The Corsican coast forms a unique spatial unit in the Mediterranean, consisting of extensive meadows of the seagrass P. oceanica, mostly spreading on sandy bottoms (Pasqualini et al., 1998; Telesca et al., 2015), and highly productive Cystoseira (sensu lato) forests inhabit rocky reefs down to 30 – 35 m depth (Sales and Ballesteros, 2010; Blanfuné et al., 2016). From 2019 to 2021, subtidal reefs were surveyed along the Corsican coast in and around three MPAs: the Marine Natural Park of Cap Corse and Agriate (MNP CCA) in the north, the Scandola Nature Reserve (SNR) in the west and the Bonifacio Straits Nature Reserve (BSNR) in the south ( Figure 1 ). In total, benthic communities were assessed at 59 sites during three research campaigns “Our planet reviewed” led by the French Muséum National d’Histoire Naturelle (8 – 23 May 2019, 13 – 30 October 2020, 14 – 27 May 2021). All sampling was conducted in late spring or fall, thus avoiding potential, biomass variation (algae), mortality, or recruitment peaks often observed in mid‐summer or mid‐winter. A relative balanced number of sites were surveyed within the three regions divided into shallow sites from 15 – 25 m depth and deeper sites from 25 – 35 m depth ( Figure 1 , Table 1 ). These depth ranges are characteristic of reefs dominated by infralittoral photophilic algae, reflecting the influence of light penetration in the water column. Shallower reefs, from 1 to 15 m depth, have been more extensively explored historically and were therefore the least explored during the 2019 – 2021 biodiversity surveys, thus limiting the depth range of the present study.

The filtering of reef taxa prior to field deployment followed a three-step selection process. First, a preliminary list of known marine benthic species was compiled based on the French National Inventory of Natural Heritage in the French Mediterranean (INPN) database (n~1000 sp.). The list was then cross-checked and reduced based on the characteristic species and determinant species for reef habitats listed in the French classification of Mediterranean benthic habitats for reefs (Michez et al., 2014; La Rivière et al., 2021) and the French ZNIEFF (Natural Zone of Interest for Ecology, Flora and Fauna; Lepareur et al., 2022) inventory species list, respectively (n~300 sp.). Finally, the list was streamlined to the finest taxonomic resolution possible according to species characteristics, expert feedback and field requirements (n=74 taxa). Indeed, taxa were selected on the basis of habitat association and depth range (reef-associated species for the 10 – 40 m depth range), species longevity and phenology (long-lived taxa with minimal seasonal variation), mobility (mostly sessile species), size (generally > 10 cm), ease of in situ identification (during dives), and ecological/structural importance as indicated by the scientific literature and from expert judgment (de Bettignies et al., 2020). The final list contains 74 taxa (61 species levels, 9 genus levels and 4 morphological groups) ( Supplementary Table S1 in the Supplementary Material ).

The VIT-Reef protocol was developed based on the Rapid Assessment Surveys (RAS; Ward and Larivière, 2004) protocol using a semi-quantitative approach adapted from the SACFOR scale (Mieszkowska et al., 2006; Lévêque et al., 2017), considering standard operating procedures for safety. For each site, a 30 m transect was deployed along a selected (relatively) homogeneous topographic reef and along bathymetric contours to maintain constant depth, avoiding extensive reef walls, overhangs and depth variations greater than 5 m. During transect deployment, the first diver deployed the tape while the second diver videotaped the topographic features 1.5 m above the measuring tape for future estimation of mean reef slope (i.e., Reef index). Once the transect was deployed, the two divers separately initiated a 25 min visual census of all listed taxa ( Supplementary Table S1 in the Supplementary Material ) along a 150 m² belt transect (30 m long × 5 m wide). While typical 0.25 m² photographic quadrats are often preferred for circalittoral coralligenous assemblages (Deter et al., 2012; Çinar et al., 2020), they can be adapted for low complexity infralittoral communities such as turf or encrusting algae-dominated barrens, as in the eastern Mediterranean (Tsirintanis et al., 2018; Nikolaou et al., 2023) but are not suitable for diverse multilayered communities such as those found throughout Corsica. Given the complexity and diversity of the Corsican reefs, the infralittoral communities are then better captured by a large belt transect with a comprehensive list of indicative species to be identified in situ, without the need for extensive image post-processing (reef coverage equivalent to 600 photographic quadrats of 0.25 m²). In addition, the area of the belt transect was chosen to balance the practical constraints of diving (air consumption and safety) with the need for a comprehensive reef coverage. To standardize the transect width (5 m), the length (m) was converted to the average number of fin kicks (per diver) along the transect line during practical dives, to allow for standardized exploration distance away from the transect line (2.5 m from each side). Additional practical taxa identification dives were conducted at the beginning of each research campaign prior to VIT-Reef deployment to ensure rapid, accurate, and consistent taxa identification among divers.

