Temporary threshold shift in bottlenose dolphins exposed to steady-state, 1/6-octave noise centered at 0.5 to 80 kHz^a)

Numerous studies have shown that marine mammals have sensitive hearing and are susceptible to noise-induced hearing loss (NIHL), as revealed by measurements of temporary threshold shift (TTS) after controlled noise exposures [e.g., see Finneran (2015) and Southall (2019)]. As in terrestrial mammals, TTS in marine mammals is heavily influenced by species-specific hearing characteristics, hearing test frequency, and exposure level, duration, frequency, and temporal pattern [see Finneran (2015)].

To account for effects of exposure frequency, recent efforts to predict and mitigate the effects of underwater noise have utilized auditory weighting functions (Southall , 2007; National Marine Fisheries Service, 2016, 2018; Southall , 2019). Weighting functions are essentially bandpass filters applied to a noise exposure before a single, weighted sound pressure level (SPL) or sound exposure level (SEL) is calculated (Houser , 2017). The goal of the weighting function is to emphasize noise at frequencies where animals are more susceptible to NIHL and de-emphasize noise at frequencies where animals are less susceptible. In humans, auditory weighting functions for NIHL were derived from equal loudness contours, which show combinations of frequency and SPL resulting in sounds perceived to be equally loud (Houser , 2017). Human equal loudness contours are obtained from perceptual loudness comparison tasks, which are difficult to mimic with animals [but see Finneran and Schlundt (2011)]. As a result, marine mammal weighting functions for NIHL have typically been based on the manner in which the SEL required for the onset of TTS (typically defined as the exposure level producing 6 dB of TTS) varies with exposure frequency (Southall , 2007; National Marine Fisheries Service, 2016, 2018; Southall , 2019). As a result, the accuracy and validity of the resulting marine mammal weighting functions is limited by the relatively sparse data showing the effects of noise frequency on marine mammal TTS (e.g., few species, individuals, exposure scenarios).

The primary goal of the present study was to measure TTS in bottlenose dolphins (Tursiops truncatus) at a number of different exposure frequencies, to increase the available data for developing an auditory weighting function for predicting TTS onset. Bottlenose dolphins are one of the few species of marine mammals for which such measurements are possible, and have been used to represent other toothed whale species with known or suspected similarities in hearing ability and susceptibility to noise [e.g., National Marine Fisheries Service (2016)].

The exposure paradigm was similar to that previously employed in some studies with harbor porpoises (Phocoena phocoena), harbor seals (Phoca vitulina), and California sea lions (Zalophus californianus), where free-swimming animals were exposed to relatively long duration (60 min), 1/6-octave noise, and exposure levels were estimated by combining estimates of animal position with sound field measurements [e.g., Kastelein (2019b), Kastelein (2020c), and Kastelein (2021)]. The present study differed from these studies in the species involved, use of netted ocean enclosures rather than pools, and use of animal-borne archival audio recording tags to supplement estimates of received exposure levels on the moving animals. The use of free-swimming animals during exposures differs from other studies where animals were (nearly) stationary during exposures [e.g., Kastak (1999) and Schlundt (2000)], and was adopted to accommodate the long-duration exposures necessary to induce TTS given the limited SPLs that could be generated at lower frequencies using the available equipment. Though not a goal of the study, the availability of received levels estimated from animal positions and sound field measurements, and those estimated from animal-borne tags allowed a limited comparison between exposure estimates from the two methods.

