Coral Communities, in Contrast to Fish Communities, Maintain a High Assembly Similarity along the Large Latitudinal Gradient along the Saudi Red Sea Coast

The Red Sea is a unique habitat characterized by several environmental conditions considered potentially stressful for coral reef growth [1]. Those occur particularly in the south, where high temperatures (up to 33°C) and high nutrient inputs (leading to chlorophyll a concentration up to 4.0 μg l-1) co-occur. In contrast, the northern Red Sea is largely oligotrophic (Figure. 1A); [2,3] and temperatures remain below 30°C [4] . The most northern region, the Gulf of Aqaba, features the lowest regional temperatures (21 27°C), highest salinity (<42 PSU) and low chlorophyll concentrations (0.1 0.4 μg chl a l-1, Figure 1A). Although, the Red Sea is at the most northwestern end of the tropical Indo-Pacific, coral diversity is high (>260 hermatypic coral species) and reef communities are complex throughout the Red Sea [5]. While a thorough study on coral community structure and diversity patterns in 1991 identified a clear zonation into a northern, central and southern Red Sea region a [6] recent study on coral communities suggested an increase in community homogenization within the last two decades throughout the Red Sea [7]. This was inter alia related to environmental changes such as ocean warming [7], which affects some species more than others [8], leading to a decimation of some species and a spread of other pecies [7]. However, it remains to be assessed whether changes in coral abundance and diversity patterns in the Red Sea alter the reef functionality and whether differences in the functionality (ecological role) exist along the latitudinal gradient.


Introduction
The Red Sea is a unique habitat characterized by several environmental conditions considered potentially stressful for coral reef growth [1]. Those occur particularly in the south, where high temperatures (up to 33°C) and high nutrient inputs (leading to chlorophyll a concentration up to 4.0 µg l -1 ) co-occur. In contrast, the northern Red Sea is largely oligotrophic (Figure. 1A); [2,3] and temperatures remain below 30°C [4] . The most northern region, the Gulf of Aqaba, features the lowest regional temperatures (21 -27°C), highest salinity (<42 PSU) and low chlorophyll concentrations (0.1 -0.4 µg chl a l -1 , Figure 1A). Although, the Red Sea is at the most northwestern end of the tropical Indo-Pacific, coral diversity is high (>260 hermatypic coral species) and reef communities are complex throughout the Red Sea [5]. While a thorough study on coral community structure and diversity patterns in 1991 identified a clear zonation into a northern, central and southern Red Sea region a [6] recent study on coral communities suggested an increase in community homogenization within the last two decades throughout the Red Sea [7]. This was inter alia related to environmental changes such as ocean warming [7], which affects some species more than others [8], leading to a decimation of some species and a spread of other pecies [7]. However, it remains to be assessed whether changes in coral abundance and diversity patterns in the Red Sea alter the reef functionality and whether differences in the functionality (ecological role) exist along the latitudinal gradient.
Hermatypic corals are the main bioengineers of coral reefs creating a structurally complex habitat for various reef associated organisms. For example, structural complexity was found to strongly favor fish abundance. Structural complexity is particularly high when the abundance of branching corals is high [9]. Many branching corals belong to the genus Acropora, which are competitvely favored due to their rapid growth, but rather sensitive to stress [10]. Potentially stressful conditions (e.g. high turbidity, high temperature or temperature fluctuations, pollutants) could lead to a community shift towards more stress-tolerant species, which are generally represented by rather bulky and slow growing corals such as most Porites, Goniastrea, Favites, Favia and Lobophyllia [10], entailing inter alia a lower structural complexity [9]. A decreased abundance of herbivore fishes may lead to proliferation of macroalgae, in particular when nutrient levels are increased, which in return lowers coral recruitment success [11,12]. When a functional group is composed by several species, this redundancy may render a given function, for example provision of structure, as well as a higher resilience to disturbances and stress [9,[13][14][15]. Therefore, it is important to assess not only coral species diversity, but also the functional components of coral reef communities (corals, fishes, non-coral benthic invertebrates) in order to evaluate the resilience of ecosystem services to disturbances and stress. Such an assessment along the latitudinal gradient of the Red Sea is of particular interest in the light of predicted global change scenarios, since reefs in the southern Red Sea already persist under conditions predicted for other geographic regions in the future (sea surface temperatures >31°C, ICPP 2007 is reference number 37; high nutrient input) [16]. Thus, the assessment of coral reef functionality, particularly in the southern Red Sea, may provide a "preview" of potential future scenarios for reefs in the northern-central region of the Red Sea and for other geographic regions.
