and physical characteristics of the seafloor. The statistical distribution of backscatter intensities is usually determined at lower spatial resolution over long sets of small patches (e.g. several rows of rectangles per the port and starboard sides along the swath track).

Backscatter intensity image of the seafloor build from five overlapping swath lines

Figure 5. Backscatter intensity image of the seafloor build from five overlapping swath lines, before correction for the angular dependence (left panel) and after (central panel). The right panel demonstrates the mean slope of angular dependence within a 5-40 0 measured at the central points of each section of swath lines, superimposed on the grey-scale backscatter image. The seafloor in the surveyed area consisted mainly of sand. Note seagrass patches of various sizes clearly visible as yellow and red coloured (dark) spots at the bottom and at the upper right corner of the area.

If the size of an instantly insonified area is much larger than the characteristic length of seafloor roughness, the number of statistically independent elementary scatterers within this area becomes large enough for the distribution of the complex amplitudes of backscattered signals to tend to a Gaussian form, which follows from the central limit theorem. Consequently, the variation of backscatter intensities tends to Rayleigh-like statistics. For high-frequency MBSSs operating with smaller footprint of individual beams and smaller areas insonified instantly, the statistics of backscatter intensities become non Rayleigh’s in form. Ol’shevskii (1967) showed that broad-spectrum spatial variability of surface roughness should lead to a product model of backscatter statistics. Jakeman (1988) utilized a K-distribution as a product model for backscatter statistics that describes a Rayleigh-fluctuating process modulated by a G -distributed term that depends on two parameters (mean and so-called shape factor) and represents local reflectivity from the relief particularities.

Figure 6 demonstrates that the K-distribution fits experimental data for rougher surfaces much better than Rayleigh’s one. Hellequin et al. (2003) show that the shape factor a eff of the K-distribution estimated from backscatter intensity data can be used to distinguish different types of the seafloor cover, such as sand, rock, gravel, and rock, as shown in Figure 7.

Texture is one of the important characteristics that can be effectively used for identification of particular regions in an image. A texture analysis of backscatter images is widely used in processing of side-scan and MBSS data for classification of the seafloor, as noted in the technical section concerning side scan sonar. The most common method of the texture analysis is based on determination of statistical features of the so-called grey-level co-occurrence matrices (GLCM).

Measured histograms and statistical distribution fitting of the seafloor cover: rock Measured histograms and statistical distribution fitting of the seafloor cover: sand

Figure 6. Measured histograms and statistical distribution fitting for two different types of the seafloor cover: rock – left panel; sand – right panel. Gray-scale images on the left of each graph are backscatter intensity images of the respective seafloor types (from Hellequin et al., 2003).

Variations of the statistical estimates of shape factor 1/ a eff versus the average incidence angle (from Hellequin et al., 2003).

Figure 7. Variations of the statistical estimates of shape factor 1/ a eff versus the average incidence angle (from Hellequin et al., 2003).

Statistical characteristics calculated from GLCM describe distinctive textural properties that show the relationship between a given pixel and a specific neighbour. In particular, the GLCM characteristics give a detailed description of contrast and correlation of the intensity pixels in a backscatter image. The GLCM analysis is a useful image processing technique for MBSS backscatter imagery, because its results do not depend on the absolute backscatter strength and hence on the absolute calibration of sonar systems. Moreover, it is widely believed that the angular dependence of the GLCM features is weak enough for the results of the textural analysis to be weakly dependent of the irregularity of MBSS directivity and the angle of incidence. The latter is not fully true, because statistics of backscatter intensity depends on the incidence angle (as clearly shown in Figs. 6 and 7), and hence the GLCM features also should depend on the incidence angle. The GLCM analysis is a quite sophisticated technique. Generally, GLCMs are determined for each N-by-N pixel patch of an amplitude-quantised backscatter mosaic. The elements of each GLCM are expressed as the number of times a pixel of value i neighbours a pixel of value j in direction q , at distance d. The dimension of GLCM depends on the dynamic range of intensity variations expressed in quantising units (typically 255 for a 8-bit grey-scale image). The normal values of direction q are 0°, 45°, 90°, and 135°, which is a unique set of directions for N = 3. For each GLCM derived from a backscatter mosaic, one can calculate a large number of different textural characteristics, such as homogeneity, dissimilarity, correlation, variance, mean, entropy, contrast, angular second moment, grey-level difference vector (GLDV) contrast, GLDV mean, GLDV angular second moment, and GLDV entropy. For discrimination of different image classes in the whole backscatter image, each GLCM characteristic can be used to create a separate layer in the map of image. However, it is impractical to treat every statistical measure as a dimension in the feature vector space, because some of the GLCM characteristics are strongly correlated with each other. Moreover, most of the GLCM characteristics are nearly uninformative with respect to seafloor classification because they are not adequately correlated with the actual physical and morphological properties of the seafloor.

