The main winter chorus, Chorus F, was characterized by melodic trumpeting sounds with rhythmic qualities and was present at all mesophotic sites (Figure 3; Table 2). While Chorus F was recorded across all mesophotic sites, variability in received levels, timing, and peak frequencies was observed between locations (Figures 3, 4; Supplementary Table S1). Chorus F was also briefly audible at the Al Fahal shallow site. The highest averaged received levels for F occurred at KA_DS3 (105.7 dB re 1 μPa²) and the inshore adjacent KA_DS4 (92.9 dB re 1 μPa²) (Supplementary Table S1). Received levels of F at AF_SH reached 102 dB re 1 μPa², due to overlap with snapping shrimp energy. On average, Chorus F lasted 6:06 h (±02:24 h) during the evening, starting shortly after sunset (18:16PM ± 00:03), and reaching an average peak of 97.8 ± 5.7 dB re 1 μPa² within 20 min. Chorus F consequently lasted until ∼01:30 AM, with a peak frequency of 568.8 ± 19.5 Hz (Table 2). On KA_DS4 where received levels were highest (Table 2; Figure 4), four harmonics were present in Chorus F at 600 Hz, 1.15 kHz, 1.75 kHz, and 2.3 kHz (Figure 3A), while on other sites, only the lowest three–two harmonics were present.

Chorus G, originating in both shallow sites was a dusk chorus consisting of many different sounds, starting at sunset, and lasting approximately 1 h (Table 2). Chorus G had a peak frequency of 232.8 ± 46.1 Hz, and average received level of 93.45 ± 7.6 dB re 1 μPa² (Table 2).

Chorus E was shorter in length than other choruses (∼10 min) but had comparable characteristics to summer Choruses C and III. The peak frequency of Chorus E was 671.3 ± 320 Hz (C: 1,039.3 ± 484.8 Hz), and its peak SPL was 90.2 ± 3.2 dB re 1 μPa² (C: 97.7 ± 4.1 dB re 1 μPa²) (Table 2). Chorus E was audible across all mesophotic sites (Supplementary Table S1) except KA_DS3.

Choruses C, D, and III exhibited differences in received levels, timing, and location (Figure 5), but shared similarities in the type of individual sound being a series of pulses (Figure 7). Chorus III was only found on KAEC sites. The highest received levels were observed for III at KA_DS3, peaking at 117.5 dB re 1 μPa², followed by the adjacent inshore mesophotic site KA_DS4 (109 dB re 1 μPa²) (Supplementary Table S1). Choruses C and D, only observed on Al Fahal sites, exhibited average peak levels of 97.7 ± 4.1 and 104.8 ± 8.2 dB re 1 μPa², respectively, with slightly higher levels at AF_DS2 than AF_DS1 (Figure 5). Choruses C and D were also audible at adjacent shallow site AF_SH, as was Chorus III at KA_SH (Figures 5, 6; Supplementary Table S1). Choruses III and C overlapped temporally.

Summer choruses Chorus C and Chorus D were most prevalent at the Al Fahal mesophotic sites but were also audible at lower levels in adjacent shallow sites (Figure 6). As Chorus C ended, Chorus D began, shifting from a higher peak frequency (1039.3 ± 484.8 Hz) to a lower one (850 ± 164.6 Hz) (Table 2; Figures 5, 6).

Individual sounds of III on KA_DS3, the site with highest received levels, showed 4-7 pulses (Figure 7C). On KA_DS4, further inshore, the sounds only had 2-4 pulses (Figure 7B). However, for C and D it was not possible to separate clean examples of individual sounds (Figures 7D, E).

Chorus II was restricted to the KAEC shallow and mesophotic sites during the summer (Figure 5; Supplementary Table S1). Chorus II, perceived as popping/pulse sounds and white noise, had a variable peak frequency of 698.3 ± 507.6 Hz and peak SPL of 101.03 ± 2.6 dB re 1 μPa² across sites.