During VIT-Reef deployment, taxa abundance was assessed using a semi-quantitative scale (0 to 4), allowing a rapid in situ assessment of (CL) large countable taxa (> 10 cm), (CS) small countable taxa (< 10 cm), (SM) space-occupying taxa which are considered as meadow/crustose species and (ST) turf/understorey taxa (Mieszkowska et al., 2006, Table 2 ). Given the relative complexity of reef habitat in these regions, less common sciaphilous habitats (depending on photophilic conditions of transect deployment) such as walls, overhangs, and small fissures were still searched for listed taxa when encountered along the transect. However, abundance values for these taxa were minimal as horizontal and more photophilous conditions were preferred. After each dive, the two divers compared and merged their scores to minimize sampling biases between operators. In total, the abundance values of 49 countable taxa (dominated by faunal species, except for Codium bursa) and the coverage values of 25 space-occupying taxa (dominated by algal species, except for the encrusting sponges Crambe crambe and Phorbas tenacior) were estimated at 59 sites, including 33 sites at 15 – 25 m depth and 26 sites at 25 – 35 m depth ( Figure 1 ; Table 1 ). Special attention was given to the estimation of the cover of Caulerpa cylindracea, considering the seasonal variation of fronds and ramets and the different sampling times. Priority was then given to the percentage cover of the stolons and rhizoids network (more perennial structure). Prior to analysis, VIT-Reef scores were converted to semi-quantitative abundances as class range means. Two Hellinger transformed matrix species were then constructed for countable taxa (CL–CS, mostly fauna) and space-occupying taxa (SM–ST, mostly algae). Hellinger pretransformation of species data is required to reduce skewness and to avoid the use of double zeros (Borcard et al., 2018).

In this study, we collected 17 explanatory variables to identify the main environmental drivers of subtidal reef communities. From field surveys, we estimated the maximum depth of the sampling site (m), the semi-quantitative abundances of Cystoseira (sensu lato) (% of cover) and the nearby Posidonia oceanica reef meadows (% of cover), and the index of reef slope (Reef index; unitless). The Reef index was estimated from the field videos recorded on each VIT-Reef transect. The degree of reef slope was used to classify VIT-Reef transects into four slope categories (1: flat < 10°, 2: gentle slope 10° – 40°, 3: steep slope 40° – 80° and 4: wall > 80°). The reef slope index was then calculated as:

where score represents the slope category i (1 – 4) and cover, the corresponding percentage of transect cover of the four slope categories (%). Reef index ranges from 0.25 to 1.

