Drawing the borders of the mesophotic zone of the Mediterranean Sea using satellite data

The 30–150 m bathymetric range is commonly adopted in the literature to constrain the mesophotic zone. However, such depth interval varies depending on sunlight penetration, which is primarily a function of solar radiation incidence and water clarity. This is especially obvious in the Mediterranean Sea with its peculiar biophysical properties. Integrating information on light regime in the estimation of the bathymetric range of the mesophotic zone would provide a more robust definition, orienting conservation actions targeting its ecosystems. We present a first assessment of the spatial and vertical extension of the mesophotic zone in the Mediterranean Sea based upon light penetration, comparing our prediction with literature data. Our study also represents a baseline to monitor future variations in the bathymetric interval associated with the mesophotic zone in the Mediterranean Sea in relation to global changes.

derives from the deepest occurrences of zooxanthellate corals documented in the late 80s (e.g., in the Caribbean at 119 m 17 , in the Red Sea at 145 m 20 , in Hawaii at 153 m 21 , and in Johnston Atoll at 165 m 22 ).
The consistency of such depth limits for tropical and sub-tropical situations has been verified by exploring patterns of distribution and composition of mesophotic coral ecosystems (MCEs) that provided evidence of a subdivision in an upper (30-60 m) and a lower mesophotic zone (60-150 m), which host communities different in composition 8,[23][24][25][26] . The depth range associated with mesophotic conditions (30-150 m), and the community break at ~ 60 m depth were proved to be robust also through meta-analysis considering information on benthic tropical mesophotic reefs from around the world 27 , and studies based on irradiances ( 28-30 among others).
Nevertheless, situations dominated by zooxanthellate scleractinian corals represent a minority in the benthic assemblages characterizing the mesophotic depths at the global scale 11,29 . At the temperate latitudes, observations of light-dependent corals appear to be sparse and mesophotic ecosystems are typically dominated by octocorals, mainly gorgonians and antipatharians and sponges (e.g., [31][32][33][34] ), but include also bryozoans, ascidians, and shade-adapted (sciaphilous) algae 14 . The strong seasonal variability that characterizes temperate systems influences the water clarity affecting the penetration of solar radiation along the water column (e.g., 35 ). Hence, the interval 30-150 m might not necessarily represent the bathymetric range of mesophotic condition occurrence at these latitudes.
However, information on Mediterranean mesophotic ecosystems (including studies using the term "twilight" to refer to mesophotic situations) is limited to few taxonomic groups (mainly cnidarians and sponges 31 ) while other assemblages have been explored only seldom (e.g., brachiopod-dominated communities and deep-sea oyster reefs 49,50 ). The low number of studies of Mediterranean mesophotic ecosystems presenting taxonomic components other than cnidarians 31 might not be a consequence of the lower interest from the scientific community. Rather, the term "mesophotic" is likely less used to define situations without a coral component. Among these are rhodolith beds and coralligenous habitats, studied widely in the Mediterranean Sea 57,58 . If the former are increasingly accounted as mesophotic (e.g., 58,59 ), opinions differ on whether coralligenous assemblages should be considered an independent biological feature or under the mesophotic domain (e.g., 31 ). In terms of light regime, these situations might be included in the mesophotic zone since they occur in sciaphilous conditions and at mesophotic depths 60,61 .
The lack of a clear definition of the borders of the mesophotic zone at temperate latitudes may not pose much of a problem as long as the term "mesophotic" intends to refer to intermediate-depth biological situations. Problems arise when scientific information must be transferred to policies that require a coherent spatial definition to plan proper management and conservation.
The mesophotic zone extends from 30 m to where photosynthesis compensation point occurs. This lower boundary is, however, variable due to differences in light penetration, and environmental and ecological drivers that can vary over short distances 12,62 . Observations of live macroalgae assemblages, mainly coralline algae, that are able to maintain active photosynthesis at irradiance levels as low as 0.0001 mol photons m −2 day −1 (corresponding to 0.0005% of surface PAR, 63,64 ), suggest that the photic zone could extend deeper than previously assumed.