Chorus A, occurring only in the summer, was very short in length (10 min) possessed a rhythmic quality similar to the winter Chorus F. It had a peak frequency of 675 ± 42.4 Hz (vs. F at 568.8 ± 19.5 Hz) and a peak SPL of 95.1 ± 0.1 dB re 1 μPa² (vs. F at 97.8 ± 5.7 dB re 1 μPa²). A lastined only ∼10 min during the summer, compared to F, which lasts 06:06 ± 02:24 h during the winter. Chorus A was only present at both Al Fahal mesophotic sites.

Chorus H was the summer dusk chorus, starting at sunset (Supplementary Table S1) comprised of a mix of low-frequency fish sounds, such as pops, knocks, and grunts (Table 2, Figure 5). The peak frequency of Chorus H was 377 ± 102.6 Hz, and received levels were 101.5 ± 8.8 dB re 1 μPa² (Table 2; Supplementary Table S1).

Choruses U1 and U2 were also present in summer at only the mesophotic sites. Both choruses sounded like white noise. Chorus U1 started at 19:47 (±0.01), one hour after sunset and ended at 21:16 (±0.05), with a peak frequency of 3,465 Hz (±92 Hz). Chorus U2, began at 01:26 (±0:05) 6 hours after sunset, and ended just before sunrise at 05:36 (±0:05). Chorus U2 had a peak frequency of 935 ± 64 Hz.

During winter Chorus F was the most prominent, consistently present across mesophotic sites. Chorus F was present during all days of recording, coinciding with a first quarter lunar phase to full moon (59%–100% illumination) (Figure 3). Chorus F reached peak levels (117 dB re µPa²) at KA_DS3 (Table 2; Figures 3, 4), a site closest to the steep drop-off near the KAEC shelf (Figure 1). This characteristic suggests that the shoal responsible for Chorus F could be located in the deeper waters adjacent to the slope. It was not possible to identify individual calls at any part of the chorus, and the rhythmic quality of the sound (Figure 3) implied coordination of calling (Picciulin et al., 2024). While temperature in the mesophotic remained between 24°C and 27°C (Supplementary Figure S3), elevated wind and currents create an iso-speed profile and near surface duct. Water temperature in the mesophotic sites was relatively stable, ranging between 24–27°C across all sites (Supplementary Figure S3; Supplementary Table S2). Sound spreading in winter would have been affected by the presence of a near-surface sound duct (0–200 m)—where sound travels faster—formed by local conditions such as cooler water, higher currents and wind during this period in the Red Sea which also give rise to an iso-speed sound speed profile (SSP). (Erbe et al., 2022; Larsen and Radford, 2018; Larayedh et al., 2024). More efficient horizontal propagation and intensification of sound at the surface would occur along the duct, as well as reflection of sound from the bottom of the duct upwards (Erbe et al., 2022). The surface duct presence in the winter could imply that this chorus may originate shallower than 200m, as sound below would be more easily attenuated (Erbe et al., 2022).

Furthermore, the location of Chorus F could also be evident in the change in frequency components recorded between the sites. While the peak frequency of 568.8 ± 19.5 Hz was fairly consistent among sites, the harmonic components differed between sites. All four harmonics of Chorus F were received at KAEC mesophotic site 3 (Figure 3; Table 2), while these are attenuated to the two lower harmonics in the recordings of KAEC mesophotic site 4, Al Fahal mesophotic sites and Al Fahal shallow site (Figure 3; Supplementary Table S1). This degradation in frequency can occur when sound is picked up from the far-field (Erbe et al., 2022). Chorus F was only audible at Al Fahal shallow site for a brief period (∼1 h from 18:15), with elevated received levels (Figure 4) more likely due to the overlapping impulsive broadband sounds of snapping shrimp (Supplementary Table S1) (Butler et al., 2021; Lillis and Mooney, 2018). This shallow-to-mesophotic acoustic range highlights the broader spatial extent of Chorus F. The absence from the KAEC shallow site, despite being adjacent to the potential origin (KA_DS3), is most likely due to the sound attenuation in the soft absorptive sandy bottom, or by obstruction due to the complex morphology of the shallow, mostly dead reef around the recorder (Erbe et al., 2022).

