Scottish Marine and Freshwater Science Volume 4 Number 5: Modelling of Noise Effects of Operational Offshore Wind Turbines including noise transmission through various foundation types

2 Technical background

2.1 Vibration and noise produced by wind turbines

Noise from wind turbines comes in two forms: the first is aerodynamic noise from the blades slicing through the air leading to the characteristic swish-swish noise; the second is mechanical noise associated with machinery housed in the nacelle of the turbine. Aerodynamic noise travels through the surrounding air to the interface between the air and water where it is almost entirely reflected due to the large impedance contrast between air and water. Little aerodynamic noise enters the marine environment. Conversely, the mechanical noise has a strong structural pathway between the drive train (where the vibration is created), through the nacelle support frame, tower, into the foundation and finally from the foundation into the surrounding water where it is released as noise. The great majority of noise in the marine environment due to wind turbines is therefore related to mechanical vibration in the drive train.

Mechanical vibrations in the drive trains of wind turbines are created by imbalances of the rotating components, the teeth in the gearbox coming into contact with each other (referred to as gear meshing), and electro-magnetic (E-M) interaction between the spinning poles and stationary stators in the generator. Each of these vibration sources occurs in discrete frequency bands related to the rotation speed of each component: the vibrations therefore tend to be tonal (as opposed to broad band). Rotational imbalances tend to occur at very low frequencies (< 50 Hz), while gear meshing and E-M interactions tend to occur at low to moderate frequencies (50 Hz to 2 kHz), Table 2-1. Other mechanical vibration produced by wind turbines during normal operation tend to be of a temporal nature with durations of seconds to tens of seconds. These include the pumping of hydraulic fluid, cooling systems and yawing of the nacelle followed by braking.

Table 2-1 Frequency bands likely to contain vibration tones produced in the drive train of wind turbines.

	Frequency
Rotational imbalance of rotor	0.05 to 0.5 Hz
Rotational imbalance of high speed shaft between gearbox and generator	10 to 50 Hz
Gear teeth meshing	8 to 1000 Hz
Electro-magnetic interactions in the generator	50 to 2000 Hz

The amplitude of the vibration of a wind turbine and related noise emitted by the foundation is controlled by the size of the excitation force, the frequency of structural resonances and the level of damping in the structure. The magnitude of the excitation of the drive train is related to the torque acting on the rotor, which is dependent on the wind speed. The amplitude of vibration of the turbine increases with the square of wind speed at the hub height. It is likely, therefore, that the noise emitted by the foundation will also rise with wind speed.

Mechanical noise can be amplified by structural resonances within the wind turbine. Structural resonances are the harmonic frequencies at which a structure vibrates when excited by a discrete event (e.g. the frequency a bell rings when struck). When an excitation frequency such as gear meshing has the same frequency as a structural resonance, the amplitude of the vibration is amplified, sometime dramatically. This becomes important in the event of frequency matching between an excitation frequency in the drive train and a resonance in the foundation as the noise emitted into the marine environment will be significantly amplified.

Understanding structural resonances is also important because resonances can be excited by multiples of excitation frequencies. For instance, a resonant mode in the steel surface of the tower at 600 Hz can be excited by a gear meshing frequency of 200 Hz. In this example the resonance coincides with the third multiple of the gear meshing (3 × 200 Hz = 600 Hz). Structural resonances can therefore produce vibration and related noise at frequencies that would not otherwise have been excited.

All structures contain some level of internal damping. Damping is the dissipation of vibration energy via processes like heat loss and has the effect of reducing the amplitude of vibration. In general, steel structures such as jackets have less damping than structures built from granular materials such as concrete foundations. The level of internal damping will therefore affect the noise emitted by different types of foundations. Damping may also be increased over time by biofouling, where the encrusted organisms begin to act as a granular aggregate with high internal friction.

2.2 Vibration and underwater acoustics

At the interface between the foundation and water, the vibration of the foundation oscillates water molecules to produce a pressure wave which radiates away from the foundation as sound. As the sound propagates away from the foundation its intensity is reduced with distance due to geometric spreading and absorption. Water absorbs high frequency sound more quickly than low frequencies; low frequency sound therefore propagates further. At the sea surface the sound is almost perfectly reflected by the high impedance contrast between water and air, though some sound may be scattered by surface waves or absorbed by near-surface air bubbles. At the seabed sound is also reflected and scattered, though its behaviour is more difficult to predict than at the surface due to the seabed's variable acoustic properties (soft sediment to hard rock) and internal layering of material with different densities and sound speeds.

