For propeller noise one can make a distinction between the noise related to the non-cavitating flow over the propeller and the noise by cavitation (Blake, 2017). If merchant vessels are not cavitating, the noise that can be identified are from the blade passage frequencies and, if present, due to flow-excited propeller blade vibrations (‘propeller singing’). These tones may occur for wide range of frequencies depending on propeller size and shaft rotation rate: Fischer (2008) shows examples in which the frequency of singing varied between 180 and 1800 Hz.

Carlton (2018) describes cavitation showing that it is an important source of shipping noise. In the presence of nuclei such as small gas bubbles, cavitation occurs when the cavity’s volume is immediately transferred downstream with the flow after formation (Figure 2). This typically occurs when the minimum pressure on the blade occurs relatively far from the leading edge of the blade. As bubble cavitation can be very erosive, most propellers are designed such that bubble cavitation is not present. When at the leading edge of the blade the pressure drops below the vapor pressure, a small region of flow separation occurs upstream of the cavity allowing the cavity to grow into a sheet cavity rather than individual bubbles. This sheet cavity can break-up into a cloud of bubbles and vortex cavities or it can merge with the tip-vortex cavitation. The minimum pressure may also occur in the flow due to the centrifugal force exerted on the flow within a vortex (Bosschers, 2018). Cavitation then starts in the center of the vortex and it is then referred to as vortex cavitation. The loading on a propeller blade varies when the blade rotates through the ship’s wake and this typically leads to the inception, growth and collapse of a cavity on the blade within each revolution. Even though any change in cavity volume leads to noise emission, it is especially the volume acceleration during the collapse phase that contributes to the underwater radiated noise (Ross, 1979; Franc and Michel, 2006). The cavity collapse is a very efficient noise source as it is a monopole in contrary to other (non-cavitating) flow noise sources which are either dipoles or quadrupoles.

Water quality can affect nuclei and therefore the onset of cavitation and the resulting cavitation noise (Brandner et al., 2022). Nuclei typically consist of very small free gas bubbles generated in the ocean by breaking waves and their sizes and numbers decrease with depth (Atlar et al., 2002). The effect of nuclei variations in the ocean on cavitation inception is considered small (Gowing and Shen, 2001) but there is very limited data available on this topic. Beyond inception, the growth and collapse of cavitation is also affected by bubbles identified as non-condensable gas. By diffusion and coalescence, the cavity volume and the growth rate increase with air content (Brennen, 1969; Briançon-Marjollet and Merle, 1996; Nanda et al., 2022). This happens especially for cavitation in separated flow or in vortices, which can increase the radiated sound due to larger cavity volume. However, in the collapse phase, the compressibility of the non-condensable gas dampens the collapse of the cavity thereby reducing the radiated sound (Moss et al., 2000; Trummler et al., 2021). The collapse and rebounds of the large cavity structure are considered responsible for sound at low frequencies ranging from blade passage frequency up to a few hundred Hertz, with the upper frequency depending on the amount of cavitation (Bosschers, 2018). At higher frequencies, it is the cloud of resonating bubbles generated by the collapse of the larger cavity structure that is considered responsible for the radiated sound. The presence of bubbles will also attenuate sound. In the top 10 m of the water column, sound is attenuated by bubbles generated by breaking waves that are connected to wind speed (Thorpe and Humphries, 1980).

