Substrate vibrations and their potential effects upon fishes and invertebrates

Investigators, regulators, resource managers, and others who have interest in the potential effects of man-made (anthropogenic) sounds upon aquatic animals are familiar with the concept of sound pressure and, to a growing degree, the particle motion that is generated in the water column. However, far fewer are aware that some anthropogenic sources, such as pile driving, dredging, and seismic exploration, may also generate vibrations within the substrate at the bottom of the water column. Thus, the purpose of this paper is to introduce a number of the concepts and ideas associated with substrate vibration and then to very briefly discuss their relevance to a number of marine species.

We start by pointing out that vibrations within the substrate take a number of forms (Table I). Moreover, one form of substrate vibration is the seismic surface wave (or ground roll) (Fig. 1) that not only propagates along the surface of the substrate, but also produces particle motion that enters the water column (Popper and Hastings, 2009; Hovem, 2014; Roberts and Howard, 2021).

As a consequence, the acoustic environment of animals living close to, on, and/or within the substrate can be highly complex and include vibration signals that are within and also emanate from the substrate, as well as sound signals that are generated in the water column. All of these signals are potentially detectable by benthic fishes and invertebrates living close to or within the substrate. Moreover, many of these anthropogenic signals often substantially overlap within the frequency range of biologically relevant signals used by these animals in many biologically critical ways (Roberts and Elliott, 2017; Popper and Hawkins, 2018).

It is also important to understand that while we know that all fishes and many invertebrates are sensitive to the particle motion components of sound and that only some fish species can detect sound pressure [reviewed in Popper and Hawkins (2018)], very little is known about the sensitivity of aquatic animals to the energy that is generated within and close to the substrate (e.g., Roberts and Breithaupt, 2016; Roberts et al., 2016; Roberts and Elliott, 2017). Far more is known about sensitivity of terrestrial invertebrates and vertebrates to substrate vibration (e.g., Salmon et al., 1977; Narins et al., 1997; Hill, 2008; Roberts and Howard, 2021). However, there have been relatively few measurements of the levels of substrate vibration within aquatic environments, especially as compared to the large number of sound pressure measurements that are generally used to monitor underwater sounds.

Before going on, it is important to define the term “vibration” as used in this paper. Vibration is generally defined to mean the motion of any object that produces sound, such as a loudspeaker or a driven pile. Once the energy produced by the vibrating object leaves the source and travels through the adjacent fluid medium (e.g., air, water) as an acoustic wave, the energy is considered to be sound. However, in this paper, we also use the term substrate vibration for the motion induced by a wave traveling through a solid medium, such as a sedimentary seabed. Vibration displacement upon the substrate has been described by Hao and Xu (2009). There are also interface waves, often termed ground roll, that travel along the surface of the substrate and that are often characterized by low frequencies and high amplitudes (Askari and Siahkoohi, 2008).

There are a number of different types of substrate vibrations, as shown in Fig. 2 and summarized in Table I. Each is discussed in a separate section below.

Several types of wave propagation are possible within and on the substrate. Solid compression waves are established within the sediments and are similar to the sound pressure waves in fluids but travel faster in the stiffer and denser solid media. Their particle motions involve to and fro movements along the direction of travel.

Solid shear waves are also generated and involve particle motion in other directions and travel slower. The particle motion in a plane compression wave is in the direction of propagation of the wave, whereas the particle motion in a shear wave is perpendicular to the direction of propagation.

A third variety of solid wave is the seismic interface wave (also referred to as ground roll). These waves carry energy that travels through the substrate layers and can be the result of any sources that generate low frequency acoustic energy. The particles, both solid and fluid in the regions close to the substrate, follow roughly circular pathways as shown in the hodographs in this paper (Fig. 2), unlike the longitudinal (“to and fro”) particle velocities associated with the sound pressure waves that travel through the water or other fluid media.

