This study was conducted in Menorca (Balearic Islands, North-western Mediterranean). Menorca became a Biosphere Reserve by UNESCO in 1993 for its natural landscape and great biodiversity (Menorca Biosphere Reserve, 2020). The coastline of Menorca is relatively well preserved with large terrestrial and marine protected areas (Comas, 2004). Most of the perimeter of the island is fringed by seagrass meadows where P. oceanica and C. nodosa species coexist in very shallow waters, and P. oceanica thrive in deeper waters down to 35 m. The northern coastline is very irregular with steep cliffs and large sheltered bays, whereas the southern coastline is straight with rocky cliffs and narrow coves (Llompart and Rosell, 2002). Sampling on the northern side of the island took place in shallow waters (between 1 to 7 m) within small, sheltered coves, whereas on the southern side sampling was performed at deeper depths (between 15 to 18 m).

A total of five study sites were sampled, with three sites (Sanitja, S’Estany and Sa Torreta) located on the northern side of the island, and two sites (Cala Galdana and Son Bou) located on the southern side of the island ( Figure 1 ). Within each study site, fish surveys and structural complexity were assessed in each habitat type: seagrass meadow, seagrass escarpment, sand, and rocky substrates, except for Sanitja site where suitable rocky substrates and bare sand habitat were not found. Escarpments sitting in front or on top of visible rock were not sampled.

We performed visual fish surveys using SCUBA in May 2022 following standard procedures (Harmelin-Vivien et al., 1985; Giakoumi et al., 2012). All fish data were collected by E.B. and E.G. For seagrass meadows, rocky substrates and bare sand within each site, at least three transects of 50 m were conducted, with a total 23 transects in P. oceanica seagrass meadows, 12 transects in bare sand, and 12 transects in rocky substrates. The transect length in seagrass escarpments varied from 10 to 50 m owing to the scarcity of long escarpments in some study sites, with a total of 18 transects in P. oceanica seagrass escarpments ( Supplementary Table 1 ). We excluded bare sand habitat from further assessments because fish counts equaled zero in all 12 transects, which in part could be linked to the lack shelter in bare sand that results in the fishes remaining beyond the observer sight.

At each transect, the diver laid out a 50 m line-transect and simultaneously identified, counted, and sized all fish (at 5 cm intervals) (La Mesa and Vacchi, 1999). The observer swam at a constant speed (~10 min/50 m transect) and all fish passing within 2 m from each side of the transect in seagrass meadows and rocky substrates were recorded, whereas at seagrass escarpments fish were recorded within 1.5 m from the escarpment wall. Transects were randomly positioned in seagrass meadows and rocky bottoms but followed the escarpment contour in the escarpments. All censuses were conducted between 10 am to 4 pm to minimize time effects on fish behavior and fish counts.

We assigned fish species to a trophic category based on diet (Moranta et al., 2006) and trophic level (Pauly and Froese, 2006): herbivores, macrophagous, mesophagous, microphagous, and omnivores ( Supplementary Table 2 ). The biomass of each fish species was estimated using allometric length–weight relationship based on FishBase (Pauly and Froese, 2006) and Morey et al. (2003): W = a * (TL^b), where parameters a and b are species-specific constants, TL is total length in cm, and W is weight in grams. Parameters a and b were selected from the study conducted in the nearest site to our study area (Morey et al., 2003). Fish abundance and total fish biomass were calculated for each transect and trophic category.

For each transect within each habitat where fish surveys were conducted, the seafloor structural complexity was estimated. The chain rugosity method (Hill and Wilkinson, 2004; Guidetti et al., 2016); was used for seagrass meadows, sand and rocky habitats, whereas the height and cavity volume were measured along the escarpments to estimate the habitat complexity. A 30 m chain was deployed three times along each transect to measure structural complexity, by deploying the chain carefully over the undulating substrate in a straight line and registering the real lineal length used. Chain surface rugosity index was calculated by dividing the 30 m real chain length by the linear distance covered (Friedman et al., 2012). In seagrass escarpment habitats, the escarpment height was measured every 5 m, together with the width, height and depth of all cavities and ledges with an opening greater than 20 cm in diameter. The volume of cavities was estimated using the formula of a pyramid (V = (1/3) * w * d * h; where w is width, d is depth and h is height), whereas the volume of ledges was estimated using the formula of a rectangular prism (V = w * h * l; where w is width, h is height and l is length). The structural complexity in seagrass escarpments was estimated using an area-based surface rugosity index, by dividing the three-dimensional topography by the projected planar area (Friedman et al., 2012). The three-dimensional topography is the result of the sum of the area of the escarpment (i.e., transect length multiplied by the average escarpment height), plus the area of cavities and ledges. The projected planar area was estimated as the area of the escarpment (i.e., transect length multiplied by the average escarpment height).

