The study site was located in a sea area approximately 3 km from the coast of Xiamen, China. OED was conducted on a standard vessel of 45 m in length and 14 m in width ( Figure 1A ) that used 4-point anchor spread to remain in a stable location. A drilling platform of 2 m in length was welded to the deck on the middle side of the hull, and the rig derrick ( Figure 1C3 ) was installed on the platform. OED rotated the drill rod to make the diamond bit ( Figure 1C5 ), at the end of drill rod) grinding the soil and rock layer to obtain the cylindrical soil samples ( Figure 1C6 ) and rock samples ( Figure 1C7 ), and used steel casings ( Figure 1C4 ) to protect the borehole. The specific steps are as follows ( Figure 1B ):

According to the field operation, underwater noises during OED on the anchored vessel were mainly categorized into three types: hammering noise (generated in step 1), vibrating noise (generated in step 2), and drilling noise (generated in step 3).

The noise monitoring was conducted on January 10, 2021, with the weather of sunny and the sea state of 2. The OED area’s seafloor is flat, and the water depth was about 7 m (measured by Base X, a sound speed profiler made by Oceanographic AML, inc.) during monitoring. Two fixed measurement stations (station 1 and 2) and one movable measurement station (station 3) were set up to monitor the three kinds of noise, as illustrated in the measurement configuration diagram ( Figure 2 ). Station 1 and 2 were set on the drilling vessel, with distances of 6 m and 18 m from the source, respectively. Station 3 was set on a movable boat with a distance of 280 m from the source during hammering and vibrating down casings, and 40 m during borehole drilling. The distances were measured by a laser rangefinder at the source. At station 1 and 3, underwater noises were recorded by a self-contained LoPAS-L recorder (Hangzhou Soniclnfo Technology Co., Ltd., China, the receiving voltage sensitivity is -192.6 dB re 1 V/μPa) at 3 m water depth, with a sampling frequency of 128,000 Hz. At station 2, noises were recorded by a B&K 8105 hydrophone at 3 m water depth (Brüel & Kjær, inc., the receiving voltage sensitivity ranges from -205.8 to -209.6 dB re 1V/μPa), and then collected by a USB 4431 multi-channel coherent data acquisition card (National Instrument, inc.), with a sampling frequency of 65,536 Hz. Additionally, background noise of the drilling area was monitored when the platform was silent.

The raw measured data (bin files) were converted into sound pressure time-series (time waveforms) in Pa and then analyzed using the custom written MATLAB (MathWorks Inc., Natick, MA, version R2022a) scripts.

In the first step, spectrum diagrams and spectrograms in sound pressure power spectral density (PSD) level (unit: dB re 1 μPa²/Hz) of OED noises were plotted. Spectrum diagrams can provide information on the distribution of noise energy in frequency, and were generated with Welch’s method of segment averaging (Welch, 1967), using 65,536 sample hamming windows with 80% overlap. Spectrograms can provide information on the distribution of noise energy in frequency and time, and were generated using a short-time Fourier transform size of 65,536 and a window size of 65,536 with 80% overlap.

In the second step, the root-mean-square and zero-to-peak sound pressure level (SPL_rms and SPL_zp ) in dB re 1µPa, and the sound exposure level (SEL_s ) in dB re 1µPa²·s were calculated over 1-second-long segments of the sound pressure time-series (for hammering noise, the 0.5 s data before and after each pulse were selected as the 1-second segments). SPL_rms , SPL_zp and SEL_s were all recommended as the key metrics for analyzing and managing underwater soundscapes (Robinson et al., 2014), and SPL_rms is defined as follows (Madsen, 2005):

where p(t) is the instantaneous sound pressure in Pa (Urick, 1983). T is the duration that comprises 95% of the acoustic energy. p_ref is the reference sound pressure, which equals 1 μPa .

