ICES Journal of Marine Science: Journal du Conseil Advance Access originally published online on April 29, 2008
ICES Journal of Marine Science: Journal du Conseil 2008 65(6):1004-1011; doi:10.1093/icesjms/fsn061
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Acoustic seabed classification: current practice and future directions
1 Department of Fisheries and Oceans Canada, PO Box 5667, St John's, Newfoundland, Canada A1C 5X1
2 School for Marine Science and Technology, University of Massachusetts Dartmouth, 706 South Rodney French Boulevard, New Bedford, MA, USA
3 CSIRO Marine Research, PO Box 1538, Hobart, Tasmania 7001, Australia
4 Fisheries Research Services, Marine Laboratory, PO Box 101, Victoria Road, Aberdeen, UK
5 Department of Fisheries and Oceans Canada, PO Box 1000, Mont-Joli, Québec, Canada G5H 3Z4, and Université du Québec à Rimouski, 310 Allée des Ursulines, Rimouski, Québec, Canada G5L 3A1
Correspondence to J. T. Anderson: tel: +1 709 772 2116; fax: +1 709 772 5315; e-mail: andersonjt{at}dfo-mpo.gc.ca
Anderson, J. T., Holliday, D. V., Kloser, R., Reid, D. G., and Simard, Y. 2008. Acoustic seabed classification: current practice and future directions. – ICES Journal of Marine Science, 65: 1004–1011.Acoustic remote sensing of the seabed using single-beam echosounders, multibeam echosounders, and sidescan sonars combined and individually are providing technological solutions to marine-habitat mapping initiatives. We believe the science of acoustic seabed classification (ASC) is at its nascence. A comprehensive review of ASC science was undertaken by an international group of scientists under the auspices of ICES. The review was prompted by the growing need to classify and map marine ecosystems across a range of spatial scales in support of ecosystem-based science for ocean management. A review of the theory of sound-scattering from seabeds emphasizes the variety of theoretical models currently in use and the ongoing evolution of our understanding. Acoustic-signal conditioning and data quality assurance before classification using objective, repeatable procedures are important technical considerations where standardization of methods is only just beginning. The issue of temporal and spatial scales is reviewed, with emphasis on matching observational scales to those of the natural world. It is emphasized throughout that the seabed is not static but changes over multiple time-scales as a consequence of natural physical and biological processes. A summary of existing commercial ASC systems provides an introduction to existing capabilities. Verification (ground-truthing) methods are reviewed, emphasizing the difficulties of matching observational scales with acoustic-backscatter data. Survey designs for ASC explore methods that extend beyond traditional oceanographic and fisheries survey techniques. Finally, future directions for acoustic seabed classification science were identified in the key areas requiring immediate attention by the international scientific community.
Keywords: acoustic, classification, echosounders, habitat, landscape, mapping, marine, multibeam, seabed, sidescan, single beam, sonar
Received 11 July 2007; accepted 15 February 2008; advance access publication 29 April 2008.
 |
Introduction |
|---|
 Top Introduction Defining ASCReview of ASCFuture ASC issuesConclusionsReferences |
|---|
The natural world is structured hierarchically, and processes within natural regions operate across a number of spatial and temporal scales (Azovsky et al., 2000; Turner et al., 2001). Managing human impacts on marine ecosystems requires that natural regions be identified and mapped over a range of hierarchically nested scales. Science in support of ecosystem-based management of marine resources will require a new generation of assessment tools for ecosystem components ranging from single populations to integrated ecosystems. Monitoring natural and anthropogenic disturbances is a key component of ecosystem-based management. Many emerging issues are spatial in nature and require new mapping initiatives before monitoring (Nichols, 1999; Frid et al., 2006). Fisheries management now includes spatial components, such as establishing marine protected areas and implementing fishery closed areas. To execute these management strategies requires the development of management objectives, decision rules, and monitoring plans with a clear appreciation of uncertainty where accurate and precise data are required at spatial scales not previously available.
We believe the science of acoustic seabed classification (ASC) is at its nascence. The rapid evolution of our knowledge based on acoustic- and data-processing technologies gives us the assurance that significant new developments will be realized in years to come. Priority areas of research include evaluating how the performance of "decision rules" protects ecosystem components such as benthic structures and biodiversity. To do this, it will be necessary to develop monitoring programmes that integrate advanced technologies with habitat attributes and link them to population productivity and biodiversity. Advanced acoustic technologies are required for high-resolution bathymetry and the seabed classification of habitats across the continuum of spatial scales. Acoustics is regarded as the most efficient remote sensing tool for mapping and monitoring the subsurface oceans over large areas. Currently, acoustic data can be used to create digital elevation models at the scale of metres over thousands of square kilometres, generating marine-landscape perspectives never seen before. Acoustic backscatter from the seabed can be used to classify surficial sediments and, in some cases, the accompanying biological communities.
Management of marine resources across multiple spatial scales requires a habitat-classification system, and the development of classification schemes is an active area of marine research. The European Nature Information System (EUNIS) classification scheme has been developed for European waters (Davies and Moss, 1999), and other schemes have been developed for global applications in the management of marine resources (e.g. Greene et al., 1999; Valentine et al., 2005). The further development and application of these classification schemes requires explicit information that will characterize marine habitats at a variety of spatial scales. Acoustics is increasingly regarded as the remote sensing tool that will provide the basic background data for classifying and mapping ocean resources.
