• Home
• Site Map
• About Us
• Contact Us
• Links  
• 

• Search Data
• Climate Change
• Conceptual Models
• Coastal Indicators
• Habitat Mapping
• Natural Resource Management
• Landform & Stability Maps
• Tropical Rivers

Comparison of different broad-scale acoustic sensors

If the use of aerial and satellite imagery is inappropriate (e.g. the water is too turbid or too deep for optics) then the acoustic remote sensing strategies that could be employed, in increasing order of cost, are as outlined in the broad-scale acoustic sensor section, and are able to be used in benefit to habitat mapping.

Acoustic systems: a comparison and application of acoustic systems

Kennedy P K. Marmion Marine Park Hydro Acoustic Field Trials 2004 and 2005 (3 MB PDF). Prepared for the Coastal CRC by Fugro Survey Pty Ltd , May 2005.

This document describes a systematic comparison of the GeoAcoustics GeoSwath against the Reson 8125 and Reson 8101 Multibeam systems. Acquisition, Processing and analysis of the data are made. The GeoSwath system shows great potential. Subject to a few outstanding issues, none of which are insurmountable, the system will provide a good alternative to the established Multibeam technologies.

Contents

1. Introduction / background
  State of Victoria Habitat Mapping Project
  Marmion Marine Park: Area A
  West End, Rottnest Island: Area B
2. Objectives
3. Equipment
4. Survey design and parameters
5. Data processing methods
  Reson Seabat 8101 and 8125
  GeoAcoustics GeoSwath
  EG&G 272 analogue sidescan sonar
  Klein 5500 multiple beam digital sidescan sonar
  C-Max CM2 digital sidescan sonar
6. Results
  Validate the Starfix interface to the GeoSwath system
  Determine the true usable swath width of the GeoSwath systems in varying water depths over varying seabed types
  Determine the maximum operating water depth of the GeoSwath
  Validate the quality of the bathymetry acquired by the GeoSwath interferometric system in comparison with
  the known quality of the Reson 8101 and Reson 8125 systems
  Determine the required hydro acoustic deliverables for habitat mapping through exchange of technical information with
  the Coastal CRC, extend the Fugro knowledge base, and derive recommendations for the Victorian Habitat Mapping Project
  Multibeam bathymetric data deliverables
  Multibeam backscatter data – snippets
  Sidescan sonar
  Classification maps
7. Conduct of a system comparison for habitat mapping purposes of the Geoswath, Reson SeaBat8101 and 8125
8.Conclusions and recommendations
  Equipment installation
  Motion system
  Raw data formats
  GeoSwath
  Multibeam backscatter
  Proposed system for habitat mapping
  Processed data formats
  Automated classification mechanisms
  Towed video system

1. Introduction/background

This section describes a systematic comparison of the GeoAcoustics GeoSwath against the Reson 8125 and Reson 8101 Multibeam systems. It has been drawn from a technical report prepared by Fugro Survey in preparation for a major benthic mapping project off the Victorian coast in Australia. The prospect of broad scale habitat mapping on a commercial scale off the Victorian coastline triggered the creation of a decision making process to identify the most appropriate hydro acoustic and video survey hardware for that project. Carried out in Marmion Marine Park, this selection process came to be known as the "Marmion trials".

Acquisition, Processing and analysis of acoustic data is discussed. A key issue addressed in this Chapter was to determine which of several possible full coverage acoustic systems would be preferable for the Victorian task; notably whether an interferometric sidescan or a multibeam system would be preferred. The GeoSwath interferometric system showed considerable potential, but the comparison work undertaken led to the selection of a Reson 8101 system for the survey. Subject to a few outstanding issues, none of which are insurmountable, the interferometric system will provide a good alternative to the established Multibeam technologies. Multibeam backscatter data from the systems is also discussed, and the prospect of this replacing traditional Sidescan sonar is suggested. Data deliverables for habitat mapping projects are described.

State of Victoria Habitat Mapping Project

A series of marine parks under the management of Parks Victoria and the Australian National Heritage Trust were targeted for broad scale habitat mapping. These parks totalling more than 70,000 hectares were generally in remote exposed locations, far from port or shelter (see Figure 1). Water depths ranged from 10 to 90m. 100% mapping coverage was required for each park, such that a full understanding of the existing habitats within those parks could be established.

Figure 1. Polygons representing the marine parks to be mapped. The remote locations and long transit times between sites imposed some constraints on the project.

Marmion Marine Park: Area A

Marmion Marine Park is located off the metropolitan region of Perth, Western Australia (Figure 2). A few minutes steam from Hillary’s boat harbour, near Perth in Western Australia, this area is both easily accessible, and has a high variability of habitats.

Figure 2. Primary area for habitat mapping trials. A 2km by 0.5km detailed site survey over ecologically sensitive seabed of coastal habitat mapping significance within the Marmion Marine Park. Surveyed with Reson 8101, Reson 8125 and GeoAcoustics GeoSwath Multibeam systems, and EG&G 272, C-Max and Klein 5500 Sidescan sonar systems, this data provides an excellent proving ground.

West End, Rottnest Island: Area B

Figure 3. A single line of data was acquired for deep water trials west of Rottnest Island with a GeoAcoustics GeoSwath interferometric sonar.

In order to establish a realistic and reliable operational maximum depth for the GeoSwath, a transit from shallow to deep water was selected off the north-western end of Rottnest Island, Western Australia (Figure 3). Depths ranged from 30-200 metres.