Differences in fish assemblage assessments between the two observers were discarded based on the lack of significant differences in nine duplicate fish surveys (t-test; P = 0.064). Data on fish abundance, total number of species, fish biomass and Shannon´s diversity Index (H’) were standardized per unit area (m²). Differences on fish abundance (number of fish m^-2), total number of species (number of species m^-2), total biomass (g m^-2), Shannon´s diversity Index (H’), biomass of fish size categories (at 5 cm intervals), trophic categories (macrophagous, mesophagous, microphagous, omnivorous and herbivorous), and structural complexity among habitats were assessed using univariate non-parametric analysis of variance (PERMANOVA), which allows heteroscedasticity of samples and non-normality of the residuals. Data from each trophic category and fish size were square root transformed, and the resemblance matrix for each variable was computed using Euclidean distance resemblance matrix. Sum of squares type III (partial) was applied after 9999 permutations of residuals under a reduced model. Any significant factors were analyzed using post-hoc pair-wise tests. In the experimental design, study site was considered as a random factor and habitat type (n = 3; seagrass meadow, seagrass escarpment, rocky substrate) was considered a fixed factor.

Differences in fish species, trophic categories, and fish size categories among habitats were also assessed using multivariate PERMANOVA. Data were square root transformed to down-weight the influence of highly abundant species, and the Bray-Curtis similarity matrix was used. PERMANOVA post-hoc pair-wise tests were conducted on significant factors. To visualize the role of fish species and trophic category in driving differences in fish biomass among habitats, Principal Coordinate Analyses (PCO) were computed (Anderson et al., 2009). Pearson correlation was used to assess potential relationships between habitat structural complexity, fish abundance, fish biomass and Shannon´s diversity Index among habitats. All statistical analyses were performed using PRIMER v6 statistical software with add-in PERMANOVA (Anderson et al., 2009).

Significant effects of the factor habitat type (seagrass meadow, seagrass escarpment, rocky substrate) on fish abundance, total number of species, total fish biomass, and Shannon´s diversity Index were detected (P < 0.001; Supplementary Table 3 ). Total fish abundance was similar between rocky substrates (mean ± SE; 0.65 ± 0.07 fish m^-2) and escarpments (0.78 ± 0.19 fish m^-2; P = 0.739), but up to 5-fold higher than in seagrass meadows (0.10 ± 0.02 fish m^-2; P < 0.001; Figure 2A ). Similarly, total fish biomass was comparable between rocky substrates (11.66 ± 3.03 g m^-2) and seagrass escarpments (5.17 ± 1.28 g m^-2; P = 0.059), but 18-fold higher than in seagrass meadows (0.64 ± 0.15 g m^-2; P < 0.001; Figure 2B ). On the other hand, total number of fish species and Shannon´s diversity Index were significantly higher in rocky substrates (0.053 ± 0.003 species m^-2; H’ = 1.73 ± 0.06, respectively) than in seagrass escarpments (0.024 ± 0.002 species m^-2; H’ = 1.08 ± 0.07, respectively; Figure 2C ). In addition, total number of species and Shannon´s diversity Index in seagrass escarpments were significantly higher than in seagrass meadows (0.016 ± 0.001 species m^-2; H’ = 0.82 ± 0.06, respectively; Figure 2D ).