TTS was assessed by comparing hearing thresholds measured in individual dolphins before and after noise exposure. Thresholds were measured using behavioral (psychophysical) methods at lower frequencies and both behavioral and electrophysiological [auditory evoked potential (AEP)] measurements at higher frequencies. The specific type of AEP measured was the transient auditory brainstem response (ABR), a short-latency AEP reflecting summed neural activity from the auditory nerve to the inferior colliculus (Ridgway , 1981). ABR measurements were limited to frequencies > 10 kHz, where they retain cochlear place specificity in dolphins (Finneran , 2016). A number of studies have used AEPs to measure TTS in marine mammals [e.g., Nachtigall (2004), Mooney (2009), Popov (2011b), and Popov (2011a)]; however, behavioral- and AEP-based TTS measurements have only been directly compared in two studies (Finneran , 2007; Finneran , 2015). Both studies involved dolphins and utilized measurements of the auditory steady-state response (ASSR) obtained when transient AEPs to rapid stimuli overlap (Stapells , 1984). Both studies showed larger amounts of TTS for equal noise exposure conditions when measured with AEPs compared to those measured behaviorally, with one study (Finneran , 2007) showing much higher AEP-measured TTS. AEP measurements were therefore included in the present study to help ensure that TTS-inducing exposures were identified; i.e., the previous data suggest that exposures might affect ABR morphology, amplitudes, or latencies at levels below those capable of causing behavioral threshold shifts. Transient ABR measurements were used, rather than ASSR measurements, because transient ABRs can be obtained very rapidly and allow identification of individual waves (corresponding to different neural generators) in the response [e.g., Finneran (2017)]. A secondary goal of the study was therefore to compare behavioral- and AEP-based TTS measures from the same exposure, to augment the limited available data on this topic in marine mammals.

A single test session (exposure or control) was conducted each day. Each session consisted of three phases: (1) pre-exposure hearing test, (2) exposure to narrowband (1/6-octave) noise presented for 1 h (exposure session), or a mock exposure consisting of exposure to ambient noise only for 1 h (control session), and (3) post-exposure hearing test(s). Noise exposures were limited to a maximum of three per week. Noise exposure center frequencies were 0.5, 2, 8, 20, 40, and 80 kHz (Table I), tested in three phases: (1) 80, 40, and 20 kHz exposure frequencies tested in descending order; (2) 8, 2, and 0.5 kHz exposure frequencies tested in descending order, (3) additional exposures at 2–80 kHz with one dolphin wearing two archival acoustic recording tags (see below).

Hearing thresholds were measured using behavioral methods (all phases) and ABR measurements (phase 1 only). At each exposure frequency, hearing tests were conducted at the center frequency of the exposure (f₀) and/or 1/2-octave above (f_0.5). Both frequencies were tested because of uncertainties as to where the largest TTS would appear: previous studies suggested that at lower exposure levels maximum TTS occurs at the exposure frequency, but for increasing exposure levels/larger amounts of TTS the maximum TTS spreads upwards towards 1/2-octave above the exposure frequency (Finneran, 2015; Kastelein , 2019a). At each exposure frequency, testing began with relatively low SPLs and gradually increased over successive exposure sessions. Testing at a single exposure frequency persisted until TTS of ∼20 dB was obtained or the maximum achievable exposure source SPL was reached.

Two dolphins were tested at exposure frequencies < 80 kHz. To reduce the total number of noise exposures, both dolphins participated in the same test session each day. When one dolphin was taking a hearing test, the other dolphin quietly waited at the surface a short distance away. After a 2–3 min time interval, the dolphins were signaled to switch places, so the dolphin that had been taking a hearing test began a wait period and the other dolphin began a hearing test. The process was then repeated as necessary. At the conclusion of pre-exposure threshold testing, both dolphins swam freely in the test enclosure while the fatiguing noise exposure or mock exposure occurred. Immediately after cessation of the exposure, post-exposure testing began in a similar fashion as pre-exposure testing.

As dolphins were free-swimming during the exposures, received exposure levels were estimated from animal spatial positions obtained from overhead video recordings and fatiguing stimulus sound field measurements conducted without the dolphins present. Phase 1 testing with higher-frequency exposures showed no TTS at relatively high estimated received levels, raising concerns that the video-based noise exposure estimates were higher than the “true” exposure levels. For this reason, two archival digital acoustic recording tags (DTAG3) (Johnson and Tyack, 2003; Johnson , 2009) were placed on the left and right sides of one of the dolphin's head (TRO, see below) during subsequent exposure sessions to improve estimates of received exposure levels. Tags were obtained late in phase 2; for exposure frequencies where testing was already complete, a limited number of additional exposures was conducted with tags present (phase 3 testing). When tags were present, received exposure levels were taken from the tag recordings. Otherwise, received levels were estimated by applying a correction to the video-based estimate. The correction was derived from differences observed between tag recordings and video-based estimates for those exposures with tags present.