In this study, we assessed a) coral assemblies based on functional groups reflecting their life-history traits and growth forms, b) fish assemblies based on functional groups reflecting feeding mode and size, and c) the abundance of selected benthic indicator species. This was conducted at off-shore reefs along the nautral latitudinal environmental gradient, as well as at near-shore reefs close to sources of pollution. The results are discussed in the context of environmental changes.

Material and Methods
Thirteen reefs were investigated along the Saudi Arabian Red Sea coast between 16°34' and 28°31'N in a total of 7 regions (Figure 1). In 5 regions, not only reefs far away from any source of pollution appearing unaffected and healthy, but also reefs close to a potential source of pollution were selected ( Figure 1). The study sites were consecutively numbered from north to south (1 to 7) and denoted with "N" for nonpolluted and "P" for polluted in superscript throughout the manuscript. The sources of pollution were within 500 m of the impacted sites crest (most sites) or perpendicular to a shallow slope (LIT 6P and FAR-S 7N ) in March 2012. Coral composition and abundances were recorded along the transects and assigned to one of the 12 pre-defined functional taxa reflecting the major life-history traits growth form, growth speed and susceptibility to bleaching/stress (Table 1) [18]. The different coral taxa were further assigned to one of the three overarching strategic groups defined by Darling [10,19] with a competitive group including fast growing broadcast spawning species, b) a weedy group including mainly brooding species with a short generation time and small colony size and c) a stress-tolerant group including slow growing broadcast spawning species with a high fecundity (presented by different colors in Table 1). Furthermore, soft corals, fleshy macroalgae and other sessile organisms (e.g. tunicates, bryozoans, anemones, sponges) as well as rock, coral rubble, sand and silt were quantified (Table 1). Along the same transect, in a belt of 4 m width, we quantified benthic invertebrates and varied by type: WAJ 2P was close to a desalination plant in front of the town Al-Wajh, YAN 3P was close to a petro-chemicals factory and construction site in Yanbu, RAB 4P was close to a cement factory, oil refinery and power plant in Rabigh, JED 5P was in proximity of a domestic waste water discharge of the major city Jeddah, and LIT 6P was close to the aquacultural waste water discharge of a large shrimp farm in Al-Lith. Regrettably, no data quantifying the anthropogenic impact at these sites were accessible. Fishing pressure is comparatively low in the Saudi Arabian waters of the Red Sea [17], but may be increased locally for example at the Farasan Islands, Farasan Banks and off Al-Wajh, where mainly edible fish is captured, but also fish for aquarium trade and sea cucumbers [17].
All selected sites were facing towards the exposed seaward side and a 100-m-long transects was laid out in 3-4 m depth along the reef   other than corals, as well as fishes (fishes between bottom and surface).
Here, indicator species / major taxa suggested by "Reef Check" [20] were used: non-coral invertebrates included sea urchins, giant clams, sea cucumbers and crown-of-thorn starfish, carnivore fishes included wrasses, snapper, grouper and sweetlips, the herbivore fishes contained surgeon fishes and rabbit fishes and the corallivore fishes contained parrot fishes and butterfly fishes ( Table 2).  a (chl a), total nitrogen (TN), total phosphorous (TP) and particulate carbon (particulate C). TN and TP concentrations for September 2011 were previously presented by Kürten [2]. All analyses were conducted using standard protocols (see supplementary information). Two seasons were chosen to capture seasonal differences in water quality.
