Thresholds for noise induced hearing loss in harbor porpoises and phocid seals

Underwater noise from human activities at sea is known to affect marine organisms that are sensitive to sound. One such activity that has received substantial attention is percussive pile driving of monopile foundations for offshore wind turbines. This concern has prompted many countries to install regulatory procedures for proper impact assessment and permitting of activities (for example, German Federal Ministry for the Environment and Nuclear Safety, 2013; Danish Energy Agency, 2016) and stimulate development of efficient mitigation measures, such as air bubble curtains (see, e.g., Dähne et al., 2017). Noise impact criteria and associated exposure limits are central to regulation and, therefore, such criteria have been the topic of significant research efforts. A key focus of discussion, initiated by the first general proposal for criteria and thresholds by Southall et al. (2007), has been the issue of auditory frequency weighting of sound for assessment and regulatory purposes. Frequency weighting is a filtering of the noise under concern with a filter shape that reflects the different impacts of different frequencies on the auditory system of a given species. Through frequency weighting, regulation and decision-making based on broadband (unweighted) sound levels can be avoided as reliance on broadband levels can lead to significant errors in impact assessments (Tougaard et al., 2015; Tougaard and Dähne, 2017).

The most recent guidance for marine mammals on the issue of thresholds for onset of temporary hearing loss and injury to the auditory system is the comprehensive presentation by Southall et al. (2019), which was based on the authoritative review of experimental data and derivation of weighting functions by Finneran (2015), followed by adoption as interim criteria by the U.S. National Marine Fisheries Service (2016, 2018). These criteria are based on measurements of noise levels required to induce 6 dB of temporary threshold shift (TTS) in captive animals, which are all derived from experimental data published before the review by Finneran (2015). From these thresholds, the higher levels required to induce minimal permanent threshold shift (PTS) in marine mammals are extrapolated. These TTS and PTS thresholds then constitute the recommended exposure limits for application in environmental assessments. However, new findings must continuously be added to the body of empirical evidence; Southall et al. (2007) and Southall et al. (2019) anticipated and encouraged an iterative process, involving regular reassessments, followed by adjustments of criteria and thresholds when appropriate. Since 2015, a considerable number of novel results from harbor porpoise (Phocoena phocoena) and phocid seals have become available. Harbor porpoise, a very high frequency (VHF) cetacean, and harbor seal (Phoca vitulina), a phocid seal with a broadband hearing range, are of particular interest with respect to pile driving as they are acoustically sensitive and among the most common marine mammals in shallow Western European waters—a center of the rapidly expanding offshore wind farm industry. These new results, therefore, call for revisiting the synthesis culminating in Southall et al. (2019). Here, we do that specifically by testing the predictions of the weighting curves against new experimental data for these two species to address whether a revision of weighting curves and exposure thresholds is needed to inform the most appropriate mitigation measures.

Although terminology required for the discussion of thresholds and frequency weighting are provided by Southall et al. (2019), some key definitions regarding impulsive sounds and frequency weighting are revisited below for convenience.

Some confusion exists in the literature regarding the use of the terms “pulse” and “impulse” as characterization of “brief” sound signals. The confusion is partly based in semantics and partly by the lack of clear definitions. What is important, however, is that this group of acoustic signals has been identified as differing from other sounds in that they are generally believed to be more likely than other sounds to inflict damage to biological tissue, including hearing organs. These impulsive sounds have some shared characteristics that make them stand out from other sounds: short rise time from start to peak pressure, often followed by a slower decay; short duration, typically not more than a second and often much less; and large bandwidth (Southall et al., 2007; National Marine Fisheries Service, 2016, 2018; Hastie et al., 2019). It is not possible to provide a more exact definition than the above, but it may be helpful to consider some examples. Good examples of sources that produce sounds with the above characteristics are underwater explosions, seismic air guns, and percussive pile driving. This group of sounds is referred to as impulses or impulsive sounds in some texts (National Marine Fisheries Service, 2016, 2018; Southall et al., 2019) and pulses in others (Southall et al., 2007), differing from other sounds, which have been referred to as non-pulses (Southall et al., 2007) and non-impulsive sounds (National Marine Fisheries Service, 2016, 2018). The non-impulsive sounds may have some of the listed properties of impulsive sounds but not all. Examples of sources producing such non-impulsive sounds are sonars (less sharp onset, often narrow bandwidth, or frequency modulated) and seal scarers (long duration, narrow bandwidth, although often with strong harmonics). Further, impulsive sounds gradually lose their impulsive characteristics through propagation over larger distances and will, therefore, no longer be classifiable as impulses beyond some range from the source (Southall et al., 2007; Southall, 2021). The range at which this gradual transition occurs depends heavily on the frequency spectrum of the sound and the environment, but for pile driving sounds, it has been estimated to occur at a range between a few km to tens of km (Hastie et al., 2019).

