41   Mitigation measures   For all species, the first issue in the proposed devices is signal shape, or rise time and peak spectra. As discussed earlier, impulsive sound has substantial potential for inducing broad spectrum, compounded acoustic trauma; i.e., an impulsive source can produce greater threshold changes than a non-impulsive source with equivalent spectral characteristics. Consequently, impulse is a complicating feature that may exacerbate the impact. Conventional suggestions for minimizing such effects are to ramp the signal, narrow the spectra, lower the pressure, and/or alter the duty cycle to allow recovery and decrease impact. Once again, however, it must be recalled which, if any, of these measures is important to the marine mammal ear has not been determined.   42   Given that impulsive noise can be avoided, the question devolves largely to the coincidence of signal characteristics with species sensitivities. High intensity, ultrasonic devices of course have enormous potential for serious impact on virtually every odontocete and their deployment in pelagic fisheries raises the greatest concern after impulse or explosive sources. Such devices are relatively unlikely to be employed, however, because they are unsuitable for longer range detection. With high frequency sonic range devices, the possibility of profound impact from disruption or masking of odontocete communication signals must certainly be considered, as well as the possibility of coincident impacts to pinnipeds. Because the majority of devices proposed use frequencies below ultra or high sonic ranges, odontocetes may be the least likely to be impacted species. Most odontocetes have relatively sharp decreases in sensitivity below 2 kHz (see fig. 3). If frequencies below 2 kHz are employed with a non-impulsive wave-form, the potential for impacting odontocetes is likely to be drastically reduced, but it must also be borne in mind that it is non-zero. In every case, the difference between some to little or no significant physiologic impact will depend upon received levels at the individual ear. For the purposes of general discussion, a theoretical comparison is shown in Figure 7 for marine mammals audiograms compared with a human audiogram and with source levels of major anthropogenic underwater noise sources. Because mechanisms and onset levels of TTS and PTS are still unresolved for marine mammals, this curve is presented largely for the purposes of gross comparisons of spectra of different sources with animal hearing ranges and is not intended to suggest mitigation guidelines.   Mysticetes and the majority of pinnipeds have substantially greater potential than odontocetes for direct acoustic impact from low to mid-sonic range devices. However, depending upon the diving and foraging patterns of these animals in comparison to the sound field propagated to detect fish, the risks to mysticetes and the majority of pinnipeds may be substantially less than a simple sound analysis would imply. That is, given that substantial numbers of these marine mammal groups are either not present or are infrequently found in the areas of tuna fisheries, there is little probability of any one animal encountering a signal with an intensity and a period of time that will induce acoustic trauma, despite their better absolute sensitivity to the signal. Mitigation, like estimation of impact, requires a case by case assessment. At this time we have insufficient data to accurately predetermine the underwater acoustic impact from any anthropogenic source. Consequently, it is not possible to definitively state what measures will ameliorate any one impact.   For the immediate future and in the absence of needed data, a best faith effort at mitigation must be founded on reasoned predictions from land mammal and the minimal marine mammal and fish data available. It is reasonable to expect, based on the similarities in ear architecture and in the shape of behavioral audiograms between marine and land mammals, that marine mammals will have similar threshold shift mechanisms and will sustain acute trauma through similar mechanical loads. Therefore, fast-rise impulse and explosive sources are likely to have greater or more profound impacts than narrow band, ramped sources. Similarly, we can expect that a signal that is shorter than the integration time constant of the odontocete, mysticete or pinniped ear or which has a long interpulse interval has less potential for impact than a protracted signal; however, simply pulsing the signal is not a sufficient strategy without considering adequate interpulse recovery time. Strategies, such as compression, that allow the signal to be near or below the noise floor are certainly worth exploring. Certainly, no single figure can be supplied for these   43   values for all species. Because of the exceptional variety in marine mammals ears and the implications of this variety for diversity of hearing ranges, there is no single frequency or combination of pulse sequences that will prevent any impact. It is however, reasonable, because of species-specificities, to consider minimizing effects by avoiding overlap with the hearing characteristics of species that have the highest probability of encountering the signal for each device deployed.   Research Needs   To that end, substantially better audiometric data are required. This means more species must be tested, with an emphasis on obtaining audiograms on younger, clearly unimpaired animals and repeat measures from multiple animals. Too often our data base has be undermined by a single measure from an animal that may have some impairment. It is equally important to obtain some metric of the hearing impairments present in normal wild populations in order to avoid future over-estimates of impact from man-made sources. To obtain these data requires a threepronged effort of behavioural audiograms, evoked potentials on live strandings, and post-mortem examination of ears to determination of the level of "natural" disease and to hone predictive models of hearing capacities. It should be noted also that equivalent auditory databases are lacking for most commercially important fish species. Again, all of the recommendations presented are applicable for the fish stocks of interest in this endeavor, and coordinated or tandem research on both the commercially targeted and protected species that may be impacted may be the most productive approach to the problem of determining an effective frequency range for a device that balances effectiveness in fish censusing against minimal impact.   The most pressing research need in terms of marine mammals is data from live animals on sound parameters that induce temporary threshold shift and aversive responses. Indirect benefits of behavioral experiments with live captive animals that address TTS will also test the hypotheses that cellular structure in the inner ear of odontocetes may be related to increased resistance to auditory trauma. Combined data from these two areas could assist in determining whether or to what extent back-projections from land mammal data are valid.   Biomedical techniques, such as ABR and functional MRI, offer considerable potential for rapidly obtaining mysticete and pinniped hearing curves. Evoked potential studies of stranded mysticetes are of considerable value but must also carry the caveat of determining how reliable is a result from a single animal that may be physiologically compromised. Post-mortem studies should be considered on any animal that is euthanized after an ABR with the goal of both providing data about the normality of the ear and supplying feedback to modeling studies of hearing ranges. Otoacoustic emission experiments are not considered to be a viable approach for cetaceans; they may provide basic hearing data in pinnipeds but are technically difficult.   Playback studies are a well-established technique but because of the uncertainties about individually received levels they may not considerably advance our knowledge of acoustic impact per se unless tied to dataloggers or very accurate assessments of the animal's sound field.   Tagging and telemetry are valuable approaches particularly if linked to field or video documentation of behavior that is coordinated with recordings of incident sound levels at the animal. Telemetric measurement of physiological responses to sound; e.g., heart rate, may be valuable, but little is currently known of how to interpret the data in terms of long term impact.   Permanent threshold shift data may be obtainable by carefully designed experiments that expose post-mortem marine mammal specimens to either intense sound and explosive sources since these effects are largely detectable through physical changes in the inner ear. These studies would also substantially increase the species diversity of the available data base because most marine mammal species will not be testable with conventional live animal audiometric techniques. Lastly, because many impact models depend upon assumptions about received levels at the ear, these projections would clearly be enhanced by basic measures on specimens of the underwater acoustic transmission characteristics of marine mammal heads and ears.