Several underwater acoustic measurements of offshore wind turbines have been carried out (Westerberg 1994, Degn 2000, Ingemansson Technology 2003, Betke et al 2004, Thomsen 2006, Nedwell 2011). Measurements recorded to date have been of turbines with different design parameters, such as foundation type, water depth, turbine size, sediment type and wind speeds - making direct comparisons difficult. However, noise related to off-shore wind turbines have common features; specifically, the sound intensity is dominated by pure tones likely to originate from rotating machinery in the nacelle with frequencies mostly below 700 Hz.

The range of water depths for previous measurements is from as little as 2 m up to depths of 15 m. The shallower measurements have a lower frequency cut off as sound can only propagate if the wavelength is less than or equal to 4 times the water depth (Urick 1983). This results in frequencies less than 60 Hz being unable to propagate in 6 m of water. The exact threshold has a dependency on the sediment type.

A comparison of key results from these measurements provides an indicator as to the sound levels produced and the sensitivity of noise generation from offshore wind turbines. Table 2-2 presents maximum noise levels recorded with their corresponding frequencies and the conditions under which the measurements were taken. Table 2-2 is presented to give the reader an overview of known operational noise produced by offshore-wind turbines. However, it should be noted that Table 2-2 shows measurements from several of published sources that used a variety of measurement techniques and present their data with different units. The reader is directed to the caption of Table 2-2 for notes on units of SPL.

Table 2-2 Summary of underwater acoustic measurements carried out and conditions during data collection. Dominant peak levels are shown with the respective frequency and the distance at which the measurement was taken from the turbine foundation. Where noted the distance is back calculated to a level at the source. Units reported are: *(dB re1 µPa), †(dB re1 µPa ²/Hz), ††(dB re1 µPa ²/Hz TOL), ‡(dB µPa 1/3 octave level).

** Values are taken from from Wahlberg and Westerberg (2005)

Source	Location	Foundation	Power ( MW)	Depth (m)	Wind Speed (m/s)	Distance from Foundation (m)	Frequency (Hz)	Received Noise Level (see table caption for reported units)
Westerberg (1994)	Nogersund	Tripod	0.2	5-15	12	100	16	113*
Degn (2000)	Vindeby	Concrete Gravity Base	0.5	2-4	13	14	150	100†
Degn (2000)	Bockstigen	Monopile	0.6	10	13	20	160	95†
Henriksen (2001)	Middlegrunden	Concrete Gravity Base	2.0	5	13	Converted to Source Level	125	115††
Betke (2004)	Mecklenberg	Monopile	1.5	10	17	110	180	112‡
Ingemansson (2003)	Utgrunden** (from 2005)	Monopile	1.5	5-10	13	Converted to Source Level	180	151*
Thomson (2006)	Utgrunden	Monopile	1.5	5-10	12	110	160	115*
Nedwell (2011)	UK	Monopile	3-3.6	5-15	3.9-7.2	20m	100	112†

In addition to the site specific variables there will also be discrepancies in recording conditions and radiating patterns as these measurements do not investigate directional spreading. Wahlberg and Westerberg (2005) present a comparison of the larger measured values back calculated to a distance of 1 m from the foundation, shown in Figure 2-1.

Figure 2-1 Wahlberg and Westerberg (2005) present a summary of source-level measurements of underwater noise generated by turbines. The measurements were caried out by Westerberg (1994: Nogersund), Degn (2000: Bockstigen 8 ms ^-1 and Vindeby), Fristedt et al. (2001: Bockstigen 5 ms ^-1) and Ingemansson (2003: Utgrunden). Noise level: background noise level measured by Piggott (1964) in 40 to 50 m water depth. The properties and conditions of recording are as in Table 2-2. The noise levels have been back calculated by Wahlberg and Westerberg (2005) to a distance of 1 m for comparison.

The highest levels presented by Wahlberg and Westerberg (2005) were achieved by Ingemansson (2003) with 151 dB re1 µPa at 180 Hz at a back calculated distance of 1 m from the foundation. Madsen (2006) expand on the review of Wahlberg and Westerberg (2005) summarising that the tonal peaks seen below 1 kHz are likely to be linked to the mechanical properties of the turbine with no direct measurements of source tonal levels exceeding 145 dB re 1µPa ( RMS). In addition, Madsen (2006) reviewed that measured levels drop below 120 dB re 1µPa ( RMS) at 100m. However measurements to date do not account for directional components or cumulative effects of multiple turbines.