Whereas the term “impulsive sound” probably is the most consistent and precise term for the abovementioned sounds with all of the distinctive features, this term is also used in a broader sense in other contexts, such as in the EU Marine Strategy Framework Directive (European Commission, 2008). Here, the term “impulsive” also includes short non-pulses/non-impulses and is used to separate these brief sounds from more continuous sources, such as ships and offshore structures (bridges, platforms, renewable energy installations, etc.). In the following, the more restrictive definition of impulsive sounds sensu Southall et al. (2019) is used.

Animals do not hear equally well at all frequencies. For humans, where enormous empirical evidence is available in the form of thousands of patients with known noise exposure history and measured hearing loss, the consensus is to weigh noise exposures with a curve roughly resembling the inverted audiogram. This so-called dBA-weighting, provides the best overall prediction of risk of injury (see Houser et al., 2017 for an extensive review). For marine mammals, the first auditory weighting curves were proposed by Southall et al. (2007), the so-called M-weighting curves, justified by the shape of equal-loudness curves for loud sounds. Although conceptually important, this justification and the M-curves themselves are now considered obsolete and have been replaced by weighting functions resembling inverted audiograms (Tougaard et al., 2015; National Marine Fisheries Service, 2016, 2018; Southall et al., 2019). Because experimental data are only available for a limited number of species—but also to simplify regulation and assessment procedures—marine mammals have been divided into a number of “functional hearing groups” within which all members are assumed to have roughly the same hearing capabilities and, therefore, similar sensitivity to noise. Out of the total of six hearing groups defined by Southall et al. (2019), two groups of particular relevance to this review are “phocid carnivores” in water (PCW) for the seals and “very high frequency cetaceans”¹ for porpoises.

The PCW and VHF weighting functions are both inverted U-shapes and derived from the TTS data by adjusting the parameters of the equation describing the curve to obtain the best possible fit to the data. See National Marine Fisheries Service (2018) for details and Tougaard and Beedholm (2019) for a practical implementation.

Empirical data on TTS onset thresholds were compiled from the literature. For studies already included in Finneran (2015), onset TTS thresholds, i.e., the sound exposure level (SEL) required to induce 6 dB of TTS measured shortly after cessation of the fatiguing sound, were taken from his review. For all of the newer studies, the TTS onset threshold was derived by the same method used by Finneran (2015), which was fitting a curve to the TTS growth function, describing the amount of onset TTS as function of the SEL of the fatiguing sound. The TTS growth function is described by

where a and b are constants that are adjusted by a least squares method to obtain the best fit to the data (Finneran, 2015). When hearing was tested at different frequencies, only the TTS at a test frequency about half an octave above the fatiguing sound was used as this was the frequency with the highest induced TTS in most of the studies. Once parameters a and b had been estimated for a given experiment, the SEL predicted to result in 6 dB of threshold shift, defined as the TTS onset threshold, was found by rearranging Eq. (1) such that

These TTS onset thresholds were compared with thresholds predicted by the exposure function from Southall et al. (2019). The exposure function is the weighting function flipped around the horizontal axis and vertically aligned to provide the best fit to TTS onset thresholds.

TTS onset thresholds for broadband, impulsive sounds cannot be compared to the exposure function directly and must be compared to the weighted TTS onset thresholds provided by Southall et al. (2019). This was performed with the same method as was used by Tougaard and Dähne (2017), where the decidecade² frequency spectrum of the fatiguing sound was adjusted in amplitude so that the total energy (sum of all decidecade bands) equaled the unweighted TTS onset threshold. The weighted threshold could then be found by weighting the individual decidecade bands with the value of the weighting function at the center frequency of each decidecade band and subsequently summing the energy of all of the bands.