Two male bottlenose dolphins (COL and TRO) participated in the experiments. Upper-frequency limits of hearing for COL and TRO were ∼80 and 140 kHz, respectively [see American National Standards Institute (ANSI) (2018) and Strahan (2020)]. Therefore, only TRO was tested at the 80-kHz exposure frequency. Thresholds for COL and TRO were similar at frequencies below 80 kHz. Tests were conducted between May 2018 and July 2020, within a 9 m × 9 m floating, netted enclosure at the U.S. Navy Marine Mammal Program facility in San Diego Bay, California (Fig. 1). Background ambient noise at the test site was dominated by contributions from snapping shrimp, other dolphins, and passing vessels and aircraft. Median ambient noise pressure spectral density levels varied from ∼77 dB re 1 μPa²/Hz at 0.5 kHz to ∼52 dB re 1 μPa²/Hz at 113 kHz (Fig. 1), which is likely high enough to mask thresholds in dolphins. However, the limited data comparing masked and unmasked TTS have failed to provide a clear relationship, and at present the effects of masked threshold measurements on the resulting TTS estimates are not clear [see Finneran (2015)].

The hearing test station consisted of a submerged frame containing a “biteplate” upon which the dolphins positioned themselves at a depth of 1.4 m. A piezoelectric, underwater sound projector was suspended 1 m in front of the biteplate. The specific sound projector depended on the hearing test frequency: either International Transducer Corp (ITC) 5446, ITC 1032, ITC 1001, or Lubell Labs LL916. Hearing test tones were calibrated before each session using a miniature hydrophone (Reson TC4013) placed at the estimated midpoint between the dolphin's lower jaws when on the biteplate. During test sessions the hydrophone was moved to the side and used to monitor background noise and the dolphin's acoustic responses.

Fatiguing noise was presented using a submerged, piezoelectric transducer placed in the center of the enclosure at a depth of 1.4 m. The fatiguing sound projector depended on the exposure frequency: either Reson TC4033 (80 kHz), ITC 1042 (80 kHz), ITC 1032 (40 kHz), ITC 1001 (20 kHz), or Lubell Labs LL1424-HP (0.5, 2, 8 kHz). The nominal source level for each noise exposure was measured before each session using a TC4013 placed at 1.4 m depth and 1-m distance from the noise projector. Source levels during the actual exposure were also recorded using an autonomous recording device (soundTrap 300 HF, Ocean Instruments, Auckland, New Zealand) placed 1-m from the noise projector, at the same depth. Sound pressures recorded from DTAGs were adjusted based on comparison calibrations with the soundTrap and TC4013 (itself calibrated with a B&K 4228 pistonphone calibrator) placed in close proximity and ensonified by the same broadband noise, so the resulting SPLs obtained from each device were within ± 1 dB.

Figure 2 shows the typical daily test sequence. Each session began with behavioral threshold measurements at either f₀ or f_0.5. The order in which the frequencies were tested was varied from day-to-day (in Fig. 2, the first frequency tested is designated f_A, the second, f_B). At each hearing test frequency, the dolphins were tested in alternating time blocks as described above, with TRO always being tested first. At the conclusion of the behavioral tests, ABR measurements were conducted (phase 1 only), again in the order TRO, COL, with f₀ always tested before f_0.5. At the conclusion of ABR testing, the electrode leads were removed and the exposure began. At the conclusion of the exposure (or mock exposure period for controls), behavioral hearing tests began immediately, and alternated between TRO and COL as during pre-exposure testing; however, behavioral testing was only conducted at the last frequency tested before the exposure (designated as f_B in Fig. 2). Post-exposure ABR tests were conducted in identical fashion as pre-exposure. Additional post-exposure sessions were conducted, in similar fashion, at approximately logarithmic time intervals (e.g., 30, 60, 120 min) until thresholds recovered to near pre-exposure values. The major deviations from Fig. 2 were (1) at 80 kHz, only TRO was tested, the test intervals were 2 min rather than 3 min, the initial post-exposure test period lasted 12 min, and only a single frequency was tested during the behavioral pre-exposure session in early testing; (2) ABR testing was only conducted during phase 1.

TTS (in dB) at specific post-exposure times was defined as the difference between the post- and pre-exposure thresholds. The primary data of interest consisted of the initial behavioral TTS measured at f_B for each subject and, if the initial TTS was sufficiently large, the recovery from TTS. ABR measurements were employed to help ensure any auditory system dysfunction was discovered, whether it resulted in behavioral threshold changes or not, and to provide comparison data between ABR- and behaviorally measured TTS.