Other variables such as water temperature, salinity and CO₂ may affect the amount and size of cavitation bubbles. Water temperature may affect cavitation because an increase in water temperature leads to an increase in vapor pressure. At 15°C, a 1°C temperature rise increases vapor pressure by 7%, but the resulting change in inception speed and emitted noise is negligible (<0.1%, ITTC, 2011). Temperature also affects the bubble clouds generated by breaking waves via air entrainment processes. In particular, a minimum critical temperature is needed for air entrainment (10°C for a water jet) and the bubble penetration depth increases with increasing temperature up to 19°C (Hwang et al., 1991). The effect of salinity on cavitation and the resulting acoustic emission has been analyzed by Ceccio et al. (1997) for an axisymmetric head form (a modified ellipsoidal shape) in a small water tunnel. They found that in saline water event rate and size of bubble cavitation were both reduced compared to those in fresh water, which was explained by a suppression of nuclei distribution in salt water. Bubbles of similar size showed comparable noise levels, independent of water salinity (Ceccio et al., 1997). However, smaller bubbles produced larger noise levels than larger bubbles, and therefore the noise of cavitation on the head form was about 10 dB larger for smaller bubbles. In the future other parameters affecting water quality such as surfactants (e.g. plankton) and the seawater CO₂ content, might need to be explored to better understand and estimate impact on cavitation. These parameters will also change due to climate change and are therefore relevant to predict future noise levels. Li et al., (2021) found that in an industrial setup varying the solution pH from 4.4 to 9.5, the cavitation inception number increases in a more acidic solution showing that hydrodynamic cavitation is easier at low pH. The effect of pH needs to be verified with a specific experiment at the relevant surface ocean pH interval (8.00 to 8.25, Jiang et al., 2019).

The term “shallow water” implies an acoustic environment such as a continental shelf shallower than 200 m. Despite shallow waters covering only 8% of the total sea areas, they host a large part of the total anthropogenic activities, including the associated noise generated. The interactions with the sea surface and seafloor make the PL in shallow waters larger and more complicated than in deep waters (Figure 7, Urick, 1979; Jensen, 1981). In general, shallow waters act as steep high-pass filters (Forrest et al., 1993), where low-frequency sound propagates poorly or not at all. For that reason, despite the large presence of broadband sources, such as shipping noise, the medium- to high-frequency contribution is larger than in the open ocean (Noise, 2003). In these environments, the major contributors to PL are cylindrical spreading, bottom attenuation and scattering due to roughness at the surface and bottom. Therefore the sound does not propagate over large distances because it loses energy after interacting with the sea surface and seafloor. The short propagation distance for low-frequency noise in shallow waters makes the contribution of sound absorption in seawater negligible but may still be important in the sediments.

The propagation of sound in shallow waters varies with season. During winter, shallow seas are generally well-mixed by wind and waves, leading to homogenous temperatures and salinities over the depth of the water column. As a result, sound speed can be considered uniform over depth and the propagation channel covers the entire water column from the sea surface to the seafloor. The constant sound speed also implies that sound travels straight until it interacts and it is reflected by the sea surface and seafloor. The seafloor does not reflect all sound but depending on the angle of incidence absorbs a portion of it (Urick, 1979). In this environment, high sound frequencies are largely attenuated by absorption and scattering by the seafloor (Kuperman and Ingenito, 1977). Since sound scattering increases with seafloor roughness, estimating PL is particularly challenging in shallow seas. An accurate estimation of the seafloor contribution to the PL therefore requires a specific survey to derive the seafloor sound speed, density and depth.

The consequent PL is connected to the frequencies of the sound source. At low frequencies, PL is largely controlled by attenuation at the seafloor where sound can leave the water column with the excitation of shear waves. However, low-frequency sound can only propagate if the effective water depth is larger than half the sound’s wavelength. The attenuation depends on the sediment properties (Kuperman and Ingenito, 1977; Urick, 1979) and the propagation of sound into the seafloor depends on critical angle ψ that is calculated using Equation 4 from the sound speed difference between the two layers:

where c_w and c_s are the sound speed in seawater and in the seafloor, respectively. In general, c_s is larger than c_w and it varies with the seafloor composition (for example, ψ for coarse silt is around 22° and for medium sand 30°). When sound travels at lower angles than ψ, sound is almost perfectly reflected. Instead, when sound travels at grazing angles greater than ψ, some of the sound is absorbed by the seafloor, reducing the reflected sound. In practice, to simplify the PL calculations the seafloor is generally considered a fluid with properties not too different from seawater. Therefore, after the interaction with the seafloor the sound field is not subject to drastic changes (Jensen, 1981). The other propagation limit is the sea surface which is considered a pressure release (Jensen, 1981) contributing to large PLs for receiver positions close to the sea surface, especially at low frequencies at<800 Hz the PL at 1 m depth is 25 dB higher than at mid-depth.

