7 March 2002



Secure Ports and Harbors

Seabed Classification and Multibeam Bathymetry:
Tools for Multidisciplinary Mapping

By William T. Collins,
Quester Tangent Corporation,
Sidney, British Columbia, Canada


James. L Galloway,
Canadian Hydrographic Service,
Department of Fisheries and Oceans,
Institute of Ocean Sciences,
Sidney, British Columbia, Canada

Multibeam bathymetric technology is rapidly becoming the tool of choice for marine surveys due largely to the efficiency in mapping swaths of seabed. In conjunction with the multibeam surveys, acoustic seabed classification provides another layer of information for an incremental expense. The following article presents results of a survey using both multibeam echo sounding and single beam seabed classification. The survey was sponsored by the Canadian Hydrographic Service. The multibeam data would be used to support undersea security inspections for the 1997 Asia Pacific Economic Conference (APEC) meetings. Quester Tangent Corporation took the opportunity to collect information on the nature of the seabed using the QTC VIEW seabed classification system.


Survey Site

The survey site comprised an area of the inner harbour of Vancouver, Canada. This location included the area directly adjacent to the Vancouver Trade and Convention Centre. Water depths ranged from near drying line to 60 metres and included all wharf and jetty areas. Sediments consisted of gravel on bedrock, silt and a range of sand with various concentrations of pebbles. The seabed near the Convention Centre had been modified by dredging activities. Otherwise, the site was sheltered from waves and was influenced only by tidal currents.

Seabed Bathymetry

Bathymetry was measured using a 300 kHz Simrad EM3000 multibeam swath sonar from a 13-metre hydrographic survey launch. The EM3000 system applies 127 beams to measure a 120� swath transverse to the vessel heading with an overlapping angular resolution of 1.5� X 1. 5�. A high ping rate, combined with pitch stabilisation for this sonar guarantee 100% bottom coverage in each swath to ensure no seabed features or navigation hazards are missed. Typically, however, surveys take place with 150-200% coverage.

In order to measure accurate georeferenced bathymetry, the EM3000 requires a vessel motion measurement package, a precise position measurement system, and a sound velocity profiling instrument. Motion and positioning data were supplied by the Position Orientation System for Marine Vessels (POS/ MV) package by ApplAnix. A combination of an inertial platform and dual differential GPS provided vessel pitch, roll, and yaw to an accuracy of 0.05�, heave to an accuracy of 5% of vertical displacement, and a typical horizontal position accuracy of two metres. Sound speed profiles accurate to 0.06 m/ s were measured with a SV Plus from Applied Microsystems Ltd. These exceptional specifications were essential to measure bathymetry in accordance with IHO standards for wide-multibeam technology.

Seabed Classification

The amplitude and shape of an acoustic signal reflected from the sea floor is influenced by bottom roughness, contrast in acoustic impedance between water and sea floor, and perturbations caused by reverberation of the volume of substrate. The remote classification of the sea bottom requires an acoustic data acquisition system combined with a set of algorithms that analyze the data, determine the seabed type and relate the results of the acoustic classification to the physical properties of the sediments.

The QTC VIEW was connected to a 38 kHz Knudsen 320M echo sounder. Sounder parameters are given in Table 1.

Table 1: Settings used by the Knudsen echo sounder and the QTC VIEW.
Parameter Setting
Pulse length 0. 5 ms
Range 100 m
Beam width 17�
Transmit power 1000 watts
Ping rate 5 per second

The QTC VIEW captured the bottom echo and generated a digital time series representing the transmission and reception of an echo trace. The signal was pre-processed to identify the sea floor and filtered to suppress noise.

Several algorithms extracted shape parameters from each trace. Statistical analysis of the series of data collected during system training reduced the information to the three most useful shape descriptors (Q-Values) for distinguishing echoes during real-time mapping.

The QTC VIEW classification technique uses the shape of a returning echo to characterize the seabed. The response from a smooth seabed, simple in character, will be a sharp rise in the signal, a peak, and a short tail (Figure 1). The response from a rough, complicated seabed will be a peak followed by a slower decay in the signal represented by a longer tail.


