Fish nursery value of algae habitats in temperate coastal reefs

Fish study species

The study focused on three common and abundant fish species of littoral rocky reefs in the Mediterranean: Diplodus vulgaris (common two-banded seabream), Symphodus ocellatus (ocellated wrasse), Coris julis (Mediterranean rainbow wrasse). These species were chosen as they occurred at sufficiently high abundances to engage with the posed research questions. Settlement of the two-banded seabream, occurs in very shallow rocky coastal areas and seagrass meadows but are thought to migrate quickly to deeper waters following settlement (Harmelin-Vivien, Harmelin & Leboulleux, 1995). Settlement peaks for this species have also been reported to occur in two settlement pulses, one in early November and the other in January-through to March (Vigliolat et al., 1998; García-Rubies & Macpherson, 1995; Harmelin-Vivien et al., 1985). Biagi, Gambaccini & Zazzetta (1998) describes the settlement periods to occur in December–January and March. Within the Balearic Islands, the authors observed the presence of recently settled juveniles in early spring (April–May).

The settlement period of the two wrasse species has been reported to occur during late summer between July to August (García-Rubies & Macpherson, 1995; Biagi, Gambaccini & Zazzetta, 1998; Bussotti & Guidetti, 2011) while Raventós (2006) reports settlement of S. ocellatus slightly earlier from June to mid-July. Within the present study, the presence of recently settled juveniles was observed at the end of August by the authors. The two species of wrasse have been reported to settle predominantly in rocky habitats with high algal cover (García-Rubies & Macpherson, 1995; Biagi, Gambaccini & Zazzetta, 1998; Bussotti & Guidetti, 2011; Félix-Hackradt et al., 2014). Juveniles of D. vulgaris have been reported to feed predominantly on micro-crustaceans (Altin et al., 2015). Adults of S. ocellatus feed predominantly on small crustaceans and molluscs with a suspected tendency towards herbivory (Kabasakal, 2001) and adult C. julis feed on gastropods, sea urchins and small crustaceans (Sinopoli et al., 2016). Presently there is no information on the diet of the juvenile stages of these two wrasse species.

Study area and sampling design

Littoral habitats were studied in the Balearic Archipelago, in the western Mediterranean. Surveys were conducted in early spring (April–May) and late summer (August–September) of 2014, on the islands of Mallorca and Menorca. Sampling in Mallorca took place around the northwestern part, while sampling in Menorca was primarily conducted around the western part (see Fig. 1). Sampling at each island and during each period was conducted in 10 consecutive day surveys. The sampling stations comprised rocky reef zones with a depth of 7–10 m. The same depth range was chosen not to introduce any depth related confounding effects. Furthermore, at this depth we encountered extensive algae communities and juvenile fish. From an operational point of view, this depth allowed for comfortable diving and sufficiently long bottom time for surveying. Stations were haphazardly chosen to cover a variety of habitats, including different physical (geomorphology) and biological (algae cover) properties. Eight stations were sampled per sampling period and island.

In Mallorca the linear distance between consecutive stations was on average 4 km and the survey area extended over a coastal strip of approximately 55 km. In Menorca, the distance between consecutive stations was 7 km on average and the survey area covered a coastal strip of approximately 75 km. Some of the survey sites were sampled both in spring and late summer, while other sites were only sampled within one season. The overall number of samples was however balanced between the two seasons and islands (see above).

Fish and habitat census

At each sampling station the fish community and the habitat were assessed by diver underwater visual census. The fish assemblage was assessed in 6 replicates of 15 × 2 m transects with a horizontal gap between each of at least 20 m. All fish sighted along the transect were recorded and their size was estimated to the nearest cm (Harmelin-Vivien et al., 1985). Benthic composition along the transects was characterized using the following categories: sand, pebbles and gravel, seagrass and rock. The habitat complexity of the substrate was also assessed, using a five-point scale index for rugosity, presence of different sized boulders and amount of refuge spaces (see Appendix S1 for field sampling protocol).

Algae cover, composition and height was measured in three haphazardly placed 50 × 50 cm quadrats along the transect. Quadrats were further subdivided into 25, 10 × 10 cm squares using nylon string. These sub-squares were used to estimate the % cover of 8 algae morphotypes (Fig. 2) and unvegetated barren patches. If a morphotype occurred in a sub-square, the square was counted. For each morphotype, the total number of squares it occurred in was recorded, as well as their average height in cm form the rock surface. Algae morphotype cover in m² within transects was calculated by taking the m² recoded for rocky reef habitats overgrown by algae (as opposed to being occupied by sand or pebble substratum) and this area was dividing proportionally according to the % cover of each morphotype obtained by the quadrate samples. The following algae morphotypes represented by various related species that are commonly encountered in the rocky littoral zone were used (Fig. 2): erect tree like 1-ET (Cystoseira spp.), soft leaf like 2-SL (Dictyopteris polypoides and Dictyota spp.), filamentous 3-FI (Dictyota dichotoma var. intricate, Dictyota spp. and occasional Hincksia spp.), tubular 4-TU (Cladostephus spongiosus), plumose 5-PL (Asparagopsis spp.), bulbous tree like 6-BT (Halopteris spp.), leathery bands 7-LB (Padina pavonica), turf forming 8-TF (dominated by Corallinaceae such as Haliption vigatum). For data from dive transect see Dataset S1.