2. Objectives

Objectives of the trials were grouped into specific commercial requirements for the Fugro Group, and generic habitat mapping requirements as follows:

• Field test a wireless radio link for construction applications
• Validate the GeoSwath XTF data format
• Field test the F180 motion sensor to attempt to replicate time jumps that have been reported in some field operations
• Train new Fugro Survey Pty Ltd field personnel on Multibeam operations, calibrations and systems
• Produce Fugro marketing material that may benefit the company and be incorporated into brochures
• Validate the Starfix interface to the Klein 5500 system in both an online driver capacity and an ability to log XTF data
• Increase our knowledge of the GeoSwath

Habitat Mapping Requirements

• Determine the true usable swath width of the GeoSwath systems in varying water depths over varying seabed types
• Determine the maximum useful operating water depth of the GeoSwath
• Validate the quality of the bathymetry acquired by the GeoSwath interferometric system in comparison with the known quality of the Reson 8101 and Reson 8125 systems
• Conduct a comparison of GeoSwath, Seabat, EG&G analogue Sidescan and Klein 5500 digital backscatter
• Determine the angular and spatial extent of the known data ‘holiday’ under the nadir of the GeoSwath system
• Exchange technical information with the Coastal CRC to determine the required deliverables for habitat mapping and further Fugro’s knowledge base in this field
• Derive recommendations for the Victorian habitat mapping project
• Conduct a side-by-side system comparison of the GeoSwath, 8101 and 8125 for the purpose of habitat mapping

Personnel
Eleven personnel from Fugro and the University of Western Australia were involved in the comparison exercise.

3. Equipment

To provide a unbiased comparison of the various hydro acoustic systems, the operating environment for each of the surveys had to be identical. The following constraints were set for all data collection during the trials;

• The same vessel
• The same bow mounting
• The same vessel driver
• The same navigation equipment
• The same motion sensor
• The same survey lines in the same direction
• The same survey operators

Vessel

The vessel used for the trials was the 12.5m long F/V Mirage (Figures 4 and 5), an ex-fishing vessel owned by Hillary’s Yacht Club.

Figure 4. The M/V Mirage – A photo of the Mirage showing the GPS antennas and the bow installation to mount the acoustic sensors.
Figure 5. Offset diagram of the M/V Mirage – showing the three dimensional offsets of the equipment installed on the vessel.

Surface Navigation Sensors

The surface navigation sensors (Figure 6) used for the trials were the same for the entire trials; this largely eliminated external error sources.

• Starfix.Seis online navigation system
• Starfix online logging and display system
• F180 inertial position, heading and motion sensor
• Fugro OmniSTAR 3000L differential corrections receiver
• Honda 2.2kV generator
• 2 Uninterruptible power supplies

• Fugro OmniSTAR 3000L differential corrections receiver
• Applanix POS M/V inertial position, heading and motion sensor

Figure 6. Surface navigation equipment.

Transducer Mounting
Traditional single beam operations using a vessel of opportunity, are typically conducted using an over the side pole arrangement. Unfortunately, this installation approach is often adopted for Multibeam operations resulting in pole wobble, which introduces motion artefacts into the bathymetry data. Rectifying these artefacts during data processing is extremely time-consuming. At best it is a workaround. More commonly, and at worst they cannot be fixed, thus compromising the quality of the bathymetric surface.

Best practice methodologies for transducer installation are direct hull mounted flanges. This requires the boat to be slipped, a flange and cable gland to be installed through the vessel hull. Divers are then used to install the transducer, and wet mate the deck cable to the transducer. Second tier practice is to bow mount the transducer with a flush mounted pole running parallel to the vessel bow (see Figure 7). This ensures the pole does not have sufficient unrestrained length to introduce wobble. As the boat travels faster, the mount becomes more stable.

Figure 7. Bow pole arrangement for fixing the hydro acoustic transducers. All systems were mounted to this common flange. This ensured no additional error sources were introduced in the systematic comparison. In this case, the freestanding section is less than 300mm in length.

In the case of MV Mirage, the hull is wooden, precluding the installation of a hull flange. Instead, an aluminium bow pole was employed. The pole was purpose designed for the shape of Mirage’s bow, to minimise the freestanding section (Figure 7.7).

Hydro Acoustic Sensors
The following hydro acoustic devices were mobilized for the trials:

• GeoAcoustics GeoSwath MBES
• Reson Seabat 8101 MBES
• Reson Seabat 8125 MBES
• EG&G 272 Analogue Sidescan sonar
• Klein 5500 multiple beam digital Sidescan sonar
• C-Max Sidescan sonar

GeoAcoustics GeoSwath
The GeoSwath (Figure 8) is an interferometric swath system that can be configured either with 125 kHz or 250 kHz transducers. The unit used for these trials was fitted with 250 kHz transducers. As this system is interferometric the number of beams across a swath is defined by a user defined sample rate. Essentially this sample rate is configured in software rather than by the physical design of the transducers.

GeoAcoustics published (GeoSwath 2004) specifications for the GeoSwath state that the system is capable of a swath width twelve times the water depth or 161 degrees with a maximum operating depth of 100m. The GeoSwath produces a travel time and vertical angle for each beam across the swath and two channels of Sidescan intensity data. Data is logged to proprietary Raw Data Files (RDF).

Due to its claims of high resolution, very wide swath, and excellent backscatter, the GeoSwath was considered a likely candidate for the Victorian habitat mapping project.

Figure 8. The GeoSwath transducer mounted on MV Mirage bow mount.

It should be noted that the GeoSwath trials were not without problems. Field trials were initially carried out on October 2004 and repeated in April 2005.

Unfortunately, the results from the first survey were unacceptable. Bathymetry quality was poor, backscatter was unusable, and there was a clear time stamping issue that could not be resolved. Most of these problems were identified as software problems, or user documentation related.