A substantial number of studies of TTS in porpoises and seals have been conducted in recent years and, thus, a large body of empirical data has become available after the review by Finneran (2015). The results of these studies are summarized below with details provided in Table IV of the  Appendix.

Four studies on harbor seals were included by Finneran (2015), two of which related to thresholds for PTS (Kastak et al., 2008; Kastelein et al., 2013a). Seven new publications have appeared after the review (all are listed in the  Appendix). One early study (Kastak et al., 1999) was not included by Finneran (2015), likely due to difficulties in quantifying the exposure and thereby, also, the derivation of a TTS onset threshold. All of the thresholds for non-impulsive (narrowband) sounds for phocid seals included by Southall et al. (2019) and more recent thresholds from the literature are shown in Fig. 1.

TTS onset thresholds at fatiguing frequencies 6.5 kHz (Kastelein et al., 2019c) and 16 kHz (Kastelein et al., 2019b) align well with predictions from the PCW weighting, whereas the thresholds were underestimated by considerable amounts at low frequencies (1 and 2 kHz, Kastelein et al., 2020c) and somewhat overestimated at high frequencies (32 kHz, Kastelein et al., 2020e; and 40 kHz, Kastelein et al., 2020g). Both seals in Kastelein et al. (2020c) did not show any TTS in response to fatiguing noise at the maximum level (211 dB SEL, unweighted) at 500 Hz and one seal did not show TTS in response to 2 kHz fatiguing noise.

Three studies have been published for phocid seals since Finneran (2015), involving impulsive (broadband) sounds as fatiguing sounds. One set of experiments on two ringed seals (Pusa hispida) and two spotted seals (Phoca largha) failed to induce TTS by exposure to air gun pulses at a SEL up to a maximum of 181 dB re 1 μPa² s (unweighted; Reichmuth et al., 2016; Sills et al., 2020). A bearded seal (Erignathus barbatus) in the same setup showed TTS to exposures between 191 and 195 dB re 1 μPa² s (Sills et al., 2020). A third study (Kastelein et al., 2018a) used pile driving noise as fatiguing sound and measured an unweighted TTS onset threshold of 193 dB re 1 μPa² s in two harbor seals. PCW-weighted exposure levels were provided for the bearded seal by Sills et al. (2020) but not for the two other studies. However, as decidecade frequency spectra are available for both of the studies, a PCW frequency weighting could be performed, as described in Sec. III (Fig. 2). The total energy of the fatiguing sound was calculated from the spectra by summing the individual decidecade bands, thereby calculating the difference between the weighted and unweighted thresholds. This difference and PCW-weighted thresholds are listed in Table I.

Four studies with non-impulsive (narrowband) pulses are included in the data for Southall et al. (2019), including one study on finless porpoises (Neophocoena phocaenoides) in which all of the fatiguing stimuli produced considerable TTS (Popov et al., 2011) and two studies with impulsive (broadband) sounds. More recent studies comprise 12 publications on non-impulsive sounds and 5 studies on impulsive sounds. All of the studies and TTS onset thresholds derived from these are listed in the  Appendix.

The auditory weighting function of Southall et al. (2019) for VHF-cetaceans is based on three data points. These are plotted in Fig. 3, together with the exposure function (inverted weighting function) and data points after 2015 from three individual porpoises. TTS onset thresholds for two finless porpoises (Popov et al., 2011) are also included. These data points, however, not only come from a different species but are also based on downward extrapolation from substantial levels of TTS induced in the animals (20 dB TTS and above). Substantial uncertainty is, thus, associated with these data points, and they are included for completeness only.

Overall, there is a good agreement between TTS onset thresholds at lower frequencies (6.5 kHz and lower) and predictions from the VHF exposure function of Southall et al. (2019). At higher frequencies, one study (Schaffeld et al., 2019) reported a TTS onset threshold at 14 kHz, which is more than 10 dB lower than the predictions, whereas several studies at frequencies from 16 to 88 kHz (Kastelein et al., 2019a; Kastelein et al., 2019d; Kastelein et al., 2020a; Kastelein et al., 2020d) reported TTS onset thresholds substantially higher than predicted, as much as nearly 40 dB. These TTS onset thresholds between 16 and 88 kHz are also substantially higher than the thresholds derived from experiments on finless porpoises (indicated by asterisks in Fig. 3) in the same frequency range.

