Introduction

In the marine environment, vast aggregations of sessile, megabenthic suspension-feeders act as important ecosystem engineers (sensu Jones et al. 1994), enormously increasing the tridimensionality and heterogeneity of the seafloor and forming the so-called animal forests (Rossi et al. 2017a). The presence of these habitat-forming species on the seafloor directly influences the local hydrodynamics and sedimentation rates (Luckenbach 1986; Buhl-Mortensen et al. 2010; Guizien and Ghisalberti 2017); consequently, forests favour retention of benthic organisms’ propagules and larvae (Eckman 1985; Ponti et al. 2014) and increase the magnitude of the pelagic-benthic coupling (Gili and Coma 1998; Rossi et al. 2017b; Cerrano et al. 2019). Smaller sessile species tend to aggregate below the canopy constituted by the main habitat-formers, creating a complex understory, which increases the attractiveness of the forest for other organisms (e.g., Husebo et al. 2002; Baillon et al. 2012; Henry and Roberts 2017). As a result, benthic and nektonic biodiversity, abundance and biomass increase in the presence of animal forests (Buhl-Mortensen et al. 2010; Cerrano et al. 2010; Miller et al. 2012; Rueda et al. 2019; Enrichetti et al. 2022; Bosch et al. 2023).

The Mediterranean Sea hosts many animal forests, mainly in the mesophotic and upper bathyal realms, where competition with macroalgae is virtually absent (e.g., Bo and Bavestrello 2019; Bo et al. 2019; Cerrano et al. 2010; Grinyó et al. 2016; Gori et al. 2017; Rossi et al. 2017a; Chimienti et al. 2019; Enrichetti et al. 2019a; Dominguez-Carrió et al. 2022). Anthozoans, and in particular antipatharians, scleractinians, and alcyonaceans, are among the primary structuring species, thanks to their erect and often branched habitus, the typically large size, the tendency to create dense forests or bioconstructions, and the ability of some species to represent a secondary substrate for the growth of opportunistic species (e.g., Bo et al. 2009, 2015; Deidun et al. 2015; Rossi et al. 2017a; Orejas and Jiménez 2019). Fishes are often associated with anthozoan forests; consequently, many Mediterranean coral aggregations suffer from high levels of fishing impact (e.g., Bo et al. 2014a; Clark et al. 2016; Ragnarsson et al. 2017; Galgani et al. 2018; Enrichetti et al. 2019b; Angiolillo and Fortibuoni 2020). Fishes can exploit these forests as nursery areas, as well as shelters or feeding grounds; therefore, fish diversity and abundance are generally greater in the forests than in the surrounding areas (D’Onghia et al. 2010, 2012; D’Onghia 2019; Mytilineou et al. 2014; Capezzuto et al. 2018, 2019; Sion et al. 2019). Among anthozoans, antipatharians are known as common habitat-forming species in the Mediterranean Sea, especially at mesophotic depths (Bo et al. 2009, 2019), and their role as nursery grounds has already been documented in previous studies (Bo et al. 2015; Cau et al. 2017). They may also host an adult fish fauna, including highly mobile demersal species, such as Scyliorhinus canicula (Linnaeus 1758), many bentho-nektonic species, including Anthias anthias (Linnaeus 1758), Macroramphosus scolopax (Linnaeus 1758), Zeus faber Linnaeus 1758 (Bo et al. 2014b, 2015; Cau et al. 2017; Chimienti et al. 2020), and close inhabitants, such as Benthocometes robustus (Goode and Bean 1886) and Lappanella fasciata (Cocco 1833) (Bo et al. 2011; Gomes-Pereira et al. 2017).

Nowadays, deep shelf and shelf break coral forests and their associated fauna are mainly investigated by remotely operated vehicles (ROVs), which supply visual information on their distribution, demographic characteristics, and megabenthic species composition (e.g., Bo et al. 2012, 2015; Consoli et al. 2016; Chimienti et al. 2020; Toma et al. 2022). However, by supplying punctiform information, ROVs do not usually allow the investigations of specific dynamics of the observed fauna, including biotic interactions and behaviours connected to diel rhythms.

Surveys can be carried out to observe the temporal dynamics of fishes associated with coral forests with the help of benthic landers equipped with temporised cameras and, occasionally, baits. These low-impact tools are considered optimal systems for non-punctual investigations in the deep sea (e.g., Grehan et al. 2017; D’Onghia et al. 2018; Prado et al. 2023). So far, in the Mediterranean Sea, such studies have mainly been carried out in white coral ecosystems. Baited landers were used in the St. Maria di Leuca and Bari coral provinces to investigate the presence and behaviour of fishes and other large benthic fauna in the white coral mounds (Capezzuto et al. 2012; Maiorano et al. 2013; Sion et al. 2013; D’Onghia et al. 2015a, b, 2018; Linley et al. 2017; Carluccio et al. 2021). Studies on diel rhythms of Mediterranean fishes are scarce and mainly concentrated on shallow-water species (e.g., Barans et al. 2005; Azzurro et al. 2007, 2012; Aguzzi et al. 2013; Francescangeli et al. 2022), even if rhythmic behaviours mediated by tides are also known for great depths (Aguzzi et al. 2010, 2011). No information exists for diel variations in the fish frequentation of mesophotic and bathyal coral forests.

The present study aims to evaluate and describe the diel pattern of fish frequentation of a Mediterranean mesophotic forest dominated by the black coral Antipathella subpinnata (Ellis and Solander 1786) and the species-specific exploitation of the tridimensional habitat with the use of an unbaited lander. A particular focus was directed to the study of the diel rhythms of A. anthias, a common, gregarious, benthopelagic species commonly observed in coral-dominated habitats both in the Mediterranean Sea and Atlantic Ocean (Bo et al. 2012, 2015; Porteiro et al. 2013; Gomes-Pereira et al. 2014; Louisy 2015; Consoli et al. 2016; Chimienti et al. 2020; La Mesa et al. 2021). This focus was carried out especially since mesophotic habitats are subjected to daily light intensity variations and respond to shallow-water fluctuations of other environmental variables, including bottom currents and turbidity plumes linked to rainfall and freshwater discharges (Coppari et al. 2020; Enrichetti et al. 2022).