If we were able to include the light regime in the definition of the mesophotic zone at temperate latitudes, we could not only constrain its bathymetric range and estimate the portion of seafloor characterized by mesophotic conditions but also appreciate variations related to local factors. Drawing the limits of the temperate mesophotic zone under current climatic conditions would be very important especially in the context of future global changes. According to IPCC predictions, nutrients supply and ocean productivity will likely decline in the near future, largely affecting phytoplankton communities and the transparency of the water column 65 . On the other hand, changes in precipitation, storm surges, and heatwaves related to ongoing climatic variations have already increased the volume of sediment loads from river runoff, affecting the turbidity of the water column [66][67][68] . Consequently, the vertical extension of the photic and mesophotic zones will change, affecting the ecosystem structure of these environments with taxonomic groups that might take advantage in the long term of a concomitant decline of more sensitive organisms 65,69,70 .
In the Mediterranean Sea, mesophotic communities are already threatened by natural and anthropogenic pressures, including seawater temperature increase, heat waves, bottom trawling and marine litter (e.g., abandoned or derelict fishing gear), which trigger a gradual but irreversible process of habitat degradation (e.g., [71][72][73]. A better understanding of the bathymetric and spatial extension of mesophotic zone in the Mediterranean Sea is, thus, urgent to orient conservation actions. However, the penetration of PAR along the water column depends on complex dynamics involving the inherent optical properties (IOPs) of the water column, the combined effects of absorption and scattering, and the amount and on the directionality of the light at the surface, despite the latter has been proven to be weak away from boundary effects, such as at mesophotic depths 74 . For such a reason, identifying the borders of the mesophotic zone that rely upon light regime might be challenging when on-site irradiance and IOPs measurements are not available.
Here we present a first assessment of the seabed portion ascribable to mesophotic conditions for the entire Mediterranean Sea based on irradiance penetration derived from satellite data. Our estimation is then validated with literature data on benthic mesophotic assemblages. By providing its current spatial and bathymetric www.nature.com/scientificreports/ extension, we provide a baseline to monitor the future variation of the mesophotic zone of the Mediterranean Sea in response to global changes.  The Kruskal-Wallis test (p < 0.01) and the pairwise comparison by the Dunn's multiple comparison test revealed significant differences in the irradiance and percentage of surface PAR at the upper limit of the mesophotic zone among the Mediterranean sub-regions (Fig. 4A,B, Table 1). In particular, the Ionian and Central Mediterranean Sea and the Aegean-Levantine Sea sub-regions are statistically different from the Western Mediterranean Sea and the Adriatic Sea. The depth of the lower limit of the mesophotic zone, defined at 0.0001 mol photons m −2 day −1 (i.e., ca. 0.0005% of surface PAR) shows values in line with the north-west-south-east gradient observed for K d490 , surface PAR and K dPAR (Fig. 4C). The mean depth for the Western Mediterranean Sea and the Adriatic Sea is 152 ± 1.3 m and 154 ± 1.2 m, respectively. The penetration of light along the water column increases towards the south-eastern part of the basin, with the Aegean-Levantine Sea and the Ionian and Central Mediterranean Sea displaying values of 198 ± 1.9 m and 190 ± 2.5 m, respectively (Fig. 4C). The Kruskal-Wallis and post-hoc pairwise tests indicate that the differences between the north-west (Western Mediterranean Sea and Adriatic Sea) and the south-east (Ionian Sea and Central Mediterranean Sea and Aegean Levantine Sea) sub-regions are statistically significant (Fig. 4C, Table 1). The mean value for the entire Mediterranean Sea is 175.5 ± 1.2 m.  (Fig. 5A). In particular, the Aegean Sea and the Italian and Croatian margins present the highest number of observations, followed by studies along the French and Spanish coasts (Fig. 5A). Out of 997 occurrences of mesophotic assemblages extracted from literature, 685 observations (69%) occur in seabed characterized by mesophotic conditions, based on criteria adopted in the present study (Fig. 5A,B). The remaining ca. 24.7% of the dataset (246 records) reports observations above 30 m depth and 6.6% (66 records) below the lower limit of the mesophotic conditions (i.e., < 0.0001 mol photons m −2 day −1 , Fig. 5B).