The extended spatial scale of the fish choruses observed is further highlighted by the presence of shallow dusk Chorus G on mesophotic sites (Figure 3; Supplementary Table S1), as well as the dusk Chorus H during summer. Chorus G at shallow sites KA_SH and AF_SH was present during winter temperatures of 24.4°C–26.5°C (Figure 3; Supplementary Table S2), at a received level of 93.45 ± 7.6 dB re µPa². This dusk chorus includes, among other sounds, identifiable pulse and knock sounds attributed to Pomacentridae (Akamatsu et al., 2018; Staaterman et al., 2013) and popping sounds characteristic of nocturnal species like Pempheridae (sweepers) (Kaplan et al., 2015; Lyon et al., 2019; Nedelec et al., 2015). Mixed-species assemblages with high site fidelity are commonly attributed to dawn and dusk chorusing (Anderson, 2021; Kaplan et al., 2015; Nedelec et al., 2015). Chorus U1 also appears to originate on the shallow reef site, and due to the high frequency nature and timing, is most likely attributed to the peak in crepuscular invertebrate (snapping shrimp) chorusing (Butler et al., 2021; Lillis and Mooney, 2018) (Figure 3; Table 2). The presence of both Chorus G and Chorus U1 in the deep soundscape recordings (Figure 3; Supplementary Table S1) affirms how shallow reef soundscapes can provide critical acoustic cues for larval settlement and reef navigation for marine fauna (Montgomery et al., 2006; Vermeij et al., 2010; Simpson et al., 2005). Previous experiments using light traps equipped with speakers playing nocturnal reef sounds, with a range of only 65 m, showed an increased catch of larval fish up to 1.6 times that of just light (Leis et al., 2003). Pelagic larvae or predators may rely on these choruses as navigation cues (Montgomery et al., 2006; Simpson et al., 2005; Remage-Healey et al., 2006), such as dolphins, with foraging echolocation visible in AF_DS1 sound levels (Figure 4) as well as communication sounds such as whistles and buzzes (Dibble et al., 2016; Janik, 2000) throughout all mesophotic sites.

Natural sources of sound underwater include wind, waves and currents, and higher wind speeds can affect ambient noise levels underwater, especially in low frequencies (>1 kHz) (Burnham et al., 2023). It is important to note that wind may also affect received levels of chorus activity (Burnham et al., 2023; Ainslie, 2005) and higher wind speeds have shown negative correlation to peak chorus levels (McWilliam et al., 2017). During winter recordings, wind speed steadily increased from a daily average of 3.76 knots (Beaufort scale 2) on 9/1/22, to reach a peak of 10.8 knots (Beaufort scale 4) on 20/1/22 (Supplementary Figure S4). A change in underwater conditions also occurs, during a period of higher wind, on the 15^th January (Supplementary Figure S4). During higher wind days, the near-surface sound duct could be disrupted (Ainslie, 2005), allowing sound to travel less efficiently horizontally, influencing the received levels of winter choruses at the shallow hydrophone, or more distant mesophotic sites. If the conditions were more calm, recordings of choruses may present different received levels and spectral qualities. The varying effects of environmental factors makes it difficult to compare other fish sounds and choruses to identify the source species.

As is common with many fish choruses and sounds globally, there is no current evidence for a species responsible for the winter choruses. The tonal, harmonic sound structure of Chorus F (Figure 3A, C) shows similarity to the characteristics of sound produced actively in courtship and/or spawning coordination, as observed in Terapontidae species (Parsons et al., 2017; Laxminarsimha et al., 2020; Mahanty et al., 2015; Parsons et al., 2016b). The harmonic structure and frequency components (600 Hz–2.3 kHz) of Chorus F share similarities with choruses linked to swim bladder-driven sound production (Parsons et al., 2017). However, while comparison of waveforms (Figure 3B) show a similar pattern to Chorus VI (Parsons et al., 2017) in Western Australia, spectral comparison reveal differences in harmonic frequencies and peak values (568.8 ± 19.5 Hz for F vs. 440 Hz for VI), either suggesting a distinct source or difference in propagation conditions, or geography (Parmentier et al., 2005). Known Red Sea soniferous fish families which spawn during winter (e.g., Haemulidae, Gerridae, and Mugilidae (El-Regal, 2013; Bogorodsky and Randall, 2019)) have not been documented to produce harmonic choruses of this type (Amorim, 2006; Parmentier and Fine, 2016; Tricas and Boyle, 2014). Overall, Chorus F likely represents a key aggregation site off Thuwal, and underscores the spatial extent of fish chorusing in both mesophotic and shallow habitats.