Twelve physico-chemical variables were extracted from modelling rasters obtained from Bio-Oracle (Tyberghein et al., 2012; Assis et al., 2017) and Ocle (De la Hoz et al., 2018). Eleven physico-chemical variables were collected from the Bio-Oracle satellite data (surface layers only): minimum temperature (T_min; °C), maximum temperature (T_max; °C), temperature range (T_r; °C), minimum salinity (Sal; PSU), maximum nitrate concentration (NO₃; mol·m⁻³), phosphate concentration maximum (PO₄; mol·m−³), maximum dissolved oxygen concentration (DO; mol·m−³), maximum photosynthetically available radiation at the surface (PAR; Einstein·m−²·day−¹), maximum chlorophyll-a concentration (Chl-a; mg·m−³), maximum primary productivity (PP; g·m−³·day−¹), and maximum current velocity (Veloc; m−¹). These data were compiled from monthly in situ measurements recorded between 2000 and 2014, previously assigned to a grid resolution of 5 arcmin (Tyberghein et al., 2012; Assis et al., 2017). The significant wave height variable (SWH; m) was collected from the Ocle satellite data (De la Hoz et al., 2018). SWH corresponds to the average height of the highest wave measured across Europe during the winter period of 2000–2015. The SWH data were transformed to match the same resolution as the Bio-Oracle variables (5 arcmin and monthly measurements). We supplemented the dataset of explanatory variables with an estimated index of the anthropogenic pressure (Impact; unitless) extracted from the IMPACT model of Andromède (Holon et al., 2015; 2018). The IMPACT index combined global anthropogenic pressures from coastal engineering, anchoring, urban runoff, land cover, coastal population and fisheries and assigned a score in each location (Holon et al., 2015, 2018). The IMPACT model considered both the distance from the sources of disturbance and the bathymetry. Our use of historical Bio-Oracle and Ocle data was driven by their ease of access and consistent spatial coverage, which minimizes the noise of short-term variability and ensures comparability across sites. Given that Mediterranean benthic communities respond primarily to long-term environmental conditions, we assumed that multi-annual climatologies would be well suited to identify the main spatial drivers of community structure.

To understand the main environmental drivers shaping the community composition of CL–CS and SM–ST, results from RDA and LASSO models were examined.

Preliminary analyses of the two depth ranges (15 – 25 m vs. 25 – 35 m) provided insights into the vertical zonation patterns for both algal and fauna taxa, in particular the depth extent of infralittoral communities dominated by photophilic algae and the emergence of pre-coralligenous communities, indicating a shift towards circalittoral communities. Such a pronounced vertical zonation has already been demonstrated for the Mediterranean Sea, despite its reduced tidal range (Ballesteros, 1992), and has rather been explained by the interactive effects of light, hydrodynamics and nutrients (Zabala and Ballesteros, 1989; Ballesteros, 1993), but the literature on this topic remained scarce. For the Corsican reefs studied, the most structuring species driving this vertical zonation (and hence differences between sites) were species or groups of species well known for their light affinities. In summary, shallow sites (15 – 25 m) were characterized by a high cover of turf-forming algal species (Halopteris spp. and Dictyota spp.) and a high number of long-lived and photophilic species such as the green balloon-like macroalga Codium bursa (life span > 15 years; Vidondo and Duarte, 1998), a solitary Mediterranean coral Balanophyllia europaea (life span > 20 years; Goffredo et al., 2004) and grey-ball sponges (Ircinia variabilis most likely). Cystoseira (sensu lato) forest did not explain such depth zonation (given its taxonomic resolution in this study), as it was found across the entire depth range. On the other hand, deep sites (25 – 35 m) attested the intrusion of typical coralligenous species (Ballesteros, 2006) into the infralittoral communities, with Eunicella cavolini being the most abundant gorgonian species, Axinella verrucosa a circalittoral-like erect sponge and an increasing cover of encrusting corallines. While the vertical zonation pattern was evident here, the transition to coralligenous communities appeared gradual as evidenced by a mix of infralittoral and circalittoral species in the 25 – 35 m depth range. This transitional area can thus be referred as pre-coralligenous communities. In the western Mediterranean, the 25 – 35 m depth range is considered shallow for coralligenous reefs, where they begin to develop mostly on vertical walls and overhangs and are referred to as “cliff morphology” (~ 20 – 50 m) (Bracchi et al., 2017; Piazzi et al., 2022). Further down (usually below 50 m), the “platform morphology” of coralligenous reefs develops horizontally (Ballesteros, 2006; Piazzi et al., 2022). Further analyses were then replicated at each depth range to avoid the confounding effect of depth ranges when discussing spatial structure, given the transition in reef communities.