In order to give the system a further opportunity, GeoAcoustics were invited to attend a second trial. With the addition of Tom Hiller of GeoAcoustics, and new software, data quality was significantly improved. Once again, this trial revealed a problem in the synchronisation of the GeoSwath time stamping algorithm to GPS 1 pulse per second (PPS). This jitter in acquisition time stamps resulted in motion artefacts visible in the resulting bathymetric surface. GeoAcoustics made attempts to post-process this error out of the data, but it was still apparent in the resulting data. Data from the second trial is presented in this document. The first trial has largely been discarded.

Reson Seabat 8101
The Reson 8101 (Figure 9) is a traditional beam forming Multibeam system that operates on a frequency of 240 kHz. This system measures 101 ranges at 1.5 degree spacing resulting in a 150 degree (or 7.4XWD) swath to a maximum depth of 200m. Whilst not at the cutting edge of technology, due to its reliability, accuracy and flexibility, the 8101 is considered one of the workhorses of the commercial Multibeam survey industry.

The 8101 produces a travel time for each beam, two channels of Multibeam Sidescan and two channels of snippets backscatter data derived from the individual beams. All these are logged to XTF files in real time.

Figure 9. The 8101 transducer and the topside 81P processor.

Unfortunately due to operational time constraints the only operational time window available for trials with the 8101 was during unfavourable weather. Notwithstanding this, the results from the 8101 are excellent.

Prior to the trials, the 8101 was considered to be the second best option for the Victorian habitat mapping project. Its 150 degree swath width was considered a limitation to cost-efficient, broad scale mapping. The system was operated on a 75m range using standard Fugro operation settings.

Reson Seabat 8125

The 8125 (Figure 10) operates at a frequency of 455 kHz to produce 240 individual beams at 0.5 degree spacing from the transducer. This system has been designed for shallow water channel and clearance surveys and as such has a relatively narrow swath width of 120 degrees and a shallow maximum operating depth between 45m and 60m depending on seabed type.

Figure 10. The Reson 8125 transducer on the bow mount of the MV Mirage. Note the requirement for the SVP probe alongside the head. The flat acoustic transducer face requires an instant and accurate value for the velocity of sound at the head. This is used for determination of direction of acoustic reception.

In a similar manner to the 8101, the 8125 produces a travel time for each beam, two channels of Multibeam Sidescan and two channels of snippets backscatter data derived from the individual beams. All these are logged to XTF files in real time.

Due to the relatively narrow scanning sector combined with the maximum depth limitation, the 8125 was not a considered a viable option for the Victorian habitat mapping project, but it is widely considered the benchmark in Multibeam echosounder systems, by which all other systems are compared.

The system was operated on a 75m range using standard Fugro operation settings.

EG&G 272 analogue Sidescan sonar

The EGG 272 analogue Sidescan sonar (Figure 11) is a conventional single frequency sonar operating at 100 kHz. For the trials, the 272 was run on a range scale of 75m per channel. Chesapeake SonarWiz was used to perform the A/D conversion, and data was logged to XTF files via a post process format conversion routine.

Traditional survey techniques for the gathering and assessment of geological and benthic presence on the seafloor commonly use Sidescan technology. As such, the Sidescan sonar was considered as a basic requirement.

Figure 11. The tow fish and its descendant, the GeoAcoustics 159D, are the real workhorses of the shallow geophysical industry. Simple to operate and maintain, they are well understood, robust and reliable.

Klein 5500 multiple beam digital Sidescan sonar

Kindly provided by the DSTO the Klein 5500 digital Sidescan sonar (Figure 12) consists a five beam Sidescan sonar (Figure 13) designed for Hydrographic applications requiring high resolution images of the sea floor and bottom obstructions while operating at tow speeds of up to 10 knots.

The system achieves this by creating five acoustic pings simultaneously at different frequencies along the length of the 2m tow fish, thereby ensonifying 2m of seabed along track with each ping and enabling the fish to move 2m before the next ping without creating data gaps/holidays. With a 100m per channel range scale the fish can ping seven times per second and therefore move at 14 knots.

The Klein 5500 is widely considered the ‘Rolls Royce’ of Sidescan sonar’s. Whilst the equipment was not available or considered for the Victoria habitat mapping project, it was included in the trials as it an ideal benchmark.

Figure 12. The Klein 5500 installed on the bow of MV Mirage. Although this is not a standard operating mode, it provided excellent geo-referencing of the data, and was the safest option in such shallow water.

Figure 13. Klein 5500 beam pattern. With the five simultaneous beams, 100% coverage at a high survey speed is possible.

C-Max CM2 digital Sidescan sonar

The C-Max CM2 digital Sidescan sonar (Figure 14) is modern, dual frequency sonar operating at either 325 or 780 kHz. A second unit operating on 100/325 is also under manufacture but was not available for trials.

For the trials the CM2 was run on a range scale of 75m per channel.

Figure 14. The CM2 digital tow fish off the stern of MV Mirage. As the fish has no fixed mounting point, it had to be towed behind the vessel. Also shown is the "CM2 C-Case" all in one acquisition and logging unit.

Towed video camera

The University of Western Australia provided a towed video camera system for the purposes of visual inspection (ground truthing) of the surveyed area (Figure 15). The system comprised a high resolution CCD colour camera in an underwater housing. A composite video signal to a topside digital video logging (DV) combined with a digital video overlay of time and position was used to geo-reference the video data. No USBL and no artificial lighting were employed for these trials.

Figure 15. The UWA towed video system. The underwater lights in this photo were not used in these trials.

4. Survey design and parameters

All coordinates are referenced to the World Geodetic System 1984 (WGS84) datum. Fugro’s Differential GPS Reference Stations are currently defined in the International Terrestrial Reference Frame 2000 (ITRF2000 Epoch 2004.75) datum. Due to the continual refinement of the WGS84 reference frame, for all cases, the transformation parameters indicate that the WGS84 and ITRF2000 reference frames are essentially identical.