Since particle motion takes place in both directions, it can best be viewed by the hodograph plots showing the vector properties of the moving particles within a medium. The arrows in the hodographs shown in Figs. 2 and 3 indicate the direction followed by each particle around its locus. For brief impulses, the particle starts at rest with zero displacement. The particle then moves from this point of origin around the locus, forming a shape that depends on the medium. The plots use the data from finite element (FE) modeling analysis. The hodograph plots also show the changes with time of directional vector properties of the moving particles within a medium. In FE analysis, the particles are called nodes, which are numbered (see Fig. 3).

Seismic interface waves are known to be strongly dispersive, meaning that different components of applied energy travel at different speeds. This effect will cause different parts of a signal waveform to separate. If the signal is a short impulse, containing a wide frequency range, the energy will then be dispersed in both space and time, varying at the locations where the sensor is placed. Jensen et al. (2011) discuss the different modes that control the propagation of seismic interface waves.

Seismic interface waves involve vibrations near the top of the solid substrate and also evanescent waves within the adjacent water. Where energy mainly propagates along the substrate, it can also result in the generation of evanescent sound within the sediment itself. Such waves can be shown using mathematical representations that are also used with other sources such as electromagnetic waves. An evanescent field of sound pressure can also be formed in the adjacent liquid medium, which decays rapidly with distance from the interface. As seen in Fig. 3, the solid particle hodographs have an elliptical outline with bigger vertical displacements than horizontal, whereas those in the adjacent water have circular outlines, as the fluid is equally free to move horizontally as well as vertically.

While usually traveling much slower than the waves within the substrate itself, the seismic interface waves often dominate the seismic energy transmission (Shearer, 2019) and involve a circulating motion of the particles. The descriptive term “ground roll” indicates this more complex motion, as experienced in earthquakes. The mathematical properties of ground roll were first described by Lord Rayleigh (1887). A similar model with overlying water, developed by Scholte (1949), propagated all frequencies at the same speed, providing a reproduction at a receiver of the vibration generated by a source. In contrast, the FE analyses used to generate Fig. 3 all use a more realistic model with graded sedimentary properties that result in the waveform distortion and filtration typical of seismic interface waves. A linear increase in the material shear wave speed with depth follows a step change at the interface and generates the simple wavelet, discussed later.

Our FE model can be contrasted with that proposed by Godin and Chapman (2001), which used a power law for the shear speed profile. Unlike our model and those of Rayleigh and Scholte, there is no initial stiffness applied to the solid by Godin and Chapman, but only a gradual increase without a tangible interface. Studies by Hazelwood et al. (2018) showed that this failed to match the measured data.

Any increase in the source power will usually result in larger vibrations. However, the waveforms produced are highly influenced by the nature of the propagation medium (Hazelwood et al., 2018). The key sediment properties, including the density, stiffness, and rigidity, are often graded with depth, affecting the shear wave speed and the compressional wave speed. FE models of simple flat sedimentary areas can provide indications of the waveforms but do not apply to complex rocky areas. Complex mechanical motion can be computed by using many small “elements” or “particles,” each with single values for their material properties, such as density and stiffness. Note that most acoustic models, including those described by Heaney et al. (2020), do not consider shear waves at all. They also ignore the sensitivity of some animals to particle motion. Instead, they treat the solid as a fluid, thus simplifying their calculations but failing to generate some of the effects discussed here.

Seismic interface waves generated by both piling and dredging were examined by Hazelwood et al. (2018). FE modeling confirmed that seismic interface waves do not radiate energy into the water above or the sediment below, but create evanescent sound waves, largely confined to the bottom 1 m of water. They travel slower than the speed of sound but may travel kilometers. For example, Jensen et al. (2011) reported measurements at over 2 km from an explosive source. More studies of the propagation of vibration energy from various sources are required to estimate the magnitude of particle motion at different distances and the environmental impact of these sources.