It has long been understood that details about the character of the seabed, e.g. roughness, rock or sediment type, grain size distribution, porosity, material density, and tortuosity, are embedded in the acoustic echoes from the seabed. As sound may penetrate into the sediments and the basement material, the echoes can also contain information about the zone below the water–seabed interface. Acoustic echosounding and related technologies are increasingly being used to assess, characterize, and map seabed environments. Scientists are now asking whether acoustic sensors might be a cost-effective means of assessing seabed roughness, substratum type, small-scale detail of the benthic habitat, and even community structure of the organisms and plants that are a part of benthic ecosystems. In addition to just knowing what is there, much attention is being directed at human impacts (e.g. fishing, pollution, dredging) that affect benthic habitats. Important questions also arise regarding the time-scales in which a disturbed ecosystem, either by natural changes or by anthropogenic influences, recovers (Collie et al., 2000). The seabed is not a static environment, even when only natural phenomena are active. The confluence of interests expressed by resource managers and fisheries scientists and the need to preserve ecosystems is leading acousticians to revisit both simple echosounding and more complex methods, such as sidescan sonar, multibeam, ship-mounted swathe systems, multiple frequency acoustic sensors, and even some passive acoustic sensors, to evaluate their utility in quantitative surveys of the seabed and what lives there. To that end, the international scientific community initiated the task of reviewing the current state of ASC science under the auspices of ICES (Anderson et al., 2007). This work builds upon the earlier work of Kenny et al. (2003) by providing a comprehensive theoretical and practical review. The purpose of this study is to provide a brief background and introduction to the work carried out by this international group of scientists, and to comment on priority areas for future research.
|
Defining ASC |
|---|
|
TopIntroductionDefining ASCReview of ASCFuture ASC issuesConclusionsReferences |
|---|
The definition of seabed-habitat features that would be acoustically classified and mapped includes both physical and biological attributes. At traditional echosounding frequencies (
10–300 kHz), most of the acoustic information for bottom classification will be from the topography and material of the immediate water–sediment interface. We defined the extent of interest of ASC to be from 1 m (or less) below the seabed to 1 m (or more) above the seabed. This zone was intended to include biogenic structures directly associated with the seabed. It also acknowledges the fuzzy boundary that constitutes the benthic layer, where some organisms are ephemeral residents at varying time-scales and benthic organisms undertaking diel vertical migrations. The results of ASC are related to the shape and geological nature of the seabed and to marine organisms, including finfish, invertebrates, and benthic species. The pelagic component of acoustic classification is not addressed here, but we recognize the importance of linking demersal and pelagic habitats. The biological component of the water column was addressed some years ago by another international team (Reid, 2000). Integration of pelagic and demersal habitats through the use of acoustic technologies awaits future consideration. Acoustic classification of subsurface geological features was not addressed, because these are better measured by low-frequency seismic systems and the subject was beyond the scope of our mandate. The science of correlating acoustic properties to marine surficial sediments dates from the early use of marine acoustics (Nafe and Drake, 1964; Morris et al., 1978). The science of ASC is more recent, largely driven by the development of commercial systems in the 1990s to classify surficial sediments and demersal habitats. ASC developed from the application of normal-incident, single-beam echosounder systems (SBESs) used for marine sciences. More recently, oblique-incident, sidescan-sonar systems (SSS) and multibeam, acoustic echosounders (MBES) are being used to acoustically classify and map marine landscapes. Today, acoustic remote sensing of seabeds is concerned with identifying, classifying, and mapping surficial geological and biological features. The basis of ASC rests on the early observation that the on-axis (i.e. nadir) acoustic echo contained information that was related to surficial-sediment properties such as hardness, porosity, and median grain size. Additionally, the roughness of the water–sediment interface causes sound to be scattered, which affects the coherency of the echo (Parrott et al., 1980). In this way, the echo can be divided into coherent and incoherent components, analogous to the specular and diffuse fields associated with roughness scattering. As the amplitude and wavelength of surface roughness increase relative to the acoustic wavelength, the amount of coherent energy decreases, and the amount and duration of the incoherent energy increases (Clay and Leong, 1974). It is convenient to categorize the combined properties of the coherent and incoherent acoustic echo into hardness and roughness components that relate to surficial seabed sediment and geomorphology. However, it remains a reality that acoustic signals returned from the seabed are complex, and that there is no simple relationship between the backscatter signal and surficial-sediment type and structure.