During the warmer season, water columns in temperate regions can be stratified with a warmer surface layer separated from a deeper cool layer by the thermocline. In this environment, noise generated at certain frequencies can propagate for many kilometers with the main PL given by the frequent interaction with the sea surface and seafloor (Jensen, 1981). The sound speed gradient in the surface layer changes the angles of the steeper rays into less steep angles, reducing the reflection loss at both sea surface and seafloor. This happens because steeper rays have higher losses than less steep rays. Another parameter affecting PL is the sound frequency that depends on the sound source. The PL for low frequency sound, defined as above the low-frequency cut off of the shallow water channel, is generally smaller because of the decrease in sound absorption and scattering at the surface (Francois and Garrison, 1982b; Kuperman and Roux, 2007). Snell’s law is frequency independent, however for a layered seafloor the reflectivity has a complicated frequency dependence affecting the reflection loss (Kuperman, 2019).

Deep seas are characterized by a minimum sound speed, of which the depth varies per region. At mid-latitudes the lowest sound speed is located at approximately 1000 m below the water surface. The depths adjacent to this minimum are known as the SOFAR or deep sound channel. The refraction in the SOFAR channel allows low-frequency noise to propagate over long distances (in some cases more than 1000 km) with no interactions with the sea surface and seafloor (Figure 8, Thorp, 1965; Chow and Turner, 1982). Consequently the largest contributor to PL here is absorption. However, at very low frequencies (e.g.< 10 Hz) noise cannot propagate in the SOFAR channel because the wavelength is larger than the SOFAR’s channel vertical extension. The propagation in the SOFAR channel happens because noise is always refracted or bent back towards the minimum sound speed (following Snell’s law). Below the SOFAR channel, temperature and salinity are generally constant and sound speed increases with depth causing the refraction of noise back to the SOFAR channel. In the SOFAR channel, PL is mainly driven by cylindrical spreading and absorption. Instead, the contribution of absorption increases with frequency, limiting the propagation distance at higher frequencies, for example contributing 36 dB km^-1 to the PL at 100 kHz (Francois and Garrison, 1982a). Noise is trapped in the SOFAR channel when it is originated within the channel’s depth or when sound originates on the continental slope and part of it enters the SOFAR channel after reflection by the seafloor (Qin et al., 2014). At high latitudes, the SOFAR channel is replaced by a surface duct formed by freshwater coming from ice melting. As at mid-latitudes, under this surface duct salinity and temperature are constant for the entire water column and sound speed increases due to pressure, causing the refraction of noise back to the surface. The continuous interaction with the sea surface does not allow the propagation of noise to the same distances as happens in the SOFAR channel (Urick, 1979).

In most cases, a sound source located at the surface does not propagate into the SOFAR channel. In this case, noise is attenuated by the same phenomena of shallow seas but with a smaller PL due to fewer interactions with the sea surface and seafloor. A sound ray that leaves the source horizontally is refracted downward at steeper and steeper angles until it crosses the axis of the SOFAR channel. Under the SOFAR channel, the ray is refracted upward until it is horizontal and may reach the surface. Rays that start more vertically can reach larger depths and may be refracted back toward the surface. The refraction of noise downward leads to a rapid decrease in sound intensity in the horizontal direction. Figure 5 shows that noise refraction creates three different zones: 1) shadow zone, 2) convergence zone located at the surface (a narrow region of very high noise level) and 3) a wider region of lower noise level. The shadow zones are characterized by no noise. Frosch (1964) showed that the typical distance between convergence zones is between 48 to 56 km and they have been observed in the Atlantic and Pacific Oceans at more than 650 km from the source. However, in the Arctic, the range of convergence zones is generally very short or even absent (Urick, 1979) and sound mainly propagate in the top 100 m (Figure 9).

The ocean carbon system has been changing since the industrial revolution (starting around 1760) by the increase of atmospheric x(CO₂) by more than 100 μmol mol^-1. This resulted in a decrease of the global ocean pH average from 8.21 to 8.10, corresponding to a 29% increase in H⁺ activity (Fabry et al., 2008; Doney et al., 2009). The increase of x(CO₂) is driven by human activities such as fossil fuel combustion and deforestation (Doney and Schimel, 2007). Future projections suggest that in the next decades, the ocean CO₂ uptake will continue, further decreasing the surface ocean’s pH.