Discussion
The mesophotic zone in the Mediterranean Sea hosts a variety of different biological components, from coralligenous formations to rhodolith beds, sponge grounds, gorgonian forests and oyster reefs, depending on bathymetry, geographic location in the basin and local factors 32,33,45,46,[49][50][51][52][53]55 . These ecosystems supply essential goods and services, acting as nursery and spawning grounds for fishes, potential carbon sinks, and represent a unique natural heritage 71,73,79,80 . Providing a spatial definition of the mesophotic zone is, therefore, of primary importance to support the implementation of conservation policies addressing such ecosystems and inform management plans 14 .
Although the bathymetric range associated with the mesophotic zone in the Mediterranean differs in the literature 31 , most of the studies so far have adopted the depth range defined for tropical and subtropical latitudes, between 30 and 150 m 13 . The lower limit of the mesophotic zone at temperate latitudes is, however, strictly related to the penetration of light in the water column that vary largely with geographic location. The gradient of www.nature.com/scientificreports/ irradiance not only defines the border of the mesophotic zone but also concurs to shape the structure of mesophotic assemblages, along with other physical and biochemical factors 30,63,81 , controlling the presence of obligate phototrophic organisms and ultimately defining the bathymetric range of the mesophotic zone. Our analysis of the irradiance at the upper limit of the mesophotic zone (30 m depth) for the entire Mediterranean Sea (3.69 ± 0.09 mol photons m −2 day −1 , i.e., 9.8 ± 0.2% of surface PAR), and the presence of live crustose coralline algae at irradiance level down to 0.0001 mol photons m −2 day −1 (i.e., 0.0005% 60,64 ) suggest that mesophotic conditions occur at irradiance between ca. 4 and 0.0001 mol photons m −2 day −1 , corresponding to surface PAR levels between ca. 10 and 0.0005%. This range agrees with observations on coralligenous and rhodolith bed assemblages, which are commonly found in dim light conditions where irradiance can be as low as 0.05% and 0.0005% of surface light, respectively 60,61 .
At the Mediterranean basin scale, the depth of the lower limit of the mesophotic zone is 175.5 ± 1.2 m, consistent with the bathymetric definition in previous studies 31 . The threshold of 0.0001 mol photons m −2 day −1 (0.0005% of surface PAR) as the lower limit for mesophotic conditions ensures to include portions of seabed characterized by low light conditions but excludes areas below 200 m depth, commonly referred as the beginning of the "deep-sea" (below 200 m depth 82,83 ), thus avoiding an overlap between different bathymetric domains.
The comparison between our estimation and the location of literature records on mesophotic ecosystems showed that 69% of literature records on mesophotic ecosystems fell within the proposed area (Fig. 3B), lending support to the reliability of the results.
Using our approach, more than 377,000 km 2 of the seafloor, corresponding to ca. 15% of the entire Mediterranean Sea surface area, is under mesophotic conditions. The attenuation of solar radiation increases with increasing depth and is affected by the amount of dissolved and particulate matter suspended in the water column, which is correlated with primary productivity, hydrodynamic conditions, and terrigenous influences 84 . Freshwater inputs (e.g., rivers and precipitations) carry substantial quantities of terrigenous organic and mineral particles, dissolved organic matter, and release large amounts of nutrients to the sea. For instance, in the Western Mediterranean Sea, the estimated portion of seabed under mesophotic conditions is lower than in the eastern subregions, although covering a larger area. Intense algal blooms occur in the Western Mediterranean Sea in spring when the surface layer stabilizes, and less frequently in autumn when the thermocline is eroded 76 . These factors can increase the attenuation of light penetration in the water column, as confirmed by the higher values of K dPAR in the north-western Mediterranean Sea.