There are a number of human activities that can result in vibration of underwater substrates, together with many natural sources of substrate vibration. Human activities that can generate vibration of the substrate underwater include (among others) pile drivers, explosives, offshore wind-driven electric turbines that are fixed to the seabed rather than floating at the surface, dredging and trawling activities, aircraft generated sonic booms, air guns used for seismic surveys, and even subsurface transportation tunnels and onshore vehicles on roads close to the water's edge or on bridges with in-water piling (e.g., Martin and Popper, 2016; Reeder et al., 2020). Natural sources of substrate vibration include volcanos, earthquakes, and breaking waves as well as animal movements/interactions and objects falling/rolling onto/on the seabed. Some of these are discussed in Secs. III A–III F.

Pile drivers are used in connection with offshore construction. The driving of the pile by a hammer generates sound in the air and the water and also produces vibration in the substrate (Fig. 2). The noise generated within the water by impact pile driving was studied by Reinhall and Dahl (2011), who modeled the sound generation and propagation. The model results were compared with measurements made at a marine construction site in Puget Sound (Washington, USA) using a vertical line hydrophone array. Results showed that the main underwater noise from impact driving was generated by a Mach wave, which is a sound pressure wave associated with the radial expansion of the pile that propagates down the pile.

In addition to the conical sound waves generated by pile drivers in the water, compressional, shear, and interface waves are also generated within the substrate and propagate outward from the pile. The seismic interface waves are generated by the interaction between the solid compression and shear waves and may propagate over long distances and potentially be detected by, and even affect, marine animals living close to or within the substrate sediment (e.g., Popper and Hawkins, 2018).

Strong vibrations radiate from the sides of the pile as it is driven into the sediment (Athanasopoulos and Pelekis, 2000). The wave propagation in a pile, water, and substrate was investigated by Bruns et al. (2016) using hydrophones and geophones. Low frequency particle motion was detected in both the water and the substrate, at frequencies between 1 and 40 Hz, and the substrate vibration waves propagated slower than the hydro-acoustic waves but were monitored up to 70 m away.

Dahl and Dall'Osto (2017) measured the sound pressures developed by a pile driver at different water depths using a line array of hydrophones. They found that the sound pressure level on the seabed was about twice that measured on their lowest hydrophone in the water. They also observed seismic interface waves that propagated along the seabed, resulting in energy transmission at frequencies well below the acoustic cutoff frequency for a water depth of 12.5 m, which was about 65 Hz. The highest spectral frequency in the interface wave was around 9 Hz. They concluded that the contribution by the interface wave to the broadband sound exposure level (SEL) in the water was negligible.

Measurements were also made by Hazelwood and Macey (2016) close to a test pile as part of a trial mounted by TNO (Den Haag, Netherlands) in the estuarine water at Kinderdijk. An instrumented sledge, containing a geophone set (Figs. 4 and 5), was placed on the sediment at a distance of 68 m from the hammer source.

The contrast between the sharp high frequency sounds produced by the blow on the pile and the narrow frequency band of the received vibrations is a consequence of the filtration of energy involved with the production of the seismic interface waves. This also creates the phase difference between the vertical response and the horizontal radial response, which can be clearly seen in the figure. The 70 kJ blows of this test pile created particle velocities of 2.5 mm/s at 68 m range. Current piling activity is understood to generate much higher energies, by a factor of 10 or more.

The signals detected by the geophone buoy, when used at in a reservoir near Wraysbury (UK), included the water particle motion, but also the sound pressure. The horizontal particle velocity waveform shape matched that of the sound pressure in the water. This showed a peak-to-peak variation of 34 μm/s horizontal particle velocity, corresponding to 22 Pa sound pressure. This confirmed that the increase in particle motion in the water created by substrate vibration can be predicted by computer modeling, which will be discussed later.

Underwater explosions produce sounds that may travel considerable distances. Because water is of high density and stiffness, the explosions may generate strong impulsive underwater sounds, some of which may travel into the seabed. Some explosive tests may also be conducted within sediments beneath the water. The effects of an underwater explosion in terms of substrate vibration likely vary with distance from the explosion, the strength of the explosion, the depth of the explosion, and the depth of the water.