Materials and methods

Study area

The study was carried out on a rocky shoal situated about 1 NM off the port of Bordighera (43°46.11’ N; 7°40.82’ E) (Ligurian Sea, NW Mediterranean Sea) (Fig. 1a). The shoal comprises large rocky boulders scattered on the seafloor at around 60–70 m depth (Enrichetti et al. 2019a; Terzin et al. 2021) (Fig. 1b). The study site is part of a larger, EW-elongated rocky elevation and is about 300 m from the flanks of the Nervia Canyon (Fig. 1b). The heavily silted shoal hosts a forest of Antipathella subpinnata, which is an Atlantic-Mediterranean species representing the most common black coral in the Mediterranean Sea; its colonies can grow up to 1.5 m and form forests on hardgrounds between 53 m and around 600 m depth (Bo et al. 2009, 2019; Chimienti et al. 2020). The Bordighera forest is the densest population of this species in the Ligurian Sea; it is represented by about 140 colonies with an average density of 4.4 colonies m− 2 and a height of more than 1 m (Enrichetti et al. 2019a; Terzin et al. 2021) (Fig. 1c). A few colonies of the red gorgonian Paramuricea clavata (Risso 1827) are interspersed in the forest (Van de Water et al. 2020) (Fig. 1c).

Fig. 1
figure 1

Sampling station. a-b) Localization of the study area (yellow star) off Bordighera in the western Ligurian Sea and multibeam map of the elongated rocky elevation (Datum WGS84). The red box in (b) indicates the position of the shoal hosting the black coral forest under investigation. c) ROV image of the Bordighera forest of Antipathella subpinnata. d) Position of the lander in front of the black coral forest at 63 m depth. e) A close-up view of the lander and its components

Full size image

In situ data collection

An autonomous tripod Lander W1 (BlueResearch Srls) was deployed at 63 m depth near the outer rim of the coral forest (Fig. 1d, e). The lander was equipped with a MOBIUS action camera (image resolution 2048 × 1538 pixels) and a 450-lumens LED light encased in two Delrin (250 × 70 mm) cylinders with plexiglass portholes. Each cylinder also hosted two NiMh rechargeable batteries (8.4 V, 2500 mAh); a control board (Arduino) was located in the first cylinder, allowing the camera and lights to synchronise. A temperature (°C) profiling sensor was also present (Fig. 1e).

The visual field included seven black coral colonies, the seafloor beneath them, and about one meter of water column above the canopy for about 4 × 2.5 m of visible foreground (Fig. 2a-d). The colonies were considered together as a single spatial unit. The lander was previously tested to ensure that, during the night, the artificial light was powerful enough to light up the field of view, even if the left portion of the image resulted darker than the right portion because of the asymmetrical placement of the light. The lander remained in place for 21 days (and 20 nights) in 2018, between June 19th and July 9th, shooting a total of 1921 rapid series of 8 images (3 in ambient light and 5 with the aid of the artificial LED light) every 15 min, for a total of 15,368 pictures.

Fig. 2
figure 2

Field of view of the lander at dawn (05:38) (a), during the day (08:59) (b), at dusk (19:56) (c) and during the night (22:22) (d). Numbers indicate the stems of the seven Antipathella subpinnata colonies visible in the visual field, while ‘sf’ indicates the seafloor, ‘ca’ the canopy, and ‘wc’ the water column. Arrows indicate the position of the three monitored resting sites

Full size image

Image analysis

All the fishes observed in the visual field of each picture were identified and counted, and their position was noted (1 − 0, presence/absence) considering three layers, namely seafloor, coral canopy, and water column. Since fishes are often rapidly moving subjects, only the highest values of diversity and number of individuals for each set of 8 images were considered. In contrast, the other photographs in the series were used to help confirm identifications; all images were enlarged and, if necessary, post-produced (e.g., increasing brightness and contrast) with Adobe Lightroom software to improve the identification of the subjects. The occurrence of large invertebrates moving in the forest was also noted.

In addition, five environmental variables were reported hourly. Two were measured semi-quantitatively, namely water transparency and current. These were evaluated based on the presence of turbid plumes in the visual field, using a 0–2 scale, and the inclination of the coral branches, using a 0–2 scale, respectively. In situ temperature (°C) was measured with the probe attached to the lander, while cumulative rain (mm) and sea surface average solar irradiance (W m− 2 h− 1) were obtained from the meteorological archives of the Regional Agency for the Protection of the Environment in Liguria (ARPAL) (https://ambientepub.regione.liguria.it) for the nearest site to Bordighera. Cumulative rain represented an indirect measure of the weather conditions and variations of river inputs in the area, while surface irradiance gave an indication of the light fluctuations in the considered period. Considering a light extinction coefficient of η = 0.112 in the site (with K = 1.7 and water transparency DS (Secchi Disk) = 15.1 m (Attolini and Coppo 2005)), then it is possible to estimate that the incident light at 60 m (Iz) is approximately 0.1% of the surface one (following the formula Iz = Ic e− a, with Ic, incident light minus reflection, considering a 7% reflection at noon, and a = z, depth * η).

Finally, three Anthias anthias individuals, clearly visible in the nocturnal visual field on the seafloor, were monitored during the night to characterise sleep (here intended as lack of motility for a prolonged period on the seafloor), evaluate sleeping site selection, sleep periodicity, and triggers and duration of awakenings (here intended as interruptions of resting and short movements around the sleeping site) (Fig. 2d).