Behavioral hearing test methods were similar to those used during previous TTS studies in San Diego Bay (Finneran , 2000; Finneran , 2002; Finneran , 2015). Behavioral hearing tests with each dolphin were conducted over 2–3 min time intervals. Each interval was divided into 2–4 “dives” where the dolphin submerged and positioned itself on the underwater biteplate and listened for hearing test tones. Within each dive, trials occurred at 2–3 s intervals. Hearing test tones were presented on 70% of the trials (signal-present trials), determined via pseudo-random schedule. The remaining 30% of the trials were signal-absent trials (no hearing test tone was presented). The dolphin's task was to produce a conditioned acoustic response within 1.25 s after onset of a hearing test tone and to remain quiet otherwise. At the conclusion of each dive, the dolphin was signaled to return to the surface to receive fish reward. The amount of reward was scaled based on the number of trials and dolphin's performance during the dive.

Hearing test stimuli were 500-ms, linear frequency-modulated tones with 5-ms rise/fall times and 10% frequency modulation. Tones were digitally synthesized then converted to analog at 1-MHz rate using a USB-6361 device (National Instruments, Austin, TX). Analog tones were low-pass filtered at 200 kHz, amplified, and applied to the sound projector. The first tone stimulus was presented with SPL ∼15 dB above baseline hearing threshold at that frequency. Tone SPLs were adjusted on each subsequent signal-present trial using a staircase method (Cornsweet, 1962) with 3-dB step size. Hearing threshold was defined as the mean stimulus SPL over the last 10 reversals—transitions from a response to no response, or vice versa, over consecutive signal-present trials. The false alarm rate was defined as the number of responses during signal-absent trials, as calculated over the same trials used for threshold determination.

ABR hearing tests were also conducted over 2–3 min test intervals divided into a number of dives. Typical dive length was ∼1–1.5 min. During each dive, the dolphin positioned itself on the biteplate and remained quiet while tonebursts were presented and ABRs measured. At the conclusion of the dive, the trainer signaled the dolphin to return to the surface for fish reward, then the next dive was begun if necessary.

Before the first dive, two 10-mm gold-plated, surface electrodes embedded in suction cups were placed on the dolphin: the non-inverting electrode was located on the midline, ∼5 cm posterior to the blowhole, and the inverting electrode was placed near the right external auditory meatus. A third, common electrode was located in the seawater near the dolphin. The electrode signals were amplified (94 dB) and filtered (0.3 to 3 kHz) using a biopotential amplifier (ICP511, Grass Technologies, West Warwick, RI). For each ABR measurement, tonebursts were presented to the dolphin for 12 s while the differential voltage between the inverting and non-inverting electrodes (the instantaneous EEG) was digitized at 100 kHz with 16-bit resolution using the USB-6361 and saved to computer disk. ABRs were obtained from the instantaneous EEG using the randomized stimulation and averaging method (Valderrama , 2012): The time interval between successive tone burst stimulus onsets was randomized from 1 to 3 ms (uniform distribution) and ABRs were obtained by synchronously averaging 10-ms epochs of the instantaneous EEG data time-aligned with the onsets of successive stimuli (Valderrama , 2012; Finneran, 2017). The mean stimulus interval of 2 ms and 12-s averaging time resulted in ∼6000 epochs for each ABR.

From the averaged ABRs, peaks P4 and N5 were identified [see Popov and Supin (1990)] and the p-p amplitude P4-N5 measured. Thresholds were estimated by first fitting linear functions to each dolphin's P4-N5 data from each session, with the slope shared across all the datasets (i.e., pre-exposure and all post-exposure data from the same dolphin that day). Fitting was done using the leastsq function in the optimize module of the python package scipy (Virtanen , 2020). Pre- and post-exposure thresholds were then defined as the peSPLs corresponding to the intersections of the best-fit lines and the x axis [i.e., the peSPL resulting in a 0-V P4-N5 amplitude, see Nachtigall (2004)].

