Analysis of data from Multibeam and Sidescan Sonars

Due to the ocean floor's inaccessibility, acoustic remote-sensing with sonars, towed from research ships or installed in their hulls, presents the only feasible way of studying ocean floor processes quantitatively and over reasonably large areas. The following describes some of our effort to develop methods for processing and analysing the different kinds of data produced by these systems.

The types of sonar

Multiple beam echo-sounding sonars are designed to measure the water depth within a series of narrow 'pencil-like' beams, therefore creating a view of the topography of the ocean floor. These systems are usually attached to ship's hulls (e.g., manufacturers Sea Beam, Simrad, Hydrosweep, Reson). Sidescan sonars are designed to produce acoustic images of the seabed by sequentially scanning it with a narrow fan-shaped beam. These images are analogous to terrestrial remote-sensing images, in particular those from side-looking airborne radar (sidescan sonars include GLORIA, SeaMARC II, MR1). The figure on the right shows a GLORIA image collected by the US Geological Survey and IOS (UK) of a lava flow and old seamounts near Hawaii, and illustrates some of the difficulties of interpretation. The backscatter from the seafloor varies because of the varying angle of incidence (hence the shading around the seamounts) and also because of varying bottom properties (hence the difference between the lava flow and the adjacent sediments). Many modern multibeam sonars are also capable of generating an acoustic image, and sidescan sonars can have the capability of measuring seafloor topography, so these systems are becoming less distinct.

Processing

Multibeam sonars

Though we now use various software for processing multibeam sonar data (e.g., MB-System of Caress and Chayes), we have previously developed our own processing software with some unique features, such as despiking algorithms which adapt to the terrain ruggedness and a binning method which reports measures of depth reliability. Abstract

Sidescan sonars

Our earlier effort to produce maps from sidescan sonars focused on ways to project the data using bathymetry information from other multibeam systems to reduce geometrical and radiometric distortions. Abstract

Analysis

Quantifying sidescan sonar

Sidescan sonars give qualitative measures of the backscattering properties of the seafloor. If the data are suitably corrected, they may provide quantitative measures ("backscatter strength") which can then be compared with the results of laboratory or field experiments on different seafloor types and hence might be useful for assessing the content of a sonar image. This study attempted to quantify GLORIA sidescan sonar data. (Collaborator: Mike Somers) Abstract

Signal penetration into the seabed

Many reconnaissance sonars use low acoustic frequencies that are able to penetrate the seafloor where it consists of fine-grained sediment. Seafloor penetration is demonstrated by lava flows which are sometimes seen in GLORIA images even where they are buried by several metres of sediment. This study used a very simple acoustic model and published attenuation measurements to assess the depth of penetration. We showed how the model could in principle be used to extract information on sediment draping lava flows from the amount of signal attenuation in GLORIA sidescan sonar data, and also to infer variations in sediment cover over a mid-ocean ridge from backscatter variations. The first of the following figures shows the varying backscatter with incidence angle from the lava flow at the top of this web page, and the slope of the graph is consistent with simple attenuation in a surface sediment layer. ("Grazing angle" is the angle relative to the horizontal, i.e. 90 degrees minus the incidence angle.) The second figure shows the decreasing mean backscatter with distance from a mid-ocean ridge, reflecting increasing attenuation due to the increasingly thick cover of sediment with age of the volcanic seafloor. Further characteristics of these fluctiations can potentially be used to infer how the sediment is accumulating, for example by ponding or draping over the new volcanic seafloor. Abstract and Further information

Seabed classification with multibeam data

Most attempts to classify the seabed have been based on data from either sidescan sonars (using measures of image texture) or from broad-scale bathymetry, but not from the combination of both sidescan sonar and bathymetry. This project, carried out in collaboration with the University of New Brunswick and the Canadian Hydrographic Service, attempted a simple seabed classification using backscatter, slope and curvature measures derived from Simrad EM1000 data. The test area was a section of the Nova Scotian shelf and the classification was used to isolate small pockets of silt in depressions. Using the classification to interrogate their surface properties, we found that the silts had small surface gradients directed away from the shoreline, which we speculated may reflect offshore transport of sediment during storms. This was a very simple demonstration but indicates that information obtained by combining bathymetry and backscatter should improve seabed classifications and allow further class-specific data to be extracted. The colour figures below show images of the bathymetry and backscatter collected during this programme. (Collaborator: John Hughes Clarke) Abstract

Cable route selection

Routes for submarine telephone cables are chosen with various competing issues in mind: the route must be safe for the cable, and hence should avoid bare rock or steep slopes, but also must be as short as possible to keep the costs of cabling low. This study, carried out mostly by an MSc student in Durham (John Spencer), developed a simple algorithm for locating routes through complex terrains. The multibeam data (left) were used to calculate a representation of potential hazards to the cable, such as due to steep slopes or bare rock. The algorithm then applied graph theory to trace a route across the hazard map, with distances between nodes artificially distorted according to the specified hazard values. Abstract


Denoising of topographic/bathymetric datasets

In this project, a denoising algorithm developed by Xianfang Sun (Cardiff University) was applied to morphologic datasets by John Stevenson, while working as a researcher in Manchester. John's algorithms and advice on how to use the denoise method can be found here
.
Relevant publications

Funding for the above work was provided by Research Fellowships from the Royal Society, the NERC and the Roy. Comm. for the Exhibition of 1851, and by a studentship from the NERC. The fieldwork off Nova Scotia was supported by the Canadian Hydrographic Service while I was an International Research Fellow of the NSERC, Canada, based in Fredericton. John and Xiangfang's work was funded by an EPSRC grant to Paul Rosin in Cardiff.


Return to Neil Mitchell home page