Significant differences in fish biomass among habitats were observed for all size categories (P < 0.001; Supplementary Table 3 ). Fish biomasses of small size categories 0–5, 5–10 and 10–15 cm were significantly higher in seagrass escarpments (0.05 ± 0.02, 2.54 ± 0.72 and 1.41 ± 0.37 g m^-2, respectively) and rocky substrates (0.05 ± 0.02, 1.39 ± 0.13 and 2.32 ± 0.36 g m^-2, respectively) compared to seagrass meadows (0.006 ± 0.002, 0.23 ± 0.07 and 0.33 ± 0.13 g m^-2, respectively). On the other hand, fish biomass of size category 15–20 cm was significantly higher in rocky substrates (2.17 ± 0.71 g m^-2) compared to seagrass meadows and escarpments (0.07 ± 0.03 and 0.59 ± 0.33 g m^-2, respectively). Fish biomass of size category 20–25 cm was significantly higher in rocky substrates (2.81 ± 1.43 g m^-2) than in seagrass escarpments (0.13 ± 0.10 g m^-2), whereas biomass of fish size 25–30 cm was similar between rocky substrates and seagrass escarpments (1.74 ± 1.32 and 0.46 ± 0.35 g m^-2; respectively; P = 0.238). Fish larger than 20 cm in size were absent in seagrass meadows ( Supplementary Table 4 ; Figure 3A ).

Significant differences in fish biomass among habitats were observed for each trophic category (P < 0.003; Supplementary Table 3 ). Macrophagous and microphagous fish biomasses were significantly higher in seagrass escarpments (1.37 ± 0.44 and 1.76 ± 0.72 g m^-2, respectively) and rocky substrates (0.55 ± 0.19 and 0.64 ± 0.21 g m^-2, respectively) compared to seagrass meadows (0.04 ± 0.02 and 0.12 ± 0.06 g m^-2, respectively). On the other hand, omnivorous and herbivorous biomasses were significantly higher in rocky substrates (2.65 ± 0.84 and 3.61 ± 1.71 g m^-2, respectively) compared to seagrass escarpments (0.17 ± 0.08 g m^-2 omnivorous biomass whereas herbivorous fish were absent) and seagrass meadows (0.10 ± 0.08 and 0.14 ± 0.13 g m^-2, respectively). Mesophagous fish biomass was significantly lower in seagrass meadows (0.23 ± 0.06 g m^-2) than in seagrass escarpments (1.87 ± 0.52 g m^-2), and significantly higher in rocky substrates (3.04 ± 0.44 g m^-2) than in seagrass meadows and escarpments ( Supplementary Table 4 ; Figure 3B ). Two principal components of the PCO showing differences in fish biomass across habitat types according to trophic categories explained 59% of the total variance ( Figure 4A ). The first component (PCO1; 39% of the variance) differentiated seagrass escarpments and rocky substrates from seagrass meadows based on the lower fish biomasses across all trophic categories in the latter. PCO2 (20% of the variance) mostly differentiated seagrass escarpments from rocky substrates based on the larger abundance of herbivorous and omnivorous fish in the latter.

The PCO based on the biomass of fish species explained 39% of the total variance ( Figure 4B ). The first component (PCO1; 25% of the variance) differentiated seagrass meadows from both seagrass escarpments and rocky substrates based on the higher abundance of Serranus scriba, Coris julis, Symphodus tinca and Chromis chromis in rocky substrates, and the higher abundance of Spondyliosoma cantharus and Diplodus annularis in seagrass meadows. PCO2 (14% of the variance) differentiated seagrass escarpments from rocky substrates based on the greater abundance of S. scriba and C. julis in seagrass escarpments, and the higher abundance of Diplodus vulgaris and S. tinca in rocky substrates.

A total of 169 cavities were measured across all seagrass escarpments in order assess habitat structural complexity. Structural complexity was different among habitats (P (perm) < 0.001, Supplementary Table 3 ), with significantly higher values in rocky substrates than in seagrass meadows and escarpments (8.3 ± 1.7, 13.1 ± 3.3, 19.7 ± 1.7, respectively). Weak positive relationships between habitat structural complexity and Shannon´s diversity Index, total number of fish species and total fish biomass across all habitats were observed (r = 0.46, 0.44 and 0.38, respectively; Figure 5 ). The abundance of fish was not correlated with habitat structural complexity (r = 0.20).

The high productivity of P. oceanica meadows together with their capacity to trap particles contributes to the accretion of carbon-rich soils over centuries to millennia at an average rate of 2.1 mm yr^-1 (Fourqurean et al., 2012; Serrano et al., 2016), resulting in the formation of the matte (Bonhomme et al., 2015). The presence of vertical cuts within P. oceanica meadows can originate from either natural or anthropogenic factors, but it is overall rare. The dimensions of the escarpments surveyed had an average length of 37 ± 1.4 m, 1.4 ± 0.2 m in height, and 0.6 ± 0.1 cavities per m² of escarpment. Overall, Posidonia escarpments are barrier reefs that resemble a rocky habitat in aspect and constitute unique formations, some of which have been recognized as a protected National Monument under the IUCN (Blas Aritio, 1990).