Data elaboration and statistics

The species diversity associated with the forest was evaluated by considering the total number of observed species, genera and families, as well as the percentage frequency of occurrence, representing the percentage of pictures in which each species occurs. To assess the habitat use by the fish fauna during the 24 h, we considered four light phases: dawn, day, dusk, and night. We used the following cut-offs of the light record to evaluate the distinction between them: 0–2 W m− 2, night; 10–60 W m− 2, dawn; 90–1000 W m− 2, day; 10–60 W m− 2, dusk. The change in fish community composition over phase transition was measured by quantifying assemblage dissimilarity. The community detected in phase transition was pairwise compared by calculating the Jaccard dissimilarity (Jac), which ranges from 0, indicating completely similar communities, to 1, in case of complete dissimilarity (Magurran and McGill 2010). Specifically, the observed assemblage dissimilarity was partitioned into two components: (i) turnover (Jtu), consisting in the substitution of species in one light phase by different species in other phases; (ii) nestedness (Jne), which implies the loss or gain of species in one phase, with the reduced community being completely nested within the larger one (Baselga 2010). Pairwise comparison and partitioning of the Jaccard dissimilarity were calculated in the R environment, using package betapart (Baselga and Orme 2012).

The existence of diel patterns was evaluated for the three species showing the highest number of observed individuals and the highest frequency of occurrence (A. anthias, S. smaris and S. cabrilla), taking into account the trend in the number of counted individuals in the considered period with respect to the light fluctuations. The trends were superimposed on the other environmental variables to evaluate other possible relationships. A schematic representation of the diel rhythm of A. anthias was created.

The diel rhythm of A. anthias in the occupancy of each layer was evaluated by building Cosinor models (Cornelissen 2014), which allowed to assess the periodicity and timing of the spatial distribution. Cosinor models are a well-established statistical tool commonly used in chronobiological studies to analyse and quantify the rhythmic patterns in time-series data (Refinetti et al. 2007) when the value of a response variable (Y) is assumed to depend on time (t) following a regular cycle. Cosinor models describe a sinusoid by incorporating a sine and cosine term into a generalised linear model. From Cosinor models, estimates are provided for (i) the rhythm-adjusted mean (MESOR - Midline Estimating Statistic Of Rhythm), representing the average value of the cosine functions fitted to the data (Refinetti et al. 2007); (ii) the amplitude, which is a measure of the extent of predictable change within a cycle; (iii) the phase, corresponding to the time at which the fitted cosine function has its peak (Fernández et al. 2009). We built a Cosinor model with binomial error distribution for each layer, assuming the occupancy of A. anthias (0/1) in each layer for each picture (n = 1921) as the response variable, while the presence of other fish species, the presence of turbid plumes and the presence of current in the visual field were employed as predictors. As a result, the probability of use P for a given layer was modelled with the following formula:

$$\eqalign{& P = Intercept{ + _1}*Presence\,of\,other\,species \cr & { + _2}*Turbidity{ + _3}*Current{ + _4}*cos\left( T \right) + {\,_5}*sin\left( T \right) \cr}$$

where β1−3 are covariate coefficients and β4−5 are the coefficients for the cosine and sine components, respectively, of time T, capturing the periodic pattern. For each model, we assessed the significance of the diel pattern in forest use by evaluating both the amplitude and the significance of the sine and cosine terms of the model (adjusted significance at α = 0.025; Barnett and Dobson 2010). The diel rhythm was also tested with the same approach for the other two species showing a sufficient number of records. Cosinor models were built in R environment, using package season (Barnett et al. 2022).

Finally, to evaluate the sleeping behaviour of A. anthias, the average (± SE) duration of all sleep moments and the average number and duration (± SE) of all awakenings were calculated.

Results

Environmental variables

The light record (Fig. 3a) accounted for 80 images falling into the category dawn (with surface irradiance values ranging between 15 and 55 W m− 2 h− 1). Dawn lasted about one hour, between 05:00 and 05:56. The record also accounted for 1038 images falling into the category day. Irradiance values ranged between 95 and 986 W m− 2 h− 1, with four exceptions ranging between 24 and 85 W m− 2 h− 1 and related to cloudy moments. The time of day varied between 06:00 and 18:59. The dusk category included 80 images (with irradiance values ranging between 14 and 62 W m− 2 h− 1). Dusk lasted about one hour each day, between 19:00 and 19:59. Finally, 723 images fell into the category night (with irradiance values ranging between 0 and 2 W m− 2 h− 1). The night time varied between 20:00 and 04:59. The irradiance variations between the four phases were always sharp and quick.

Altogether, the survey period was sunny and dry, with the initial five days characterised by high irradiance levels. These were followed by three days (24th, 25th, and 26th of June) of instability concomitant to a peak of rain (1.4 mm h− 1) on the 24th of June (Fig. 3a). Similarly, on the 28th of June, low irradiance levels were due to a series of rainfalls (up to 0.8 mm h− 1) (Fig. 3a). Other variations in irradiance concomitant to two rain episodes were appreciated on the 6th and 7th of July when 1.0 mm h− 1 of cumulative rain was registered (Fig. 3a).

Seafloor temperature ranged on average between 14.5 °C and 14.7 °C with a minimum of 14.2 °C registered on the 25th of June and a maximum of 15.3 °C registered on the 4th of July (Fig. 3b). The visual field was invested by various turbid plumes, with one series occurring during the first week of the survey and another one occurring in the last week (Fig. 3c). Two of these events lasted 4 to 5 continuous hours, the one registered on the 22nd of June and the one on the 4th of July, respectively. No intense turbid plumes (category 2) were recorded in the survey period (Fig. 3c). On average, the site was interested by frequent moderate to strong currents, representing 49.5% of the records, with strong currents registered in about 9% (Fig. 3d). Almost continuous events of currents could last from 12 to 29 h, with the longest reported between the 20th and 22nd of June, the 25th and 29th of June, and the 6th and 9th of July. The most intense ones (with the highest number of continuous 2 records) were registered on the 28th and 29th of June and on the 9th of July.