Explosions are often used to deal with an unexploded ordnance lying on the desired path for pipelines and cables. A study of the waves created by seabed explosions was reported by Jensen et al. (2011). This trial in 1983 deployed 180 g seabed charges in 20 m water depth, and recordings of both vertical and horizontal particle velocities were made over 2 km away. It was evident that a seismic wave was propagated along the interface. It was shown that the propagation speed and attenuation were closely related to the shear properties of the sediment.

Seismic surveys involve the deployment of air guns, which produce impulsive sounds close to the water surface but are designed to penetrate hundreds of meters into the substrate to explore for geological features such as oil and gas deposits (e.g., Gisiner, 2016). These air-gun shots at the surface of the water column generate seismic interface waves at the surface of the seabed. Kugler et al. (2007) monitored the shear waves generated at the seabed, using both ocean-bottom seismometers and geophones buried within the seabed. Sensitivity analysis revealed the maximum penetration depth of the shear waves to be about 40 m below the seabed, although it was concluded that the densities of the substrate sediments may be an important parameter in determining the depth of penetration in addition to the strength of the air-gun sound.

Offshore wind farms generate underwater noise during construction (see Sec. III A), operation, and decommissioning (e.g., Amaral et al., 2020). Vibration, together with underwater sound, is produced by the rotation of the wind turbine blades, the gear box, and the electricity generator, with the vibrations moving downward and being transmitted into the water and seabed by the tower wall and seabed support structure. The substrate vibration can be continuous when the wind turbine is operating. Although the way the wind farm affects the substrate is understood, few measurements have been made of the levels of substrate vibration and particle motion that are generated by operational wind turbines.

Early measurements of the noise generated in water by operating wind turbines were reviewed by Madsen et al. (2006), who concluded that the underwater sound is limited to low frequencies (below 1 kHz). Although the sounds are of low intensity, the recent increase in the number of turbines in coastal waters (Offshore, 2021) may result in quite high levels, especially where there are turbines present in large numbers. It was pointed out by Bailey et al. (2014) that over 2000 wind turbines had been installed then in 69 offshore wind farms across Europe, with the greatest installed capacity in UK waters.

Sigray and Andersson (2011) monitored particle motion close to a wind turbine using a neutrally buoyant sphere. The measurements of particle motion were made at different distances from the turbine as well as at varying wind speeds. The relationship between mechanical vibrations and the particle motion was established using simultaneous records from a three-axis accelerometer as well as the particle motion detector in the water. The results showed that a wind turbine generated broadband vibrations, with amplitudes increasing at higher wind speeds. The blade rotation gave rise to low frequency vibrations over the frequency range 1–6 Hz, and the particle motion sensor located in the sea showed the same characteristics as the mechanical vibrations. Close to the foundation, the particle motion was comparable with the levels observed in studies on the behavioral reactions of fishes. However, at a distance of 10 m, the amplitudes declined to levels comparable with hearing thresholds, suggesting that the area affected by the particle motion may be restricted to an area close to the wind turbine. Most recently, Tougaard et al. (2020) have pointed out that the noise is radiated into the water through the turbine foundation, and there will be differences in the noise from different turbines, depending on the types of foundations used.

Bottom trawls, seines, and shellfish dredges all create disturbance of the seabed and are likely to generate substrate vibration as well as sound in the water column (Daly and White, 2021). It was suggested by Bagenal (1958) that the efficiency of particular trawls in catching fishes may be attributed to low frequency vibration of the otter boards and net wings. Berghahn et al. (1995) monitored the responses of shrimps to vibrations from the rollers of commercial shrimp trawls. The vibration signals from the rollers were directly recorded with an accelerometer. Other sources that that are “connected” to the substrate, such as hydropower dams, can also generate substrate vibration. Substrate vibration can also be generated by onshore vehicles passing close to rivers, lakes, and harbors.

