7.3. Modelling Approach for Vessles and Continous Sources

  1. For the noise field model, relevant survey parameters were chosen based on a combination of data provided by the Applicant combined with the information gathered from the publicly available literature. These parameters were fed into an appropriate propagation model routine, in this case the Weston Energy Flux model (for more information refer to Weston, 1971; 1980a; 1980b), suited to the region and the frequencies of interest. The frequency-dependent loss of acoustic energy with distance (TL) values were then evaluated along different transects around the chosen source points. The frequencies of interest in the present study are from 20 Hz to 80 kHz, with different noise sources operating in different frequency bands.
  2. The propagation loss is calculated using one of four regions, depending on the distance of the receiver location from the source, and related to the frequency and the seafloor conditions such as depth and its composition.
  3. The spherical spreading region exists in the immediate vicinity of the source, which is followed by a region where the propagation follows a cylindrical spread out until the grazing angle is equal to the critical grazing angle. Above the critical grazing angle in the mode stripping region an additional loss factor is introduced which is due to seafloor reflection loss, where higher modes are attenuated faster due to their larger grazing angles. In the final region, the single-mode region, all modes but the lowest have been fully attenuated.

7.4. Modelling Approach for Impact Piling

  1. In the case of offshore pile installation using an impact hammer, the noise source can be thought of as a “line source” extending through the water column (or in the case of installations using a submersible hammer, a line source through a lower portion of the water column). The hammer strike at the top of the pile produces a compression wave in the pile resulting in radial displacement of the pile walls which is transmitted into the surround media (water and sediments) as noise waves. These compressional waves travel through the pile at circa 5,000 m/s, resulting in a conically shaped wavefront in the water column.
  2. Underwater acoustic propagation modelling for this project will be undertaken using a combined distributed line-source array normal mode model for low frequencies (<1 kHz) complimented by a line-source energy flux model for high frequencies (>1 kHz). The line source normal mode model is based on the KrakenC solver (Porter 2001) implemented for a line array over the pile length. The line-source energy flux model is based on an implementation of the energy flux model for a directional source set out in de Jong et al. (2019).
  3. The normal-mode method involves solving a depth-dependent equation based on the assumption of a set of modes of vibration which are roughly akin to the modes of a vibrating string (Jensen, 1994). The complete acoustic field is constructed by summing up contributions of each of the modes weighted in accordance with the source depth. The KrakenC solver finds the normal modes in the complex wavenumber plane, which allows it to deal with elastic seabed layers, and to include the effects of leaky modes, making it a good choice for calculation of low frequencies for both close and long range noise fields. The method is, however, slow at higher frequencies and has therefore only been implemented for low frequencies (<1 kHz).
  4. The line-source energy flux model (de Jong et al. 2019) used for higher frequencies includes the effect of directionality of the cone shaped wavefront associated with piling noise (circa 17 degrees). This results in damped cylindrical spreading at shorter ranges and mode stripping behaviour at more distant ranges. At even more distant ranges, once the ‘mode stripping’ has eliminated the contribution of all waveguide modes except the lowest mode, propagation is evaluated according to a single mode regime.
  5. For estimation of propagation loss of acoustic energy at different distances away from the noise source location (in different directions), the following steps were considered:
  • The bathymetry information around this chosen source points will be extracted from the GEBCO database in 72 different transects.
  • A geoacoustic model of the different seafloor layers in the survey region will be calculated based on the British Geological Survey (BGS) borehole database and Emodnet sediment database.
  • A calibrated line-source propagation model will be employed to estimate the transmission loss matrices for different frequencies of interest (from 25 Hz to 80 kHz) along the 72 different transects.
  • Source levels for the line-source array will be determined based on a back-calculation from the received noise level and spectrum shape at 750 m (based on the scaling laws set out in von Pein et al. (2022).
  • The calculated source level values will be combined with the transmission loss results to achieve a frequency and range dependant RL of acoustic energy around the chosen source position.
  • The TTS and PTS potential impact distances for different marine mammal groups will be calculated using relevant metrics and weighting functions (from (Southall et al. 2019) and by employing a simplistic animal movement model (directly away from the noise source) where appropriate (as agreed with NatureScot on 27 September 2023; refer to section 7.6.1 for more detail). For assessing marine mammal disturbance using single pulse SEL and for assessing effects on fish, no frequency weighting is applied.
  1. The approach to pile modelling using a line source array model has been consulted on with NatureScot and agreement was received via email on 05 December 2023 (pers. comms., 2023).
  2. The level of detail presented in terms of noise modelling needs to be considered in relation to the level of uncertainty for animal injury and disturbance thresholds. Uncertainty in the noise level predictions will be higher over larger propagation distances (i.e. in relation to disturbance thresholds) and much lower over shorter distances (i.e. in relation to injury thresholds). Nevertheless, it is considered that the uncertainty in animal injury and disturbance thresholds is likely to be higher than uncertainty in noise predictions. This is further compounded by differences in individual animal response, sensitivity, and behaviour. It would therefore be wholly misleading to present any injury or disturbance ranges as a clear line beyond which no effect can occur, and it would be equally misleading to present any noise modelling results in such a way.