Interestingly, most deep sites (25 – 35 m) were heavily colonized by Caulerpa cylindracea, a non-indigenous species endemic to southwestern Australia with temperate warm to pan-tropical affinities (Verlaque et al., 2003) and reported as the most serious invader in the Mediterranean (Piazzi et al., 2005; Montefalcone et al., 2015). Since its first report in Libya in 1990 (Nizamuddin, 1991) and according to its spread kinetics (Verlaque et al., 2000; Piazzi et al., 2005), C. cylindracea has probably affected all the Mediterranean coasts on all types of substrates and communities from 0 to at least 50 m depth, indicating a wide ecological niche (Verlaque et al., 2000; Piazzi et al., 2005). While algal turf has been proposed as a facilitating factor for the establishment of C. cylindracea on Mediterranean reefs (Piazzi et al., 2003; Bulleri and Benedetti-Cecchi, 2008), here the opposite pattern was observed with the spread of C. cylindracea and its co-occurrence with encrusting corallines in deep sites (25 – 35 m), opposite to algal turf depths.

We hypothesized that the facilitation provided by algal turf was overcome by biotic inhibitors to the establishment and/or spread of C. cylindracea in the 15 – 25 m depth range (infralittoral). In particular, the presence of a dense Cystoseira (sensu lato) canopy has been proposed as the main limiting factor for C. cylindracea invasion (Bulleri et al., 2010, 2011), but the exact mechanisms remain unresolved (Piazzi et al., 2016). Extensive C. cylindracea mats were repeatedly observed in the coralligenous habitat (between 40 and 65 m depth), often covering more than 50% (Le Gall L., pers. com.). This suggested that coralligenous communities may be more vulnerable to C. cylindracea invasion than infralittoral communities, which are protected by a dense and more resistant macroalgal canopy. Moreover, the circalittoral habitat may serve as an important source of vegetative dispersal for C. cylindracea to nearby infralittoral communities. If a shift from macroalgal canopy to algal turf occurs (as observed in disturbed Mediterranean reefs), the recipient habitat may become more susceptible to successful establishment of C. cylindracea. In this context, C. cylindracea can be considered as a “passenger” rather than a “driver” of habitat degradation (Bulleri et al., 2010). However, once established, it has the potential to become a “driver” of change, preventing the recovery of native macroalgal canopy species (once disturbed) by facilitating turf growth, which in turn increases sediment deposition (Bulleri et al., 2010).

Covariations (in the same or opposite direction) among structuring taxa provided relevant ecological information for each depth range. At shallow depths (15 – 25 m), Codium bursa and Petrosia ficormis co-varied and were predominantly found in lighted conditions. Both species are known to reach large sizes only under sufficient light exposure. Indeed, the photosynthetic capacity of C. bursa decreases with increasing size (Geertz-Hansen et al., 1994) and the trophic source from the cyanobacterial symbiotic association (symbiocortex) of P. ficiformis contributes to the increase in size with light conditions (Bavestrello et al., 1994). At deeper depths (25 – 35 m), the two gorgonian species Eunicella cavolini and Paramuricea clavata covaried, as expected, as we usually observe them in sympatric populations (Ponti et al., 2014; Valisano et al., 2019). Both species are characteristic of coralligenous communities in the northwestern Mediterranean where they increase habitat complexity (Harmelin and Marinopoulos, 1994; Linares et al., 2008).

