9

The advent of offshore exploration for oil and gas during the 1970s created a need for very accurate underwater position fixing for survey activities such as searching, mapping and photography, and for work tasks such as pipe laying, surface rig positioning, underwater structure positioning, towfish positioning, remotely operated vehicle (ROV) navigation, subsea construction, mining, drilling and a vast range of other applications. The precision required dictates the use of underwater acoustic navigation and tracking techniques, and the best known of these employ arrays of hydrophones, beacon (also called pingers), transponders (receiver-transmitters) and responders (transmitter- receivers). For most of the tasks listed above the positioning is carried out only after a series of procedures to fix the position of the array on the sea bed; these may include transponder deployment, array baseline calibration and possibly absolute calibration in terms of geodetic co-ordinates by integrated satellite, surface and underwater data telemetry. In this paper, we consider how techniques and systems developed for industrial applications may be adapted for tracking echo-locating cetaceans (dolphins, porpoises and whales), in particular the principles of position fixing in three-dimensional space in real time. This is a problem that presents a serious engineering challenge.

Computer classification of dolphin whistles requires a number of steps from detection of a whistle in background noise to classifying the whistle among a database of whistle types. Detection was achieved by using a broadband noise reduction technique and a filter for tracking relatively slowly changing FM tones. After whistle detection, its characteristic time-frequency-intensity contour was extracted using a tracing algorithm with an 'inertial' following rule to avoid crossover when multiple whistles were present. The data requirements for each extracted contour were reduced using an encoding technique that splits each whistle into segments based on frequency slopes and using a curve fitting routine to represent the contour within each segment. Classification is achieved using hidden Markov models to represent the possible segment sequences within each class, and three distance measures based on the segment curves. In this way class similarity percentages can be calculated for each class, and a candidate whistle can be assigned to the most probable class.

In gill-net fisheries around the world, very large numbers of small cetaceans are killed each year as bycatch (IWC report 1994). If commercial fishing is to continue in areas where the incidental catch of cetaceans is significant then technical improvements to the method of fishing are required to mitigate the impact. Modifying fishing gear by the addition of acoustic alarms which signal the position of the net by the transmission of low level sounds was pioneered by Lien and his colleagues in Newfoundland and this technique successfully reduced baleen whale interactions with set fishing nets and traps. These simple devices, which were constructed by the fishermen, were also tested in a bottom set gillnet fishery in the Bay of Fundy where harbour porpoise mortality is high. Encouraging results from this Canadian research resulted in an improved device being developed in the USA for use in a subsequent Gulf of Maine study. The low frequency signals developed as baleen whale deterrents were also tested on captive harbour porpoises at Harderwijk in Holland where it became very evident that the frequencies used were inappropriate. The technology used in these early devices is electro- acoustically inefficient and the operating (battery) costs rather high. In Europe studies of a variety of potential acoustic deterrent devices were carried out by Loughborough University which included tests with a harbour porpoise of a wide variety of signal frequencies and waveforms. These sounds were synthesised digitally and the behaviour of a 9ee swimming animal, contained within a large floating net enclosure, were observed. The new generation of micro-controller based beacon-mode alarms developed at Loughborough University is discussed here. These devices synthesise the sounds shown to be most aversive to a porpoise and implement new features intended to minimise habituation rates and maximise battery life. A preliminary field test with wild harbour porpoises in Scotland during September 1996 showed that these devices induce a dramatic avoidance behaviour, displacing the animals in a short test to a range greater than 640 m. The design and engineering of this new technology device is discussed in the context of preparing them for a large-scale commercial fishery trial in Denmark.

As part of the EU funded CETASEL project recordings were made of small cetaceans on the edge of the continental shelf between SW Eire, Biscay and North Spain. Various passive acoustic systems were attached to a pelagic trawl operating in relatively deep water (200m). These systems were then used in conjunction with a surface observer program and remotely operated television systems to study small cetacean behaviour around a moving fishing net. A multiple hydrophone system has been developed to track echolocating cetaceans. Comparison of the arrival times of echolocation 'clicks' on a number of spatially separated hydrophones were made allowing estimation of the positioning of the cetacean in relation to the hydrophone array. Additional lower frequency (3-20 kHz) recordings were made on a single hydrophone also attached to the trawl. A number of examples of lower frequency 'whistles' signals including multi-path (bottom and surface reflections) were observed. With knowledge of the hydrophone and water depth, comparison of the arrival time of the various multi-path components of a signal has also allowed the estimation of the range and depth of its source from the receiver. A simplified ray path model has been developed to simulate various source, receiver geometrys. Arrival times of the multi-path signals were calculated and compared with those seen during sea trials. A number of assumptions have been made in initial models including a constant sound velocity depth profile and the treatment of the surface and seabed as a simple reflecting surfaces. Initial results have shown a number of examples with a reasonable correlation between estimated position of a submerged cetacean and the associated surface observations. Examples of multiple (positioning) solutions were however found, these are in the main due to imprecision in the knowledge of the hydrophone and water depth geometry and inaccuracies in the initial timing measurements. The use of correlation techniques and stand alone depth measurement devices is therefore proposed for future measurements and analysis using this technique. It is felt that within certain constraints this technique can provide valuable additional information regarding cetacean behaviour in the wild. The addition of more complex time measurement techniques and better ray path modelling will hopefully provide a useful analysis tool in the study of cetaceans.