Fewer studies are available for porpoises with impulsive (broadband) sounds. A seminal study is that by Lucke et al. (2009), which showed that TTS, surprisingly, could be induced in a harbor porpoise by exposure to a single pulse from an airgun at a received unweighted (broadband) SEL of only 164 dB re 1 μPa² s, adjusted to 163 dB re 1 μPa² s by Finneran (2015). Recently Lucke et al. (2020), in a reevaluation of the data from Lucke et al. (2009), provided VHF-weighted levels of the exposures used in the original study. From the relationship between unweighted and VHF-weighted levels of the air gun pulse (Table II in Lucke et al., 2020), a weighted threshold for the 2009 experiment was found by a linear regression between weighted and unweighted levels (see the supplementary material³). The VHF-weighted TTS onset threshold from Lucke et al. (2009) was thereby found to be 138 dB re 1 μPa² s.

Since Finneran (2015), several other studies using impulsive sounds on porpoises have appeared. Kastelein et al. (2015a) measured TTS in a porpoise after exposure to a 1 h sequence of pile driving pulses and reported a threshold of 180 dB re 1 μPa² s, unweighted, cumulated over all of the pulses. A subsequent study with the same animal and fatiguing sounds but with variable exposure duration (Kastelein et al., 2016) reported a slightly lower threshold for TTS: 175 dB re 1 μPa² s, unweighted. Two additional studies on the same animal (Kastelein et al., 2017c; Kastelein et al., 2020f) with airgun pulses as fatiguing sounds reported unweighted TTS thresholds of 191 and 199 dB re 1 μPa² s, respectively.

As the frequency spectra of the signals from the studies by Kastelein et al. (2015a),, Kastelein et al. (2016), Kastelein et al. (2017b), and Kastelein et al. (2020f) are known, it is possible to weight them with the VHF weighting function as was done for the harbor seal results (Fig. 4). All of the weighted and unweighted thresholds are listed in Table II.

The weighting functions derived by Southall et al. (2019) were based on limited empirical data, except for the HF-cetaceans, where substantial data were available from bottlenose dolphins (Tursiops truncatus) and beluga (Delphinapterus leucas). This, in concert with the recognition that collection of new data will always be ahead of integrative reviews and generalizing synthesis, Southall et al. (2007) and Southall et al. (2019) anticipated and encouraged revision of their recommendations in light of the emerging new data. Pointing to discrepancies between predictions of their work and data not available in 2015 is, thus, not a critique of the work already performed but simply an extension of their work along the very same lines laid out in their reviews.

For harbor seals, only two data points from two different animals were used to establish the recommendations of Southall et al. (2019), and for harbor porpoises, only three data points, all from the same individual, were available. Inclusion of the results from a significant number of new studies makes it possible to critically review the proposed weighting functions and generalized TTS onset thresholds proposed by Southall et al. (2019). Several systematic discrepancies between the predictions of Southall et al. (2019) and new experimental data are observed in some cases and also a good agreement between the predictions and new results were observed in other cases. These findings are summarized in Table III.

Southall et al. (2019) provided weighted TTS onset thresholds for impulsive sounds for phocid seals and VHF-cetaceans of 170 dB re 1 μPa² s and 140 dB re 1 μPa² s, respectively. Neither of these two thresholds were based on empirical evidence, but they were based largely on extrapolation of the results from TTS onset thresholds for non-impulsive sounds and available evidence from bottlenose dolphins and belugas. For phocid seals and VHF-cetaceans, there are now data from experiments with airguns and pile driving noise as fatiguing sounds (Tables I and II, respectively).