7.4.1. Seiche Line Source Model Calibration

  1. The Seiche Ltd line array model has been benchmarked against the COMPILE benchmark workshop for numerical models for pile driving acoustics (Lippert et al., 2016). The COMPILE workshop included modelling results from a number of different organisations in an attempt to compare the performance of acoustic models for piling against pre-defined input parameters. The models included in the benchmarking exercise included those developed by Seoul National University (SNU), Netherlands Organisation for Applied Scientific Research (TNO), Hamburg University of Technology (TUHH), Jasco Applied Sciences, Curtin University and Bundeswehr Technical Centre for Ships and Naval Weapons, Maritime Technology and Research (WTD 71).
  2. A comparison between the Seiche model and the benchmark workshop model results is presented in Figure 7.2   Open ▸ . The results of the benchmarking exercise show good correlation with the other models with the results most closely matching the TNO model. The Seiche model predicts slightly higher received levels compared to the other models at 20 km range for this particular benchmark scenario (10 m water depth, sand substrate). Nevertheless, it is considered that the results of the benchmarking exercise demonstrate a good degree of agreement with other noise propagation models for piling.

Figure 7.2:
Comparison of Seiche Underwater Acoustic Model Against COMPILE Benchmarks

Figure 7.2: Comparison of Seiche Underwater Acoustic Model Against COMPILE Benchmarks


7.5. Geo-acoustic and Noise-speed Input Parameters

  1. Based on BGS core data in the vicinity of the Array, the geo-acoustic model is based on the parameters presented in Table 7.1   Open ▸ .

 

Table 7.1:
Geo-Acoustic Model Used in Propagation Model

Table 7.1: Geo-Acoustic Model Used in Propagation Model

 

  1. The sound speed profile has been based on the mean summer temperature and salinity profile for the region, as presented in Figure 7.3   Open ▸ . To produce a representative sound speed profile, conductivity, temperature, and depth (CTD) data was obtained from the NOAA service WODselect for the closest sample point to the development[12] (NOAA 2023).

Figure 7.3:
Noise Speed Profile Used in Propagation Model

Figure 7.3: Noise Speed Profile Used in Propagation Model

 

7.6. Batch Processing

  1. To improve the performance and reduce the time taken to process and evaluate multiple TL calculations required for this study, Seiche Ltd.’s proprietary software was employed. This software iteratively evaluates the propagation modelling routine for the specified number of azimuthal bearings radiating from a source point, providing a fan of range-dependent TL curves departing from the noise source for each given frequency and receiver depth. In-house routines are then employed to interpolate the TL values across transects, to give an estimate of the noise field for the whole area around the source point.
  2. Once the TL values were evaluated at the source points, in all azimuthal directions, and at all frequencies of interest for various sources, the results were then coupled with the corresponding SL values in third octave frequency bands. The combination of SL with TL data provided us with the third octave band RL at each point in the receiver grid (i.e. at each modelled range, depth, and azimuth of the receiver).
  3. The received levels were evaluated for the SPLpk, SPLrms or SEL metric, for each source type, source location, and azimuthal transect to produce the associated TL. The broadband RL were then calculated for these metrics and from the third octave band results. The set of simulated RL transects were circularly interpolated to generate the broadband RL maps centred around each source point. Representations of these RLs are provided in volume 2, chapter 9 in the form of contour maps.
  4. For impact piling, the far-field received peak sound pressure level was calculated from SEL values via the empirical fitting between pile driving SEL and peak SPL data.
  5. RMS sound pressure levels were calculated assuming a typical T90 pulse duration for impact piling (i.e. the period that contains 90% of the total cumulative noise energy) of 100 ms. It should be noted that in reality, the rms T90 period will increase significantly with distance which means that any ranges based on rms sound pressure levels at ranges of more than a few kilometres are likely to be significant overestimates and should therefore be treated as highly conservative.