Oceans form two-thirds of the surface area of our planet. Our understanding of this environment even after several decades of research is best described as rudimentary. The watery abyss is home to the world's largest and most diverse biological ecosystems. Monitoring of these ecosystems and the study of the physical and biological evolution of the oceans is of paramount importance to our increased comprehension and understanding of the oceans. Efficient monitoring relies on the collection of data over long periods of time and in diverse sea conditions. Stationary or moving sonar systems have already proven invaluable for this task. However, most of these systems are fixed in their mode of operation, possessing little or no flexibility. Thus, when the environment changes data retrieval often becomes difficult or impossible. Changing the characteristics of the monitoring device to match the local sea conditions is a possible solution to this problem. In this paper the hardware and software implementation of a system that realises this function is presented together with some preliminary transmission tests in the Mediterranean Sea. The system, VERTLINK is designed to collect, store and transmit physical and bio-acoustical data to an end-user. The system, a half-duplex underwater communication link, com- prises two units-a sea-bed unit and a surface unit. The sea-bed unit, under the control of the surface unit, records oceanographic data and then transmits it to the sea-surface unit. The latter is linked to a computer station on-board a survey vessel via a physical cable or to shore station by means of a radio link. Using this system the end-user is able to change the way the sea-floor unit acquires and transmits data. The modulation rate/technique can be changed from 2-DPSK to 8-DPSK, allowing data retrieval to be changed according to sea noise conditions. The power level can be varied from 20% to 100%, to achieve more efficient use of power resources and the sea-floor unit can be told to release.

A mobile video/acoustic system (MVA), previously developed (Dudzinski et al. 1995), was modified with new hydrophones and a new hydrophone mount to increase the recorded frequency bandwidth and to decrease low frequency noise due to wobble in the physical mount, respectively. The MVA permits real-time synchronous recording of vocal and behavioural activities of individual dolphins. The system is swimmer-propelled and facilitates localisation of dolphin sound sources by associating video data of animal distributions with audio data from two hydrophones spaced relative to the human interaural distance as scaled to sound speed in water. A recent addition to the original design was a small second housing containing a Sony TCD D8 digital recorder together with a pre-processor circuit which detects the highly directional part of a dolphin's echolocation 'clicks' and makes these audible and recordable. The pre-processor comprises a preamplifier with band pass filtering which selects only signals between 90 kHz and 130 kHz. The filter output is then rectified and the low frequency envelope extracted by a further filter. This technique produces a low frequency pulse accurately representing the original signal amplitude which can be recorded on the R- DAT sound track. As the directionality of the original signal, the inter-pulse intervals and the pulse amplitudes are retained this band-limited data carries significant information and the loss of the high frequency spectrum is acceptable in this application. This 'click detector' data is recorded onto one track of the recorder while unprocessed low frequency signals from the hydrophone are captured on the second. The R DAT stores real-time information (date and time) in the sub-code of recorded signals so that video image data, with in-picture timecode, can be correlated accurately with the echolocation behaviour. Data is captured when dolphins swim directly towards the MVA and click detector unit. The click detector housing mounts to the dorsal surface of the MVA housing, and externally measures 37cm by 9cm by 5cm. The modified MVA system (including click detector) has been used to document the behaviour and vocalisations of interacting Atlantic spotted dolphins Stenella frontalis in the Bahamas and bottlenose dolphins Tursiops truncatus in Japan. Example data sets will provide evidence of this system's utility for examine the social lives and signal exchange among interacting, free-ranging dolphins.

Underwater acoustic tracking techniques mostly rely on an active acoustic source triggering replies from an array of transponders. If the tracking equipment must remain passive, the transponders are replaced by hydro- phones. This research is to develop systems which can passively locate and track echo-locating cetaceans, with a view to reducing the by-catch of cetaceans in fishing nets, particularly in pelagic trawls. A 'flat' array of four hydrophones in two parallel streamers can be used to produce these tracks. The streamers are attached to the top of the trawl and 'fly' in a reasonably stable manner as the trawl is towed through the water. A fifth hydrophone is placed on the trawl headline, slightly above the plane of the two streamers, so as to obviate any ambiguities in the position-fixing computation. The bandwidth limitations of the coaxial cable transmission link to the towing vessel led to the use of a look-up table to determine positions. This paper describes the technology of the system and the position-fixing procedure associated with it.