Datum
Reference Spheroid: World Geodetic System 1984
Semi Major Axis: 6378137.000m
Inverse flattening: 298.257223563

Grid projection
Projection: Universal Transverse Mercator
Latitude of Origin: 0°
Central Meridian: 117° E (UTM Zone 50)
Central Scale Factor: 0.9996
False Easting: 500000m
False Northing: 10000000m
Units: Metres

A primary factor in the survey line design was the purpose of the survey. Commercial surveys typically utilise a line pattern with optimally spaced parallel and an agreed line overlap. For the Marmion area depth range of 5m to 16m (average 10m), and the specific project requirements, a line spacing of 20m was used (see Figure 16). This provided a heavy degree of overlapped data, which could be utilised in the analysis of the outer beams with respect to both inner and nadir beams of adjacent lines.

For a commercial operation, 40m line spacing would be more appropriate.

Figure 16. Survey line plan in Marmion Marine Park. Double density line spacing was used to provide excellent overlapped Multibeam bathymetry. This assisted in the analysis and comparison of the various systems.

5. Data processing methods

A primary aim of the trials was to process the data from the various systems as fairly as possible, such that sensible comparisons could be made.

In all cases, data was processed using industry standard ‘best practice’ as much as possible. In many instances, the Fugro Starfix suite was utilised. This provided a single platform to process the various datasets. This ensured common processing procedures and algorithms were followed. In some instances, such as the GeoSwath system, a comprehensive processing suite from the equipment vendor was used.

In all instances care was taken to perform similar de-spiking and smoothing processes. This provided a realistic product from the systems under trial. All Multibeam bathymetry was reduced to LAT Hillary’s boat harbour using observed tides for Hillary’s from the Department of Land Information (DLI). All Multibeam bathymetry was gridded to a 0.5m pixel size using the same gridding engine. This ensured the gridding algorithm was not a factor in the final data comparisons. All Sidescan sonar backscatter data was gridded (mosaiced) to a 0.5m resolution. This permitted perfect overlap and alignment with the Multibeam bathymetry. In each case, the Sidescan was mosaiced using the same software. On completion of processing, all data was imported into the Marine Survey Data Model (MSDM). This is a relational data model residing within the ESRI ArcGIS 9.0 environment.

Reson Seabat 8101 and 8125

Both Reson systems were processed using Starfix.Proc. Proc provides both an automated batch processing engine and interactive manual editing framework to marine survey data sets. Heavy emphasis is placed on the batch processing of data, as this ensures consistency, efficiency and repeatability. Manual interaction is largely reduced to quality control of the automation process.

The basic processing flow is as follows (Figure 17):

Figure 17. Data processing pipeline for the Reson Seabat systems. Both Reson data sets were processed without any difficulties. Processing rates exceeding real time capture rates were easily achieved.

GeoAcoustics GeoSwath

The GeoSwath was processed using the GS+ software version 3.08. This is a manual interactive application which permits the user to replay the acquired data, apply filters, tides, calibration corrections and reduce the data to both random XYZ swath (bathymetry) files, and XYI swamp (backscatter) files in a single integrated environment.

The basic processing flow is as follows (Figure 18):

Figure 18. Data processing pipeline for the GeoAcoustics GeoSwath system

It should be noted that the significant problems in the October 2004 trials data resulted in it being discarded. These problems are discussed below. Processing rates for the GeoSwath data did not exceed data capture rates. This was largely due to replay approach to data processing, in which the user is required interact with the processing filters during processing. On a flat seafloor, unattended GeoSwath processing may be possible, but in the reef zone of Marmion Marine Park, this was not feasible.

EG&G 272 analogue sidescan sonar

The EG&G system was logged and processed using the Chesapeake SonarWiz software by Andy Bickers at UWA. The results of the mosaicing engine are geo-referenced tiff files, which are imported into the MSDM (see Figure 19).

Figure 19. Data processing pipeline for the EG&G 272 Sidescan sonar.

Klein 5500 multiple beam digital sidescan sonar

The Klein 5500 system was logged with the Klein SonarPro acquisition system into .SDF files. These were converted into XTF files using a utility from Klein. The XTF files were then mosaiced using both CODA and Starfix.SonarMap. The results of the mosaicing engine are geo-referenced tiff files, which are imported into the MSDM (see Figure 20).

Figure 20. Data processing pipeline for the Klein 5500 Sidescan sonar.

C-Max CM2 digital sidescan sonar

The CM2 digital logging system was used to acquire data, and Chesapeake SonarWiz was used to convert into regular XTF format. The XTF files were then mosaiced using SonarWiz, CODA and Starfix.SonarMap. The results of the mosaicing engine are geo-referenced tiff files, which are imported into the MSDM (see Figure 7.21).

Figure 21. Data processing pipeline for the C-Max Sidescan sonar.

6. Results

Due to the objectives covering many areas each goal is addressed in turn for clarity.

Validate the Starfix interface to the GeoSwath system

The GeoSwath system is an interferometric system using phase measurements to calculate the angle of the incoming seabed return. As such, the number of individual depth measurements taken across the swath is purely dependant on how frequently the angle measurement is taken. Given that there is no physical limitation to the number of soundings that can be sub-sampled for each acoustic ping the manufacturer calculate as many angles as they deem necessary.

GeoAcoustics have decided to take advantage of this ability and have designed the system to take an enormous number of angle measurements per ping. As the system is taking an angle measurement at each epoch the number of measurements increases with the user selected swath width.

Due to the large quantity of data the system produces with each ping, several modifications to the Starfix suite were required to interface and log the raw data from the GeoSwath.

As the system has separate transducers for port and starboard, which each have there own set of pitch, roll and yaw/alignment corrections, the data is published as two separate Multibeam devices from the one IOWin DLL.