The interpretation of the results with impulsive sounds as fatiguing stimuli is difficult, especially the transfer from experimental settings to full-scale exposures on free-swimming animals—and is associated with significant uncertainty. To reach cumulated SELs sufficient to induce TTS, one can either use the actual source (possible with small air guns and impossible with pile driving noise) or a recorded signal, which is played back through an adequate transducer. In both cases, it is difficult to replicate the full frequency spectrum of the actual source, either because the frequency response of the transducer does not cover the lowest part of the spectrum (see, for example, Fig. 5 in Kastelein et al., 2013b) or because of problems with propagation of the lowest frequencies in the shallow waters typical of experimental settings (see, for example, Lucke et al., 2020).

Bearing these caveats in mind, the results from porpoises are remarkably consistent across experimental paradigms and individuals and cluster well around the 140 dB re 1 μPa² s threshold proposed by Southall et al. (2019), thus, lending substantial support to this weighted threshold, at least for pile driving and airgun sounds. The results from phocid seals are somewhat more scattered and TTS onset thresholds range from 167 to 183 dB re 1 μPa² s, measured across four different species and in two different facilities. The scatter in the results can possibly be attributed to differences between the stimuli (airguns vs playback of pile driving noise), differences between experimental facilities and protocols, or genuine differences between the four different species or seven different individuals studied. While it likely will require additional experiments to identify the source of variation, it is worth noting that the TTS onset threshold measured in the bearded seal exposed to ten or fewer impulses is within 3 dB of the current threshold from Southall et al. (2019), whereas the threshold for the free-swimming harbor seal, exposed to several thousand weaker pile driving pulses, is more than 20 dB higher than the TTS onset threshold of Southall et al. (2019). This difference is consistent with an overestimation of the received SEL in the case of the free-swimming seal. The precautionary approach to setting TTS onset thresholds would, therefore, be to retain the current threshold and at the same time, test the validity of the low susceptibility to TTS found by Kastelein et al. (2018a) through new experiments.

The current phocid weighting function in water is based on two data points in the frequency range 3–5 kHz. Two newer data points at 8 and 16 kHz are consistent with the thresholds predicted from the exposure function, lending support to the weighting function in this frequency range. Two additional experiments at higher frequencies (32 and 40 kHz) resulted in thresholds that were 7–10 dB lower than the prediction of the exposure function, indicating a higher than predicted sensitivity to TTS in the two tested harbor seals. This leads to the conclusion that the shape of the weighting function should likely be revisited for frequencies above 20 kHz and adjusted to better reflect the new data.

Additional data are also available at frequencies below 3 kHz from the same two seals that have provided most of the other data points. At frequencies of 0.5, 1, and 2 kHz, predictions of TTS onset thresholds are substantially higher than those predicted from the exposure function. For 500 Hz, no significant TTS at exposures 20 dB above the predicted threshold was observed. As the seals were reported to spend between 17% and 73% of the exposure time (between 1 and 6 h) with the head at the surface, it is possible that the cumulated SEL that the seals actually received was well below the level estimated from the sound field measurements deeper in the pool. The exposure was quantified by a meticulous mapping of the sound pressure levels at different locations and depths in the tank (in the absence of the seals) and from these, an average exposure was estimated. Such an estimate, however, relies critically on the mean being representative of the sound pressures that were actually present at the location of the seal's ear. The details of sound reception under water are still largely unknown in seals, but sound in water likely enters the head through an acoustic window below the outer meatus (Møhl and Ronald, 1975; see also Kastelein et al., 2018b). If the seal swims very close to the surface, the sound pressure at the acoustic window of the seal is likely to be well below the level predicted from measurements made at 0.5 m depth (the shallowest depth of measurements in the 2 m deep pool) due to pressure release from the surface and destructive interference by reflections from the surface (Lloyd's mirror effect). Because these effects are more pronounced at low frequencies due to the longer wavelengths, this could explain why the largest discrepancy from the expectations occurs at the lowest frequencies. Whether this is the case or the seals are in fact less prone to TTS than predicted at low frequencies can be tested with an animal exposed to noise while stationary in a calibrated sound field under water (as performed in the studies by Kastak et al., 2005; Kastak et al., 2008; Reichmuth et al., 2016; and Sills et al., 2020) or by direct measurements of the sound exposure of the swimming seals, for example, by monitoring the sound level directly on the side of the seal's head with a small hydrophone attached to a suitable recorder on the back of the seal.