Large numbers of small cetaceans are caught each year as incidental catch in gill-net fisheries around the world. The 1994 EC ASCOBANS agreement stresses the importance of reducing marine mammal bycatch, in particular of the harbour porpoise Phocoena phocoena. With the uncertainty of the severity of the problem in other fisheries it has become necessary to research into the scale of the problem in the different fishery types. Two 3-year EC (AIR DG XIV) projects have been initiated to analyse the scale of the problem in the pelagic trawls and to advise on possible methods of reducing cetacean bycatch in this fishery, one with the remit to analyse the scale of the problem in commercial fisheries (BIOECO), the other to analyse the reasons for cetacean bycatch and suggest methods of reducing it (CETASEL). As part of CETASEL, trials have been taking place both at sea with pelagic trawls and in dolphinaria in Europe. A cetacean rehabilitation centre in Neeltje Jans, Holland, has been used to examine the behaviour of a single wild harbour porpoise to different forms of acoustic disturbance produced by electronic means. The results of such tests provide valuable information as to the animal's tolerance to sound pressure levels at various frequencies and to different signatures. This information can be used further to design effective deterrents which only produce the signals which are known to deter the relevant species. The porpoise was housed in a floating net pen in minimum (tidal) of 4 metres of water providing a shallow, controlled environment in which the animal can be monitored as it re-acclimatises to the open sea. Signals introduced to the animal via four transducers produced the required aversive effect, however patterns of behaviour emerged which could not easily be explained. The porpoise did not move to the furthest point from the source, instead preferring the furthest point in line with the transducers. Mathematical modelling of the propagation of the signal in the water, showed a possible reason for the behaviour. The complicated signal pressure level pattern around the projectors showed the near field effects of the sparse array (a feature confirmed by measurements in the field), and highlighted the end-fire of the array as the single point in the enclosure where a large stable area was present. If the spacings of the projectors in the array were non integer, which was highly likely, a form of null was formed along the line of the transducers, the exact preferred position of the porpoise. This paper gives the details of the tests and results, and shows the situations modelled in post processing.

Inter Pulse lnterval (IPI) measurements on clicks recorded from diving sperm whales have been confirmed to be useful in assessing the whale's body size (Gordon 1991, Goold 1996a, 1996b). Goold (1996a, 1996b) developed a cepstrum-based method to accurately measure the IPIs and to assess the animal size by taking into consideration the acoustic transmission properties if the spermaceti oi1 under different temperature and pressure conditions. To extensively apply IPI measurements, we developed a program, based on our custom real-time Digital Signal Processing Workstation (Pavan et al. 1997), to show in real-time both the spectrogram (spectrum vs time) and the cepstrogram (cepstrum vs time), optimised for this special purpose, of the click sequences received by means of a towed array. The cepstrogram based method resulted very sensitive and capable of displaying low level clicks otherwise difficult to see on the spectrogram and the real-time approach proved very helpful in browsing long recordings as well as in discriminating and counting different whales clicking at the same time. We extended the analysis on sighted and tracked whales to consecutive dives. Our results were consistent throughout whole dives and never showed the scattering of IP1 values previously by other authors, even though we confirmed IPIS slightly varied according to the variation of spermaceti properties with the progression of the dive. Therefore, spot measurements on short sequences of clicks should be considered reliable. Even though the method proved very effective, some problems still remain unresolved, in particular those concerned with the influence of long range sound propagation paths and of the relative position and orientation of the clicking whale in regard to the hydrophone. This research was carried out within a project granted by the Central Inspectorate for Sea Protection of the Italian Ministry of the Environment (1989-1996).

Classification of dolphin whistles has generally been a manual process requiring a trained observer, and calling for subjective and qualitative decisions. As a result it has been difficult to determine the differences and similarities between dolphin whistles. Software developed in the Underwater Acoustics Group at Loughborough University automates many of the aspects of classification of dolphin whistles, and provided quantitative probabilities for class memberships. Using the software, whistles may be detected on recordings, enhanced and background noise removed, their frequency-time contour traced and then encoded into a more compact data structure. Pattern recognition techniques can then more easily be applied to such encoded whistles. Examples of the use of the extraction, encoding, and classification methods are shown using the software, demonstrating the current levels of automation. Quantitative class probabilities for the signals are calculated which can be used to determine identities of different groups of dolphins in the field.