No overlaps were observed among rainfall events at surface and temperature peaks, turbid plumes, or strong currents at depth. The peak in temperature at the beginning of July corresponded to a turbid plume and strong currents but to no rainfall events; therefore, it was probably associated with deep hydrological phenomena.

Fig. 3
figure 3

Patterns of environmental variables in the study period. a) Average surface solar irradiance (W m− 2 h− 1) (black line) and average cumulative rain (mm h− 1) (dashed red line). White arrows indicate significant anomalies in solar irradiance during the day. b) In situ temperature (°C). The red box indicates the average range. c) Turbidity, attributed using a 0–2 scale. White arrows indicate moments with prolonged turbid events. d) Current, attributed using a 0–2 scale (grey circles, 1, moderate currents; red circles, 2, strong currents; 0 not shown for clarity). White arrows indicate moments with prolonged strong currents

Full size image

Fish diversity

The survey identified 17 fish taxa belonging to 15 genera and 10 families (Table 1; Fig. 4). Sparidae was the most represented family, with eight species (Table 1). In terms of fish counts, a total of 43,369 individuals were recorded in the visual field during the entire investigation, of which only 0.2% were undetermined.

Table 1 Fish taxa observed in the survey. Da, dawn; D, day; Du, dusk; N, night. References for main diet information: 1 Whitehead et al. 1986; 2 Weinberg 2013; 3 Sartor et al. 2017
Full size table

Anthias anthias resulted the most frequently observed species in the visual field, being present in 82.9% of the pictures, followed by Serranus cabrilla (Linnaeus 1758) (4.7%), Spicara smaris (Linnaeus 1758) (4.7%), Diplodus vulgaris (Geoffroy Saint-Hilaire 1817) (1.1%), and Phycis phycis (Linnaeus, 1766) (0.8%). The other species were present in less than 0.5% of the pictures (Table 1). Anthias anthias resulted by far also the most recorded species, representing 97.3% of the counted individuals, followed by S. smaris (2.0%) and S. cabrilla (0.2%) (Table 1).

Besides fishes, other large invertebrates were recorded, including one crab, seven pelagic shrimps, one squid, one seastar Peltaster placenta (Müller and Troschel 1842), and one octopus Octopus vulgaris Cuvier 1797.

Fig. 4
figure 4

Fishes observed in the black coral forest. a) Anthias anthias, b) Trachurus sp., c) Ariosoma balearicum, d) Mola mola, e) Mullus surmuletus, f) Muraena helena, g) Phycis phycis, h) Serranus cabrilla, i) Dentex dentex, j) Diplodus puntazzo, k) Diplodus sargus, l) Diplodus vulgaris, m) Pagellus sp., n) Sparus aurata, o) Spicara smaris, p) Spondyliosoma cantharus, q) Chelidonichthys lastoviza

Full size image

Forest occupancy: temporal and spatial patterns

The diversity of the fish community inhabiting the investigated black coral forest changed among the four light phases, with the highest number of species observed during the day (14 species), followed by night (10 species) and then the two transitions (6 and 5 species for dusk and dawn, respectively) (Table 1). As measured by Jaccard dissimilarity, the overall change in community composition during the transition from day to dusk (Jac = 0.53) was entirely driven by species loss (Jne = 0.53; Jtu = 0.00). The transition from dusk to night implied a more evident change in assemblage composition (Jac = 0.71) and was mainly driven by species turnover (Jne = 0.11; Jtu = 0.60). The phase transition from night to dawn resulted less evident (Jac = 0.45) and was entirely driven by species loss (Jne = 0.45; Jtu = 0.00). Finally, the transition from dawn to day resulted in a substantial change in assemblage composition (Jac = 0.45), entirely driven by species gain (Jne = 0.60; Jtu = 0.00).

Undetermined fishes were reported in all light phases, but mainly during day and night. Two components represented the day community: (i) resident species, with a high frequency of occurrence during the day but also observed in the other phases, namely A. anthias, S. cabrilla, and S. smaris (80%, 9%, and 3.6% of the day records, respectively) and (ii) visiting species, observed solely or almost solely during the day. Visiting species could be recurrent (D. vulgaris, 2.1% of the day records) or occasional (observed less than five times). These latter included Mola mola (Linnaeus 1758), Dentex dentex (Linnaeus 1758), Diplodus puntazzo (Walbaum 1792), Pagellus sp., Sparus aurata Linnaeus 1758, Spondyliosoma cantharus (Linnaeus 1758), and Chelidonichthys lastoviza (Bonnaterre 1788), and could be occasionally recorded also during a transition phase. The night community was represented by two resident species, A. anthias and S. smaris (87% and 5.4% of the night records, respectively) and various visiting species. These included recurrent fishes, namely P. phycis and Trachurus sp. (1.2% and 1.0% of the night records, respectively), and occasional ones (Ariosoma balearicum (Delaroche 1809) and Muraena helena Linnaeus 1758). Two species, Mullus surmuletus Linnaeus 1758 and Diplodus sargus (Linnaeus 1758), overall occasional, showed no clear preference (Table 1).

Regarding the sharing of the forest environment, the large majority of the pictures (72.4%) included only one species (mainly A. anthias), followed by 18% of the records showing no species at all, 12% showing two species (mostly the couples A. anthias-S. smaris and A. anthias-S. cabrilla), about 1% showing up to three species (mainly the triplet A. anthias-S. smaris-S. cabrilla occurring during the day and the triplet A. anthias-S. smaris-P. phycis during the night).

The fish community also changed among the three considered habitat layers in terms of frequency of occurrence (Fig. 5). Ten species exploited the seafloor, mainly during the night (62-100% of the records), with some sightings in the transitions and occasional reports during the day (such as D. vulgaris). Three species, namely A. balearicum, M. helena, and P. phycis, were solitary sightings solely or almost solely patrolling the forest floor during the night. The canopy was exploited by nine species, mainly during the day (43-100% of the records). Two species, P. phycis and D. sargus, also frequented the canopy at night, while A. anthias, S. cabrilla, and S. smaris were regularly observed in the canopy in all light phases with a higher contribution of day records (Fig. 5). Finally, 12 species were observed in the water column above the forest, almost exclusively during the day (35-100% of the records), with two exceptions represented by Trachurus sp., observed only at night, and D. sargus, observed only at dawn. As for the canopy, A. anthias and S. smaris were regularly observed in the water column above the forest in all light phases, but a higher contribution of day records was reported only for A. anthias (Fig. 5).