The extension of seabed receiving sunlight and the amount of irradiance reaching the seabed have also been proved to be influenced by the morphology of the seafloor 30,81 . For example, some sectors of the Western Mediterranean Sea are characterized by vertical or highly steep bottom whose extent is not properly represented on a horizontal projection. Consequently, the portion of seabed under mesophotic conditions appears restricted. This, nevertheless, does not necessarily mean that mesophotic ecosystems are not present in such situations but rather that they might distribute along highly steep bottoms. Several mesophotic communities have been, indeed, documented colonizing rocky vertical walls in the Western Mediterranean (e.g., 42,45,85 ). On the contrary, the large (more than 50%) area predicted as under mesophotic conditions in the Adriatic Sea does not imply more abundant mesophotic communities, but it is likely related to the geomorphology of the subregion. The Adriatic Sea is, in fact, characterized by gentle slopes, presenting an average bottom depth of about 35 m that increases moving to the south (140 m on average, central Adriatic), with the two Pomo/Jabuka Depressions reaching 260 m 86 . Estimating the spatial extent of the mesophotic zone using the penetration of light in the water column can only identify areas with mesophotic light regime. A complex combination of variables concurs in determining the presence of mesophotic communities, such as biogeochemical factors and type of substrate along with biotic and stochastic processes. Consequently, the extent of seabed under mesophotic conditions does not necessarily represent the probability of finding mesophotic communities.
Our results, however, suggest that the quantity of light at seabed calculated from satellite K d490 concentration is a valid tool to spatially constrain the mesophotic zone, helping in the identification of areas suitable for mesophotic ecosystems, and their related biodiversity and ecological services. Although the depth range 30-150 m has been proved to be associated with mesophotic conditions at tropical latitudes, the greatest strength of using a physical approach based on satellite data is that it guarantees to appreciate variations in the bathymetric extension of the mesophotic zone related to local factors and seasonal variability, especially when dealing with non-tropical situations. Estimating the penetration of light along the water column assures also to identify differences in the depth of occurrence of obligate phototrophic organisms with geographic position, ultimately providing information on the structure of mesophotic assemblages and the ecological functions they provide.
By including information on the light regime, our approach can also estimate the spatial and bathymetric extension of the mesophotic zone in situations characterized by data deficiency. For instance, the southern Mediterranean would appear devoid of mesophotic assemblages from an analysis of the literature due to the few accounts available to date. Our method, instead, highlighted that large portions of this sector of the Mediterranean Sea might be under mesophotic conditions, enabling their inclusion in the quantitative assessment of the mesophotic zone at the basin scale.
Direct measurements of irradiances and turbidity along depth would surely represent the best method to validate the penetration of light obtained from satellite data. Our results, nevertheless, show that literature records can be successfully used as groundtruthing surrogate when on-field irradiance measurements are not available, providing a degree of confidence associated to the spatial definition of the mesophotic zone. The method presented is not meant to be a "one-size-fits-all" solution but an approach to estimate the depth range associated with mesophotic conditions based solely on satellite data. This is particularly important to furnish robust information to be included in management plans and to orient future explorations. In this context, by comparing the distribution of areas estimated as presenting mesophotic conditions with information on local and www.nature.com/scientificreports/ regional anthropogenic impacts, our methodology might help in identifying areas potentially hosting mesophotic communities threatened by human activities.
Finally, an approach that integrates information on the light regime will also permit to explore the variations in the spatial and vertical extension of the mesophotic zone in response to future climatic conditions of the oceans. The IPCC prediction foresees that the vertical extension of the photic zone might vary in the near future as a consequence of the decline of nutrient supply, marine productivity and phytoplankton community, leading to an increase in water transparency and, thus, light penetration 65,69 . Irradiances associated with the mesophotic zone might occur deeper than at present by the end of the century, forcing a physiological response in benthic mesophotic species and surely influencing the structure, distribution of mesophotic ecosystems. By drawing its current extension, our estimation represents a baseline for studying how the depth range of the mesophotic zone in the Mediterranean Sea will vary in relation to future conditions of the oceans, permitting the identification of areas most affected by climate changes.