SPL_zp provides the peak energy information of the noise (Hawkins et al., 2014; Merchant et al., 2015), and is defined as follows (ISO, 2017):

SEL_s reflects the energy exposure level of a single signal (impulsive sound) or signal in unit time (non-impulsive sound), and is defined as follows (ISO, 2017):

where T equals 1 s, t_ref is the reference time and equals 1 s.

The statistics including means, standard deviations (SD), and ranges (minimum–maximum values) of the three metrics were then calculated, and note that the means were calculated in Pa and then converted to dB.

In the third and final step, source levels (SL_s ) in the three metrics were calculated. The propagation of underwater acoustic signals in marine environments is complex, especially in shallow water where OWFs are commonly located. Urick (1983) gives an equation to simplify the solution of SL_s :

where RSL_s are the received sound levels of each statistic at measurement stations, and TL_s are the transmission losses in dB, which can be defined by the equation:

where r is the propagation range in m. A is a distance-dependent factor, and it equaled 20 (spherical spreading) to estimate the SL_s (r = 1 m) based on the RSL_s at measurement station 1 (r = 6 m). 𝛼 is the frequency-depended absorption coefficient in dB/m, and it was ignored in this paper as the dominant frequencies of OED noises are lower than 10 kHz (Fisher and Simmons, 1977).

This paper further evaluated the potential hearing damage and behavior alteration of OED noises on marine mammals. Hearing damage, also called noise-induced threshold shift (Finneran and Jenkins, 2012; Finneran, 2015), can be divided into temporary threshold shift (TTS) and permanent threshold shift (PTS). TTS means the animals’ hearing thresholds return to normal when the noise exposure disappears, while PTS means the hearing thresholds remain elevated eventually (Southall, 2021). This paper assessed the PTS and TTS risk of OED noises with reference to the marine mammal noise exposure criteria (hereinafter referred to as the criteria) developed by Southall et al. (2019). The criteria divide marine mammals into six hearing groups: Low- (LF), High- (HF), and Very High- (VHF) frequency cetaceans, sirenians (SI), and otariid (OW) and phocid (PW) pinnipeds in water. OW group (sea lions, walruses, and polar bears) was not analyzed in this paper as relatively few conflicts have been reported between these animals and OWFs.

The criteria take the auditory weighted cumulative sound exposure level (SEL_w ) in dB re 1µPa²s as the main assessment metric. SEL_w is an important indicator for evaluating the overall energy exposure level of underwater noise on marine organisms (Martin et al., 2019), and can be expressed as follows:

where f_s is the sampling frequency in Hz, and to measure the SEL_w for all marine mammal hearing groups, the f_s should be 64 kHz or higher. (f) is the mean PSD level of each 1-second-long segment data. T_d is the total exposure time (or called cumulative time) of underwater noises in s. W_aud (f) is the auditory weighting function in dB/Hz, which aim to emphasize the frequencies where the animals are more sensitive and de-emphasize the frequencies where the animals are less sensitive, and it is expressed as follows (National Marine Fisheries Service, 2018):

where f is the frequency in Hz. The values of gain parameter C in dB, cut-off frequencies f ₁ and f ₂ in kHz, and frequency exponents 𝛼 and b all vary with the hearing groups. Eq. (6) reduces to the unweighted cumulative SEL (SELuw) when W_aud (f) = 1.

The criteria establish different PTS and TTS risk thresholds for different hearing groups and noise types (impulsive sound and non-impulsive sound). This paper calculated the SELw for the five hearing groups, then evaluated the PTS and TTS risk of OED noises at three measurement stations referring to the corresponding thresholds. The potential ranges of PTS and TTS risk were estimated by calculating the distance from the source to the point where SELw attenuated to the thresholds.