Ocean acidification is affecting the ocean environment by lowering the calcium carbonate saturation state (Ω_carbonate) and carbonate ion concentration. This impacts shell-forming organisms such as plankton, benthic mollusks, echinoderms and corals (Doney et al., 2009; Hofmann et al., 2011). Another effect is the reduction of α at low frequencies (<10 kHz, Hester et al., 2008). The climate projections predict by the middle of this century a decrease of pH up to 0.3 (Brewer, 1997), leading to a decrease in α by almost 40% (Hester et al., 2008). Figure 6 shows that at frequencies between 10 kHz to 100 kHz α is not affected by pH because α changes are controlled by the chemical relaxation of magnesium sulphate. At frequencies higher than 100 kHz α is driven just by water viscous absorption.

The most applied algorithm to calculate α was derived by Francois and Garrison (1982a) and its inputs are sound frequency, temperature, salinity, depth and pH. The pH dependency at low frequencies is influenced by a pH-dependent chemical relaxation between B(OH)₃/ B ( OH ) 4 − (Simmons and Fisher, 1975). Acoustic relaxations occur due to pressure-dependent volume changes. When a sound wave encounters a charged molecule such as borate a resonance can occur and the sound energy is lost. After the passage of the sound wave, the molecule returns to its normal state. In this case the B ( OH ) 4 − molecule is bigger than B(OH)₃ and because of its charge is associated with water molecules as a loose assemblage. A sound wave can temporarily compress this weak complex into B(OH)₃ which is a lower-volume form (Brewer and Hester, 2009). The ratio between B(OH)₃ and B ( OH ) 4 − is set by seawater pH (Zeebe and Wolf-Gladrow, 2001), therefore in a more acidic environment B ( OH ) 4 − will decrease in favor of B(OH)₃ leading to a smaller α.

For simplicity, most of the α algorithms (Francois and Garrison, 1982b; Ainslie and McColm, 1998; van Moll et al., 2009) consider at low frequencies (<10 kHz) just the chemical relaxations between B(OH)₃/ B ( OH ) 4 − . However, other studies have shown that at these frequencies α is also affected by other chemical species. Mellen et al. (1980) proposed a simple mechanism that involves the chemical relaxations between B(OH)₃/ B ( OH ) 4 − and HCO 3 − / CO 3 2 − . The final algorithm was not satisfactory because the inclusion of these two mechanisms suggested a lower α than expected. In a later study, Mellen et al. (1983) hypothesized a four-state exchange including Ca²⁺ but they suggested that more research is needed to quantify the effect of Ca²⁺ and other ions on sound adsorption. Fischer (1979) hypothesized another mechanism involving the relaxation of MgCO₃ and Mg(HCO₃)⁺. All these hypotheses have not led to a final equation that relates seawater chemistry to sound absorption. Therefore, further research is necessary to identify all the chemical species affecting α and combine these dependencies within one single algorithm.

Several studies analyzed the effect of ocean acidification on PL showing a final effect of <2 dB (Joseph and Chiu, 2010; Udovydchenkov et al., 2010; Ainslie, 2011). For example, Hester et al. (2008) used the Global Ocean Data Analysis Project (GLODAP) (Key et al., 2004; Sabine et al., 2005) to estimate the decrease of α at 0.44 kHz since the industrial revolution. They found that in parts of the North Atlantic Ocean in the top 400 m α decreased by over 15% and by more than 10% in other parts of the Atlantic and Pacific Oceans. In the same study, they applied a conservative pH decrease of 0.15 for the coming decades to calculate sound absorption, they found a decrease of α by over 20% that increases to 60% for a pH decrease of 0.6. A similar decrease in α was found by Ilyina et al. (2010) that using climate models found a decrease in α up to 60% between 0.1 to 10 kHz for a decrease of ocean pH by 0.6. These changes in α depend on the region considered since ocean acidification has smaller effects at high latitudes (e.g. North Atlantic Ocean and Arctic Oceans) and regions of deep-water formation (Ilyina et al., 2010). Despite the projections showing a significant reduction in α the absolute change at >100 km from the source are projected to be less than 2 dB (Joseph and Chiu, 2010; Reeder and Chiu, 2010; Udovydchenkov et al., 2010; Ainslie, 2011). Sound absorption is not the only effect contributing to PL, it acts together with refraction, seabed attenuation, stratification, surface scattering and geometrical loss (spherical or cylindrical) (see section 3.1.1 and 3.1.2). In shallow seas, the changes in α will not be visible as PL is dominated by the interactions with the sea surface and seafloor. The contribution of α is larger in the deep sound channel where the largest components are geometrical spreading and α. Initially, the main mechanism controlling PL is geometrical spreading but at a certain distance (cross-over range) α becomes important. α has a frequency-squared dependence and therefore the cross-over distance is smaller with the increase in frequency.