The angular dependence of backscattering strength is an important characteristic that distinguishes different types of the seafloor cover. Figure 8 demonstrates the angular dependence of backscattering strength measured for two different types of seafloor cover (seagrass and sand) using the Reson SeaBat 8125 MBSS.

Angular dependence of backscattering strength from seagrass

Figure 8. Angular dependence of backscattering strength from seagrass (1) and sand (2) measured in Cockburn Sound, Western Australia within the CWHM project [Gavrilov et al., 2005].

In practice, it is difficult to define replicas of the angular dependence for every type of the seafloor and classify the bottom surface by searching for the best-fit replica for the measured backscatter data. Therefore the whole angular range is usually divided into a small number of specific domains according to the physical peculiarities of acoustic scattering at different angles. The bottom backscattering model formulated by Jackson et al. (1986) distinguishes three domains. At near vertical incidence, backscattering from large smooth roughness dominates the volume scattering and backscattering from small-scale roughness. The tangent plane (Kirchhoff) approximation is an appropriate approach to modelling the angular dependence within this domain. At moderate incidence angles, Bragg scattering from small-scale roughness and volume inhomogeneities is the primary mechanism that can be modelled using a composite model approach based on the small-perturbation approximation. At small grazing angles below the critical angle, the volume scattering becomes negligible, which reduces the backscatter intensity especially at lower frequencies.

The most comprehensive procedure of angular dependence (AD) classification involves determination of the domain boundaries and calculation of certain characteristic values, such as the mean backscatter intensity, AD slopes, and second derivatives, within each domain. Figure 9 gives an example of AD classification with 10 selected characteristics. In practice, some of the AD parameters are not robust enough for adequate recognition of the seafloor type and therefore only a few parameters are used for seafloor classification, which usually are the mean backscatter intensities and slopes of the angular dependence measured within certain angular intervals belonging to different domains.

Three main domains (D1, D2, D3) of angular response curves and the parameters extracted to describe each domain Parameters estimated:

a. Mean BS intensity for D1 (0-10)
b. 2-d derivative at c
c. Location of boundary D1-D2
d. dB range of D1
e. Mean BS intensity for D2 (15-50)
f. 2-d derivative of g
g. Location of boundary D2-D3
h. Slope D2
i. Mean BS intensity for D3 (55-70)
j. Slope D3



Figure 9. Three main domains (D1, D2, D3) of angular response curves and the parameters extracted to describe each domain (from Hughes Clarke et al., 1997).

Back to top of page


Seafloor classification procedure

The analysis of MBSS bathymetric and backscatter data produces a large number of characteristics, as discussed in previous paragraphs. These characteristics are determined at mesh points of a grid that samples rectangular patches of the surveyed seafloor area. If the mesh size of the primary analysis is different for the parameters selected for classification, the obtained estimates are interpolated into points of a common grid. The grid spacing is usually selected such that the spatial resolution remains the maximum possible, avoiding excessive ambiguity of interpolation. As a result, N parameters selected for classification constitute N characteristic layers on the gridded map and form an N-dimensional vector space of variables. The main constituent of the classification procedure is clustering of the obtained vector space. It is impractical to apply clustering for a vector space of dimensions as high as several tens or even hundred of variables. Therefore the number of variables is usually reduced by searching for the parameters which, in combination, contribute most to the total variance over the gridded area, so that the rest of the parameters and their combinations could be disregarded. A principal component analysis (PCA) is the most common method for selecting appropriate combinations of the classification parameters ( Reed and Hussong, 1989). Usually the top two or three combinations are enough to model variations over the mapped seafloor area.