2.3 Noise detection by marine species and background noise

Marine fauna exposed to anthropogenic sound may experience detrimental effects that include physical injury, behavioural disturbance and displacement, masking of biologically important signals, and other indirect effects. These are defined by the proximity of the animal to the sound source, the sound level received by the animal, the hearing sensitivity and acoustic characteristics of the vocalisations of the animal, and the acoustic characteristics of the anthropogenic noise.

The key potential impacts of operational turbine noise on marine species are:

• Disturbance as a result of underwater noise arising from operational offshore wind turbines.
• Potential longer term avoidance of the development area by marine mammals
• Potential reduction of the feeding resource due to the effects of noise, vibration, and habitat disturbance on important prey species

Assessment of the likely extent and significance of such impacts should be quantitative wherever possible and all uncertainties explicitly included. Marine species hearing sensitivities cover a broad frequency range and as such the same sound source may elicit different behavioural and physiological effects in different species. In addition, within-species responses may vary depending on individual traits of exposed animals and the context in which they are exposed. Nevertheless, based on previous published studies and established noise threshold recommendations, generalised predictions of how individual species may be impacted by certain noise sources are possible.

The frequency band over which different species can sense noise varies greatly ( Table 2-3). In general fish sense noise in the 10s to 100s of Hz range, though some clupeiform fish including shad may hear into the 100s kHz range. Marine mammals such as seals, dolphins and whales are capable of hearing noise between 10s of Hz to 100s of kHz. Wind turbines produce vibration and related noise between 0.5 Hz to 2 kHz ( Table 2-1) which overlaps frequency bands that are detectable by species living in Scottish waters ( Table 2-3).

While marine species are capable of hearing noise from wind turbines, in many cases this will be masked by the background noise of the ocean. The ocean is inherently a noisy place with background noise contributed by wind interaction with the surface, rain, industrial activity and shipping, explosions and earthquakes and biological activity. At some range from a wind turbine the noise produced will be less than the background noise at which point marine species will no longer be able to detect it. Of particular importance in relation to wind turbines is the relationship between wind speed and noise production. As has been noted above, the vibration and noise produced by wind turbines increases with wind speed. There is a similar relationship between wind speed and background noise (Urick, 1983). Thus, while wind turbine noise increases with wind speed, so too does the masking effect of background noise, Figure 2-2. For the purpose of this study a worst-case scenario is taken where the lowest applicable background noise is used so that masking effects are minimised. Thus, the background noise is taken as the lowest SPL associated with low frequency shallow water noise from 0 to ~100 Hz (lower boundary of the brown field in Figure 2-2) combined with the SPL associated with the relevant sea state conditions at higher frequency (>100 Hz). The relevant sea states to wind speed are taken as: sea state 2 for wind speeds of 5 ms ^-1; sea state 4 for wind speeds of 10 ms ^-1 and; sea state 6 for wind speeds of 15 ms ^-1. Continuing to follow the worst-case for masking; temporal increases in SPL associated with shipping have been excluded from the background noise.

Table 2-3 Approximate sound detection frequency range for some species in Scottish waters.

Species	Hearing range	Reference
Allis shad	10 Hz to 180 kHz	(Mann, et al. 2001)
European eel	10 to 300 Hz	(Jenko, et al. 1989)
Salmon	32 to 400 Hz	(Hawkins and Myrberg 1983)
Sea trout	100 to 1000 Hz	(Horodysky, et al. 2008)
Minke whales	7 Hz to 22 kHz	(Southall et al. 2007)
Bottlenose dolphin	150 Hz to 160 kHz	(Southall et al. 2007)
Harbour porpoise	200 Hz to 180 kHz	(Southall et al. 2007)
Grey seal	75 Hz to 75 kHz	(Southall et al. 2007)
Harbour seal	75 Hz to 75 kHz	(Southall et al. 2007)

Figure 2-2 Wenz curve showing typical sound levels in the ocean (Wenz 1962) with sound level in units relative to 1μPa.