The current VHF weighting function is based on three data points between 1.4 and 7 kHz, all of which were obtained from the same animal. A substantial number of additional data points have become available since 2015 on three other individuals and at higher and lower frequencies. For frequencies between 500 Hz and 10 kHz, there is a good correspondence between the new experimental data and predictions of the exposure function. In the intermediate range, 10–20 kHz, there are now two studies on three different animals with substantial scatter in the TTS onset thresholds. The lowest threshold was measured with the auditory evoked potentials (AEP) technique at 14 kHz (Schaffeld et al., 2019), and this threshold is nearly 30 dB lower than the least sensitive of two animals tested with behavioral methods at 16 kHz (Kastelein et al., 2019d). Although the average of the TTS onset thresholds of these two experiments is close to the predicted threshold from the VHF exposure function, the source of the large variation should be further investigated. In particular, it needs to be understood if the discrepancies relate to evoked potential vs behavioral methodology: The large difference may possibly be explained by the different experimental protocols as TTS is known for bottlenose dolphins to occur at lower exposure levels when hearing thresholds are measured by evoked potentials rather than if measured by behavioral techniques (Finneran et al., 2007). Another possible explanation is the difference in the duration of the fatiguing noise in the two studies: 1 h of low-intensity noise vs 0.5 s of high-intensity noise (Kastelein et al., 2019d; Schaffeld et al., 2019, respectively), where the short duration exposure is more likely to generate a TTS for the same summed SEL as the long exposure.

Above 20 kHz, the measured TTS onset thresholds for harbor porpoises are consistently higher than the predicted thresholds from the VHF exposure function, by about 35 dB for the highest frequencies (64 and 88 kHz), and even higher than the extrapolated TTS onset thresholds for finless porpoises in the same frequency range. A difference in the TTS onset threshold of 35 dB reflects an expected comparison between one pulse with the intensity predicted from the exposure function to elicit TTS and an exposure more than 3000 times longer actually being required. Such a large deviation between predictions and measurements at relatively high frequencies is puzzling and deserves careful scrutiny in future experiments. As noted by Kastelein et al. (2020d), it is known that dolphins and porpoises can reduce their hearing sensitivity by up to 40 dB when anticipating loud sounds in the frequency range 20–80 kHz (Nachtigall and Supin, 2014; Nachtigall et al., 2016; Finneran, 2018), which can potentially explain the discrepancies between the prediction and measurements in full or in part. Whether the ability of porpoises to raise their hearing threshold is larger at higher frequencies is unknown and should be tested by experiments. If possible, such experiments should also address the large discrepancy between the results from harbor porpoises and the results of Popov et al. (2011) on finless porpoises. From a regulatory point of view, a precautionary approach would be to retain the current VHF weighting function until the source of the discrepancies between the results from harbor porpoises and finless porpoises have been identified. At the same time, however, it should be explored whether the possible reduced risk of TTS and PTS following noise exposure—if the exposure is anticipated by the animal—could be exploited in mitigation of the impact. Such mitigation could be in the form of warning sounds immediately before loud noise emissions (Nachtigall and Supin, 2013).

All of the available data between 16 and 88 kHz were obtained from the same two porpoises in the same facility and with the same experimental protocol. As the two animals are consistent in their responses and have shown normal hearing in other experiments (Kastelein et al., 2017b), it appears unlikely that they are both atypical in their susceptibility to TTS. This, again, points to the experimental protocol, where animals are free to swim around in the pool during exposure as a possible explanation as to why the exposure of the animals might be underestimated, which, in turn, would lead to the higher than expected observed TTS onset threshold. It is noteworthy, however, that the same noise exposure protocol resulted in lower than expected TTS onset thresholds at 32 and 40 kHz for seals. In any case, it would be straightforward to test for a possible effect of the protocol by conducting experiments with a porpoise trained to station in a fixed position while exposed to narrowband noise at frequencies in the range 16–88 kHz. If the exposure function of Southall et al. (2019) is correct, an exposure to, for example, a 32 kHz fatiguing noise of 1 s at 160 dB re 1 μPa (SEL = 160 dB re 1 μPa² s) should induce a small amount of TTS at 50 kHz, half an octave higher. In contrast, according to the results of Kastelein et al. (2019a), a continuous exposure of the same frequency for 2 min (SEL = 181 dB re 1 μPa² s) would be required to induce a minimum TTS. Alternatively, the received noise during exposure of a free-swimming animal could be measured directly, as suggested for the seals, by equipping the porpoise with a sound dosimeter during long duration exposures to fatiguing noise.