Lastly, Chorus E present on all mesophotic sites except for KA_DS3, was short in length (10 min, Table 2) and sounded similar to the summer Choruses C and III. Although individual sounds could not be separated, Chorus E appeared to be comprised of pulsed calls, with a peak frequency within the range of III and C, of 671.3 ± 320 Hz. If attributed to the same source species, this may signal either migration or year-round residency with fluctuating numbers of individuals (Parsons et al., 2017). However, as fish choruses and sounds can show very similar characteristics (Hawkins et al., 2023), there is a chance this chorus can be attributed to a distinct species.

Summer recordings revealed the majority of choruses seen across the two seasons (n = 8). Al Fahal sites recorded adjacent Choruses C and D, at respective peak frequencies of 1039.3 ± 484.8 Hz, and850 ± 164.6 Hz), both characterized by dampened pulses or “croaks” (Table 2; Figures 5, 6). On KAEC sites, Chorus III exhibited a similar sound, at a peak frequency of 848 ± 19.6 Hz. The variation in received levels (Figure 6) and spectral differences (Supplementary Table S1) and timing were enough to label them different choruses.

Notably, Choruses C and D were present at both mesophotic and shallow sites at Al Fahal, whereas Chorus III was restricted to KAEC mesophotic sites (Figures 5, 6). These spatial differences could reflect site-specific shoaling behavior, differences in reef topography, orientation relative to the hydrophone, and sound propagation factors as discussed in the previous Section (4.1.1). In the summer, the Red Sea possesses a downward-refracting SSP where sound speed increases with depth, and there is a reduction in surface reflection of sound without the presence of a duct like in winter (Larayedh et al., 2024). This potentially means that soundwaves refracted downward are eventually absorbed or scattered by the seafloor, causing propagation loss (Erbe et al., 2022). Due to these conditions, it would be less likely for a chorus originating deeper than 200 m to be received at such high levels on KA_DS3 at 73.2 m (117.5 dB re µPa², Supplementary Table S1). Chorus C, shares similar temporal characteristics as III (21:32 h ± 00:23–00:40 h ± 00:09, vs. 21:43 h ± 1:10–00:42 h ± 00:50, respectively), but exhibits a much lower received level of 97.7 ± 4.1 dB re µPa² (III = 106 ± 13.2 dB re µPa²). Spectrally, Chorus C has a higher peak frequency (1,039.3 ± 484.8 Hz) than Chorus III (848 ± 19.6 Hz), the latter which is more similar to Chorus D (850 ± 164.6 Hz) (Table 2). As previously noted (Section 4.2), wind can also affect sound propagation (Ainslie, 2005) and contribute to ambient noise levels in the low frequencies. The average wind speed declined steadily from 21/7/22 at 8.8 knots (Beaufort scale 3), down to 4.3 knots (Beaufort scale of 2) on 31/7/22 (Supplementary Figure S4), and could have potentially scattered frequencies to reduce propagation of deep choruses to the shallow reef sites.

Alternatively, a short, consistent temporal gap between Chorus C and D suggests they may originate from two distinct species, although vertical movement within the water column by a single species remains a possibility (Hawkins et al., 2023; McCauley and Cato, 2016). A previous study observed frequency shifts in fish choruses as shoals moved higher in the water column, consistent with swim bladder-driven sounds (McCauley and Cato, 2016). If multiple species are responsible for these choruses, this pattern (C to D chorusing) may exemplify the acoustic niche hypothesis, wherein organisms partition vocal frequencies to avoid overlap while maximizing communication efficiency (Krause, 1993; Bertucci et al., 2020).