Initial work on ASC was based on normal-incident systems that categorized seabed hardness and roughness components. These SBESs have the advantage of insonifying the seabed from the same incident beam, and providing the same acoustic information from the unresolved footprints. However, a significant limitation is their narrow footprint (i.e. a narrow swathe with no angular resolution) typically sampled across-track, where large areas of the seabed remain unsampled between track lines. The use of SSS greatly increases the spatial coverage, but these data are restricted to the off-axis roughness component and have relied largely on subjective visual interpretations of texture as opposed to objective image processing and classification. The use of multibeam systems has extended significantly the classification and mapping of seabeds through fine-scale and continuous coverage. Combined use of these acoustic systems is providing an opportunity to classify and map seabed features from the scale of boulders (<1 m2) to banks (>10 000 km2) and shelves (>100 000 km2).
|
Review of ASC |
|---|
|
TopIntroductionDefining ASCReview of ASCFuture ASC issuesConclusionsReferences |
|---|
It was felt that a review of existing knowledge and technologies was now necessary (Anderson et al., 2007). Acoustic technologies reviewed include SBES, MBES, sidescan sonar, and calibrated phase-difference (interferometric) bathymetric SSS. We note that SBESs sometimes are referred to as acoustic ground-discrimination systems. The ICES Cooperative Research Report (Anderson et al., 2007) begins with a review of physical models for sound-scattering from the seabed (Holliday, 2007). It is important for researchers, managers, and stakeholders to appreciate the limitations of these models, the acoustic-measurement process, and the resulting limits to seabed classification. As understanding and instrumentation improve, we can expect advances. The limitations of current acoustic technologies to classify seabeds in the context of the precision, repeatability, and comparability among systems and frequencies were addressed by Kieser et al. (2007). Data quality and machine–operator interface are important issues in ASC. To this end, acoustic-data collection and the quality at which it is displayed were addressed, and standard methods are proposed in relation to seabed-scattering theory (Kloser, 2007). The next step is to use objective methods to classify acoustic data for interpretation and mapping. This classification can stand alone (unsupervised) or be linked to interpretations of seabed habitats (supervised). The general statistical approach and methods available to classify seabeds are reviewed by Simard and Stepnowski (2007), with emphasis on their possibilities and limitations and the importance of clearly defining the classification objectives and quality control along all the processing steps. Defining the relevant spatial and temporal scales of observation with respect to fisheries conservation, ecosystem-based management, and biodiversity issues is critical in relation to defining the types of management questions that may be addressed with ASC methods (Reid, 2007). ASC technologies were reviewed by Michaels (2007), and a summary of their capabilities to classify marine habitats objectively with respect to scales and relevant questions is provided by the same writer, including a summary of commercial ASC systems. Acoustics is a remote sensing technique that requires verification (ground-truthing). Existing techniques used to verify acoustic data were reviewed and evaluated by Brown and Coggan (2007). Designing surveys for ASC, particularly with respect to the development of habitat maps, requires rigourous evaluation of survey techniques that bridge the disciplines of hydrography, marine geology, and fisheries science (Simmonds, 2007).
|
Future ASC issues |
|---|
|
TopIntroductionDefining ASCReview of ASCFuture ASC issuesConclusionsReferences |
|---|
During the review process, as part of the ICES Study Group on Acoustic Seabed Classification (Anderson, 2006), we identified a number of issues that tended to recur in the discussions. These issues were specifically discussed and eventually prioritized into a top ten list of "burning issues" that tended to centre on two topics. The first is one of the standardization of instruments and methods. Fisheries acousticians have a long history of developing internationally accepted methods that standardize how we measure and map fish and plankton in the water column and near the seabed. These methods include such things as the standard calibration of echosounders, parameterizing the target strength of fish, standard formats for data exchange, standard definitions for acoustic variables, and standard survey design. The second topic relates to measuring variability in seabed attributes that determine the natural variability of the seabed at different spatial scales, independent of measurement error. These include measuring things such as along-track (i.e. small spatial scale) variability patterns to establish survey-line spacing, structured survey designs to assure unbiased observations, and optimal coverage for uniform spatial uncertainty, measuring, and taking into account the directionality of seabed characteristics that can affect the acoustic response, effects of range and bottom slope, and the complex issue of verification (ground-truthing) of data and how seabeds are classified towards establishing representative acoustic datasets for supervised classification in a training context. Addressing these topics will ensure that our measurements are objective, repeatable, and comparable among areas and over time. Collectively, we felt that these top ten issues (detailed below) require immediate and future attention by the international scientific community to advance meaningfully the field of ASC. Other issues raised but not ranked in our top ten are listed for reference.
Statistical vs. interpretative classification
Statistical classification is based on objective criteria and data processing, to partition the seabed variability (Simard and Stepnowski, 2007). Interpretative classification is based on subjective analysis often of backscatter mosaics of sidescan (SSS) or multibeam (MBES) data. Subjective interpretation also applies to verification (ground-truth) data when various datasets (e.g. grabs, cores, trawls) are used to measure sediment composition and to identify flora and fauna to characterize a region according to classes or biotopes described in a classification scheme (Brown and Coggan, 2007). Members of the study group were much in favour of statistical, objective classification procedures over interpretative, subjective classifications. The major issue here was one of being able to repeat and generalize results, and this is best done using objective procedures. To do this requires special attention at all steps in the classification process, including the input data quality and control of interfering factors, the usefulness of extracted seabed acoustic attributes for the classification, the discriminating power of numerical classification methods, and the classification probability reflecting expected and ground-truthed variability at the scale observed. Repeatability of results will depend on the acoustic-instrument stability, settings, processing algorithms, range, environmental conditions, and survey methods. Verification of the instrument performance using stable reference sites is one way to check that instrument effects are not altering the repeatability of the classifications.