Toneburst stimuli consisted of 8-cycle, cosine-enveloped sine waves centered at the test frequency. Tonebursts were digitally synthesized, then converted to analog at ∼741 kHz (1.35-μs interval) with 16-bit resolution (National Instruments PCIe-7852R). Analog tonebursts were filtered (low-pass at 200 kHz), amplified, and used to drive an underwater sound projector. ABR measurements were conducted at peak-equivalent SPLs (peSPLs) from ∼100 to 140 dB re 1 μPa in 5 to 10-dB steps.

The fatiguing noise signal with 1/6-octave bandwidth was synthesized by digitally filtering Gaussian white noise using custom software. The digital signal was then converted to analog at 500-kHz rate using a USB-6251 (National Instruments). Analog noise was filtered (low-pass at 200 kHz), amplified, and input to the fatiguing sound projector (Table I). The sound field produced by each fatiguing source was estimated by measuring the SPL over the 1/6-octave noise bandwidth, within horizontal planes at depths of 0.5, 1.0, and 1.5 m. Measurements were made over the entire enclosure area, using a rectangular spatial grid with 0.5-m spacing. Figure 3 shows representative examples of the normalized received levels (RLs), defined as the received SPL relative to the source level at 1 m distance and 1.5 m depth, for each noise frequency. At the maximum exposure levels, SPLs at harmonic frequencies were at least 40 dB below those at the fundamental frequency for exposure center frequencies of 8, 20, 40, and 80 kHz. At 0.5 and 2 kHz exposure frequencies, harmonics were at least 25 and 30 dB, respectively, below the fundamental.

Dolphin spatial positions during each exposure and mock-exposure were estimated from overhead video recordings. For each session, still images were extracted from the video at 5-s time intervals. An observer then marked the position of each dolphin in each frame using custom software. Each dolphin was given a distinctive, temporary marking on the dorsal surface to facilitate identification. If a dolphin could not be clearly identified in a frame, its position was estimated by interpolating over previous and successive frames. Video observers were blind to the session type (exposure or control).

When possible, SELs for the actual noise exposures were calculated from both the soundTrap located near the source and the DTAGs worn by TRO. The calculation procedure was identical: The noise mean square pressure was first calculated every 5 s by integrating the pressure spectral density over the 1/6-octave noise bandwidth of the noise. Data from the two hydrophones on each DTAG tag were averaged. The mean square pressures were then integrated over the duration of the exposure to obtain the exposure source SEL. If recordings were obtained from both tags, the mean SEL was utilized (the average SPL differences between the two tags was always < 2 dB).

The noise SEL received by a dolphin during an exposure session was estimated in one of three ways. If DTAGs were present, the received SEL calculated from the DTAGs was used. If tags were not present, the received SEL was estimated by combining estimates of dolphin spatial positions, normalized RL measurements at 0.5-m depth (based on the dolphins' typical depths during sessions), and the source level. The source level was based on the soundTrap recording, if available. If not, the nominal source level (based on the daily calibration) was used. The video-based exposure estimate was then adjusted using the average difference (at that same frequency) between DTAG exposure measurements and video-based estimates for those exposures where DTAG s were present. Correction factors (see Table II) were always negative (i.e., DTAG measurements were lower than video-based estimates) and tended to increase in magnitude as frequency increased. As a consistency check, received SELs from video-based exposure estimates were compared to the actual DTAG data (when available). The comparison showed agreement within ±2 dB at all frequencies except 80 kHz, where agreement was within ± 4 dB.

Figure 4 shows representative examples of dolphin positions during control and exposure sessions. During both control and exposure sessions, the dolphins tended to swim around the periphery of the enclosure, with more time spent along the north and west sides of the enclosure—the sides facing the other enclosures at the facility (as opposed to the East and South sides facing San Diego Bay). This tendency increased during the 8-kHz exposures. During some higher-level exposures at 2, 8, 20, and 40 kHz, the dolphins also tail-slapped the water surface while they swam about the enclosure.

Figure 5 shows examples of the SPL recorded during exposures using the soundTrap and the DTAGs worn by TRO. Source levels were generally constant during exposures except during bouts of tail-slapping, where it was common to see drops in the measured SPLs (seen in the 8-, 20-, and 40-kHz examples), presumably due to entrained air bubbles near the projector. The 20-kHz data represent an extreme example where the dolphins often tail-slapped and swam fast around the enclosure, dramatically affecting the fatiguing sound levels.