Multivariate analyses were performed using PRIMER v6 and PERMANOVA add-on [21,22]. In order to identify trends in the assembly structure of reef communities, Principal Coordinate analyses (PCO) were conducted using the full data set. PCO, similar to Principal Component Analysis (PCA), projects samples (here sites) based on their dissimilarities on PCO axes that minimize residual variation in the space of the resemblance matrix, thereby explaining most of the variation between samples by the first view PCO axes. While PCA is based on Euclidean distance between samples, PCO allows using any distance measure. Prior to running PCO, the data was log (x+1)transformed down-weighing the contribution of quantitatively dominant taxa and a resemblance matrix was created based on Bray-Curtis similarity.

Environmental conditions
The environmental parameters chl a, particulate C and TP concentrations, but not TN concentration, increased from north to south and were generally higher in March than in September (Figure1B), concurrent with previous descriptions of the Red Sea conditions and with satellite derived data (NASA Giovanni, Ocean Color Radiometry). This can mainly be explained by the current regime [23,24,3]: Nutrients are mainly introduced into the Red Sea from the Indian Ocean through the Bab el Mandeb and distributed towards the central Red Sea during the winter monsoon. During spring and summer, wind and surface currents are generally weaker and directed southward, while some locations in the northern Red Sea are fertilized by eddy-driven upwelling. Other nutrient sources, particularly in the northern Red Sea including the Gulf of Aqaba, are aerosols deposition and fixation of 'new nitrogen' [2,25,26]. We note that sampling of seawater at different depth (3 and 10 m) in the different seasons may have introduced some bias in our seasonal comparisons, however, we are confident that the general pattern remains representative, as our nutrient related parameters can well be explained by the current regime.
Within regions, near-shore (close to sources of pollution) chl a, particulate C and TP concentrations were higher compared to the offshore sites ( Figure 1B). This was particularly evident in the central Red Sea. Here, RAB 4P , close to the oil refinery and cement factory, revealed 10-fold increased values of particulate C, and JED 5P , close to a domestic waste water inlet, showed 2-fold higher concentrations of TN, when compared to their corresponding off-shore reefs (Figure1B).
Logged temperature at 3 m depth ranged from 21 to 27.5°C in the most northern region (MAQ 1N ) and from 28 to 33°C in the southern region (DOG 6N and FAR-S 7N ) ( Figure 1C) from February 2011 to March 2012. The most southern region, the Farasan Islands, featured the longest continuous warm water period with temperatures above 31°C from May to October.

Reef community assemblies along the latitudinal gradient
Taking all investigated taxa (coral, fish and non-coral invertebrates) into consideration, the assembly structure of the reef community roughly followed a latitudinal shift, which is mainly represented by the x-axis of the PCO (Figure 2A). This shift, however, was largely explained by the assembly structure of the coral associated community rather than by the coral community itself ( Figure 2B). This suggests that the coral assemblies changed surprisingly little from north to south, despite the strong environmental gradient. Hence, corals thereby provide comparable functionality throughout the latitudinal gradient of the Red Sea. Structurally complex and fast growing Acropora species (competitive group), short branched and brooding Stylophora (weedy group) and rather bulky and slow growing species of the families Poritidae, Faviidae and Musiidae (stress-tolerant group) co-occurred throughout the gradient ( Figure 2B). Only among some strategic groups, some latitudinal shifts were identified. Within the weedy group, the abundance of bushy Pocillopora (taxon 4) increased northwards. Within the competitive group, the abundance of compact bushy Acropora (taxon 1) increased northwards, while the abundance of rather long-branched and table-like Acropora (taxa 2 and 3) increased southward ( Figure 2B). This demonstrates certain variability among growth forms, with more physically robust corals in the northern half of the Red Sea. The reason for this is most likely found in hydrodynamic forces, rather than in temperature or nutrient-related changes, because wave action and currents were generally stronger in the northern half than in the southern half of the Red Sea ( [6] personal observation).