Fig. 5
figure 5

Percentage contribution of each species to the community during each light phase in the three considered habitat layers

Full size image

Anthias anthias diel rhythms and sleeping behaviour

The number of Anthias anthias individuals in each picture showed a rhythmic pattern during the survey, related to the irradiance variation (Fig. 6a). The number of fish peaked rapidly during the day (up to 335 individuals per picture series), while only a few individuals were generally recorded in a single series during the night (Fig. 6a). This pattern was particularly evident starting from day four of the survey period (22.06.2018), while no major peaks were evidenced for the first few days. Images showed that, at dawn, the individuals started to aggregate in the canopy (Fig. 2a), rapidly rising in the water column with the rising sun (Fig. 2b). The fish school remained above the shoal and the forest for the entire day phase, occasionally rising and descending, hence moving out or in the visual field of the lander. At dusk, the individuals descended again within the dense coral ramifications (Fig. 2c) and then positioned on the seafloor for the nocturnal rest (Fig. 2d). The entire school was consistently present in the habitat, but its entity was appreciable only during the day; in the other light phases, in fact, the individuals were scattered in the canopy or on the seafloor; therefore, only those individuals in front of the camera were visible.

Fig. 6
figure 6

Trends in numbers of Anthias anthias, Serranus cabrilla and Spicara smaris individuals (red lines) counted in the picture series in relation to the patterns of solar irradiance (black lines) during the investigated period. Stars indicate rain episodes, circles indicate periods of prolonged turbidity, and triangles indicate periods of prolonged strong currents

Full size image

The Cosinor models of forest use confirmed these findings. The diel rhythm of A. anthias resulted to be significant (Wald test: P < 0.001) for all three considered layers (Table 3; Fig. 7). The probability of use of the seafloor on a 24-hour cycle had an amplitude of 1 (on the probability scale) with a function phase at 01:30 and was not influenced by other predictors. The probability of use of the canopy showed a reduced amplitude (0.1), with a peak at 15:50 and resulted positively affected by the presence of other fish species (β = 0.59; Wald test: Z = 3.94; P < 0.001; Fig. 7). Finally, the probability of use of the water column had an amplitude of 0.34 with a function phase at 13:30 and resulted negatively affected by turbidity (β = -2.39; Wald test: Z = -4.86; P < 0.001) and positively affected by current (β = 0.79; Wald test: Z = 4.63; P < 0.001).

Table 2 Sleep pattern of three target Anthias anthias individuals. * refers to particularly short night sleeps. Names of fish species in parentheses refer to the nearby passage of a potentially disturbing species on the seafloor, causing no alert reaction
Full size table

At night, three individuals of A. anthias were consistently observed resting on the seafloor (Fig. 8a). It was not possible to ascertain the sex of the individuals nor other morphological features suggesting that they were the exact same specimens; however, their size over the nights remained comparable. The individuals also assumed the exact same position while resting on all monitored nights (Fig. 8a). The position was held still unless some disturbance occurred; in such case, the individuals moved nearby, and when the disturbance ceased, they regained their position and continued to sleep. No individual was observed resting on the seafloor during dawn, day or dusk. The observed individuals slept without awakenings five to seven nights out of the 20 monitored, and on three occasions (nights n° 3, 12, and 20), they all slept all night (Table 2). The average length of sleep was 05:50 h ± 00:11, ranging between 00:30 and 07:30 h. The sleeping time of the fish in the three resting sites significantly differed (Kruskal-Wallis test, H = 10.5, P < 0.001), with one individual (n°2) significantly sleeping more consistently (Mann-Whitney pairwise test, P < 0.01) all monitored nights (on average 06:42 h ± 00:07), with respect to the other two fish (on average 05:35 h ± 00:22 and 05:13 h ± 00:22 for n° 1 and n° 3, respectively), which showed a few particularly short nights (00:30 − 03:45 h of sleep) (Table 2). The average number of awakenings of an individual per night was 1.3 ± 0.15 (ranging between 0 and 5). Despite individual n° 2 showed 18 overall awakenings with respect to 27 for individual n° 1 and 33 for individual n° 3, and never more than 3 per night (Table 2), no significant differences were observed between individuals (Kruskal-Wallis test, H = 2.9, P = 0.23). The duration of the awakenings was, on average, 00:29 h ± 00:03 (ranging between 00:15 and 03:15 h). Although the awakenings were on average shorter for individual n° 2 (00:23 h ± 00:02) with respect to the others (on average 00:28 h ± 00:04 and 00:33 h ± 00:08 for n° 1 and n° 3, respectively) (Table 2), no significant differences across individuals were observed (Kruskal-Wallis test, H = 0.7, P = 0.72). On average, a typical sleep pattern of A. anthias included a pre-sleep (two hours) and a post-sleep (one hour) period, in which the individuals moved above or within the canopy before going to the resting sites (after dusk) or aggregating in a school waiting to ascent in the water column (before dawn) (Fig. 8f).