The message manager system is designed to deal with data packets of a maximum 32 kilobyte size. If the GeoSwath system is set to operate on a swath greater than 25m per channel, the output data packets exceed this maximum message size. MMDataLib.DLL was modified to publish a header for each ping and then follow this header with as many separate packets that are required to handle the ping. This modification allowed other applications to subscribe to the messages.

When configuring the system to display in real time for waterfall and coverage displays both port and starboard transducers must be configured in Multibeam. The corrections for the roll angle of each head must be entered into the Multibeam setup, the port transducer roll correction being approximately -30 ° +/- patch test result and starboard being +30 ° +/- patch test result.

RTGraph was modified to allow the intensity window above the waterfall display to show up to 2,500 beams per Multibeam swath. Both Multibeam messages can be displayed in real time waterfall and one shot displays. These real time displays have been modified to display only one depth per pixel, which is the mean depth within that across track distance. GeoSwath intensity data can be displayed by selecting the palette and intensity options in RTGraph; these will also be decimated in real time display to one intensity value per pixel.

As the displays in the real time coverage package decimate the displayed data by binning it to a grid size this application required no modification.

The data is currently logged in STS format in IOWin. No facility has been made to log Fugro Binary Format files (FBF) as this package is designed to use fixed packet files and does not use MMDataLib. There is also no ability to log the data in XTF format as there are no packages Fugro presently operate that could read GeoSwath XTF files. This could be incorporated if there was a use for this format.

The GeoSwath logs Raw Data Files (RDF) in real time that contains all data in a range and angle pseudo raw format. An import wizard to bring this data into Proc in POS format has been written. As the system is capable of collecting data to twelve times water depth even though it only receives a seabed return from eight times at best, the import wizard has been configured to allow the user to import a requested swath angle rather than all the data.

The GeoSwath Plus software will process the data to produce SWATH files containing bathymetry data and SWAMP files containing intensity data. SfxXYZ has been modified to read these files allowing them to be displayed in DataView, converted to other standard file formats and gridded using DTM to produce bathymetry grid or intensity mosaics. As DTM has no de-spiking or TVG controls the data must be processed via the GeoSwath Plus playback software prior to gridding.

Determine the true usable swath width of the GeoSwath systems in varying water depths over varying seabed types

As with all swath bathymetry systems the usable swath width of the system is dependant on the seabed sediment type, seabed topography, water depth and prevailing weather conditions. An example for the GeoSwath system is shown in Figure 22.

Figure 22. GeoSwath swath width examples. The green soundings above indicate the soundings that fall within eight times the water depth. The red soundings are those that fall between the eight and twelve time’s water depth.

Data was logged in very calm weather offshore of Hillary’s marina over a flat seabed consisting of coarsely grained sand in approximately 9m depths. This ensured almost no motion, minimal seabed artefacts or slopes with a strong seabed return well within the operating depths of the system.

Given optimum system operating conditions the maximum usable swath width achievable with the 250 kHz supplied was eight times the water depth. It is envisaged that less than ideal weather conditions will see the useable swath width reduce.

Determine the maximum operating water depth of the GeoSwath

To determine the maximum operating water depth of the 250 kHz system data were logged over a sandy seabed to a water depth of 100m, west of Rottnest Island. The system was configured with maximum power setting and pulse length.

Figure 23. GeoSwath single beam maximum operational depth was not as good as expected.

The GeoSwath uses a narrow single beam echosounder mounted between the two transducers to assist in depth determination at nadir. This echosounder lost track of the seabed at approximately 50m depths and did not record any usable depths beyond this depth (Figure 23). The sounder did return to normal operation at almost exactly the same depth upon return to shallow water.

The interferometric swath data continued to track the seabed with a continual narrowing in usable swath width. At approximately 75m depths the usable swath width reduced to approximately two times the water depth and the ambiguity between soundings had increased to approximately 2m in height making the true location of the seabed questionable (Figure 24).

Figure 24. GeoSwath cross profile in 82m water depths. This cross profile is indicative of the data collected off Rottnest.

Validate the quality of the bathymetry acquired by the GeoSwath interferometric system in comparison with the known quality of the Reson 8101 and Reson 8125 systems

To provide a valid comparison, the operating environment of close each system was as to identical as possible. Each system was operated using the same vessel, mount, positioning system and motion sensor over the same patch of seabed. This makes the Marmion dataset somewhat unique.

Figure 25. Reson 8101 survey weather conditions were less than ideal. Credit should be given to the F180 GPS system which performed admirably underwater!

Unfortunately the only variable beyond operator control was the weather. The 8101 data was collected in less than ideal weather conditions and the data suffered as a result (see Figure 25).

Quality of individual soundings across the swath

Figure 26. A cross profile of a single swath of filtered GeoSwath data. The soundings here are all considered good.

Several operators have noted the "bow tie" effect across the swath of the GeoSwath data (see Figure 26). This artefact is common to all data logged during these trials. As seen in a single swath of GeoSwath data below (clipped to the eight times water depth as mentioned in Section Determine the true usable swath width of the GeoSwath systems in varying water depths over varying seabed types ), the uncertainty in depth at nadir spans approximately 0.2m from 12.4m to 12.6m depths (Figure 26). The uncertainty in depth increases with the distance from nadir, at a distance two times water depth from nadir (four times water depth total swath width) the spread of uncertainty increases to 0.8m from 11.8m to 12.6m on the starboard channel and 12.5 to 13.3m on the port channel. Note that this data has already been de-spiked for gross outliers.

This profile was measured over a hard, relatively flat seabed (slight side slope) with minimal motion in shallow water and is representative of the dataset.

Computation of a Multibeam error budget to IHO specifications requires an integration of error budgets for all the devices used in the survey, including the motion system, the positioning system, the tide gauge etc. Every system put in place for these trials was to industry best practice, and each system exceeds the IHO special order specifications.