The cliffs along a coastline are structurally very similar to seagrass escarpments and thereby, it is intuited that the formation of escarpments is analogous to that of vertical cliffs. The erosion of the escarpment, owing to wave and current action, likely erodes the base of the matte creating a ledge where the upper portion overhangs from the wall. Then after, the overhang collapses resulting on the retreat of the escarpment over variable time scales. Field observations of ledges towards the base of the escarpment and collapsed fragments of matte (up to 1 m³ in size) in front of the escarpments support this hypothesis ( Figures 6A, B ).

The geomorphology of seagrass escarpments differed among sites, likely according to the interaction of physical variables (e.g., water depth, exposure to hydrodynamic forces and dominant winds, water flow, and terrestrial inputs), biological variables (e.g., seagrass primary production and meadow structure) and anthropogenic disturbances. Typically, the deep escarpments surveyed were more scarce, shorter, and narrower than shallower escarpments ( Figures 6C, D ). It seems plausible that the relatively higher seagrass productivity, biomass, and soil accumulation rates in shallow meadows results in higher soil accretion and matte build up (Serrano et al., 2016), thereby likely allowing the formation of higher escarpments than in deeper areas. All escarpments observed, except those located in waters deeper than 12 m, were oriented in parallel to the coast, which could indicate a natural mechanism of formation linked to hydrodynamic forces. Shallow seagrass escarpments likely form by either i) exposure to continuous hydrodynamic forces that results in the opening of channels across the meadows to facilitate water outflow (Bonhomme et al., 2015); ii) extreme storm events that rework sediment and dislodge portions of the matte (Gera et al., 2014; Oprandi et al., 2020); iii) anchor damage followed by erosion of the edges of the meadow; or iv) dredging to open a channel into shallower areas. Indeed, hydrodynamic energy and anchoring are typically more intense in shallow environments making shallow meadows more susceptible to erosion. This hypothesis is supported by multiple observations of large portions of matte detached from the escarpment wall in shallow settings compared to the scarcity of detached matte in deeper settings, which could also indicate faster rates of formation and evolution of escarpments in the former. At the deeper escarpments surveyed, in addition to occasional extreme hydrodynamic events (e.g., storm surges), it seems plausible that anchors deployed by big boats opened a scar within the continuous meadow (typically found at depths >10 m), favoring the expansion of the scar, resulting in the large and linear escarpments observed. It may also seem plausible that fish could actively erode the escarpment when digging to create shelters, as previously observed following deep sea oil and gas piping deployment (Twichell et al., 1985). This hypothesis is supported by observations during field surveys of species from genus Gobius and Symphodus eroding the escarpment wall to maintain or create shelter.

Erosion likely exposed the ancient matte along the edges of the meadows at S’Estany and Sa Torreta, whereas the escarpments at Sanitja likely originated from channels carved in the meadow for boats to enter the coastal settlement during Roman times (Contreras et al., 2006). However, establishing a cause-effect relationship around the origin or evolution of escarpments is difficult owing to the multiple climatic and anthropogenic processes that occurred since millennia, which further accelerated and intensified during the last decades in coastal areas. Physical and geomorphological variables are site-specific and multiple biotic and abiotic factors may have interacted over millennia to shape the currently observed geomorphology of escarpments (Bonhomme et al., 2015).

The walls of the escarpments were colonized by organisms that typically inhabit rocky substrates, and the taxa and their abundance seemed to vary according to differences in irradiance, escarpment slope, rhizome density and orientation. Escarpments located in shaded areas (e.g., north-oriented) with vertical profiles or overhangs were colonized by sciaphilic algae, including Flabellia petiolata and Peyssonnelia squamaria. On the other hand, those exposed to higher irradiance were colonized by photophilic algae including Halopteris scoparia and Dictyota spp., and also by invasive species such as Caulerpa cylindracea and Lophocladia trichoclados. Escarpments were also colonized by other organisms typically found in rocky substrates including ascidians (e.g., Halocynthia papillosa) and bryozoans (e.g., Schizoporella errata). The colonization of living and dead P. oceanica rhizomes by fauna and flora typically inhabiting rocky substrates has been previously reported (Boudouresque, 1974; Ballesteros, 1992; Mabrouk et al., 2014). However, the intrinsic characteristics of escarpments, which differ from seagrass canopies, requires further studies to understand the distribution of fauna and flora assemblages according to differences in escarpment geomorphology.