Spatial scales and sampling resolution
The issue of natural variability and hierarchical spatial scales and our ability to resolve these using ASC ranked second in importance. Much of the discussion revolved around the acoustic footprint, what it was, and what it meant. These included theoretical issues (Holliday, 2007; Kloser, 2007), as well as practical issues of feature resolution, and the acoustic footprint as a function of range from the acoustic gear (Kieser et al., 2007; Reid, 2007). It is generally assumed that seabed structure becomes less complex as one moves from the continental shelf to greater depths, but is it, or does this simply reflect our lack of knowledge? ASC surveys using SBES and MBES typically use transducers mounted at the sea surface, so their range is a function of water depth. Greater depths result in larger footprints and a reduced ability to discriminate small-scale features, while introducing a disparity in observation sizes that complicates data-processing and confuses interpretation (Reid, 2007). Sampling range for ASC can be held constant using towed systems or autonomous underwater vehicles. Resolving this issue before there is significant investment in mapping by coastal nations is a priority.
Considerable effort is being invested in the development of hierarchical classification systems for ocean habitats (e.g. Davies and Moss, 1999; Greene et al., 1999; Madden and Grossman, 2004; Valentine et al., 2005). At the broadest scales of shelves and basins, acoustic data on depth and topography have been the primary determinant of these classes. As the spatial scales decrease, much more information is required to comply with the classification criteria. In some cases, the concepts of hardness and roughness have been incorporated into the classification scheme (Valentine et al., 2005), which probably increases the utility of acoustic-seabed-classification products. However, the future utility of ASC to match these existing classification systems remains to be demonstrated.
Verification of ASC
The primary issue here is one of the mismatch of spatial scales between acoustic- and verification (ground-truth)-sampling techniques. In one instance, the resolution of acoustic data is coarse compared with verification methods, for instance, SBES classification where adjacent pings are normalized (stacked) into a single observation on a scale of, say, 100 m2, compared with verification by grab samples which typically sample just 0.5 m2 (Brown and Coggan, 2007; Reid, 2007). Alternatively, high-frequency (>300 kHz) acoustic SSS and MBES can produce seabed backscatter with accurate georeferenced locations at scales of 0.25 m2 and 1–2 m2, respectively, over vast areas of the seabed. Accurately positioning and interpreting the verification data on the acoustic-backscatter maps with known precision remains a challenge. Other issues raised here included the fact that verified data can include physical data, biological data, or both, and classification of these data is often subjective. Which classification scheme provides the best fit to the acoustic information extracted from the seabed echo? Often, species profiles from a number of grab samples are classified into communities based on a predetermined sampling effort for a given area. However, there is seldom any evaluation of how adequately these communities have been sampled using, for example, species-area curves. Moreover, temporal variability in benthic communities is seldom incorporated into biological classifications.
Temporal variability: the fourth dimension
Current mapping efforts typically produce three-dimensional spatial maps of the seabed topography, as well as covering physical and biological attributes. Despite there being considerable evidence demonstrating significant temporal variability in seabed properties and reflectivity (Holliday, 2007; Reid, 2007), there are few examples of this in ASC. In some instances, short-term temporal variability could be erroneously interpreted as spatial pattern when making three-dimensional maps. There are also reasonable questions about how long a map remains a valid representation of the seabed. Most important is the understanding of cyclical variation, especially between seasons, where a survey could provide quite different results, dependent on time of year. Understanding how habitats vary with natural physical and biological processes, and determining the frequency of such changes, requires immediate attention before the accuracy and spatial stationarity of seabed habitat maps can be assessed.
ASC reference areas
Reference areas were often referred to as "patch tests", in which a known and previously sampled area of the seabed is resampled at frequent intervals with the same, or different, acoustic systems. In many ways, this is a "poor man's" corroboration, where current technologies or research capabilities prevent formal system calibration beyond the instrument calibration on physical standards as used in fisheries acoustics. It is acknowledged that reference areas can change over time. However, proper verification sampling should detect such changes and this, in turn, can contribute to our understanding of natural variability vs. instrument drift. Reference areas will remain an important aspect of verifying the whole system operation in conjunction with more-detailed instrument calibrations. This need for standards in ASC, to allow inter-comparability of results, will remain an important area of research that will involve a suite of acoustic instruments and other sampling tools, to address the difficulties systematically.
Calibration of acoustic systems
Some might argue that the calibration of acoustic systems should have ranked higher. It did not because of the difficulties arising from the variability in the angular response and in matching backscatter to existing schemes of seabed classification. It was regarded, however, as a critical step necessary in standardizing measurements among types of acoustic gears and areas over time, to make comparisons possible and to build the necessary knowledge for a better understanding of the processes of seabed backscattering (Kloser, 2007). Calibration difficulty, which is a function of the complexity of the acoustic system used, may be low for narrowband, single-beam echosounders, but much higher for systems such as multifrequency, wideband, interferometric SSS, and MBES (Simard and Stepnowski, 2007). In particular, the angular response of the acoustic system needs to be calibrated, and this is problematic for both SBES and MBES. Traditionally, SBESs are calibrated for on-axis sensitivity, whereas the time-dependent, seabed-backscattering response requires calibration of the single-beam's angular response. Similarly, MBES are difficult to calibrate, and the common method for ensuring system repeatability is to use reference areas. More advances are needed to provide routine calibration procedures. There are many advantages associated with the use of systems whose measurements can be traced to international standards. Until calibrated systems are used to make the scattering measurements, it is unlikely that statistically based interpretations relating acoustic-backscattering measurements to the seabed's physical surface and underlying volume can be achieved reliably.