The cross profile display presented above indicates a significant variability in depth, increasing with across track distance. Taking each beam on an individual basis these trials initially suggest the achieved accuracy was sufficient to meet IHO Special Order guidelines for the data collected in less than 20m depths for a swath width of approximately one times water depth. Any data collected beyond the one times water depth swath width or in water deeper than 20m met IHO first order specifications to a maximum swath width of approximately six times water depth at which point the data quality drops to IHO Order 2.

Indeed, as often noted by GeoAcoustics, beams should not be taken on an individual basis. The true depth will not be represented from any individual observation. The true depth is represented by a statistical model of the observations, typically generated by a gridding process. This is based on sound principals and is a valid approach.

There are no operator configurable filters that can be used to smooth the data but there are several configurable gate and de-spiking settings that greatly improve the data. The figure below shows the same swath of data shown in Figure 27 illustrating the online de-spiking, where the individual red data points have been rejected and the green accepted. The large red and green blocks are the extents of the gating values.

Figure 27. A single swath of de-spiked GeoSwath data.

Given the large volume of data in each swath and the large vertical spread between individual soundings across the swath the filters do a very good job of de-spiking, thereby limiting the volume of data written to the processed Swath files.

The GeoSwath does suffer on slopes where there is ambiguity in the grazing angle for a given epoch. As with all interferometric systems when the slope on the seabed reaches a stage such that the seafloor is perpendicular to the acoustic transmission, the phase calculation begins to fail.

Use of the de-spiking and gating in the GeoSwath Plus software can aid in the tracking of the seabed over steep terrain, but require constant attention over an undulating seabed. In the case of the Marmion data, this proved a successful technique, and resulted in a good clean dataset. The cost impact of this approach is discussed in the conclusions below.

For purposes of Multibeam bathymetry comparison, we gridded the three datasets using the same parameters (Figures 28 to 30). Gridding was carried out to a 0.5 metre pixel resolution, which is appropriate in the 5-16m water depth range. Following gridding, interpolation across empty pixels using a search radius of 5 * 5 pixels (2.5m) was employed to provide a seamless surface. Finally, the surface was smoothed using a median 3 * 3 filter. This largely removed any remaining noise in the surface, whilst retaining the seafloor features.

By clipping the outer beams to between six and eight times water depth, the GeoSwath gridded bathymetry compared well with the Reson systems. The slight drop in sharpness of seafloor features is to be expected, and does not significantly impact the final result.

Figure 28. Gridded Reson 8125 bathymetry over a shallow reef. Note how clearly the features are defined. Grid axes spacing is 50m.

Figure 29. Gridded 8101 bathymetry over the same shallow reef. Note the small artefacts in the data due to adverse weather conditions

Figure 30. Gridded GeoSwath bathymetry data over the same shallow reef. Whilst all the features are still in place, and have similar depths, there is smearing of the features. This is caused by the higher variance in the point cloud. Note the artefacts due to the time stamping issue. Determine the angular and spatial extent of the data ‘holiday’ under the nadir of the GeoSwath system Prior to these trials it had been reported that the GeoSwath system suffered from a data gap under nadir and determining the extent of this gap was one of the objectives of the field trials.

This gap no longer exists. Even if the Sidescan data is logged without any slant range correction the bathymetry still applies a slant range correction. Initially it was assumed this bathymetry slant range correction was supplied by the echosounder but during the deep water phase of the trials after the echosounder failed at 50m depths the bathymetry was still slant range corrected.

Determine the required hydro acoustic deliverables for habitat mapping through exchange of technical information with the Coastal CRC, extend the Fugro knowledge base, and derive recommendations for the Victorian Habitat Mapping Project

The primary goal for habitat mapping is to accurately identify the spatial location, extent and characteristics of differing habitats on the seafloor. Traditional approaches of diving, still camera or towed video surveys provide the most direct mechanism to observe the characteristics of habitats. Unfortunately, these fall short on the spatial location and extents of the habitat in question. When costed on a coverage basis, these traditional techniques prove to be prohibitively expensive, prone to omissions and relatively unsafe survey techniques. A more cost effective and rigorous approach is required to identify all habitats at a broad scale as efficiently as possible. This would then be followed up with targeted towed video camera and stills photography in identified areas of interest. Data from this follow up survey is used for the actual habitat classification of the areas identified in the hydro acoustic survey.

For the purposes of seafloor mapping, broad scale hydro acoustic survey acquisition generally fall into the realm of Multibeam bathymetry and Sidescan sonar. Whilst it is also possible to include single beam echo sounding in this category, it is not considered to be a cost effective solution where full coverage of the seafloor is expected, so is not included in this discussion.

Multibeam bathymetric data deliverables

A typical deliverable from a Multibeam survey is a bathymetric surface. This surface is the result of a complex series of processes, including calibration, filtering, reduction, gridding and smoothing. Once a seamless bathymetric surface achieved, it is possible to derive other surfaces such as slope, aspect, and rugosity.

Presented below are examples of surfaces generated from Multibeam bathymetry (Figures 31 to 34).

Whilst not a direct classification mechanism, surface of shaded relief are intuitive mapping techniques. The use of slope often highlights subtle changes in relief as low as 5cm. These can often easily be identified as sand waves, outcropping rock, reef, and even biota such as seagrass.

Figure 31. Surface of seafloor depth from the Reson 8101. A simple transform of depth to colour provides useful information. It is widely understood that habitats change with depth. As such, seafloor depth is a significant factor in automated classification algorithms.

Figure 32. Surface of seafloor slope from the Reson 8101. Depiction of relief via the use of shaded relief maps provides immediate visual cues to the location and spatial extents of potential habitats.