2.4 Impacts of noise on marine mammals

Marine mammals spend most, or all, of their lives at sea, and for the majority of that time they are submerged. Sound propagates efficiently through water and marine mammals rely on the use of sound to communicate with conspecifics, for predator avoidance, to locate and capture prey, mate selection and social interactions (Akamatsu et al. 1994; Au et al. 2004; Goodson and Sturtivant 1996; Hafner et al. 1979; Hastie et al. 2006; Janik 2000; Janik 2009; Madsen et al. 2005a; Madsen et al. 2005b; Rendell and Whitehead 2004; Schulz et al. 2008). Coupled with this, they have an acute sense of hearing with a high sensitivity over a wide frequency range (Nedwell et al. 2004; Richardson et al. 1995; Southall et al. 2007). This reliance on sound in their general ecology makes marine mammals particularly vulnerable to the effects of underwater noise.

Any anthropogenic noise could impact a marine mammal if the sound falls within its audible range; noise disturbance can have a range of effects depending on the sound type or source level. Loud, intense noise sources such as explosions have the potential to cause lethal physical non-auditory injury to marine mammals, while other noise sources can cause auditory damage, elicit behavioural responses (e.g. displacement and/or habitat exclusion), induces stress and/or mask biologically important signals (Richardson et al. 1995).

Given the relatively low noise levels emitted from operational wind turbines, it is not likely that sound levels are high enough to cause auditory injury beyond a few metres of the device and only if animals remain there for extended periods of time. We therefore only consider the impact of noise from marine operational wind turbines on the behavioural response of five priority species of marine mammal:

• Harbour porpoise ( Phocoena phocoena)
• Bottlenose dolphin ( Tursiops truncatus)
• Minke whale ( Balaenoptera acutorostrata)
• Harbour seal ( Phoca vitulina)
• Grey seal ( Halichoerus grypus)

2.4.1 Behavioural response

The introduction of noise into the underwater environment may impair an animal's ability to detect calls or may disrupt its normal behaviour in some way. Noise impacts can be thought of in terms of 4 zones of influence (Richardson et al. 1995). The zone of audibility is the range at which animals can only just detect the anthropogenic sound source. The zone of masking is the range at which the sound exposure interferes with the signals produced by the animal, at a given frequency, and thus lowers the probability of the animal's signal being detected. This means the distances over which animals can communicate will be greatly reduced. The zone of responsiveness is smaller and is the impact range around the sound source where animals are expected to show physiological or behavioural responses to the sound. The zone of injury is the smallest zone but potentially with the highest impact. This is defined as the range at which the received sound levels are high enough to induce either direct physical injury or loss of hearing sensitivity (hearing damage).

The likelihood of an animal experiencing one or more of these effects is defined by the spatial relationship of the receiver and the sound source, the hearing sensitivity and acoustic characteristics of the vocalisations of the receiver, and the acoustic characteristics of the anthropogenic noise. Consequently, for many sound sources, responses are poorly described and predictions of potential effects can be challenging. Nevertheless, based on previous published studies and noise threshold recommendations, generalised predictions of how individual species may be impacted by certain noise sources are possible.

While the physical process of detecting or being damaged by a sound can be predicted from a combination of empirical studies and acoustic models, this is generally not the case for behavioural responses. The behavioural response of animals to sound appears to be influenced by a number of factors including food motivation, the context of exposure, and the animal's previous exposure history. This means that the way in which an individual responds to sound can vary between both individuals and sound exposure events.

2.5 Hearing sensitivity of marine mammals

The hearing ability of marine mammals is commonly described using audiograms; this is a plot of the hearing sensitivity of a species at different frequencies, which indicates the range of frequencies detectable by a species and can highlight where hearing is most sensitive. The hearing threshold can be defined as the received sound level in the vicinity of the ear that is just audible to an animal. Hearing thresholds depend on the frequency of the sounds and can vary strongly across species. An audiogram displays hearing threshold as a function of frequency. A lower sound pressure level value on an audiogram display reflects a low hearing threshold at a given frequency and hence a high auditory sensitivity. Audiograms for mammals are typically V- or U-shaped reflecting the fact that hearing sensitivity declines towards the edge of the hearing range (i.e. at both ends of the V/U-shape).

Audiograms are typically derived experimentally and can be based on behavioural or electrophysiological responses ( AEP/ ABR) to sound stimuli. Both approaches are considered robust methods for collecting audiogram data. However, it is unclear how the AEP measurements compare with audiograms derived from behavioural methods. Some studies have indicated that behavioural hearing thresholds are generally lower (i.e. more sensitive) than AEP/ ABR thresholds (e.g. Szymanski, et al. 1999; Yuen, et al. 2005). It is also important to consider that audiograms may be generated using different stimuli and therefore not all audiograms may be directly comparable. However, given the paucity of audiogram data for marine mammal species- considering the suitability of all available data is important.