Most surprisingly, the alien green alga Caulerpa cylindracea and the fleshy red alga Phyllophora crispa were both found in high abundance at deeper sites. The regression analysis further supported a strong association between these two species, with P. crispa emerging as the only significant predictor of C. cylindracea abundance in the continuous model. The refitted linear regression confirmed this positive relationship, suggesting that P. crispa may play a facilitating role in the establishment of C. cylindracea. Dense mats of P. crispa may create structural conditions that facilitate the colonization and persistence of C. cylindracea, as has been observed in other species (Bonifazi et al., 2017). The architectural complexity of P. crispa mats (Bonifazi et al., 2017) increases microhabitat heterogeneity, potentially providing more anchoring points for C. cylindracea and promoting sediment accumulation. This, in turn, could facilitate stolon elongation and the lateral spread of C. cylindracea (Bulleri et al., 2010; Piazzi et al., 2016). While our binary LASSO model initially identified additional potential species associations, none remained statistically significant after refitting the logistic regression, underscoring the particularly strong association between P. crispa and C. cylindracea abundance rather than mere presence. Across both depth ranges (15 – 25 m vs. 25 – 35 m), Cystoseira (sensu lato) forests showed opposite trends compared to the shrubs of the encrusting red algae Peyssonnelia spp. The later thrive only in low light conditions, typically found on overhangs, vertical walls, crevices and caves in the infralittoral zone; and is one of the most representative genera of coralligenous algae in the circalittoral (Barberá et al., 2012). On the contrary, extensive Cystoseira (sensu lato) forest are usually found in high light conditions on low-sloping reefs. Such changes in community structure were indicative of differences in both light conditions and/or reef substrate slope for all sites, regardless of depth range.

Overall, photophilic and sciaphilic species drove variation in spatial structure at shallow and deep depths ranges (15 – 25 m vs. 25 – 35 m) respectively, which the algal community being a key explanatory factor. Co-inertia analyses showed a strong correlation between the countable fauna and space-occupying algae matrices (m² > 0.46). These results strongly suggested a bottom-up control of benthic primary producers towards higher trophic levels, probably not through trophic dependence, as most of fauna taxa listed in this study are sessile filter-feeders, but rather through competition for space and/or habitat provisioning. Further down in the circalittoral zone, the competition release with algal forests allows the development of complex and well-structured faunal communities (mainly composed of sponges, gorgonians, corals, and bryozoans), also called “animal forests”, which will then play a key role as ecosystem engineers (Rossi, 2013).

Given the different sampling times, a potential confounding effect of time (or season) may have been introduced into our study, weakening further comparisons of spatial patterns between regions. However, we believe that our results are not significantly biased by time. In fact, the VIT-Reef protocol minimizes such temporal bias by prioritizing long-lived and sessile taxa with minimal seasonal variation, avoiding extreme seasons (i.e., mortality or recruitment peaks observed in mid-summer or mid-winter), paying special attention to estimating Caulerpa cylindracea cover from stolon/rhizoid rather than frond/ramet, and buffering moderate fluctuations in taxa abundance and cover thanks to the coarse categorical approach of the semi-quantitative method.

Reef communities showed geographical contrasts along the northern, southern and western Corsican MPA regions, but these patterns varied with both depth ranges and communities studied (fauna vs. algae). The northern and western MPA regions showed similar fauna communities across both depth ranges but contrasted with the southern region. In particular, most southern sites showed higher abundance of specific structuring species, i.e., the scleractinian Balanophyllia europaea in shallow sites and the two sponge species Acanthella acuta and Axinella verrucosa in deeper sites. In southern shallow sites, the shift from erect to crustose coralline algae may in turn have favored the photophilic solitary scleractinian, Balanophyllia europaea (Goffredo et al., 2004), probably by releasing competition for space and/or improving substrate under light exposure. In addition, extensive development of Posidonia oceanica on southern coastal reefs (over 20% coverage in 60% of sites) resulted in limited areas (mostly rocky outcrops of large boulders and overhangs tops) for large erect algae, dominated instead by coralline, probably due to higher disturbance regime. The combination of these biotic and abiotic conditions appeared to be particularly favorable for the colonization of B. europaea, as evidenced by its high numbers in these reef areas. On the contrary, many southern deep sites displayed significant structural difference compared to northern and western sites, due to the high abundance of the two sciaphilic sponges, A. acuta, A. verrucosa (Santín et al., 2019). Low light conditions may then be more prevalent for southern deep reef communities, given the complexity of deep reefs with numerous small walls, overhangs, crevices and caves (mean reef slope index > 0.75), which physically reduce incident light on reefs and favor sciaphilic species. Indeed, such abiotic (depth, as a proxy for incident light reduction) and biotic (cover of Cystoseira s. l. and Posidonia oceanica on reefs) factors best explained the regional partitioning of fauna communities. Within this depth range (15 – 35 m) in the Mediterranean, while the downstream effect of Cystoseira (sensu lato) canopy on surrounding communities (recruitment and development) is well established (Benedetti-Cecchi et al., 2001; Mangialajo et al., 2008), the effect of seagrass meadows thriving on reefs has rarely been documented. It is interesting to observe the opposite trend between extensive cover of Posidonia oceanica and Cystoseira (sensu lato) which might indicate a relative competitive exclusion of these two engineer species under certain conditions, particularly in the southern region of Corsica, where extensive cover of P. oceanica on reefs was documented alongside minimal cover of Cystoseira (sensu lato).