Substantial interest from regulatory bodies and other stakeholders surrounds the question of estimating generalized, weighted thresholds for TTS onset in marine mammals as they form the basis of legislation and regulation of noise-generating activities in many countries. The present review of available data leads to conclusions regarding impact on phocid seals and harbor porpoises⁴ from exposure to some of the most common types of noise sources: pile driving, airguns, sonars, and acoustic deterrent devices.

The TTS onset thresholds proposed by Southall et al. (2019) for impulsive (broadband) sounds are in good agreement with direct experimental results for pile driving playback and airgun sounds as fatiguing stimuli, which lends substantial support for the use of the suggested thresholds in regulation of the impact from pile driving and seismic surveys. No experimental results are available for other loud impulsive sounds, such as those generated by sparkers and other geoacoustic equipment, but until such data may become available, it seems reasonable and precautionary to conclude that the current thresholds would be applicable to these sources as well. Shock waves from explosions, however, are acoustically very different from the other impulsive sounds and criteria other than weighted SEL may be more appropriate, such as peak pressure or acoustic impulse.⁵ See Southall et al. (2019) and Lance and Bass (2015) for further discussion.

Low- and mid-frequency sonars, all in the range below 10 kHz, have been the subject of several of the studies reviewed above due to their high relevance to the use and possible regulation of military sonar exercises. The updated results for this group of non-impulsive sounds are very different in seals and harbor porpoises. Whereas there is strong support for the use of the VHF-weighted threshold proposed by Southall et al. (2019) for harbor porpoises (153 dB re 1 μPa² s), there is substantial discrepancy between the predictions and experimental results for seals. However, as the new experimental data indicate a lesser sensitivity to low-frequency sounds than the current criteria, the current criteria are precautionary as they are.

The results from seals appear to indicate a higher sensitivity to TTS at high frequencies than predicted by Southall et al. (2019), lending support to a revision of the phocid weighting function for frequencies above 20 kHz, making the criteria more precautionary for high frequency sounds. The situation is more complex for harbor porpoises when it comes to non-impulsive sounds above 20 kHz as there are major and systematical deviances between the predictions and experimental results. However, sound exposures, if weighted by the current VHF weighting function, will overestimate impact and thereby be precautionary.

The initial (Southall et al., 2007) and recent recommendations (Southall et al., 2019) stressed the need for regular data driven revisions of the weighting functions and thresholds if required. The present review of new experimental data for porpoises (VHF-cetaceans) and phocid seals has highlighted the importance of such an iterative process. Despite significant discrepancies between newer data and the existing weighting functions, especially at the higher and lower edges, the current thresholds and weighting functions for VHF-cetaceans and phocid carnivores are still precautionary and, therefore, if anything, will tend to overestimate impact in assessments. The only exception is for narrowband signals above 20 kHz for phocid seals, where the impact may be underestimated and a revision of the PCW weighting function is warranted. Most importantly, however, is that the thresholds for onset of TTS for pile driving noise, seismic air guns, and military low- and mid-frequency sonars are well supported by the experimental evidence that has become available after the current criteria were established. These three sources of underwater sound are among the most important anthropogenic sources from a regulatory viewpoint.

Substantial uncertainties remain when it comes to the impact of low-frequency noise exposure in seals and high frequency noise exposure in porpoises. The observed discrepancies between predicted and measured TTS onset thresholds can only be resolved through new and dedicated experiments.

This work was supported financially by the Danish Energy Agency and is based on the report “Thresholds for noise induced hearing loss in marine mammals. Background note to revision of guidelines from the Danish Energy Agency” by J.T., Aarhus University, Danish Center for Energy and Environment, Roskilde, Denmark, 2021. Colleen Reichmuth, Dorian Houser, and an anonymous reviewer are thanked for extensive and constructive comments to an earlier version of the manuscript.

The results of all studies included in this review are summarized in Table IV.