To date no criteria have established behavior thresholds for different hearing groups to different underwater noises (Southall, 2021). Given the hearing capability of marine mammals’ receiving system (hearing audiogram) is normally slightly higher than the background noise level, a simple method to assess the range of behavior alteration is turning to estimate the distance that the noise propagates from the source to the point where its energy attenuates to the background noise level (assume the animal responds to the noise once received). For instance, Wang et al. (2014) estimated the impact range of vibration piling noise on Sousa chinensis by this method. This paper combined Eq. (5) and the RSL_s in SPLrms at each measurement station to calculate the TL_s , then estimated the distances between the source and the points where the RSL_s attenuated to the background noise level, that are, the behavior reaction ranges of marine mammals to OED noises.

Hammering noise was generated during hammering down the casing. Since a large amount of gravitational potential energy (approximately 3000–4000 J) of the heavy hammer was instantaneously released on casings, hammering noise exhibits the significant characteristics of short duration and broadband, and is a typical impulsive sound ( Figures 3A1–C1 ). Besides, repeated hammer strikes cause hammering noise appears in the form of pulse trains. The dominant frequency of hammering noise is below 10 kHz (take the PSD level of 100 dB re 1μPa²/Hz at station 1 as the threshold), and its peak energy appears at 1 kHz with the PSD level of 140 dB re 1μPa²/Hz. The intensity of hammering noise is the highest among the three kinds of noises, with the SL_s of 197.1 dB re 1μPa rms @ 1 m and 212.9 dB re 1μPa peak @ 1 m ( Table 1 ). Hammering noise strongly resembles the noise of offshore impact piling during OWFs construction phase, and the diameter and material of pile (pipe) are the main factors affecting the noise levels (Reinhall and Dahl, 2011; Zampolli et al., 2013; Lippert and von Estorff, 2014). The pile (pipe) used in offshore impact piling is in various diameters, usually 0.3–2.0 m (Lippert et al., 2016), and in various materials, such as concrete, steel shell, and steel core (Guan and Miner, 2020), so the noise levels vary substantially with pile parameters. In contrast, the casing used in OED is usually steel pipe with a small and relatively fixed diameter, and the intensity of hammering noise normally does not change significantly due to casing parameters.

Vibrating noise was generated during vibrating down the casing. Because the vibration energy was released smoothly from the vibrator to casings, vibrating noise, as one of the products of the energy conversion, is a typical non-impulsive sound. The intensity of vibrating noise is low, with an SL of 158.9 dB re 1μPa rms @ 1 m ( Table 1 ). Both vibrating noise and hammering noise have a strong narrow-band component at the center frequency of 14 Hz with the bandwidth of 10 Hz ( Figures 2C1 , 2C2 ). Considering the generation processes of the two noise types, the narrow-band component may be the low-frequency vibration generated by the interaction between the casing and seabed when driving down casing. In addition, vibrating noise has two strong single-frequency components of 41 Hz and 124 Hz, which may be the resonance signals of casings that driven by vibrator and related to the inherent frequencies of casings. The noise of vibrating down piles during the OWFs construction phase is similar to vibrating noise, but the vibratory hammers used in vibrating down piles are generally heavier and oscillate at a much higher rate (Guan and Miner, 2020), which results in a higher noise level. For instance, the SPLrms of a typical noise during vibratory pile driving of a 1 m diameter steel pile is 175 dB re 1μPa @ 10 m (Buehler et al., 2015).

Drilling noise was generated during the drilling bit grinding the soil and rock layer, which is a non-impulsive sound. The noise came primarily from inside the seabed, and its energy would be greatly attenuated as it transmitted from the soil and rock layer into water. Besides, casings on the outside of drill rod acted as a sound barrier and further impeded the noise propagation. The SL of drilling noise is 155.9 dB re 1μPa rms @ 1 m ( Table 1 ), and the peak energy appears at 45 Hz with a sound level of 136 dB re 1μPa²/Hz. Unlike the vibrating noise, drilling noise still has a high energy in the frequency band above 1 kHz ( Figure 3B3 ), and there is a series of clear irregular stripes with broadband on the spectrogram of drilling noise ( Figure 3C3 ), which is likely to be generated by the collision of drilling rod with the inner wall of casing during rotation. The measured SPLrms of drilling noises on the jack-up platforms are 142–145 dB re 1μPa @ 1 m (Erbe and McPherson, 2017) and 120 dB re 1μPa @ 41 m (Todd et al., 2020), which are lower than the value of 155.9 dB re 1μPa rms @ 1 m measured in this study. The difference in the noise levels may be related to the platforms, the anchored vessel in this paper coupled well with the water and led to a good leakage of the equipment noise on the vessel into the water, while the jack-up platforms are well above the water line.