White noise Choruses U1 and U2 add further complexity to the mesophotic soundscape. Like in winter, Chorus U1 is most likely the attenuated sound of the shallow invertebrate chorus (Butler et al., 2021; Lillis and Mooney, 2018), as it coincides with the timing of greatest energy present in the shallow invertebrate chorus, just after dusk (Figures 3, 5). Chorus U1 is slightly more evident in winter than summer, likely due to the different sound propagation conditions, with the near-surface sound duct potentially increasing sound propagation from the shallow chorus (Larayedh et al., 2024). This difference is most likely not related to the difference in lunar phase: snapping shrimp sound pressure levels have been seen to increase during new moon (Gordon et al., 2019), as was observed during summer recordings (third quarter to new moon; 0%–34.5% illumination. Figure 5; Supplementary Table S4). Chorus U2, unlike U1 is lower in frequency (peak of 935 ± 63.6 Hz vs. 3,465 ± 91.9 Hz) and is only present on mesophotic sites. Starting at the end of Chorus III, one potential reason for Chorus U2 may arise from hydrodynamic shoal movements, possibly reflecting changes in position within the water column. Chorus U2 appearing at the end of Chorus III for 4:10 h, could suggest the source as hydrodynamic noise from diel vertical migration (DVM) by nocturnally active species like Myctophidae. DVM is whereby individuals stay at mesopelagic depths during the day to avoid predation, and travel to surface layers at night to feed (Dypvik and Kaartvedt, 2013; Klevjer et al., 2012). Alternatively, the farfield effect on distant, overlapping calls of a chorus, whether passive through feeding or active for courtship/spawning, may create a diffuse acoustic haze, or white noise (Cato, 1978; McCauley and Cato, 2016; McCauley and Cato, 1998). Echogram studies near the KAEC canyon (22.3°N, 39.03°E) detected deep scattering layers (DSL) dominated by Benthosoma pterotum (Myctophidae) ascending to feed near the surface at dusk (Dypvik and Kaartvedt, 2013; Klevjer et al., 2012; Sobradillo et al., 2022). The synchronized timing of Chorus III with this migration and the white noise Chorus U2 (Figure 5; Table 2) aligns with the white noise, post-chorus (Moulton, 1960). More likely, these sounds could potentially be by-products of foraging, or used for keeping loose structure in the school (McCauley and Cato, 2016). A long-term soundscape study explored potential Myctophid choruses, spanning 1–5 kHz, peaking in levels similarly between 0.75–5.25 h after sunset (Chorus III: 4 h after sunset), and composed of dampened pulses (McCauley and Cato, 2016). Furthermore, these choruses were not correlated with the moon phase, but rather periods of high primary productivity. Further studies would be required to assess their sound production characteristics and their link with primary productivity in the Red Sea.

An additional potential source for Choruses C, D and III includes the Carangidae family, which is known to produce grunt-like sounds (Parsons et al., 2016a; Carvalho et al., 2019). Species such as Caranx ignobilis (giant trevally) are important commercial fish in the region and have been observed aggregating at mesophotic depths (40 m) during summer spawning seasons in the Red Sea (Rice et al., 2020; Bogorodsky and Randall, 2019; Sadovy De Mitcheson et al., 2008). This species’ vertical shoaling behavior (Daly et al., 2018; Zhou et al., 2024) aligns well with the spatial and temporal features observed in these choruses.

Chorus II bears striking similarities to a chorus described in multiple studies (McCauley, 2012; Parsons et al., 2017) and attributed to crepuscular, planktivorous fish families Holocentridae, Priacanthidae and Apogonidae, and most likely related to feeding activity. As noted by the authors, this chorus shows spectral peaks between 600 and 900 Hz and contains a notch around 1,000 Hz, which suggests the shoal’s position near the surface (Parsons et al., 2017). The origin point of this chorus in our study is unknown. Chorus II shows variable spectral qualities in the mesophotic sites as compared to the shallow sites, contributing to the variable average low frequency of 362.5 ± 109.6 Hz to high frequency of 1,646 ± 848.7 Hz, and a peak frequency of 698.3 ± 507.6 Hz (Table 2).