There are, in general, two different approaches to clustering MBSS data for seafloor classification. The first, realized in Quester Tangent’s QTC Multiview system (Preston et al., 2001) is the most comprehensive and ambitious. It involves a simultaneous analysis of all characteristics that can be obtained from both bathymetric and backscatter data of MBSS. Specifically, the QTC Multiview system treats initially over 130 different features extracted from MBSS data. Then the PCA procedure is applied to reduce the number of features to three combinations of parameters, which are used for classification of the seafloor. The second approach is to perform separate clustering for the parameters extracted from different categories of MBSS data, such as bathymetry, backscatter mosaic, and the angular dependence of backscattering. Then the clustered combinations of the most informative features defined by PCA in each category are used for building a number of different characterisation maps of the seafloor. The obtained characterisation maps are compared to each other in order to determine the correlation between the seafloor classes defined from the data of different categories.

For building the final classification model, both approaches need ground-truthing by comparing the seafloor classes defined from the MBSS data to the results of sediment core sampling and underwater video recording made within a number of training patches representing the most characteristic types of the seafloor. The efficiency and adequacy of seafloor classification by using either one or another approach to interpretation of MBSS data is still a subject of serious discussion. Although the multi-parameter method realized in some MBSS processing systems, such as QTC Multiview, attempts to automatically interpret the maximum information that can be extracted from MBSS data, a separate analysis of specific bathymetric and backscatter features obtained with MBSS sometimes gives better results with respect to the correctness of seafloor discrimination. Figure 10 demonstrates that the seafloor classes defined from the textural analysis of backscatter mosaic do not always distinguish the actual grain size of the sediment. Only two of five classes (1 and 2) represented in this plot by large numbers of core samples have noticeable peaks distinguishing certain domains in the grain size scale. However these domains are too broad for classification of the sediment. The example in Figure 11 shows the correlation between the grain size of sediments and the seafloor classes defined from the AD analysis of MBSS backscatter data. The correlation is better, but the AD classes distinguish the sediments by their grain size only as a very broad trend, even if AD classes 3 and 4 are merged. These results are not promising for these particular techniques, however, median grain size is only one of the sediment or habitat properties that influence acoustic backscatter.

An example of seafloor mapping using the data of a MBSS survey in the Morinda Shoals area off the Queensland coast is given in Figure 12. This example clearly demonstrates that the backscatter intensity data from MBSS (right-bottom panel) amplify the bathymetry data with additional information that can be used for better discrimination of the seafloor habitats.

Frequency of textural classes plotted against grain size determined from core 
samples

Figure 10. Frequency of textural classes plotted against grain size determined from core samples (from Diaz, 1999).

AD classes plotted against grain size from core samples

Figure 11. AD classes plotted against grain size from core samples (from Diaz, 1999).

3-D views of the seafloor across a coral reef in Morinda Shoal, Bowling Green Bay, Queensland. 3-D views of the seafloor across a coral reef in Morinda Shoal, Bowling Green Bay, Queensland. 3-D views of the seafloor across a coral reef in Morinda Shoal, Bowling Green Bay, Queensland.

3-D views of the seafloor across a coral reef in Morinda Shoal, Bowling Green Bay, Queensland. 3-D views of the seafloor across a coral reef in Morinda Shoal, Bowling Green Bay, Queensland.

Figure 12. 3-D views of the seafloor across a coral reef in Morinda Shoal, Bowling Green Bay, Queensland. The images show four colour-coded attributes extracted from the MBSS data: bathymetry (upper-left), slope (upper-right), TVI (bottom-left), and backscatter intensity (bottom-right). All of them are draped over the 3-D bathymetric map. Colour spots indicate the location of sampling stations for assessing fish abundance.

Back to top of page