Figure 33. The combination of shaded relief and colour depth surfaces provides a primary presentation mechanism for the bathymetric data component of any Multibeam survey. It is easily understood by a casual observer and skilled technician alike.

Figure 34. (a) Bathymetry quality surface provides a useful metadata layer. This can be used as a masking layer to omit noisy or unreliable data from classification systems; (b) NSamples surface provides a metadata surface indicating how many bathymetric observations were utilised in order to determine the depth for each pixel. This surface is very useful when analysing the survey to determine how efficiently it was undertaken.

Multibeam backscatter data – snippets

As discussed above, in section 5: data processing to determine seabed types, an emerging by-product of Multibeam bathymetric systems is backscatter acoustic intensity from the seafloor. This data can be used as a surrogate to infer impedance and roughness of the seafloor (Beaudoin et al., 2002). This data is logged simultaneously with the Multibeam bathymetry. For each ping, each sample of backscatter is associated to the relevant beam number. This permits correct co-registration of the backscatter with respect to the seafloor (Fugro 2005: Backscatter Bible.Doc " Fugro Survey internal technical note ).

Configuration of the Multibeam hardware to simultaneously acquire high quality bathymetry and backscatter is still the subject of review. There is a trade off between the two datasets. A perfect configuration for bathymetry can have a detrimental impact on the backscatter and vice versa. From field trials, the work done by the CWHM Group has established optimum operational parameters. Additional options within the Multibeam system to configure the type of backscatter logged (eg flat bottom or uniform) have significant impact on the quality of backscatter data. This can be noted in the difference in the 8101 and 8125 data presented below (see Figures 35 to 39).

Figure 35. Snippet backscatter data from the 8101 using the UNB1 processing algorithm. All major features are identified within the backscatter data, and co-location of adjacent swaths is correct. It should be noted that the "uniform snippet" backscatter was used in this trial (Reson 2000). Subsequent to these trials it is suggested that "FlatBottom snippets" is more appropriate (Gavrilov 1 , personal communication).

Figure 36. Snippet backscatter data from the 8101 using a Per Beam Variance processing algorithm. All major features are identified within the backscatter data, and co-location of adjacent swaths is correct. It should be noted that the "uniform snippet" backscatter was used in this trial. Subsequent to these trials it is suggested that "FlatBottom snippets" is more appropriate (Gavrilov, personal communication).

Figure 37. Backscatter data from the GeoSwath. All major features are identified within the backscatter data, and co-location of adjacent swaths is correct. Nadir effects in the data possibly caused by interferometric bottom tracking problems can clearly be seen. Overall, the GeoSwath provided the highest resolution backscatter of all the Multibeam systems used.

Figure 38. Snippet backscatter data from the 8125 using the UNB1 processing algorithm. All major features are identified within the backscatter data, and co-location of adjacent swaths is correct. Minor features were not as clearly identifiable as the GeoSwath. Reef structure was not as clearly defined in the 8125 backscatter. Nadir effects are not as pronounced as the GeoSwath data.

Figure 39. Snippet backscatter data from the 8125 using the Per Beam Variance processing algorithm. All major features are identified within the backscatter data, and co-location of adjacent swaths is correct. Minor features were as clearly identifiable as the GeoSwath. Reef structure was well defined in the 8125 backscatter. Nadir effects are not as pronounced as the GeoSwath data, but still a serious impediment to automated classification.

Sidescan sonar

Mosaics for each of the Sidescan sonar’s are presented here (see Figures 40 to 42). It should be noted that the EG&G and C-Max were towed behind the vessel rather than bow mounted. This resulted in improved data quality, but degradation in the co-registration of data with respect to adjacent survey lines. This poor geo-referencing causes significant processing delays, and has implications on the consistency of automated classification systems.

Classification maps

The ultimate goal of a habitat mapping project is to generate a classified map of the various habitats in the survey area. A long standing goal is to create a rigorous automated classification system. The success of such a system requires good quality data inputs. At present the bathymetric surfaces consistently provide such quality surfaces (see Figure 43).

Research using the backscatter surface for the purposes of classification is underway. Nadir artefacts in the backscatter data are the subject of much scrutiny. These must to be resolved in the classification algorithms are to prove robust.

7. Conduct of a system comparison for habitat mapping purposes of the Geoswath, Reson SeaBat 8101 and 8125

The best hydro acoustic system for habitat mapping is a system that can comprehensively map the seafloor to identify all significant features with minimal vessel time and equipment cost. In the first respect, all Multibeam systems tested identified all significant bathymetric and textural features in the trial area. Similarly, all the Sidescan sonar systems identified all significant textural features. Validation of these features was carried out using the towed video system. This confirmed the existence and correct identification of the features.

Although the Reson 8125 system has the highest resolution bathymetry, and very good backscatter, depth range limitations preclude the system from being an option for the Victorian habitat mapping project. Had the depth range not been an issue, the 8125 would still have been rejected on the basis of the relatively narrow (120-degree) swath width. The relatively narrow sector would increase the survey duration beyond economic limitations in comparison to the other systems available.

Given the depth range provided by the Marmion site, both bathymetry and backscatter from the GeoSwath was more than adequate for the purpose of habitat mapping. The hull mounted transducers make this a good option over the combination of 8101 and towed Sidescan, as this ensures excellent backscatter registration with respect to the bathymetry. As discussed earlier the GeoSwath trial was conducted on two occasions. Data from the second trial was much improved, but the significant time stamping problems identified in the first trial were still unresolved. Following the 2005 trials, GeoAcoustics have advised us a hardware modification to the GeoSwath system has resolved these time stamping issue. A third trial had been proposed, and undertaken in 2006, to confirm the system is fully operational. The second unresolved problem was of maximum operational depth. The useable cross profile in 75m water depths was reduced to a relatively narrow swath. It is considered that this was due to software problems in the first trial and should no longer be an issue, but field trials need to confirm this.