Table 3 Estimates obtained from the Cosinor model for three species (Anthias anthias (AA), Serranus cabrilla (SC), Spicara smaris (SS)) and three habitat layers. Cosinor-derived estimates are provided on a probability scale for the amplitude and on a time scale for phase and low point (24 h time format). n.a. indicates that no estimates are returned for a given parameter since amplitude = 0 and no significant diel rhythm has been detected. GLM estimates are provided on the link scale (logit), and covariate effects represent the beta coefficient of the generalised linear models. Statistical significance (Wald test p-value) of model coefficients is represented by the following graphical scale: * = p < 0.05; ** = p < 0.01; *** = p < 0.001
Full size table

In 11 of the 20 monitored nights, 14 out of 62 awakenings were concomitant for the monitored fish (Table 2). In five of these occasions (nights n° 1, 2, 5, 18, and 19), the alert reactions were caused by the nearby passage of a Phycis phycis patrolling the seafloor beneath the canopy (Fig. 8b-c) and, in one case (night n° 19), by the co-occurrence of both P. phycis and Muraena helena (Fig. 8d). This latter species alone was responsible for another shared alert between the three individuals on night n° 16 (Table 2). Phycis phycis was also reported on two occasions (nights n° 3 and 12) without triggering any reactions from the individuals, while Ariosoma balearicum triggered an alert reaction in only one case on night n° 13 for individual n° 3 (Fig. 8e; Table 2). In at least six nights (n° 7–11, 14), it was not possible to attribute the awakenings to any visible trigger (Table 2). Still, the sleep during the last night, which resulted particularly short for individuals n° 1 and n° 3, was concomitant to very strong currents.

Among the other species with a high number of counted individuals, S. cabrilla always showed a maximum of one individual per picture series with no diel rhythm (Fig. 6b), while S. smaris showed peaks of counts during the day (up to 95 individuals per picture series) (Fig. 6c), and a few individuals were pictured at night on the seafloor with a nocturnal colouration (more mottled and less silvery, with lateral dots less visible). However, no clear indication of a residency of the school over the shoal or in specific sleeping sites on the seafloor could be ascertained for this latter species. No significant rhythmic patterns in the probability of use for any considered layer were detected for these two species (Table 3). Although for S. cabrilla the cosine term of the model was found to be significant for two of the considered layers (i.e., forest canopy and water column), the estimated amplitude from the Cosinor model was consistently zero. This either suggests that a significant circadian pattern of use can be excluded or indicates that a larger sample size may be required to properly assess rhythmicity in habitat use (Table 3).

Fig. 7
figure 7

Anthias anthias probability of use, for each layer, over a 24-hour cycle as estimated by Cosinor models. Dark blue lines represent the estimated probability of use, light blue lines represent the 95% confidence intervals. For the forest canopy layer: the dark blue line represents the baseline estimated probability of use when A. anthias is the only species present in the frame, while red line represents the estimated probability of use by A. anthias when other fish species are present; 95% confidence intervals are represented in light blue and orange for the two levels, respectively

Full size image
Fig. 8
figure 8

Sleeping behaviour of Anthias anthias. a) Sleeping sites of the three target individuals on the seafloor below the black coral canopy. b-c) Alert reactions of the target individuals to a patrolling Phycis phycis. Hp, a large ascidian Halocynthia papillosa. d) Co-occurrence of two predators (P. phycis and Muraena helena), e) Ariosoma balearicum. f) Schematic representation of a typical sleeping pattern of A. anthias

Full size image

Discussion

Although the close relationship between fishes and structured marine environments such as animal forests has always been highlighted (e.g., Rossi et al. 2017a), there was virtually no prolonged evidence of the so-called refuge effect of the forest on the mobile fauna. From this point of view, this work highlights a short-term stable relationship of numerous mesophotic species with the three-dimensional canopy of the black coral Antipathella subpinnata, confirming the importance of this facies, highly common on Mediterranean hardgrounds between 50 m and 100 m depth (Bo and Bavestrello 2019). Despite the relatively small field of view of the lander, the survey identified a high fish diversity (17 taxa) comprising three resident species and 14 visiting, three of which are recurrent, for a total of 35% of species stably associated with the habitat. An ROV study conducted on a mesophotic A. subpinnata forest in the South Adriatic Sea between 52 m and 80 m depth identified 22 species, of which eight shared with the present study, including the three resident species (namely, Anthias anthias, Serranus cabrilla, and Spicara smaris), one recurrent (Phycis phycis), and four occasional (Dentex dentex, Diplodus sargus, Muraena helena, and Spondyliosoma cantharus) (Chimienti et al. 2020). No data are provided in this study to elucidate the relationship of the other species with the forest, although three are epipelagic (therefore plausibly occasional), and the remaining eleven are benthopelagic or benthonic (therefore plausibly with some type of relation). This confirms the overall attractiveness of this highly tridimensional habitat, and it is reasonable to hypothesise that forests, albeit dominated by the same structuring species, might be associated with different fish communities when located in different environmental conditions. Interestingly, a few species, such as Conger conger (Linnaeus 1758), are reported in very different bathymetric ranges, being known from mesophotic A. subpinnata forests (Chimienti et al. 2020) as well as bathyal white coral reefs (D’Onghia et al. 2012). In these cases, the occurrence of any structured habitat where to find shelter and food is a driving factor; this is also valid for artificial habitats such as wrecks (Şensurat-Genç et al. 2022). Similarly, Anthias anthias, despite being the species most stably associated with the habitat, is restricted neither to A. subpinnata forests (Fig. 9a) (Porteiro et al. 2013; Ingrassia et al. 2016) nor to black coral forests in general, being reported also associated to gorgonians (Fig. 9b, c) (Bo et al. 2012; Porteiro et al. 2013; Enrichetti et al. 2019b), deep-water scleractinians (Angiolillo et al. 2021), and wrecks or platforms (Andaloro et al. 2013; Consoli et al. 2015; Sinopoli et al. 2015). In this case, the zooplanktivorous fish (Porteiro et al. 2013; Consoli et al. 2015) exploits the heterogeneity of the habitat for protection purposes and may also take advantage of the high water turbulence and secondary productivity enhanced by the presence of the rocky relief or the wreck.