Studies on the hearing sensitivity in marine mammals are usually carried out on captive animals and as a consequence, audiograms have not been measured for the majority of species. Furthermore, audiograms have been calculated for a limited number of individual animals of each species and consequently may not capture the variation in auditory ranges or most sensitive frequencies across the entire species. However, despite the lack of data on the hearing sensitivities of many marine mammals at the species level, it is possible to make some generalisations about hearing across higher taxonomic levels.

2.5.1 Compiling species audiograms

In compiling audiogram data for the species of interest, where possible only data collected for that species were used in the study. However, due to the paucity of audiogram data available it was necessary to sometimes use data from other homologous species to build complete audiograms given the nature of noise sources being considered here (e.g. low frequency continuous noise). Studies where the absolute values from hearing sensitivity experiments were presented were used. In many studies, a visual plot of the audiogram was shown, but no empirical data presented. However, many of these plots were reviewed in the generation of 'composite audiograms' to ensure important studies/findings were not being overlooked.

For each species, once the data had been assimilated, an assessment was made of how many studies and individual study animals a suitable audiogram could be compiled from. For each species a composite 'most sensitive animal' audiogram was constructed. This was a precautionary approach to try to avoid an underestimation of the potential impact zones.

2.5.1.1 Bottlenose dolphin

In general, small- to medium-sized odontocetes (e.g. dolphin species) have good hearing across a broad range of frequencies (4-100 kHz) and are most sensitive to sounds above 10 kHz (Richardson et al. 1995), but can hear sounds below this level ( Figure 2-3). Available studies of bottlenose dolphin hearing thresholds only went down to 8 kHz (Houser & Finneran, 2006; Popov, et al. 2007; Houser, et al. 2008) and so data from beluga whales (from White, et al., 1978) were used as a proxy for bottlenose dolphins below this frequency.

Figure 2-3 - Audiograms for bottlenose dolphin and the composite audiogram derived and used here. AEP indicates that auditory evoked potentials or auditory brainstem responses were used to calculate the audiograms.

2.5.1.2 Harbour porpoise

Harbour porpoises echolocate at high frequencies (125-150 kHz) and have excellent mid-high frequency hearing (Goodson & Sturtivant, 2002; Kastelein, et al., 2002). The composite audiogram for the species was constructed from behavioural and AEP studies (with animals more sensitive in the behavioural audiograms) and have good hearing down to ~4 kHz (Andersen, 1970; Kastelein, et al. 2002) ( Figure 2-4). Below this frequency the species hearing capability is predicted to be low.

Figure 2-4 - Audiograms for harbour porpoises and the composite audiogram derived and used in this study. B - behavioural audioram; AEP indicates that auditory evoked potentials or auditory brainstem responses were used to calculate the audiograms.

2.5.1.3 Minke whale / Baleen whales

Although there are no empirical data on minke whale (or any baleen whale species) hearing, there is theoretical evidence (from the frequency of their vocalisations) that they are more sensitive to low frequency sounds than odontocetes. Specifically it is reasonable to assume that baleen whales are sensitive to the same frequencies they vocalise at, i.e. primarily below 1 kHz, but sounds up to 8 kHz have been documented. Although baleen whales react behaviourally to low frequency calls from conspecifics, observations of these reactions do not provide accurate indications of hearing thresholds (Erbe 2002). Therefore baleen whale species are assumed to hear sounds at low and medium frequencies (20 Hz to >3 kHz), with the likely highest sensitivity between 100 and 200 Hz (Erbe, 2002).

Due to the paucity of audiogram data for baleen whales in general, it was necessary to generate predicted audiograms using the best information available for minke whales. Here, the "high sensitivity humpback whale" audiogram (from Erbe, 2002) was used between 0.1 - 1 kHz. Beyond these frequencies the audiogram was extended following the change in hearing sensitivity from a 'baleen whale' audiogram (predicted) presented in New Technology Inc. (2006). As a result we use a conservative audiogram for baleen whales in this study, ( Figure 2-5).