Characterization of the acoustic signal
There was some concern among members of the international team about the lack of information on how acoustic-backscatter data are processed and interpreted using commercially available systems. This is a problem with SBES where, although commercial systems use known acoustic properties of the seabed, the processing algorithms are not always available and may not be replicable. With the development of MBES, where the beams encounter the seabed at a wide range of angles, the analyses used in commercial systems are likely to be more complex and even less easy for the user to understand or replicate. Although it is recognized that a vendor must protect their investment, it is clear that systems for which the details of operation are not available will not be useful for researchers whose aim is to understand scattering from a seabed habitat at an elementary level to interpret properly the results for their specific objectives. Until such an understanding is achieved, it is likely that there will be considerable, justifiable concern that results obtained from "black-box" processors may at best contain results that cannot be uniquely associated with a single kind of bottom type. At worst, they may simply be unreliable indicators of the character of the seabed.
Single vs. multiple frequencies for ASC
The use of multiple frequencies will increase our ability to classify seabeds, because both surface (near nadir) and volume backscatter vary with frequency. A major difference is that lower frequencies penetrate the seabed to greater depths, whereas higher frequencies can resolve smaller spatial structures. We realize that the frequency palette considered penetrates seabeds only approximately centimetres, or possibly up to 1 m, depending on substratum. Combining frequencies is equivalent to optical-satellite remote sensing, where multiple wavelengths measuring different physical characteristics are combined to classify and map land and ocean areas. Currently, two or more frequencies have been combined for SBES to improve seabed classification (Kloser et al., 2002; Fösa et al., 2005). We expect that incorporating two or more frequencies through the range of 10–300 kHz will result in significant improvements to ASC. Incorporating multiple frequency SBES with single frequency MBES during single surveys may be a cost-effective way of improving ASC.
Survey design
Variability is present at all temporal and spatial scales, and this should be handled properly with a survey-design strategy adapted to the required seabed-classification scheme. More consideration needs to be given to survey designs that are weighted towards systematic line transects with no randomization and little or no prior knowledge; or to multibeam surveys based on the need to sample along bathymetric contours to obtain a uniform swathe width and to minimize outer-beam variability (Simmonds, 2007). Most discussion on this topic centred on minimizing the errors and bias associated with spatial interpolation of line-transect data, and stratifying and randomizing surveys, when necessary, to generate unbiased datasets over the region being mapped. Currently, we know very little about the small-scale spatial variability in seabed attributes measured by acoustic systems across the range of depths being considered by national mapping strategies. Even within a given depth range, we suspect that there are spatial scales to the variability in seabed attributes that are currently unknown and likely are an intrinsic property of the habitat. Adaptive and nested survey designs should be considered in directed research programmes that address such issues.
ASC design in national habitat programmes
Several countries have now embarked on national programmes to classify and map their marine resources with particular emphasis on the seabed. Discussions by the international team of scientists often reflected interest, and sometimes concern, in the use of acoustic technologies for classifying and mapping marine resources. There is a genuine excitement generated by these emerging technologies that is generating a new family of data products that allow scientists to image, classify, and map the marine environment at a variety of spatial scales not possible up to now. We are, however, concerned that ASC is often considered a "black box", where the technology has matured sufficiently and been adapted to the diverse needs of the user community so that it can now be applied unquestionably to habitat-mapping projects. We felt that we do not yet have that complete understanding of ASC, for two reasons. One was the overall lack of connections between existing seabed-scattering theory and many current ASC applications. We felt it was important to link theory with application to understand and interpret acoustic data better. This is an ongoing process as the theory continues to evolve. The other reason was our awareness of the rapid evolution in acoustic technologies, both in the capacity to generate vast volumes of data and the ability to process this with ever-increasing sophistication. This evolution is driving the science behind ASC, both in renewed interest in the theory of sound-scattering from the seabed, and in the generation of a new, spatial science at the marine-landscape scale. To this end, we recommend that formal mechanisms be established to integrate ASC research and applications into national classification and mapping programmes.
Other issues
Defining fish habitats
Fish habitat is typically defined as some function of the biotic and abiotic variables where a fish lives. Water depth is often important, but so is structural complexity, greater complexity often being associated with preferred fish habitats. However, associating marine fish with specific habitats has proven to be difficult (Rose, 2000; Beck et al., 2001; Minns and Moore, 2003; Mitchell, 2005). Among the reasons for this is our poor understanding of life histories and habitats in the marine environment, compared with understanding of terrestrial and lotic systems. The weak explanatory and predictive power of existing relationships is thought to arise largely from the application of easily measured, or available, habitat variables (e.g. depth) as opposed to relevant variables that are independent of each other (Lipicius et al., 1997; Stoner, 2003). Increasingly, marine-landscape variables are recognized as an important component of fish habitats, and issues such as spatial pattern (size, shape, fragmentation, connectivity) and relative location (e.g. to larval supply, other juvenile habitats, adult habitats) are recognized as important determinants of fish habitat (Beck et al., 2001). We believe that acoustic mapping of seabeds from fine scales, where fish can exhibit site fidelity, to landscape scales, where fish live through one or more life history stages, will be a critical component in defining and mapping fish habitats.