Fig. 9
figure 9

Anthias anthias in mesophotic coral forests of the Ligurian and Tyrrhenian seas. a) Antipathella subpinnata (Favazzina, 70 m, MoBioMarCal 2008), b) Paramuricea clavata (Bordighera Canyon, 60 m, ARPAL 2016), c) Eunicella verrucosa (Bordighera Canyon, 80 m, ARPAL 2016). d-f) Large schools of A. anthias and Trachurus spp. swarming during daytime over a Sardinian wreck (Tavolara, 90 m, Project Sardinian Red Coral 2013), detected using a Multibeam Echosounder EM2040 Kongsberg (400 kHz) and elaborated with the Software Fledermaus QPS Midwater (version 7.6.4). Schools in (d) are indicated in red. g) A male individual (arrow) sleeping at night on the bottom of an aquarium tank below gorgonians (Aquarium SZN, Naples)

Full size image

An ecological aspect highlighted in this study is that the fish community associated with this habitat significantly changes during the light phases. A substantial decrease in diversity is observed in the transition from day to night, while a significant shift in species composition is observed at night. This result is in accordance with the majority of studies exploring day-night differences in fish assemblages (Azzurro et al. 2012; Harvey et al. 2012; Ferguson et al. 2013; Vianna et al. 2013; Myers et al. 2016). A better understanding of the finer temporal changes of the fish community, the different utilisation of the habitat, and the interactions among the trophic guilds in different phases of the day is pivotal to estimating the true diversity and complexity of the fish assemblage (Myers et al. 2016). This has been indicated as valuable information to optimise biodiversity and fisheries management plans (Harvey et al. 2012).

Additionally, a different exploitation of the habitat layers emerged during the light phases. The study of complex marine habitats as multi-layered environments is a common approach, especially for what concern benthic invertebrates (Gatti et al. 2012; Moccia et al. 2022), while it has been seldomly considered for fishes in coral forests (Gomes-Pereira et al. 2017). For the most frequent species, the connection to a specific layer appears here to be mainly related to refuge or feeding habits, indirectly influenced by light. The monitored fish community mainly accounts for species feeding on benthic invertebrates (13 out of 17 species), of which five are known to feed also on benthic fishes (Table 1); three species are plankton feeders, including the two most frequent taxa (A. anthias and S. smaris). Trachurus sp. is the only species known to feed on pelagic fishes and invertebrates. While the coral canopy and the water column above it are mainly exploited during the day by planktivorous species or invertebrates’ feeders, the seafloor is frequented almost exclusively at night, mainly by solitary predators or sleeping fishes. However, it remains unclear why some rarely observed species frequent the forest or its vicinity at a certain moment of the day.

Among all species, A. anthias is the sole one regularly occupying all three layers. This species spends most of the day (90% of the photo series) in large and dense schools in the water column above the forest, making contact with the canopy only occasionally (32%), probably when some disturbance occurs (possibly large predators such as D. dentex). The influence of other species in the occupancy of this layer is also confirmed by the Cosinor models. On the other hand, at night, it spends 97% of the time taking refuge in the canopy or resting in the basal layer of the forest. During the transitions, fish are mainly observed aggregating within the canopy (72% during dawn) or just above it (89% during dusk). This novel ecological information adds to the very poor literature available for A. anthias (www.fishbase.se; Tortonese 1986; Louisy 2015), although this species has always been reported as a dominant component of fish assemblages in the Mediterranean Sea (e.g., Consoli et al. 2016). It can also be added that these large schools, often observed over topographic reliefs or wrecks, provide an important ecosystem service in these environments since the daily vertical movement of these large biomasses (Fig. 9d-f) and faecal pellet plausibly largely contributes to enhancing the pelagic-benthic coupling occurring in the coral forests. In this regard, it has been reported for many planktivorous tropical reef species that their faeces, poorly processed, convey significant nitrogen and phosphorous fluxes to the seafloor (Pinnegar and Polunin 2006; Schiettekatte et al. 2023).

This study reports for the first time a clear diel rhythm in a Mediterranean fish including sleeping behaviour. The diel rhythmicity of fish activities in relation to availability of food, salinity, temperature, light intensity, and risk of predation is well known in the Mediterranean basin and has been largely investigated mainly through fish catches to provide valuable ecological information implementing the management of commercial fish stocks (Carpentieri et al. 2005; Bussotti et al. 2018). In situ data, instead, are far rarer in this basin. Significant photoperiodic regulation of behaviour was detected, for example, for D. dentex in a long-term monitoring at an artificial reef in Mediterranean waters, but no sleep was observed (Francescangeli et al. 2022). Different occurrences of fish species suggesting diurnal and nocturnal habits have been pointed out also in a recent lander monitoring over two colonies of the scleractinian Dendrophyllia ramea (Linnaeus 1758) in mesophotic Sicilian waters (Salvati et al. 2023). Changes in fish behaviour and movement associated with the diel rhythm, instead, have been largely documented in temperate and tropical reef environments (e.g., Hobson 1972, 1975; Reebs 2002; Rickel and Genin 2005; Harvey et al. 2012; Myers et al. 2016). It has been reported that diurnal planktivorous reef fishes, due to a limited sight acuity in twilight conditions, seek refuge 30 min before to 15 min after sunset, in the so-called quiet period, and hide in cavities alone during the night (Hobson 1975). This behaviour was never reported in a Mediterranean fish, while it perfectly fits what was observed for A. anthias. Interestingly, the tropical fish Pseudanthias squamipinnis (Peters 1855), belonging to the same family of A. anthias, displays a similar diel pattern, living permanently associated with a specific reef zone, feeding during the day near the canopy and seeking refuge at night in crevices (Popper and Fishelson 1973). This supports a phylogenetic base for this behaviour.