Algal communities better explained the geographical contrasts between the three MPA regions, as indicated by the higher proportion of variance explained by the multivariate analyses, especially in the 25 – 35 m depth range. In summary, at shallow depths (15 – 25 m), the western region had the highest cover of Cystoseira (sensu lato) together with Dictyopteris spp. thriving in low slope reefs characterized by abundant light, while Dictyota spp. and crustose corallines thrived in the northern and southern regions, respectively. As expected, Peyssonnelia spp. increased in shallow southern and western reefs with more vertical features (low light conditions). At deeper depths (25 – 35 m), the southern region exhibited a higher abundance of Caulerpa cylindracea and Phyllophora crispa than the northern region, while the western region had more Cystoseira (sensu lato) and/or Peyssonnelia spp. depending of reef slopes. As for the fauna communities, the presence of dense stands of Cystoseira (sensu lato) has been proposed to structure the associated algal communities (Benedetti-Cecchi et al., 2001; Mangialajo et al., 2008), together with higher water movement and nutrient concentration, probably responsible for the extensive development of Cystoseira (sensu lato) forests themselves along the Corsican coast. To a lesser extent, and as mentioned above, the extent of Posidonia oceanica on the reefs is another crucial determinant of the algal communities, restricting reef areas for large erect algae development.

This study of the spatial structure of infralittoral communities on Corsican reefs provided important insights into depth zonation (between 15 and 35 m depth), species covariation, regional variations in community structure, environmental drivers and potential future disturbances mediated by Caulerpa cylindracea. Not surprisingly, algal communities explained most of the variation in spatial structure and Cystoseira (sensu lato) was found to be a driving factor together with light (depth), nutrient and hydrodynamic conditions. More surprisingly, Posidonia oceanica, which also develops on reefs, acted as another influential driver, its extensive coverage confining reef communities to specific areas and thereby structuring the observed fauna and algal communities. Well-developed Cystoseira (sensu lato) forests, which represent the climax and reference state of Mediterranean infralittoral communities, were found on most western reefs (especially in the Scandola Nature Reserve) and on some northern reefs away from the Gulf of Saint-Florent (near the Agriates desert and the Centuri shoals). On the contrary, most of the northern and southern reefs were characterized respectively by a high cover of Dictyota spp. and crustose corallines, with a worrying cover of Caulerpa cylindracea in the southern deeper reefs.

These findings raised concerns, as they did not align with the characteristics of undisturbed infralittoral reefs in the region. Further comparisons with historical data and appropriate calibrations against known pressure gradients will be required to propose ecological assessments of sites from the VIT-Reef protocol. The few reef-EBQI sites in Corsica have been classified as high or good ecological status (Thibaut et al., 2017). The reef-EBQI is based on a broad set of functional compartments from a conceptual ecosystem model and allows the assessment of the quality of the ecosystem functioning, but therefore requires more measurements and extensive diving (i.e., at least 2 days for 2 divers) than VIT-Reef. While the depth range is different between the current deployment of the reef-EBQI (1 – 10 m) and VIT-Reef (15 – 35 m) protocols, it is important to test the VIT-Reef protocol in other reef-EBQI ecological status frameworks (moderate, poor, and bad) (Thibaut et al., 2017). Deployment in known stressed environments will allow disentangling the relationship between reef community variation and pressure, and to relate VIT-Reef to reef-EBQI scores.