SBES vs. MBES
Single-beam echosounders are the standard instrument of fisheries acousticians. Multibeam echosounders have been adopted by the hydrographic community as their standard instrument for detecting the seabed, and an evolved multifaceted version is now emerging as the new tool of fisheries acousticians for simultaneous water column and seabed investigations. There are advantages and disadvantages to both systems. For SBES, the advantages include: relatively low acquisition costs; ready availability and wide use; calibration (on-axis) for scientific users; efficiency of data processing because of low data volumes and standard procedures; relative ease of understanding and operation; water-column backscatter availability and readily processing; use of multiple frequencies during a single survey. SBES disadvantages include the narrow, low-resolution footprint and sampled volume across-track, which necessitates inter-track interpolation. For MBES, the advantages include: wide swathe with high angular resolution over a range of normal and oblique incident angles for a single ping that typically span 160°; three-dimensional imaging with minimum interpolation; the option for complete coverage (mosaic). MBES disadvantages include: high acquisition costs; large data volumes; the steep learning curve; complex calibration; the greater complexity of backscatter processing; the fact that water-column backscatter data are not always available. Combining SBES and MBES data would provide significant value-added information, at least until new systems under development integrating both approaches become operational.
Directionality of the seabed
There is natural structure in seabed morphologies that generates different acoustic responses for different orientations of the survey track, and these structures can change with time. This phenomenon is referred to as the "directionality" of the seabed. Such different responses are particularly apparent in SSS (Michaels, 2007). They also occur in single-beam and multibeam data (Holliday, 2007), but are seldom measured. We propose that ASC surveys be designed to measure seabed directionality (Simmonds, 2007) specifically, and that temporal variability in seabed structure be incorporated into interpretative habitat maps.
Acoustic diversity
Interpretation of seabed properties in terms of surficial geological structure and biological communities can often be subjective and can vary with the observational scale. What constitutes fish habitat is still poorly defined, and it may vary between species or within the life cycle of a single species. Therefore, relating acoustic backscatter and topographic relief to an interpretation of what constitutes the seabed can be subjective and highly variable. Of potential use in its own right is the concept of acoustic diversity where distinct classes of acoustic properties occur contiguously over meaningful spatial scales. A map of unsupervised acoustic classes may provide insight into ecological processes without requiring an explicit interpretation of why the differences occurred. It may be difficult to relate the acoustic diversity to an interpretation of the seabed using standard verification techniques (see Verification of ASC). However, patterns in acoustic diversity may ultimately provide useful information for characterizing and mapping seabed habitats. The requirement is, however, that such acoustic-diversity patterns are repeatable, ideally with diverse types of acoustic gears and independent of data-processing algorithms, and ultimately that they can be related to natural processes that are both understood and universal.
Classification of biota
Acoustic classification of infauna and epifauna is regarded as an important area of future research. Will it be possible to identify different benthic communities based purely on their different acoustic signatures? Is it possible to detect acoustic differences that reflect functional communities, as opposed to taxonomically derived communities? Increased use of emerging acoustic technologies by researchers to address these questions will be a high priority. When acoustic surrogates can be found for biological species and communities, then our ability to classify and map marine habitats will advance significantly. Currently, the research focus is to use optical systems for surveying epifauna and seabed relationships remotely.
Data management and information dissemination
The imbalance of data collection vs. processing and interpretation is often referred to as the "information conundrum", in that we have developed the ability to collect vast amounts of data but our ability to process, analyse, and interpret these data lags far behind. In fact, the ability to manage large, complex, georeferenced datasets is a significant issue facing the scientific community. The volume of data generated by acoustic systems can be immense, approximately terabytes for a single survey. We anticipate an evolution towards the use of multiple frequencies and increased use of multibeam systems, so datasets can only increase in size and complexity. It is necessary to incorporate verification data to generate supervised classifications, but these data increasingly comprise ever more large and complex optical datasets. Future issues will include developing standard data formats for archiving, retrieving, and analysing ASC datasets. We feel that too little effort is being put into data management and analysis, and the effective dissemination of results.
Bentho-pelagic coupling
Integrating the benthic and pelagic environments is considered an important area for future research. Physically and biologically, there is a strong link between the benthic layer and the overlying water column. We defined the seabed to include the water column directly above the detected water–sediment interface. This definition was intentionally open-ended, because we regard the benthic layer as "fuzzy" (i.e. poorly defined). Ontogenetically, many marine species have both benthic and pelagic life history stages. In this context, there are ephemeral uses of benthic habitats dependent on life history stage or benthic organism. Biogenic structures attached to seabeds may range from centimetres to metres above the seabed, and such structures may vary over relatively short time-scales. Dense aggregations of fish settling on, or into, the seabed may alter its acoustic properties. Surface-dwelling phytoplankton can generate gas bubbles in a diurnal cycle that will alter the acoustic reflectivity of the substratum. The diel emergence of benthic organisms into the water column at night also changes seabed reflectivity. The ability of acoustic systems to measure simultaneously from the near-surface of the ocean to up to 1 m below the water–sediment interface will allow marine researchers to study and understand many aspects of bentho-pelagic coupling that should contribute meaningfully to our understanding of how ecosystems function.