The Reson 8101 produced excellent bathymetry and good backscatter. Questionable backscatter at nadir, together with immature backscatter processing algorithms and associated applications at one stage made the proposition of only using the 8101 (i.e. without a towed Sidescan) a high risk venture. Further developments during 2005 have led however to a revision of this conclusion and at the time of writing, only the Reson 8101 will be deployed, with processing to yield both bathymetry and backscatter data.

Figure 40. EG&G 272 Sidescan sonar. Good across track resolution and a high degree of discrimination make this an invaluable tool. Poor geo-referencing of the towed fish caused by lack of accurate navigation, pitch, roll and heading reduce the absolute accuracy of the resulting mosaic, but relative positions of features are maintained.

Figure 41. C-Max Sidescan sonar suffered from bottom tracking problems. This caused significant AVG artefacts in the across track direction. The bottom tracking algorithm built into the C-Max is still under development.

Figure 42. The Klein 5500 has the best along track and across track resolution. Together with the accurate co-registration via the bow mount, this provided the highest backscatter from the Sidescan sonar’s in the trial.

Figure 43. An ISO unsupervised classification based on the bathymetric surface and its derivatives (Holmes 1, personal communication). The region has been classified into areas of low reef, high reef, deep sand and sand inundated. Correlation to the bathymetry and backscatter surfaces can clearly be seen.

Dr. A.N. Gavrilov is a Professor in Underwater Acoustics at the Centre for Marine Science and Technology in the Curtin University of Technology, Perth, Western Australia.

Dr. K. Holmes is a Postdoctoral Research Fellow at the School of Earth and Geographical Sciences in the University of Western Australia, Perth, Australia.

8. Conclusions and recommendations

Equipment installation

Critical to any Multibeam project is a robust, rigid mounting point for the Multibeam onto the vessel. Over the side poles should never be an option. The bow mount used on this project provided a good platform for these trials. Changing equipment over was both easy and safe.

Motion system

The F180 continues to provide excellent inertial motion correction data, and behaved impeccably throughout the trials.

Raw data formats

The acceptance of XTF as the de-facto standard transfer format in hydro acoustic surveys is established and provides an adequate and well understood mechanism.

GeoSwath

The GeoSwath provided a tantalising dataset. If the reported timing problems are satisfactorily resolved, then it should be re-considered for habitat mapping operations.

Questions of GeoSwath processing efficiency still need to be resolved. Broad scale mapping projects will require highly efficient batch processing of data. The current replay facilities in the GeoSwath need to be augmented with a more automated approach. The maximum operating depth range for the GeoSwath is still to be established by field trials. It is noted that a trial with the 125 KHz system should be undertaken.

Concerns over the ping rate of the GeoSwath have been raised. Due to the fact that the GeoSwath has two transducers operating on the same frequency, it alternates the ping cycle between the port and starboard channels resulting in half the along track resolution of the Reson systems. This will have an impact of maximum survey speeds.

Multibeam backscatter

Cost efficient mapping of habitats on a broad scale is essential to any successful project. In many cases, the most expensive line item in a project is the vessel and its day rate. Therefore making the most efficient use of ship time is essential in any project. Line turn time, online survey speed and swath width (coverage) are major factors affecting survey efficiency. Traditional survey speeds of 4 knots are required to maintain a Sidescan sonar fish altitude close to the seafloor, especially in deep water. Deeper water requires more tow cable behind the vessel, which in turn increases line turn time. Accurate positioning of the tow fish requires the use of an Ultra Short Baseline System (USBL), increasing the survey cost even further. With the approaching maturity of Multibeam backscatter acquisition and processing, the prospect of habitat mapping without the use of Sidescan sonar is becoming a reality. This removes the requirement for a USBL, winch, topside acquisition system, systems engineer and post processing of tow fish navigation, whilst simultaneously improving safety aspects associated with fish deployment and winch operations. Additional trials should be carried out at high survey speeds of 12 knots or more. This has a significant impact on the efficiency of survey operations.

Current backscatter processing algorithms retain a nadir artefact. This causes significant problems to the current suite of automated classification tools widely used in the field of remote sensing.

Proposed system for habitat mapping

From the data examples shown in this report the most appropriate equipment for the Victorian habitat mapping project is a combination of the 8101 bathymetry and the EG&G 272 (or equivalent) Sidescan sonar.

Ensure the Multibeam backscatter is acquired for the Victorian habitat mapping project, and compare the resulting backscatter maps with those derived from the Sidescan sonar. As the Multibeam backscatter processing system matures, re-consider the cost benefit of the Sidescan sonar and assess if it, and its dependencies can be withdrawn from the project.

Processed data formats

Processed data formats should be delivered in the Marine Survey Data Model (MSDM) This is an ESRI personal GeoDatabase, and can be easily understood using the built in metadata. Raster surfaces should also be delivered in ERMapper ECW and ESRI ASC formats.

Automated classification mechanisms

Several approaches to classification are being developed. Rugosity, and Topographic Position Index (Rinehart et al. 2004) is an example of such systems that need to be investigated and tested with actual data. Software to perform classification will be the subject of research over the coming months. With advances in automation in the processing of Multibeam bathymetry and backscatter, performing classification whilst at sea is has clear economic advantages over land based processing.

Towed video system

The towed video system worked very well. Outcomes from the trials were the lack of USBL. It is recommended that a USBL always be employed when conducting towed video. Without accurate positioning, the value of the video for habitat classification is significantly degraded. The towing arrangement for the trials was to manually hold the video umbilical. For broad scale mapping a capstan is required. Subsequent to these trials a new video system and capstan winch have been manufactured.

Figure 44. A new towed video system for habitat mapping operations.

Figure 45. A capstan winch for towing and hauling the video system.