Figure 2-5 - Predicted audiogram data for baleen whales and the composite minke whale audiogram used in this study. Two audiograms were taken from Erbe, 2002 - one predicted 'high sensitivity' and one 'low sensitivity'. The more sensitive of the two was chosen as a more precautionary approach.

2.5.1.4 Seals (Grey and Harbour seals)

Seals do not echolocate but do utilise acoustic communication both in and out of water, and are considered to hear best at frequencies between 1-30 kHz (Richardson et al. 1995) ( Figure 2-6). Seals have markedly different hearing capabilities in air and underwater (Kastak and Schusterman 1998), and studies have shown that pinnipeds are sensitive to a broader range of sound frequencies in water than in air (Southall et al. 2007). In-air sound exposure is not considered within the scope of this project and so aerial hearing abilities will not be considered further.

In this study, audiogram data compiled for grey and harbour seals were used to generate a single composite audiogram for the species. Data from these species (Ridgeway & Joyce, 1975; Kastelein, et al. 2009; Götz & Janik, 2010), and other phocid seals found in the northeast Atlantic (e.g. harp seal - Terhune & Ronald, 1972), were also considered when compiling the seal composite audiogram ( Figure 2-6).

Figure 2-6 - Pinniped audiograms used to derive composite audiogram used in this study. B - behavioural audioram; AEP indicates that auditory evoked potentials or auditory brainstem responses were used to calculate the audiograms.

2.6 Modelling overview

Xi Engineering Consultants have developed a structural-acoustic interaction model using a finite element method that has successfully modelled air-borne noise produced by onshore wind turbines (Marmo & Caruthers 2011, Marmo 2011) and marine noise produced by tidal stream generators (Caruthers & Marmo, 2011). These models were developed using the commercially available modelling package COMSOL Multiphysics (Comsol) and have been validated using field evidence. Here three different foundation types have been modelled to determine the effect that different foundations have on the noise propagation from wind turbines into the marine environments.

1. Jacket with pin piles connection to the seabed

2. Monopile piled into the seabed with a transition piece

3. Gravity base structure sitting on the seabed

There is a complex range of variables that will affect the noise radiating from the foundations. These include:

• Vibration spectra produced by the wind turbine (frequency and amplitude)
• Wind speed
• Water depth
• Seabed sediment type and thickness
• Biofouling

In practise, the selection of foundation type and its design is based on a range of environmental and economic factors. Environment factors overlap with those that effect noise radiation, such as water depth and seabed type. It may not make sense to use a jacket in shallow water nor a gravity base in deep water. The purpose of the proposed work is to compare the effect of foundation type on noise, so each foundation will be modelled using the same input forces from the wind turbine and same environmental conditions with the exception of water depth. Both the jacket and gravity base can be modelled in the deeper 50 m of sea water, whereas the monopile is generally not used in depths exceeding 30 m and so will use this shallower level with all other variables remaining unchanged. Also, biofouling is assumed to not have occurred to simplify the model.

The process of modelling for this project is two-fold. Firstly a near-field model of a single turbine-foundation system is formed where the structural - acoustic interaction is quantified. The output of this is the soundscape radiating from single turbines, repeated for each of the three different foundations. The sound field consists of spatial variation of sound intensity and its frequency and characterises the foundations source term up to 30 m.

The near-field source term is then used as the input for a far-field acoustic model. Maintaining the radial and frequency dependency signal produced from the structural - acoustic interaction the source term is applied to multiple locations forming wind farm configurations, ( Figure 4-1):

1) Diamond

2) Square

This enables the far-field to be investigated using a beam trace model, including the cumulative effect from multiple sources. The 3-dimensional acoustic far-field soundscape produced is compared to background noise, accounting for the sea state under relevant environmental conditions, so that the range at which the turbine noise is masked by background noise can be determined. Audiograms for the chosen species, Table 2-3, are then imposed onto the modelled sound field and detection levels for the relevant frequencies are ascertained. The potential effect of noise output from off-shore turbines on marine species were examine in three zones of interest:

1. Audibility zone

2. Behaviour response zone

3. Auditory injury zone

The definitions of these zones and the parameters used to examine them are described below.