In contrast to the generally low stress-tolerance of Acropora species, in particular to temperature stress (Great Barrier Reef) [27,28,19], Acropora appear particularly successful throughout the Red Sea, including in the adverse warm waters conditions of the southern Red Sea. This infers specific adaptations to Red Sea conditions, which may be supported by the findings of a post-bleaching study in the northerncentral Red Sea, where certain Acropora species exhibited a strong bleaching resistance [8]. Riegl [3], likewise, described the presence of rather robust Acropora species, which mainly occur in the south, but seem to expand toward the north (e.g. A. muricata, A. nobilis). This said, the question however remains, what maintains the functionality of coral assemblies on a similar level along the Saudi Red Sea coast. Is it rather due to the adaption of single species to the entire range of environmental conditions? Or is it due to the large functional redundancy (high species richness within a functional group; [14], which allows selection for the most capable species within a given environmental setting? The second scenario may be supported after revising the differences in species richness within the strategic groups or coral taxa. Here, it was evident that large functional taxa including numerous species (e.g. 35 Acropora spp., 27 Poritidae spp., 61 Faviidae spp.) [5], barely show a north-south shift in abundance, while the smaller groups / taxa partly do. Latter was evident for the taxon Millepora within the competitive group (< 3 species, exact number unknown), a hydrozoan genus which features a high bleaching susceptibility and was most likely therefore restricted to the northern Red Sea. Furthermore, the taxon Oculinidae within the stress-tolerant group (2 species) [5] is known for a high tolerance towards turbidity, which most likely caused a shift in abundance towards near-shore and southern reefs. A last example is given by the weedy group (2 Pocillopora spp., 6 Stylophora spp.) [5], which, at least in the case of Pocillopora, seemed to be less stress-tolerant towards turbidity and high temperature, therefore occurring more in the northern Red Sea. These functional taxa, however, were generally less important for the structural complexity of reef systems and may therefore also be less important in the context of reef resilience. Thus, the latitudinal change in environmental conditions seems to be hardly reflected by the coral assembly pattern and its ecological functionality.
Not only the taxonomic and functional composition of the coral assembly, but also the composition of the non-coral substrate is important for habitat structure and functionality. Here, the coverage by rubble (0-33%), sand (0-28%) and macroalgae (0-5 %) increased towards the south thereby decreasing structural complexity, while the coverage by structurally supportive rock (stone and dead massive corals) was higher in the north (15-42%) (Figure 2, Supplementary Information Table S1). Furthermore, sea urchin and sea cucumber abundance increased towards the south, where sea urchins feed on algae helping to keep algae coverage low (<5%; see also [29,30]")", and where sea cucumbers find more nutrient-enriched sediment (e.g. sand) [29]. A decrease in soft corals towards the south is likely to be related to a decrease in water clarity [31,32].
Given that habitat structure (mainly formed by the coral assembly, but also by the hard-substratum rock) is comparably high along the entire gradient, fishes should be able to find suitable habitats throughout the Saudi Red Sea coast, independent of latitude. Yet, the fish abundance and assembly structure shifted considerably from north to south (Figures 2A, 2B). Various larger fishes, the carnivore grouper, sweet lip and snapper, as well as the herbivore rabbit fish, increased in abundance towards the south ( Figure 2B). This shift can mainly be ascribed to a higher nutrient input, where higher primary production and larger deposition of organic material supports the benthic food web und ultimately the larger carnivores [2]. A shift towards rather small fishes in the north, namely an increase in carnivore wrasses and corallivore butterfly fishes ( Figure 2C), may have the following reasons: A lower abundance of potential predators (larger carnivores) in the north, a higher temperature sensitivity of wrasses and butterfly fishes and/or a preference for the higher energy reefs [33] as provided in the north personal observation). A limitation of our study is that we sampled fish data only in March, which may disguise potential difference in fish assemblies between seasons.
Data of percentage coverage of coral, other sessile organisms and non-living substrate (Table S1), densities of fishes and motile invertebrates ( Figure S1 and S2), as well as photographs showing typical community structures at selected sites are available as Supplementary  Information (Figure S3).