Figure 1: Comparison of echo traces from two representative seabeds.

The shape of the returning waveform is related to the nature of the target (in this case, sea floor sediments). There are numerous seabed characteristics which account for the variability of the signal (Figure 2).


Figure 2: Features of a typical seabed influencing the acoustic response.

The physical properties of the sediments are of prime importance. These include textural information (e. g., grain size) and the condition of state (e. g., porosity, density). Larger scale features such as bedforms or large sedimentary particles (e. g., boulders) will also influence the acoustic response. Flora and fauna will change the shape of the echo. Seabed topography will influence the signal and is more pronounced in areas of rugged terrain.


Data Acquisition

Because no previous seabed data were collected using this particular QTC VIEW unit, the system was calibrated with the collection of a training data set. This required the acquisition of a series of sample echoes while remaining stationary over a known seabed. Training data from nine representative sites were collected across the study area. All locations were evaluated for acoustic diversity and surficial geology. There were both acoustic overlap and geological similarity in many sites. Four of the nine training data sites were chosen to capture the range of seabed types expected in the harbour (Table 2). These were combined to form a Vancouver Harbour catalogue for real-time surveying. When surveying, QTC VIEW generated a confidence value associated with each classification. This value is an indication of the similarity of the new echo with those from the training set. These data were used to assess the reliability of the classification results and the accuracy of the catalogue in capturing the range of acoustic classes.


Table 2: Summary of the sites used to calibrate the QTC VIEW.
Site/ Class Latitude Longitude Depth Description
1 49� 17.8398 123� 06.5566 10 m Algae encrusted pebbles ( 4-6 cm),
bivalve, minor coarse sand, rock


49� 17.7070 123� 06.2891 26 m Fine sand, minor silt, shell hash, 5%
pebbles 1-3 cm in diameter


49� 17.2998 123� 06.4688 17 m Coarse sand, 15% pebbles 1-2 cm in
diameter, medium sorting


49� 17.3433 123� 06.7939 16 m silt, <5% very fine sand, organic
veneer 3


To ensure high data quality, raw echo traces from the four sites chosen for the final catalogue were captured and saved (Figures 3 to 6). The QTC VIEW system analyzed the first bottom echo only; differences in its shape from the four sites were apparent. Features such as amplitude and slope of the back side of the peak and length of the tail following the peak were measured and used as shape descriptors. The QTC VIEW measured more than 150 features from each echo and used that information as the basis of echo description.

Two or more grab samples were collected at each site to validate the surficial geology. A 40 lb. Ponar grab sampler collected bottom samples. The sampler penetrated approximately 15-20 cm into the substrate. In medium sands, penetration of the 38 kHz echo sounder signal was 20-30 cm. The average water depth during the survey was 18 m, and with a 17� beamwidth on the trans-ducer, gave a seabed footprint of 4.5 m. Therefore each ping insonified a volume of sediment measuring 20-30 cm thick by 4.5 m in diameter. For applications where the biota is significant, video or still photography would be required to supplement the sediment sampling.

Figure 3: An example of an echo from Class 1.

Figure 4: An example of an echo from Class 2. 4

Figure 5: An example of the echo from Class 3.

Figure 6:
An example of the echo from Class 4. 5



System training and ground-truthing were completed in three hours. For real-time mapping the mode of operation, including line spacing and orientation, was determined by the requirements of the multibeam system. The lines were run contour parallel at a spacing of between 30 and 50 m. The survey took six hours to complete.

The bathymetric data were processed and gridded for final presentation as a shaded relief plot (Figure 7). The data show a series of complex depressions directly in front of the Vancouver Trade and Convention Centre. The seabed was flat to a depth of 18m; at 20 m depth a slope break occurred. Bathymetric highs occurred to the northwest and north of the area.

Figure 7: Shaded relief image of the survey site in Vancouver Harbour.