It seems clear that the rhythmicity of A. anthias is regulated by light, showing a clear fluctuation despite being largely attenuated at mesophotic depths (Fig. 10). The contact with the canopy starts at dusk, after 19:00, with a period analogous to the tropical quiet period, and rest usually does not take place until 22:00 (Figs. 8f and 10). Fish are observed moving in the canopy, and the aggregation is scattered between the coral ramifications. Rest usually ends after 05:00, at dawn, when fish are seen again within the canopy forming the school and ascending in the water column after 06:00 (Figs. 8f and 10). The Cosinor models support the influence of at least two other environmental variables (Table 2) on the daily rhythm of A. anthias, specifically altering the time spent in the water column. Turbid plumes, associated with dim ambient light, induce individuals to move towards the canopy, while strong currents increase the time spent high above the canopy in the water column, feeding. Similarly, it has been observed that the shallow-water species D. dentex significantly responds to solar irradiance, water temperature, and wind speed (Francescangeli et al. 2022). In situ environmental measurements of light, current, and primary production might provide additional information on the specific constraints occurring within mesophotic habitats.

Fig. 10
figure 10

Schematic representation of the diel cycle of Anthias anthias in the black coral forest

Full size image

Finally, an important aspect that emerged from this study is the sleep in A. anthias. Sleep in fishes is a known and well-investigated topic (Meddis 1975, 2017; Helfman 1986; Reebs 1992, 2002). It can manifest through different behavioural patterns. Although there are many exceptions, usually, it can be recognised by prolonged quiescence in a typical rest posture, use of a shelter and high arousal thresholds (Reebs 1992). It is usually regulated by light with a circadian rhythm under the effect of melatonin (Hur et al. 2012). With respect to what is typically reported in the literature, A. anthias clearly shows two of the three parameters of sleep, namely prolonged immobility and limited sensibility to disturbances during sleep. Anthias anthias displays a very regular diel sleeping pattern, making contact with the forest (between dusk, night and dawn) for up to eight hours, of which, on average, nearly six are dedicated to sleep on the seafloor (Fig. 8f).

The forest canopy may be considered a general shelter for this fish, but no physical refuge, such as a crevice, is involved during sleep. Crevices are indeed known to be used by A. anthias when they rapidly seek refuge during the day (Louisy 2015), but not at night, unless maybe on rocky shoals without the coral canopy. Despite not having proper shelters, the observed sleeping individuals always selected three specific resting sites, which were maintained throughout the entire survey. Based on the analysed images, in A. anthias, not only the site is maintained, but also the position of the fish, even after a disturbance, maybe reflecting the best escape route. The same behaviour was also reported in aquaria (Fig. 9g). The three Spicara smaris, observed on three occasions on the seafloor with a night colouration, might indicate that also this species may exploit the shelter offered by the canopy, even if with less fidelity and no immobility detected. In any case, Spicara with a pyjama colouration on the seafloor was reported only for Spicara maena (Linnaeus 1758) (Fischer et al. 2007).

The continuity of sleep of A. anthias can be interrupted, generally one to two times per night, with short awakenings, even if, in 30% of the observations, no movements were detected. This situation suggests that this fish has high arousal thresholds triggering the awakenings, generally associated with the close passage of large patrolling predators of small benthic fishes, while predators of benthic invertebrates (Diplodus sargus and Mullus surmuletus) and the plankton feeder Spicara smaris were almost never actively causing alert responses when seen on the seafloor at night. Piscivorous mainly include Phycis phycis, known to be a predator of A. anthias (Morato et al. 1999). It is plausible that this predator adapted to the life cycle of A. anthias by preferring to move on the seafloor of the forest at night when its preys are more vulnerable because they are asleep. Nevertheless, no active predation event was documented. The passage of non-piscivorous fishes and large benthic invertebrates (such as Peltaster placenta) did not trigger any response, even if observations in this regard are too limited. A certain number of awakenings was attributed to no specific cause, and on some nights, a very limited overall sleep was detected for two individuals. While other biological causes not detected by the lander could be involved, an in-depth analysis of the in-situ environmental parameters may provide additional evidence. On one occasion, strong water turbulence could be considered a plausible trigger. A certain degree of interspecific variability was detected among the three sleeping individuals, with one (n° 2) consistently sleeping more and displaying no reactions to the triggers at times. The occurrence of a physical shelter near its resting site (the large sessile ascidian Halocynthia papillosa (Linnaeus 1767) may have provided a better refuge, avoiding awakenings (Fig. 8b-c). Variability may also reflect different arousal thresholds, hence individual reactions to risk, with the three individuals probably always being the same ones. No phenomena of interspecific competition for the resting spots were observed in A. anthias. At the same time, this has been occasionally recorded for species using a physical shelter and was generally attributed to the necessity to quickly take refuge in case of threat (Robertson and Sheldon 1979).

In conclusion, this work provided additional evidence of the importance of unbaited landers and other similar platforms in the investigation of fish behavioural patterns (Aguzzi et al. 2012, 2013, 2015, 2020a; Francescangeli et al. 2022; Grinyó et al. 2023). Behavioural traits have been proven to contribute substantially to the interpretation of connectivity data in deep-sea fishes (Aguzzi et al. 2010, 2011), and, for certain species of commercial interest, they may also be used to elaborate better conservation strategies for fish stocks (Aguzzi et al. 2020b; Grinyó et al. 2023). Overall, the lander proved to be an effective tool to characterise short-term habits and daily dynamics based on the frequency of occurrence. At the same time, the lack of spatial or temporal replicas is a limitation, and the counts of individuals can be partially biased when fishes are moving in and out of the visual field, far in the background, at night, or during particularly intense turbid plumes, influencing the perception of residency in the area. Such limitations have also been reported in previous monitoring surveys using cabled observatories (Francescangeli et al. 2022). The disturbance caused by the presence of the lander, the use and intensity of lights (especially at night) and the resolution of the images may also have a partial influence on the data collection. Nevertheless, the study provided a good baseline for similar investigations that want to better comprehend the relationships between mobile species and habitats dominated by structuring species and the way that environmental constraints beat the life of the organisms that inhabit the mesophotic and deep zones, providing a dynamic, rather than static, view of biodiversity. This information is relevant to highlight the ecological importance of coral forests and support specific conservation measures.