Figure 4 illustrates the spectrum diagrams of hammering noise, vibrating noise, and drilling noise that measured at the three measurement stations. It can be seen from the figure that the energy of hammering noise decay significantly with the distance, and the decay rate is low in the frequency band below 40 Hz and high in the frequency band above 1 kHz. The energy decay rate of vibrating noise with distance is lower than that of hammering noise, but as the low SL_s , the energy of vibrating noise in the frequency band above 50 Hz would be closed to the background level before reaching station 3 (280 m from the source). Unlike hammering noise and vibrating noise, the energy decay rate in the frequency band below 200 Hz of drilling noise is significantly higher than that in the frequency band above 200 Hz ( Figure 4C ), which may be related to the noise sources and the propagation paths. Drilling noise has two sources: one was inside of seabed and transmitted through the soil (rock) layer to the water, and another was on the casing and transmitted directly into the water. The first source was in the low-frequency band and occupied the primary energy, and the second source was broadband with low energy, but its decay rate was lower than that of the first source. Besides, the decreased water depth (8.5 m depth during hammering down casings decreased to 6 m during borehole drilling because of the tide) was also expected to contribute to the difference in decay rate of the two frequency bands. Lower frequency acoustic signal with longer wavelength in relation to the water depth cannot propagate efficiently in shallower water, because of the “low-frequency” cut-off (Etter, 2018; Guan and Miner, 2020).

Table 2 illustrates the auditory weighted cumulative SEL for the five hearing groups (e.g., SELw,LF refers to the weighted cumulative SEL for the LF group), and the cumulative times (T_d in Eq. (6)) of hammering noise, vibrating noise, and drilling noise were 600 s, 1800 s, and 72,000 s according to the field operation. TL_s between the three measurement stations in the table were calculated based on the unweighted cumulative SEL (SEL_uw ) at each station, and the TL equations (Eq. (5)) for hammering noise and drilling noise from the station 2 to station 3 were modelled to TL ₁(r) = 9.1lg(r) and TL ₂(r) = 9.5lg(r), respectively.

It can be seen from Table 2 that hammering noise has a high hearing damage risk to marine mammals, and there is still a TTS risk to the VHF group at station 3. Based on the TL ₁ equation and the thresholds of PTS and TTS risk, the hearing damage range of hammering noise to marine mammals (VHF group) was estimated to be up to a radius of 300 m from the source. Vibrating noise has low hearing damage risk, and its SEL_w at station 1 is already lower than the thresholds of PTS and TTS risk for each hearing group. Though the SL of drilling noise is lower than vibrating noise ( Table 1 ), the cumulative time of 72,000 s resulted in a high SEL_w , and the maximum damage distance of drilling noise to marine mammals (VHF group) was estimated to be 40 m from the source.

The potential ranges of behavior alteration of OED noises on marine mammals exceeds that of hearing damage. Based on the SPLrms of the three kinds of noises and background noise ( Table 1 ), hammering noise was estimated to be detected by marine mammals up to 1.9 km from the source, and drilling noise was estimated to be detected at a distance of 170 m. The risk of vibrating noise was ignored because its noise energy was likely to approach the ambient level before reaching station 3 (40 m from the source). Comprehensive analysis of the potential risk of hearing damage and behavior alteration, it is recommended to set up a warning zone with a radius of 1.9 km from the source to observe and avoid the presence of marine mammals during OED.