Materials and methods
Quantity of PAR at seabed. The quantity of light at seabed was calculated following the methodology proposed by 87 that derives the diffuse attenuation coefficient for PAR light (K dPAR ) as a function of the diffuse attenuation coefficient for light at 490 nm (K d490 ) according to the following formula: The intensity of light (E z ) at a given depth (z) is then calculated from the intensity of light entering the ocean (E 0 ) and K dPAR by applying the Beer's Law 35 : The diffuse attenuation coefficients K d490 and K dPAR are also influenced by the directionality of the light field, but this dependence has proven to be weak at mesophotic depths 35,74 . We assumed K d490 as a good representation of the optical water quality for the purposes of our study.
Daily data on K d490 in m −1 were used to calculate the 17-years mean for the Mediterranean Sea for the period 2002-2018. The average PAR at surface for the same period was estimated from daily average values in mol photons m −2 day −1 . The data were obtained from the open-access repository NASA Ocean Colour (https:// ocean color. gsfc. nasa. gov/), at a resolution of 4 km. Bathymetric data were downloaded from open-access repository EMODnet (https:// www. emodn et-bathy metry. eu/) with a horizontal resolution of ca. 100 m. Data on K d490 and surface PAR were processed to match the resolution of the bathymetry using an upscaling approach that approximates conditions at the seafloor (sensu 88 ) and that has been demonstrated to work effectively in different scales and situations (e.g., 89 ). The quantity of PAR reaching the seabed was, then, calculated following the equations described above. Finally, the percentage of PAR at seabed was estimated as the ratio between light at seabed (E z ) and surface PAR in each pixel, multiplied by 100. The processing was performed in R software using the packages "raster", "sp" and "rgdal" 90 . The estimated irradiance at seabed is available at https:// doi. org/ 10. 5281/ zenodo. 58631 03.
Bathymetric definition of the mesophotic zone in the Mediterranean Sea. The upper limit of the mesophotic zone in the Mediterranean Sea was set at 30 m, according to literature evidence constraining zooxanthellate corals mainly above this water depth (e.g., 32 ). The portions of seabed above 30 m depth were then excluded from the subsequent analysis.
The lower limit of the mesophotic zone was identified based on the minimum value of PAR that is known to support photosynthetic metabolism. In 60 and more recently 64 , live crustose coralline algae have been observed at PAR level down to 0.0001 daily photon dose (mol phot m −2 day −1 ) corresponding to 0.0005%, which is the lowest value recorded so far. This value was used as the lower limit for mesophotic conditions. The seabed with estimated irradiance at seabed lower than 0.0001 mol photons m −2 day −1 was excluded. The remaining dataset was converted into a polygon using the "Raster to Polygon" tool ArcGIS 10.5.
Setting the limits at 30 m depth and at the minimum daily photon dose documented as sustaining net positive photosynthesis, however, omits information on light conditions at the upper and on depth at the lower. Consequently, we generated a point every 20 km along both borders and extracted the corresponding irradiance, percent PAR and depth values to provide ranges of light (average ± standard error) and depth for the mesophotic zone in the Mediterranean Sea. Their variation across the basin was tested by dividing the dataset into subsets corresponding to the sub-regions outlined by the European Union (EU) Marine Strategy Framework Directive (2008/56/CE): Western Mediterranean Sea, Adriatic Sea, Ionian Sea and Central Mediterranean Sea and the Aegean-Levantine Sea (Figs. 2, 3). Spatial distribution of benthic mesophotic communities. We compiled a list of 112 bibliographical records on benthic mesophotic ecosystems in the Mediterranean Sea, through an extensive bibliographic search in ISI Web of Science, Scopus, and the mesophotic.org website (http:// www. mesop hotic. org/). A first screening aimed at removing studies not related to benthic ecosystems (e.g., targeting fish fauna), then we selected only those reporting the exact location of the investigated assemblage. Out of 112 documents, 69 fulfilled the criteria. We integrated this information with the distribution of coralligenous and rhodolith bed habitats from the MEDISEH database based on literature reviews 57 . Overall, 1407 relevant locations to our study were selected for the Mediterranean basin. Among these, 38 records wrongly located on land and 317 falling outside the extent of satellite data (e.g., too close to coastline) were removed. Finally, 55 sites deeper than 500 m were not considered, (1) K dPAR = 0.0665 + 0.874 * K d490 − 0.00121 * K d490 −1 (2) E z = E 0 e −K dPAR z