The Vancouver Harbour catalogue was used for seabed classification. Figure 8 shows the seabed class assignments along the vessel track. All four classes were represented and ranged in distribution from homogeneous to highly variable. Class 1 occurred to thenorthwest and was associated with the bathymetric high. It also occurred near the shore at the center of the survey area. Class 2 was predominant in the east of the survey area. Class 3 occurred in association with Class 1 near the northern bathymetric high and along the shore. Class 4 was predominant near the jetty and at the center of the survey and to the west.

Recorded confidence values were used to generate a contour map which was laid under the classification (Figure 8). Results showed that the catalogue successfully captured the variety of seabed types from the mapped area. There were anomalies where confidence dropped below 50%, indicating that the seabed was in transition between the two classes or there was a new seabed type altogether.

Also plotted were the four calibration training sites. As expected, each occurred in areas of extremely high confidence except for Class 1 where the survey line did not cross the training site.

Figure 8: Results of QTC VIEW seabed classification shown as coloured crosses. The background shows a grey scale contour plot of confidence. The location of the training sites and assigned class are super-imposed as a cross with a coloured border.

A multi-layered plot captured the range of information collected during the survey (Figure 9). The uppermost layer was a contoured bathymetric map. Contours were filled with various shades of grey and the interval was 5 m.

The middle layer was the surficial geology as interpreted from the QTC VIEW classification data. The area was divided into five geological units based on the classification of specific echoes and variability of their distribution. Unit 1 was gravel of varying thickness over bedrock including patchy areas of coarse and medium sand. Unit 2 was medium sand, minor silt and shell hash and patchy concentrations of coarse sand and pebbles. Unit 3 was coarse sand with pebbles. Unit 4 was silt with very minor amounts of fine sand. Unit 5 was highly variable and consisted of a mixture of Units 3 and 4 which comprised silt with sand and pebbles.

The bottom of Figure 9 is a surface relief plot of the bathymetry. The vertical aspect of the plot was exaggerated to enhance the topography. This plot is useful in the evaluation of bathymetric control on the sediment distribution.

Figure 9: Multi-layered plot of the bathymetric contours and surface relief with interpreted surficial geology.


Summary and the Future

The multibeam system worked well during the survey, providing details on the bathymetry to support diving operations for site security clearance before the APEC leaders meeting in Vancouver.

During the survey, seabed classification data were collected, a catalogue was created and contained classes which were acoustically discrete and encompassed the range of seabed types appropriate to the mapping objective. The training data sets were stored and constitute a library of seabed types which can be combined with new calibration sites to form an alternate catalogue depending on the application.

The collection of seabed classification data alongside the multibeam data provided value-added information to the survey results. The seabed catalogue compiled for use in the harbour captured the range of seabed types and can now be used for future survey work.

Seabed classification using multibeam sounders is a topic of research at Quester Tangent and the Canadian Hydrographic Service. Quester Tangent has successfully post-processed Simrad multibeam echo sounder data for the U. S. Naval Oceanographic Office and integrated a QTC VIEW with a Reson Seabat for real-time classification capability. In both instances vertical or near-vertical beams are involved. The challenge for the future is to develop the capability to classify echoes at non-normal angles in real-time.



Mr. Bill Collins is product manager for seabed classification at Quester Tangent. He has a MSc. in Earth Sciences and has been working as a Marine Geologist since 1981. Bill has extensive

experience in mapping seabeds for a variety of applications ranging from sediment stability for oil production platforms on the Grand Banks to coastal development in the South Pacific. Bill has been responsible for the development of seabed classification applications at Quester Tangent for the past two years and has published widely on the subject.

Mr. James L. Galloway completed MASc. in Electrical Engineering and Oceanography and joined the Department of Fisheries and Oceans in 1974. As head of the Sonar Systems Group at the Institute of Ocean Sciences, he has worked to solve sonar and other acoustical problems in hydrography and fisheries research. Currently his interests are in application of multibeam technology to hydrographic surveying, fisheries stock assessment and seabed classification.