In addition to human contributions to substrate vibration, there are some natural sources of vibration, including earthquakes, thunder, and breaking waves on the shoreline that generate vibrations in the intertidal and subtidal areas. The falling of objects through the water onto the substrate generates vibration, for example, when a large animal carcass hits and impacts the seafloor. It was shown by Klages et al. (2002) that the intensity and transmission characteristics of the vibration depend upon the properties of the sediment.

As already described, underwater substrates can propagate some seismic interface waves well, with associated particle motion in both the water and the sediment. The particle motion is generated close to the sediment interface, declining exponentially with distance from the interface. The evanescent sound pressures can be measured with a vertical hydrophone array, set close to the seabed, and will show this reduction with height. The particle motion can be monitored using sensors that respond to particle acceleration or particle velocity.

The evanescent sound pressures are smaller in relation to the particle motion than those found normally for plane waves. The phase relationships are also important. For plane waves, the sound pressures are in phase with the particle velocity, indicating a flow of energy in the direction of propagation. For the rotary action of the water particles near the seabed, there is a phase difference between the vertical motion and the horizontal motion. The sound pressures are out of phase with the vertical particle velocity component but in phase with the horizontal component, indicating a propagation of energy along the interface.

FE analysis models can be used to examine both the water particle motion and the sound pressures at different distances from the source. Computer simulations provide some detailed features of these waves relevant to aquatic life. However, of necessity, they involve simple circumstances only relevant to a flat uniform sedimentary substrate rather than one with rocky outcrops. One of the main findings is that the seismic interface waves show strong particle motion, as compared with the particle motion accompanying sound pressure waves.

While these properties will be general to this class of wave, the waveforms described in detail, such as in Fig. 6, show an optimal propagation due to a lack of dispersion (Hazelwood et al., 2018). This minimal dispersion keeps the energy concentrated into a short period and maximizes the intensity. Dispersion increases the wavelet length in both time and space, with complex hodograph plots following multiple successive circuits. When dispersion is rapid, the energy quickly disappears into the background noise.

The ratio of sound pressure to particle velocity can be compared with the ratio of sound pressure to particle velocity in a free field where all reverberation is ignored. This factor is the specific acoustic impedance of the medium (Kinsler et al., 1999), which for water is about 1500 kPa/(m/s), dependent on the temperature and salinity. The ratios from the FE analysis shown in Fig. 6 are smaller by a ratio well over 12. While more research is required to determine what factors influence this change, it is clear that water particle velocity sensors are likely to be more suitable for monitoring the stimulus levels, especially when considering the effects upon benthic organisms, given the increased particle velocity compared with the sound pressure.

The water particle velocity vector magnitude, here described as the particle speed, for this model, with its linear increase in material shear wave speed with depth, follows a bell-shaped curve that reduces as the energy is spread over a greater circumference (cylindrical spreading). The upward velocity component is at a peak when the outward velocity component (dashed) is zero. The two particle motions are thus out of phase.

The sound pressure in Pa (dotted), closely follows the form of the outward horizontal particle velocity in mm/s (dashed). Their ratio is 121.3 kPa/(m/s), which is less than the corresponding ratio for plane pressure waves with no reverberation by a factor of around 12. The particle velocity is thus much higher in comparison to the sound pressure.

Figure 3 shows the displacement as modeled for two particles, represented by nodes in the FE analysis. Node 600 is a solid particle at the interface, and the locus of its displacement gives a horizontal displacement that peaks at just over 0.1 mm. In contrast, node 8266 is a water particle at the same point, which shows over 0.3 mm maximum horizontal displacement. Note that the vertical motions are identical, as is enforced by continuity. The model does not include any viscosity for the water, but estimates of the Reynolds number suggest that any boundary layer will be thin and that differential motion will exist in practice.