Reef community assemblies at polluted reefs
The reefs close to a source of pollution revealed no consistent shift in the assembly structure of reef communities, when compared to their corresponding off-shore reefs. In fact, shifts differed in direction and intensity (Figure 2A). The two most extreme situations were found in Rabigh 4 and Al-Lith 6 , where the variation between the non-polluted and the corresponding polluted reefs was highest. Here, the dissimilarities were mainly explained by the y-axis of the PCO, indicating a general loss of coral cover, a small shift towards more massive and stress-tolerant species and an increase of non-living substrate, such as rock (mainly dead massive corals), sand and silt ( Figure 2B). While at RAB 4P , such indicators of reef deterioration most likely relate to the high particle load (particulate C) [34], the reason of reef deterioration at LIT 6P cannot clearly by assigned to any of the measured water parameters and might therefore be caused by other stressors. Stressors, related to the shrimp farm may include the support of harmful/toxic algae and increase abundances of chain-forming diatoms and filamentous Cyanobacteria (Trichodesmium sp.) [2], as well as the hypothetic presence of pathogens in the aquaculture effluents. Furthermore, the high abundance of coral-feeding Acanthaster plancii observed in the year prior the survey (personal observation) may have caused substantial coral loss, as well. Although some structural diversity is still maintained by the presence of rocks and dead coral boulders, habitat destruction already led to a strong reduction in fish abundance ( Figure 2B).
The shift from non-polluted to polluted reefs off Yanbu 3 and to a lower degree also off Jeddah 5 followed the x-axis of the PCO, i.e. these polluted near-shore reefs were more similar to the southern reefs than their corresponding non-polluted off-shore reefs (Figure 2A). This suggests that the coral assembly structure changed little, whereas higher nutrient / food supplies near-shore (indicated by increased values of  nutrient-related parameters) led to higher fish abundances. Conversely, fishes seemed to be able to control for potentially increased growth of algae and other sessile organisms (low abundance, Figure 2B), which may otherwise hamper coral recruitment or coral growth [35,11].
Concluding, our results show that the assembly structure of coral communities is surprisingly similar, despite strong changes in environmental conditions along the latitudinal gradient of the Saudi Red Sea coast. Complex reef habitats are formed by corals of various functional taxa, with most taxa featuring a high functional redundancy (species richness). This, together with potentially Red Sea specific adaptations of some corals, seems to maintain reef functionality throughout the gradient. In contrast, shifts in fish community assembly and abundance, as well as in non-coral invertebrates, seem to follow nutrient / food availability rather than habitat complexity. However, these fishes and non-coral invertebrates in return, control the sessile benthic community structure and reduce the abundance of potential space competitors for corals thereby maintaining coral recruitment success. In the context of global change, these results indicate that reef communities might be able to maintain its ecological functionality under predicted future scenarios, given that functional redundancy is high and changes are slow enough to allow adaptation processes. In case fast and dramatic changes occur (major nutrient increase, sedimentation, pollutants), reef functionality will be strongly compromised as it was evident at the most deteriorated reefs close to a source of pollution.

Measurement of environmental parameter
For chl a, 1.5 l water sample was filtered through a GF/F filter (0.7 µm, Whatman, USA) after pre-filtration through a 50 µm mesh to remove the majority of zooplankton. Chl a was extracted from the frozen filters with 90 % acetone and determined fluorometrically (specro-fluorometer SFM 25, Kontron Instruments, Switzerland) after calibration with a spinach chl a standard [36]. For particulate matter, 1.5 l water samples were filtered through a pre-combusted GF/F filter and the filter were stored frozen. The filters were dried and the concentration of particulate C was measured with an elemental analyzer (Flash 2000, Thermo Scientific, USA, calibrated with Acetanilid). Untreated 50 ml water samples were taken for total N (TN) and P (TP) analyses, stored in PE-bottles and frozen until analyses. Analyses were the same as for PO 4 and NO 3 after all P and N was oxidized to the corresponding inorganic form as described by Grasshoff et al. [36,37,38]. Measurements were conducted with a spectra-photometer (U-2900, Hitachi High-Tech, Japan) and a 10 cm quartz cuvette.