7.6.1. Exposure calculations

  1. As well as calculating the unweighted noise levels at various distances from different source, it is also necessary to calculate the received acoustic signal in terms of the SEL metric (where necessary and possible) for a marine mammal using the relevant hearing weighting functions. For different operations related noise sources, the numerical SEL value is equal to the SPL rms value integrated over a one second window as the sources are continuous and non-impulsive. These SEL values are employed for calculation of SELcum (cumulative SEL) metric for different marine mammal groups to assess potential impact ranges.
  2. Simplified exposure modelling could assume that the animal is either static and at a fixed distance away from the noise source, or that the animal is swimming at a constant speed in a perpendicular direction away from a noise source. For fixed receiver calculations, it has generally been assumed (in literature) that an animal will stay at a known distance from the noise source for a period of 24 hours. As the animal does not move, the noise will be constant over the integration period of 24 hours (assuming the source does not change its operational characteristics over this time). This, however, would give an unrealistic level of exposure, as the animals are highly unlikely to remain stationary when exposed to loud noise, and are therefore expected to swim away from the source. The approximation used in these calculations, therefore, is that the animals move directly away from the source. Nevertheless, in the case of fish exposure, calculations have also been undertaken based on a static receiver assumption.
  3. It should be noted that the noise exposure calculations are based on the simplistic assumption that the noise source is active continuously (or intermittently based on source activation timings) over a 24 hour period. The real world situation, however, is more complex. The SEL calculations presented in this study do not take any breaks in activity into account, such as repositioning of the piling vessel, or downtime due to mechanics, logistics or weather.
  4. Furthermore, the noise criteria described in the Southall et al. (2019) guidelines assume that the animal does not recover hearing between periods of activity. It is likely that both the intervals between operations could allow some recovery from temporary hearing threshold shifts for animals exposed to the noise (von Benda-Beckmann et al. 2022) and, therefore, the assessment of sound exposure level is conservative.
  5. In order to carry out the moving marine mammal calculation, it has been assumed that a mammal will swim away from the noise source at the onset of activities. For impulsive noises of piledriving the calculation considers each pulse to be established separately resulting in a series of discrete SEL values of decreasing magnitude (see Figure 7.4   Open ▸ ).

Figure 7.4:
A Comparison of Discrete SEL Per Pulse, and Cumulative SEL Values

Figure 7.4: A Comparison of Discrete SEL Per Pulse, and Cumulative SEL Values

 

  1. As an animal swims away from the noise source, the noise it experiences will become progressively lower (more attenuated); the cumulative SEL is derived by logarithmically adding the SEL to which the mammal is exposed as it travels away from the source. This calculation was used to estimate the approximate minimum start distance for an animal in order for it not to be exposed to sufficient noise energy to result in the onset of potential auditory injury. It should be noted that the noise exposure calculations are based on the simplistic assumption that the animal will continue to swim away at a fairly constant relative speed. The real world situation is more complex, and the animal is likely to move in a more complex manner: at varying speed and direction.
  2. The assumed swim speeds for animals likely to be present across the Array are set out in Table 7.2   Open ▸ .

 

Table 7.2:
Assessment Swim Speeds of Marine Mammals and Fish that are Likely to Occur Within the North Sea for the Purpose of Exposure Modelling

Table 7.2: Assessment Swim Speeds of Marine Mammals and Fish that are Likely to Occur Within the North Sea for the Purpose of Exposure Modelling

 

  1. As an additional sensitivity analysis, modelling was also carried out for fish assuming a swim speed of 0 m/s (i.e. stationary).
  2. To perform the cumulative exposure calculation, the first step is to parameterise the m-weighted sound exposure levels (or unweighted in the case of fish) for single strikes of a given energy via the 95th percentile line of best fit against the calculated received levels from the model. This function is then used to predict the exposure level for each strike in the planned hammer schedule (periods of slow start, ramp up and full power).
  3. In addition to the single source pile driving, simplified situations of simultaneous pile driving from two piling rigs have been considered. The response has been approximated as moving directly away from the point on a line equidistant between the two sources. For simplicity, the sources are considered to be omnidirectional and the piling schedules (soft start, ramp up, etc.) are synchronised, entering each stage of the schedule at the same time.