Many species that are most likely to detect (and potentially be affected adversely by) substrate vibration live in, on, or just above the substrate. At the same time, since many other species might come close to the substrate at times, such as to feed, they too may detect sounds emanating from the substrate. The concern then is that anthropogenic substrate vibrations have the potential to interfere with the detection of biologically important vibrations, just as anthropogenic sounds in the water column potentially interfere with detection of biologically important sounds (e.g., Popper and Hawkins, 2019; Popper et al., 2020).

All bony and cartilaginous fishes are able to detect low frequency particle motion [e.g., Cahn et al., 1969; Hawkins and MacLennan, 1976; Fay, 1984; reviewed in Popper and Hawkins (2018)] using the otolith organs within their ears [reviewed by Hawkins and Popper (2018); Schulz-Mirbach et al., 2019; Schulz-Mirbach et al., 2020]. Thus, it is very likely that fishes that come close to or live on (or in) the substrate detect the particle motion associated with substrate motion, and it is possible they detect other motions as well, although this has not yet been tested.

As an example, measurements made by Chapman and Sand (1974) showed a sensitivity of a flatfish, the plaice (Pleuroectes platesa) to water particle velocities of as little as 0.3 μm/s at around 20 Hz. This can be compared with the 2500 μm/s particle velocity recorded, at Kinderdijk, at 68 m from the test pile.

There is much less information available on the hearing of invertebrates than fishes, but it has been pointed out that they have a number of organs that are probably sensitive to particle motion [reviewed by Popper et al. (2001)]. Considerably more data are needed on the wide array of invertebrate species near, on, or within the substrate, but there is evidence that at least some species, including mussel and hermit crab, are sensitive to physical vibration at frequencies of 5–410 Hz (Roberts et al., 2015; Roberts and Elliott, 2017; Roberts and Laidre, 2019). Sensitivity to vibration of water, sand, and the buried animal itself was shown for the North Sea shrimp (Crangon crangon) and was maximal at 170 Hz (Heinisch and Wiese, 1987). In a series of field-based experiments, Day et al. (2017) investigated the effects upon scallops of exposure to seismic surveys. The ground roll acceleration was measured using geophones placed on the seabed. It was concluded that the impacts resulted from high seabed ground accelerations driven by the air-gun sounds. Thus, while mainly speculative at this point, it is likely that many invertebrates are able to detect substrate vibrations.

We have only the most limited knowledge of the way in which fishes and invertebrates detect and respond to substrate transmission even though it is well known that they generally detect particle motion. The sound-detection organs vary widely among fishes and invertebrate species, and it is likely that detection capabilities and sensitivities may differ substantially between species. However, there is a dearth of data that address the issue of substrate vibration detection, and we need to have more studies that look at this in a very specific way to understand which animals detect substrate vibration, how they do it, their sensitivity, and the behavioral implications of such vibration signals.

There is little information available on the ambient levels of particle motion close to the seabed and the substrates of lakes and rivers or on the levels and characteristics of the particle motion generated by impulsive or continuous anthropogenic sources, which may have effects upon fishes and invertebrates. Measurements of the sound level around the sources of substrate vibration need to be done in terms of particle motion as well as sound pressure to provide more information on the potential effects on marine species that live near the substrate or in shallow water, as particle motion cannot be calculated from sound pressure in the near field or close to the substrate. There are numerous variables regarding measurement of the source of the motion (e.g., substrate type, layering of substrate, type of contact of the source with the substrate, duration of the operation, effects of multiple overlapping sources) that complicate the detection and management of sources. However, by gathering data on the levels of particle motion and by undertaking experiments to ascertain effects upon relevant species, it is possible that the likely impacts upon populations, communities, or ecosystems might be assessed.

Thanks are due to TNO (Den Haag, Netherlands) for the opportunity to make measurements at Kinderdijk and the National Physical Laboratory (Teddington, UK) for the opportunity to make measurements at Wraysbury.