Figures 6 and 7 show behaviorally measured TTS as a function of SEL for 0.5, 2, and 8 kHz (Fig. 6) and 20, 40, and 80 kHz (Fig. 7). In general, measured TTS was small and there were no clear patterns with respect to hearing test frequency, however mean TTS measured at the exposure frequency tended to be higher than that measured 1/2-octave above the exposure frequency. The exception was 8 kHz, where mean TTSs exceeded 10 dB for both COL and TRO and TTS at 11 kHz was larger than at 8 kHz. Few effects were seen at the highest exposure frequencies (40 and 80 kHz). Figure 8 compares “onset-TTS” exposure levels from the present study to those measured in Tursiops by Finneran and Schlundt (2013) and specified in current U.S. National Marine Fisheries Service (NMFS) guidelines (National Marine Fisheries Service, 2016, 2018). Onset-TTS SELs from the present study at 2, 8, and 20 kHz were within ±6 dB of those previously measured in a quiet pool at similar frequencies. Present values were also within ±10 dB of the predicted TTS exposure function, except for the values for COL at 500 Hz, where the present data were ∼24 dB below the predicted curve.

Figure 11(a) shows an example of ABR waveforms measured at 28 kHz in TRO, before and after a 20-kHz exposure with SEL of 190 dB re 1 μPa²s. ABR signal quality was generally good (i.e., peaks were typically easily discernable above residual background noise), and waveform morphologies were as expected. Figure 11(b) shows ABR P4-N5 amplitude as a function of stimulus SPL for the waveforms in Fig. 11(a), along with the best-fit lines with shared slope.

Figure 12 shows the TTS obtained from ABR measurements as a function of exposure SEL. As with the behavioral data (Figs. 6 and 7), data were pooled across SELs using the Fisher-Jenks algorithm to determine SEL bin-edges (GitHub, 2023). Within each bin containing multiple exposures, exposure SELs and TTSs were averaged and datasets with maximum TTS ≥ 6 dB were fit with Eq. (1). Measured ABR TTS was highly variable and generally small, with mean TTS ≥ 6 dB only seen after 20-kHz exposures. Differences between TTS measured at the exposure frequency and 1/2-octave above were small, in part due to the overall small amounts of TTS. Correlation between TTS measured behaviorally and with ABRs was weak (Fig. 13) and ABR-measured TTS was typically lower and more variable than that measured behaviorally for the same exposure.

The main goal of the present study was to determine the SEL required for 6 dB of TTS at a number of different exposure frequencies in bottlenose dolphins. The resulting data allowed estimates of TTS onset SELs in at least one dolphin for frequencies of 0.5, 2, 8, and 20 kHz. From 2 to 20 kHz the present data are similar (within ± 6 dB) to previously reported TTS onset SELs for dolphins exposed to 16-s pure tones in a quiet pool [Finneran and Schlundt (2013); Fig. 8], and generally follow the predicted values for TTS onset as a function of frequency [i.e., within 5–10 dB of the predicted TTS exposure function (National Marine Fisheries Service, 2016, 2018)]. At 0.5 kHz, however, the TTS onset obtained from COL in the present study (193 dB SEL) is much lower (∼24 dB) than the predicted TTS exposure function, which was derived from TTS data in dolphins and belugas (Finneran, 2016; National Marine Fisheries Service, 2016). At that time, TTS onset data did not exist below 3 kHz for these species.

At 0.5-kHz, TTS ≥ 6 dB was only observed for COL; however, TTS growth with increasing SEL can be seen for both TRO and COL. The maximum TTS for TRO was 5 dB at an exposure of 195 dB SEL, thus it seems likely that if higher exposure levels could have been tested, onset TTS would have been reached in TRO at SELs only slightly higher than those for COL (and still much lower than the predicted exposure function). Exposure sound fields at 0.5 kHz were “well-behaved” in terms of spatial fluctuations and showed the smallest deviations between exposed levels estimated from video and those measured with the DTAGs (Table II). Therefore, there is no reason to suspect that exposure levels at 0.5 kHz were significantly under-estimated. The present data at 0.5 kHz therefore indicate that the predicted exposure function significantly over-estimates TTS onset at 0.5 kHz and should be adjusted accordingly in future applications.