2.7 Determining zones of interest

2.7.1 Audibility zones

The range out to which the noise generated by operational wind turbines was audible to the species of interest was assessed. The sound field produced by each wind farm can be examined to determine where the sound pressure level ( SPL) is equal to or greater than the hearing threshold of each species for any given frequency. The audiograms described above were applied in this way to the modelled far-field sound field and the maximum range at which marine mammals could hear the wind farm determined as a function of frequency. It is assumed that if the background noise exceeds the SPL produced by the wind farm that the noise from the farm is masked and cannot be detected by marine species. Thus, the maximum range at which the wind farm is audible to marine species is less than or equal to the range shown, depending on the hearing sensitivity of the species in question.

2.7.2 Behavioural response zones

The observed variability in the behavioural response of marine mammals to sound exposure makes it difficult to identify an exposure threshold at which animals will respond. Instead behavioural response should be thought of as probabilistic and is best described with dose-response relationships between the onset of a behavioural response and the received sound level. Although the development of dose-response curves for marine mammal response to sound exposure is currently underway these curves are not yet available and as there are still discussions over the best metric for assessing behavioural impact zones, we present the range of metrics that are currently favoured by the scientific community:

• Audiogram + Sensation levels
• Weighted SPLs
  • M-weighting
  • Reverse audiogram weighting

Table 2-4 - Sensation levels and behavioural response sound pressure levels ( SPL) ( RMS) for each of the species/groups. The behavioural response SPLs correspond to the sound levels at which 10%, 50% and 90% (see section 2.7.2.2) of animals that experience the SPL are predicted to respond.

	Sensation level		Behavioural response SPLs ( RMS)			Auditory Injury ( SEL)
Species	Lower	Upper	10%	50%	90%	PTS
Seals	45 1	59 2	--	135 2	144 2	203 4
HP	49 3	--	90 4	120 4	140 4	215 4
BND	49	--	120 4	140 4	160 4	215 4
MW	49	--	120 4	140 4	160 4	215 4

¹ Kastelein et al. 2006; ² Götz & Janik 2010; ³ Kastelein et al. 2005; ⁴ adapted from Southall et al. 2007.

2.7.2.1 Audiogram and sensation level

Behavioural response zones were calculated using sensation level thresholds. The sensation level is a pressure level in dB by which a sound exceeds the hearing threshold. Equal sensation levels can be expected to roughly cause similar loudness perception. Although there are limited data on which sensation levels generate a behavioural response in marine mammals, there are a few studies that do provide empirical data for marine mammal response to sound exposure. Sensation levels for harbour porpoise response to underwater data transmission sounds were 49 dB re 1 µPa at low frequencies (Kastelein, et al. 2005), sensation levels for grey seal response to '500/530 square' noise stimuli, a 'rough' sound perceived to be unpleasant by humans, were 59 dB re 1 µPa (Götz and Janik, 2010), and sensation levels for harbour seal response to underwater data transmission sounds were 45 dB re 1 µPa (Kastelein et al. 2006).

For the purpose of this assessment the sensation levels described above were used to predict behavioural response ranges for pinnipeds and harbour porpoise. The sensation levels for grey seals and harbour seals were combined and treated as an upper and lower sensation level ( Table 2-4). Due to the lack of empirical data for bottlenose dolphins and minke whales, the harbour porpoise sensation level was used to predict behavioural response ranges for all cetaceans as a conservative approach.

For each species, the predicted SPLs were compared to the composite audiograms to calculate the sensation level. To determine the sensation level that may result in a behavioural response in seals the sensation parameters were added to the sound level integrated across the two neighbouring one-third octave bands either side of the band of interest. In the case of cetaceans the sensation level was found by adding the sensation parameters to the SPL integrated across the four neighbouring bands on the dominant side of the band of interest.

If the sensation level was equal or higher to the sensation threshold presented in Table 2-4 then a behavioural response is predicted to occur.

2.7.2.2 Weighted sound pressure level (m-weighting and reversed audiogram)

The most complete review of behavioural responses by marine mammals to date is found in Southall et al (2007). Southall and colleagues reviewed available studies on behavioural responses to sound exposure in cetaceans and pinnipeds and proposed a severity scaling on which the level of the response could be measured. They did not, however, present explicit step-function thresholds for behavioural response. Animals may exhibit behavioural responses of varying magnitudes to noise exposure depending on species, sound type, exposure level and other contextual factors. Southall et al. (2007) provides a range of sound levels at which animals have shown to exhibit a behavioural response. Given that behavioural response should be thought of as probabilistic and is best described with a dose-response relationship that describes the proportion of animals that may expected to respond to a given sound level, for the purpose of this assessment we apply a probabilistic metric at which 10%, 50% and 90% of individuals exposed to these range of sound levels are predicted to show a behavioural response. We used Southall et al. (2007) and more recent literature to ascertain what sound levels had resulted in a behavioural response for each of the marine mammal hearing groups.