The final effect has been quantified by Udovydchenkov et al. (2010) that applied a simple propagation model to predict underwater sound in the coming 100 years. Udovydchenkov et al. (2010) quantified that propagation loss will decrease by 1.5 dB at 500 Hz and about 2 dB at 1000 Hz. This happens because the absolute contribution of α at low frequencies is small (at <100 Hz is less than 0.002 dB km^-1). They found that a sound increase due to decreasing α will only be detectable in quiet regions far away from shipping lanes and other anthropogenic activities.

The changes in α are important when sound is trapped in a sound channel (duct) without interacting with the ocean boundaries (seafloor and sea surface). In the Arctic Ocean where in the next 30-50 years pH is projected to decrease by 0.2 (from 8.1 to 7.9, Ciais et al., 2013), sound can be trapped in a duct located at around 150 m (Duda, 2017). Considering a sound source in the Beaufort Sea’s duct, a 0.2 pH decrease is expected to allow the propagation of 900 Hz sound 38% further (100-300 km) leading to an increase in sound of 7 dB at 200 km. However, most sound is generally not trapped in a duct and a 1 kHz sound source outside the duct is expected to increase by 0.5 dB only. All these studies estimated the increase in sound level assuming a constant pH decrease for the entire water column. However, the decrease in pH is expected to be larger at the surface compared to the deeper waters (Caldeira and Wickett, 2005). This implies that the projected changes in α might generally be overestimated.

To conclude, further research is needed to refine Francois and Garrison (1982b)’s algorithm. The inaccurate pH values used to derive the current algorithm results in potentially large uncertainties in the calculated α. To improve the uncertainty to less than 15%, pH needs to be measured using the proper pH scale (total scale) with an accuracy of <0.05 (Brewer et al., 1995). Also, these studies need to elucidate which seawater components contribute to α and include these components in the new algorithm. Even if the effect of OA in the future soundscape is expected to be small (<2 dB) the new algorithm could help to elucidate the size of this change.

Climate change is globally increasing wind speeds. Young et al. (2011) used 17 years (1991-2008) of satellite altimeter measurements and observed a wind speed increase of 0.25 to 0.5% per year. This increase in wind speed is affected by the rise in the number of extreme events by at least 0.75% per year. The increase in wind speed will be larger in the Southern than in the Northern Hemisphere. At the sea surface, the rise in wind speed will increase PL due to rough surface scattering loss and the interaction with near-surface bubble clouds (Ainslie, 2005). The bubbles produced by wind play an important role, not just by scattering or absorbing sound but also by refracting the sound upwards of the sea surface. This refraction enhances the scattering loss associated with the rough air-sea boundary. Therefore, an increase in wind speed will lead to an increase in ambient noise but the PL is also expected to increase. Driven by wind speed the largest increase in ambient noise is expected in the tropics (Duarte et al., 2021). Also, the increase in wind speed will enhance the generation of more bubbles due to breaking waves (Thorpe and Humphries, 1980) with consequences for propeller noise, sound speed and PL. The PL is expected to increase and the sound speed to decrease but the size of these changes has not been quantified yet. Hence, the effects of wind and storms on PL require further research. The effect has not been studied in detail, making it challenging to quantify the contribution to the future soundscape.