Posidonia oceanica seagrass meadows are a sink for undiscovered ancient shipwrecks and other archaeological items (Krause-Jensen et al., 2019). The vertical accretion of the meadow can cover entire wrecks after sinking thousands of years ago, and the visual assessment of seagrass escarpments could be used as a preliminary step towards identifying archaeological remains and reveal the presence of cultural heritage. During field work we identified portions of ceramic amphoras embedded in the escarpment wall and other ceramic remains at the blowout near the escarpment, which seem to belong to Roman times (~100 BC; amphora Dressel 1 type, based on Dressel (1899) classification) and showcases the extraordinary capacity of seagrass meadows to colonize new substrates and to conceal ancient history ( Figure 6E ).

The choice of habitat by fish species typically responds to specific shelter and food requirements, and seagrasses are well known to fulfil both needs for a number of species (Rozas and Odum, 1988). The results obtained suggest that the higher structural complexity provided by P. oceanica escarpments enhances fish abundance, number of species diversity and biomass compared to homogeneous seagrass meadows. The presence of cavities within the vertical walls provides additional shelter to fish assemblages compared to seagrass canopies and resembling rocky substrates. However, the relatively higher number of species, diversity and size of fish found in rocky substrates compared to seagrass escarpments could be explained by the overall higher structural complexity of rocky habitats, with larger number and size of cavities that can provide shelter to relatively bigger fish (Ferreira et al., 2001). However, the extremely high variability in the structural complexity across the escarpments studied (ranging from 2 to 35%) points to the need to conduct further studies aiming at assessing how different escarpment types (e.g., composition, orientation, depth, fetch, and geomorphology) could affect fish assemblages. Fish diversity and abundance in escarpments seemed to be enhanced by the presence of big cavities and ledges, including overhanging and detached portions of matte, and by reduced incidence of irradiance (i.e., orientated to the north, in our case). However, it was not possible to test these hypotheses due to the heterogeneity of escarpment features along each transect and the unbalanced sampling design.

The presence of fish species linked to cavities from rocky substrates within seagrass escarpments, such as Apogon imberbis (Apogonidae) and Epinephelus marginatus (Serranidae) ( Figures 6F, G ), supports the hypothesis that escarpments provide suitable habitat for fish similarly to the one supplied by rocky substrates. For example, it seems plausible that A. imberbis, which typically congregates along reef caves and overhangs and has a nocturnal feeding habit (Corbera et al., 1996; Bussotti et al., 2018), uses the escarpments as habitat for residence during daytime and feed in the nearby P. oceanica meadows at night. Owing to the night feeding behavior of A. imberbis, it seems possible that some individuals were not counted because they were hiding inside the cavities of escarpments and rocky substrates. On the other hand, E. marginatus use caves as permanent places of residence but adult individuals are solitary and prefer deep rocky substrates compared to juveniles that are mainly found in shallow waters close to shore (Heemstra and Randall, 1993). E. marginatus is a targeted species for fishers and catalogued as vulnerable by the International Union for Conservation of Nature (IUCN) (Pollard et al., 2018); the location of escarpments within shallow and sheltered bays can constitute suitable nursery grounds for juvenile groupers (Harmelin and Harmelin-Vivien, 1999; Bodilis et al., 2003). These facts are also supported by the observation of multiple juvenile and adult E. coioides individuals hiding along P. australis escarpments in Australia (Serrano et al., 2017), a region that is less overfished than the Mediterranean Sea.

The large number of macrophagous fish found in seagrass escarpments (e.g., S. scriba; Figure 6H ) compared to the other habitats surveyed, suggests that escarpments constitute a key habitat for the shelter and likely reproduction and hunting ground for this species. The absence of herbivore fish within escarpments could be explained by limited food availability (i.e., seaweed) and by the risk of exposure to predators when feeding along the escarpments. Seagrass escarpments can be surrounded by either seagrass, sand, or rocky habitats, and the results obtained showed that fish species inhabiting fringing habitats could use the escarpment as a corridor among the heterogenous coastal seascape or even adjust their life cycle and use it as permanent residence. This is a novel hypothesis that brings seagrass meadows into scope for a new-fangled extension of the enhanced biodiversity ecosystem service.