At 40 and 80 kHz, present data showed no substantial TTS (i.e., mean TTS < 6 dB). At 40 kHz, maximum exposure SELs (185 dB SEL) were only slightly above TTS onset SELs previously measured for TYH in a quiet pool (182 dB SEL). At 80 kHz, maximum SELs were 191–192 dB SEL, well-above the predicted exposure function. Recent TTS studies with harbor porpoises using a similar paradigm [free-swimming animals exposed to narrowband noise for 60 min (Kastelein , 2019b; Kastelein , 2020a; Kastelein , 2020b)] have also resulted in relatively high TTS onsets at 32–88 kHz compared to predictions based on behavioral data, and AEP-measured TTS onsets in a similar species (Yangtze finless porpoise) that were stationary (Popov , 2011a). It is not clear whether the harbor porpoise data indicate a lower susceptibility to NIHL at these frequencies, or if the exposure estimates were affected by the free-swimming paradigm (see below). Given the relatively high deviations between video- and DTAG-based exposure estimates in the present study, and greater possibility of directional effects and shadowing at the higher frequencies, there is more uncertainty as to the “true” exposure levels, and it is possible that the exposure levels at the highest frequencies were over-estimated. Therefore, adjustment of the exposure function for dolphins and related species upward in light of the present data at 80 kHz does not appear justified at this time and would require additional data to support the higher TTS onset. The most useful data would be comparisons of received levels estimated from sound field maps and animal positions and those estimated from animal-borne tags, and comparison of TTS induced in stationary animals compared to that in free-swimming animals with the same estimated exposures.

TTS recovery curves derived from behavioral data followed the expected pattern (Finneran, 2015), with the amount of TTS decreasing proportionally with the logarithm of time post-exposure. Recovery rates—magnitudes of the slopes of linear-log fits to the recovery data—typically increased with initial TTS, and were similar to those previously reported: m ≃ 5–10 dB/decade for TTS₄ = 5–15 dB (Finneran, 2015). The exception was the 8-kHz data, where recovery rates were ∼5 dB/decade and independent of TTS₄. The reason for this difference is not known. The 8-kHz data were unique in requiring 24-h post-exposure testing, which resulted in a large gap in post-exposure test times (e.g., no tests between ∼180 and 1400 min post-exposure). Accurate determination of final recovery times was also affected by the inherent variability in thresholds resulting from ambient noise. It is possible that these factors contributed to producing shallower slopes for the fits to the recovery data with higher TTS₄.

Previous measurements showed changes in ABRs at exposure levels below those causing behaviorally measured TTS (Finneran , 2007; Finneran , 2015). ABR measurements in the present study were therefore viewed as a supplement to the behavioral data, and the ABRs could be used as an “early warning” and indicate that exposure levels were approaching those capable of causing behavioral TTS, even if the ABRs were measured at longer post-exposure times. In practice, however, ABR measurements in the present study provided only limited value. ABR data were highly variable, making it difficult to evaluate differences between pre- and post-exposure ABR thresholds from individual measurements. Only when the data were aggregated (e.g., Fig. 12) could trends with noise SEL be identified. It is not clear to what extent the high variability is simply a characteristic of ABR amplitude metrics, caused by ambient noise, or related to the specific ABR test methodology.

Comparison of ABR- and behaviorally measured TTS after 20-kHz exposures yields mixed results. For TRO, ABR TTS onsets were 186 and 183 dB SEL at 20 and 28 kHz hearing test frequencies, respectively. Behavioral TTS onset was 181 dB SEL at 20 kHz, lower than the ABR value. For COL, no behavioral TTS was observed up to 190–193 dB SEL; however, ABR TTS onset was 189 dB SEL. In contrast to the prior study with narrowband exposures where TTS measured with ASSRs was much larger than that measured behaviorally (Finneran , 2007), no consistent relationship was seen between the ABR- and behaviorally measured TTS onset, and ABR-measured TTS magnitudes were typically lower than behaviorally measured TTS (Fig. 13). Some differences were likely caused by the difference in post-exposure times (2–5 min for behavioral, ∼20–25 min for ABRs), which likely reduced the amount of ABR TTS. The ABR methodology was also different than that utilized in prior studies. Here, transient ABRs were measured using tone burst stimuli presented at rapid rates (and thus were influenced to some extent by adaptation). Prior studies utilized repetitive tonebursts or sinusoidal amplitude modulated tones, which produce the ASSR rather than a transient ABR [e.g., Nachtigall (2004), Finneran (2007), Mooney (2009), Popov (2011b), and Finneran (2015)]. It is possible that the relatively high stimulus levels required for testing given the ambient noise resulted in broad cochlear excitation patterns near threshold, which made the ABR measurements less affected by low-levels of auditory fatigue. Post-exposure ABR peak latencies often showed little change, or even decreased relative to pre-exposure values, despite large changes in ABR amplitude [see Fig. 11(a)]. This may indicate that post-exposure ABRs were primarily arising from a more-basal cochlear region (Finneran , 2016). Future studies comparing TTS measurements using different methodologies, or measurements made in conjunction with high-pass masking noise (to restrict the ABR cochlear place) would clarify this.