Integrating acoustic and optical data
Optical data include high-resolution photographs of the seabed, digital-video mosaics, laser-line scanners, and bathymetric lidar. These technologies are often used to verify acoustic backscatter from the seabed as a move to developing supervised classifications and interpretative habitat maps, but they can be regarded also as providing remotely sensed data in their own right. Fine-scale mapping using optical methods provides an interpretation of both physical and biological seabed attributes that can be resolved to species for large organisms. These data, when appropriately georeferenced, form an important record of the seabed environment and a means of monitoring both natural and anthropogenic impacts at a fine scale. Extrapolation of these fine-scale data to larger scales usually relies on the use of acoustic data with interpreted physical seabed attributes, such as roughness and hardness.
A general issue with optical data lies in interpretation. It is possible to measure grain size with physical samples (e.g. grabs) but this ability is limited with optical samples (i.e. photographs). Optical systems tend to have short ranges and hence small observation footprints. The tendency for mobile species to avoid or be herded by the instrument or its lighting system can present a problem in obtaining unbiased samples. Some optical systems can be difficult to operate in a line-transect mode, and navigating a camera close to the seabed over long distances is clearly a more complex operation than running a survey line using a vessel-mounted acoustic system. This being the case, optical systems work best at providing detailed ground-truth data at a point, and are particularly useful for information on biota, rather than just the physical aspects of the substratum. The information can then be used to ground-truth and to test the habitat classification process. We recommend further work integrating acoustic and optical data, as well as physical samples, in seabed classification and mapping.
|
Conclusions |
|---|
|
TopIntroductionDefining ASCReview of ASCFuture ASC issuesConclusionsReferences |
|---|
ASC is an important new area of marine science and will contribute significantly to both scientific research and ecosystem-based management of the marine environment. We predict that there will be a rapid, ongoing evolution in technology and applications that will continue to provide new opportunities and challenges for marine researchers. ASC and benthic-habitat mapping necessarily cross discipline and jurisdictional boundaries. Hydrographers, geologists, benthic ecologists, physicists, and fisheries scientists need to work together to generate habitat maps. ASC techniques will provide the underlying data that will allow this to happen, but organizational structures must be modified to permit it. Standards for conducting ASC must be developed by the international community to promote quality, consistency, and an inter-comparability in ASC outputs. Standardization will allow for comparisons among areas and over time. Anderson et al. (2007) attempted to capture the state of ASC science now. However, we propose that the international community revisit this subject soon to update and to maintain the relevance of our understanding and our progress in addressing the issues outlined herein.
|
Acknowledgements |
|---|
We thank R. Kieser, A. Orlowski, W. Michaels, E. J. Simmonds, J. Preston, and all members of the ICES Study Group on Acoustic Seabed Classification for their enthusiastic participation and constructive criticism of our ideas on future issues in acoustic seabed classification. C. Lang and R. Courtney patiently guided JTA through the theory of sound-scattering from seabeds. R. Coggan kindly provided a careful and constructive review of this manuscript. Funding to pay the Open Access publication charges for this article was provided by the Fisheries and Oceans Canada.
|
References |
|---|
|
TopIntroductionDefining ASCReview of ASCFuture ASC issuesConclusionsReferences |
|---|
-
Anderson J. T. Report of the study group on acoustic seabed classification (SGASC). ICES Fisheries Technology Committee. ICES Document CM 2006/FTC: 03, Ref. MHC, ACE, WGFAST (2006) 7.
Anderson J. T., Holliday V., Kloser R., Reid D., Simard Y. Acoustic seabed classification of marine physical and biological landscapes. ICES Cooperative Research Report (2007) 286:198.
Azovsky A. I., Chertoprood M. V., Kucheruk N. V., Rybnikov P. V., Sapozhnikov F. V. Fractal properties of spatial distribution of intertidal benthic communities. Marine Biology (2000) 136:581–590.[CrossRef]
Beck M. W., Heck K. L., Able K. W., Childers D. L., Eggleston D. B., Gillanders B. M., Halpern B., et al. The identification, conservation, and management of estuarine and marine nurseries for fish and invertebrates. Bioscience (2001) 51:633–641.[CrossRef][Web of Science]
Brown C., Coggan R. Verification of acoustic classes. ICES Cooperative Research Report (2007) 286:127–144.
Clay C. S., Leong W. K. Acoustic estimates of the topography and roughness spectrum of the sea floor southwest of the Iberian Peninsula. In: Physics of Sound in Marine Sediments, pp. 373–446—Hampton L., ed. (1974) New York: Plenum Press.
Collie J. S., Hall S. J., Kaiser M. J., Poiner I. R. A quantitative analysis of fishing impacts on shelf-sea benthos. Journal of Animal Ecology (2000) 69:785–798.[CrossRef][Web of Science]
Davies C. E., Moss D. The EUNIS classification. European Environment Agency. ITE project T0809219 (1999) 124.