1. Pinnipeds appeared to exhibit only mild avoidance responses to non-pulse (continuous) noise at received levels between 90 and 140 dB re 1µPa. However, Götz & Janik (2010) showed sustained avoidance responses in wild grey seals at received levels of 135-144 dB re 1µPa; for the purposes of this assessment, the range at which these levels (135-144 dB re 1µPa) were exceeded were used as a threshold where 50% and 90% of animals respectively were predicted to show a strong behavioural response.

2. Harbour porpoise appear to be relatively sensitive to noise levels as low as 90-120 dB re 1µPa (Southall et al. 2007) although there also appears to be considerable variation between individuals. For the purpose of this assessment we have chosen 90 dB and 120 dB as a step-function threshold where 10% and 50% of animals respectively are predicted to show a behavioural response. However, data also suggests that harbour porpoise and other high-frequency cetaceans are likely to show a behavioural response when exposed to noise at much lower received levels than other species. Therefore, in contrast to other species, we use a lower level of 140 dB as a threshold where 90% of harbour porpoise are predicted to behaviourally respond.

3. Bottlenose dolphins have shown moderate level changes in behaviour to non-pulse noise at received levels of 120-180 dB re 1µPa (Southall et al. 2007). A more recent study reported moderate level responses to non-pulse noise by bottlenose dolphins at received levels of 140 dB re 1µPa. For the purpose of this assessment we have chosen 120 dB, 140 dB and 160 dB as step-function thresholds at which 10%, 50% and 90% of animals respectively are predicted to show a behavioural response.

4. Minke whale response to noise remains largely unknown as there are very few empirical studies on minke whale behavioural response to noise. Southall et al. (2007) suggest moderate level changes in behaviour by another baleen whale species (humpback whale) to non-pulse noise at received levels between 120 -150 dB re 1µPa. For the purpose of this assessment we have chosen 120 dB, 140 dB and 160 dB as step-function thresholds at which 10%, 50% and 90% of animals respectively are predicted to show a behavioural response.

There are currently two types of weighting functions that are proposed for marine mammals. The first is the M-weighting that is similar to C weighting for humans (Southall et al. 2007) and the second is the species-specific audiogram-weighting that is based on the absolute hearing sensitivity of the species in question (Verboom and Kastelein, 2005; Nedwell et al. 2006; SOI, 2011). These weighting functions are defined as:

2.7.2.2.1 M-weighting

Southall et al. (2007) developed a series of weighting functions (M-weightings) that could be used to take account of the hearing sensitivities of four different marine mammal groups (low-frequency cetaceans, mid-frequency cetaceans, high-frequency cetaceans and pinnipeds). The premise being that the sound levels are frequency weighted to account for the sensitivity of the animal to the frequency of a given sound. The M-weighting functions essentially de-emphasise sounds at frequencies to which a given marine mammal hearing group is not particularly sensitive.

2.7.2.2.2 Reverse audiogram weighting

The composite audiograms discussed in section 2.5.1 were inverted and normalised at the most sensitive frequency to obtain a species-specific weighting function (De Jong & Ainslie, 2008; Li et al. 2011; Miller et al. 2011). There are few studies that have assessed marine mammal hearing at very low frequencies and the audiogram measurements only go as low as 100 Hz for seals (Gotz & Janik 2010) and 250 Hz for harbour porpoise (Kastelein et al. 2002). Therefore the existing audiograms needed to be extrapolated in order to look at the audibility of the turbine noise at these very low frequencies. Mammalian audiograms appear to share a common characteristic in that there is a gradual increase in thresholds for low frequencies with a slope of approximately 35 dB per decade (Tougaard et al. 2009). Using the approach described in Tougaard et al. (2009) the audiograms were extrapolated by a straight line with a slope of 35 dB per decade for frequencies below 100 Hz for seals, 250 Hz for harbour porpoise, 40 Hz for bottlenose dolphins and 4 Hz for minke whales.

Both sound weightings described above were applied to the sound pressure levels modelled at varying distances from the sound source. The weighted SPL was derived by subtracting the relevant weighting filter from the centre of each one-third octave band then taking the power sum over the broadband.