Hearing tests were conducted at both the exposure frequency (f₀) and 1/2-octave above the exposure frequency (f_0.5) to ensure any meaningful TTS was identified. Although TTS at f₀ was often slightly higher than at f_0.5, overall there were few differences between the amounts of TTS measured at the two frequencies, and TTS at f_0.5 was small (Figs. 6, 7, and 12). The main exceptions occurred at 8 kHz, where TTS at f_0.5 was relatively large and exceeded that at f₀. Given the relatively low amounts of TTS, the present data are consistent with previous observations: TTS at f₀ is typically higher at lower exposure levels but as exposure level increases, TTS at f_0.5 eventually dominates (Kastelein , 2014).

The main methodological challenge in the present study was estimating exposure levels received by the moving dolphins. In our previous work, maximum exposure durations were either short enough that animal movement was not an issue [e.g., Schlundt (2000)], or dolphins wore hydrophones so the actual received level could be computed [e.g., Finneran and Schlundt (2013)]. Here, the 1-h exposure duration and free-swimming animals created significant issues in determining the effective exposure level. For most frequencies—especially the higher frequencies—the measured exposure sound fields (without dolphins present) were well-behaved; however, as frequency increased the discrepancy between the exposure levels received on the DTAG and those estimated from video increased. It seems likely that depth had a major impact on the received level. Based on observations of the dolphins, we used to shallowest depth (0.5 m) at which measurements were made to estimate received level. However, actual depth could not be determined from the overhead video, and it is possible that the dolphins spent significant time at depths < 0.5 m. Since measured sound levels increased with depth, the true exposure levels may have been lower than the video-based estimates. Shadowing effects from the dolphin's body could have also affected received levels at each ear (see Fig. 5). Finally, it is likely that free-swimming animals will attempt to mitigate noise exposures by avoiding areas of relative high sound level and instead spending their time where received levels are lower. There is evidence of such behavior in a previous study, where dolphins turned their heads just before each impulsive noise exposure, presumably to mitigate the effects of the exposure (Finneran , 2015). At 20 kHz in the present study, dolphin rapid swimming and tail-slapping behaviors near the fatiguing source appeared to dramatically reduce measured source levels and received DTAG levels (Fig. 5), suggesting the swimming movements may have been intended to lower the received sound level. These factors all make estimating received exposure levels from moving animals problematic, thus future studies should attempt to keep animals stationary to the greatest extent possible. If this is not possible, animal-borne hydrophones should be used to estimate received level.

• Current regulatory guidelines for underwater noise exposures likely under-estimate impacts to dolphins and similar species at low frequencies (i.e., below ∼ 1 kHz), and should be updated to incorporate new data.

• Relationships between ABR- and behaviorally measured TTS are not well-understood and may depend on the specific ABR measurement methodology.

• Accurate estimation of received noise levels in free-swimming animals is difficult, and future TTS studies should attempt to utilize stationary animals.

The authors thank H. Bateman, R. Breitenstein, K. Christman, L. Crafton, R. Dear, C. Espinoza, G. Goya, M. Graves, J. Haynesworth, D. Ram, M. Wilson, T. Wu, and the animal care staff, training staff, and interns at the Navy Marine Mammal Program. R. Jones and M. Strahan assisted with the sound field mapping, and K. Christman and M. Strahan helped analyze video recordings. Dr. Robert Burkard provided helpful comments on the ABR data. Financial support was provided by the U.S. Navy Living Marine Resources (LMR) Program.