Fösa J. H., Lindbeg B., Christensen O., Ludälv T., Svellingen I., Mortensen P. B., Alvsvåg J. Mapping of Lophelia reefs in Norway: experiences and survey methods. In: Cold-Water Corals and Ecosystems—Freiwald A., Roberts J. M., eds. (2005) Berlin: Springer. 337–370.
Frid C. L. J., Paramor O. A. L., Scott C. L. Ecosystem-based management of fisheries: is science limiting? ICES Journal of Marine Science (2006) 63:1567–1572.
Greene H. G., Yoklavich M. M., Starr R. M., O'Connell V. E., Wakefield W. W., Sullivan D. E., McRea J. E., et al. A classification scheme for deep seafloor habitats. Oceanologica Acta (1999) 22:663–678.[CrossRef][Web of Science]
Holliday D. V. Theory of sound scattering from the seabed. ICES Cooperative Research Report (2007) 286:7–28.
Kenny A. J., Cato I., Desprez M., Fader G., Schüttenhelm R. T. E., Side J. An overview of seabed-mapping technologies in the context of marine-habitat classification. ICES Journal of Marine Science (2003) 60:411–418.
Kieser R., Preston J., Orlowski A., Chapman R. Acquiring and preparing acoustic data. ICES Cooperative Research Report (2007) 286:29–46.
Kloser R. Seabed backscatter, data collection, and quality overview. ICES Cooperative Research Report (2007) 286:47–63.
Kloser R. J., Keith G., Ryan R., Williams A., Penrose J. Seabed-biotope characterization in deep water—initial evaluation of single-beam and multibeam acoustics. In: Proceedings of the Sixth European Conference on Underwater Acoustics, Gda
sk, June 2002—Stepnowski A., ed. (2002) 81–88. (ISBN 83-907591-8-7 © Gdask University of Technology, 2002. Printed in Poland. Printed by: Z.U.P. Astra sp. Z o. o. 80-761 Gdask).
Lipicius R. N., Stockhausen W. T., Eggleston D. B., Marshall L. S., Hickey B. Hydrodynamic decoupling of recruitment, habitat quality and adult abundance in the Caribbean spiny lobster: source–sink dynamics? Marine Freshwater Research (1997) 48:807–815.[CrossRef]
Madden C. J., Grossman D. H. A Framework for a Coastal/Marine Ecological Classification Standard. (2004) Arlington, VA, USA: NatureServe. www.natureserve.org.
Michaels W. Review of acoustic-seabed-classification systems. ICES Cooperative Research Report (2007) 286:101–126.
Minns C. K., Moore J. E. Assessment of net change of productive capacity of fish habitats: the role of uncertainty and complexity in decision making. Canadian Journal of Fisheries and Aquatic Sciences (2003) 60:100–116.
Mitchell S. C. How useful is the concept of habitat? A critique. Oikos (2005) 110:634–638.[CrossRef][Web of Science]
Morris H. E., Hamilton E. L., Bucker H. P., Bachman R. T. Interaction of sound with the ocean bottom: a three-year summary (1978) San Diego, CA: Naval Ocean Systems Center. Technical Report, 242.
Nafe J. E., Drake C. L. Physical properties of marine sediments. In: The Sea, 3, pp. 794–814—Hill M. N., ed. (1964) New York: John Wiley.
Nichols K. Coming to terms with "Integrated Coastal Management": problems of meaning and methods in a new arena of resource regulation. Professional Geographer (1999) 51:388–399.[CrossRef]
Parrott D. R., Dodds D. J., King L. H., Simpkin P. G. Measurement and evaluation of the acoustic reflectivity of the seafloor. Canadian Journal of Earth Science (1980) 17:722–737.
Reid D. Report on echo-trace classification. ICES Cooperative Research Report (2000) 238:107.
Reid D. Accounting for spatial and temporal scales and interpolation in ASC surveys. ICES Cooperative Research Report (2007) 286:77–100.
Rose K. A. Why are quantitative relationships between environmental quality and fish populations so elusive? Ecological Applications (2000) 10:367–385.[CrossRef][Web of Science]
Simard Y., Stepnowski A. Classification methods and criteria. ICES Cooperative Research Report (2007) 286:64–76.
Simmonds J. Survey design for acoustic seabed classification. ICES Cooperative Research Report (2007) 286:145–152.
Stoner A. W. What constitutes essential nursery habitat for a marine species? A case study of habitat form and function for queen conch. Marine Ecology Progress Series (2003) 257:275–289.[CrossRef][Web of Science]
Turner M. G., Gardner R. H., O'Neill R. V. Landscape Ecology in Theory and Practice: Pattern and Process (2001) New York: Springer.
Valentine P. C., Todd B. J., Kostylev V. E. Classification of marine sub-littoral habitats, with application to the northeastern North America region. American Fisheries Society Symposium (2005) 41:183–200.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  J. A. Koslow The role of acoustics in ecosystem-based fishery management ICES J. Mar. Sci., July 1, 2009; 66(6): 966 - 973. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||