Sampling of algae associated fauna and prey communities

To determine prey availability for juvenile fish, the associated fauna of dominant macroalgae was sampled for each station. At each site, the three most dominant algae types were sampled by collecting 6 × 113 cm² samples. A circular tube of 5 cm height and 11.5 cm diameter was used to define the surface area for algae cuttings (0.01 m²). A 0.55 µm meshed sampling bag was draped over the piece of tube to collect algae cuttings and to retain associated fauna. The samples were transferred to plastic bags and stored frozen. Once defrosted, the samples were pooled and rinsed into a receptacle using fresh water. The algae were dryed with kitchen paper and the wet weight recorded from each station. The water from the receptacle was passed through a 300 µm mesh sieve and any retained fauna was processed. All large fauna (>5 mm) were preserved in 70% ethanol. Using a Folsom plankton sample splitter, the remaining sample, was split up to 4 times depending on its overall volume. The final split fraction of the sample was re-sieved over a 300 µm mesh and preserved in 70% ethanol. The number of sample splits was recorded for each faunal sample to later enable calculation of faunal abundance and biomass of the whole sample.

All fauna retained from the samples was sorted into large taxonomic groups (see Appendix S2 for a full list of groups). A photograph was taken of each taxonomic group using a standard digital camera (Canon powershot G7; Canon, Tokyo, Japan) for the large fraction and a digital camera (EC3) attached to a Leica stereo microscope (MZ16; Leica, Wetzlar, Germany) for the small fraction. The photograph served to measure the size of the organisms and to quantify their abundance. Image analysis software, ImageJ (Rueden et al., 2017), was used to measure the size of all organisms using the polyline measurement tool. The pixel length of each measurement line was converted to µm using values attained from calibration images. Length measurements were furthermore converted to biomass using published length-weight relationships of respective taxonomic groups (see Appendix S2 and Dataset S2 for more details on the size-mass conversion and Dataset S3 for data on associated invertebrate fauna size measurements).

Since not all invertebrates collected can be consumed, by juvenile fish, due to their size, stomach content analyses were carried out on a subset of fish. This allowed to determine the taxonomic composition and size of the fauna consumed to more accurately estimate prey densities within algae at each station. Approximately 50 fish of each species were used. These were selected at random, from different stations at which fish samples were available (see protocol of fish sampling below). The taxonomic group and size of taxa were determined from photographs in the same manner as for the determination of algae associated fauna. Only prey items that had an intact body outline were measured. For data of prey sizes found in stomachs see S6.

To determine the upper size limit of suitable prey for juveniles of each study species (those below 60 mm total length) we used quantile regressions analysis (Cade & Noon, 2003). Only significant upper quantiles were considered, starting from the 95th quantile moving down in steps of 5. Since some invertebrate taxa had an elongated, and others had compact body shape, upper prey size limits for respective prey body shapes were calculated for each fish species. The cutoff point for prey sizes to be considered for the determination of prey availability in algae was the value where the upper quantile regression line intersected with the 60 mm total length of the study species (i.e., set length of a juvenile fish) (Cade & Noon, 2003). Furthermore, we only considered prey taxa that cumulatively contributed to at least 90% of the prey taxa found in stomachs thus excluding rare species that were only occasionally or accidentally been ingested. As not all algae morphotypes were sampled in both islands, the data on the abundance and biomass of prey per algae morphotype were therefore presented using pooled data from both islands for the season in which juvenile fish of respective study species occurred i.e., D. vulgaris spring and summer for C. julis and S. ocellatus.

Total abundance and biomass of prey within a transect were estimated using the area cover (m²) by different algae morphotypes multiplied by the abundances and biomass of prey items found (per m²) within respective algae cuttings (see above). Organisms were only considered prey if they were of the relevant taxa and sizes as identified by the stomach content analysis described above.

Sampling of juvenile fish

Using hand-held nets, a sample of approximately 30 juvenile fish of each of the study species were collected from each of the 32 sampling stations (see Dataset S5 for fish size data). In addition to the stomach content analysis, the fish sample was used to evaluate the size distribution of juvenile fish.

Wave exposure data

As juvenile fish may experience higher dispersal from habitats with high wave action, and algae may be influenced by physical wave stress, wave exposure was considered in the study (Spatharis et al., 2011). Wave stress data for each station was calculated from a dynamical coastal wave model provided by the Coastal Ocean Observing and Forecasting System located in the Balearic Islands (SOCIB) using past real weather events. The mean wave exposure, as well as the wave exposure of the 4 months prior to sampling, was used as an environmental parameter in the analysis of data.

Statistical analysis

Potential prey availability in different algae morphotypes for juvenile fish

Potential prey availability was calculated considering prey sizes, taxonomic groups and the season relevant to the juveniles of the respective fish species (<60 mm). Potential prey availability was compared between morphotypes using a one-way ANOVA and post-hoc Tukey pairwise comparison tests. The data of both islands were pooled to compare all algae morphotypes since for some algae morphotypes insufficient specimens were sampled to be able to consider both islands separately in the analysis. Potential prey density was log-transformed prior to analysis to meet model assumptions regarding normality and homogeneity of variance.

Relationship between different algae morphotypes and juvenile fish abundances

The relationship between juvenile fish abundance and algae morphotype cover (m²), as well as total prey biomass per transect, was modeled using a Generalized Linear Mixed Model (GLMM) with a negative binomial error distribution and a log-link function. Within the model the station was considered a random factor, to address the dependency structure of dive transect from the same station. The cover by morphotypes, as well as total prey biomass per transect, were considered as fixed factors. All variables introduced to the model were correlated prior to analysis to identify any collinear variables to be removed or pooled prior to modelling. Model selection procedures were adopted whereby non-significant variables were removed until only significant variables were contained in the final model, using the drop-one procedures in R. Models were checked for over dispersion and the residuals were visually examined (Zuur et al., 2009; Zuur, Ieno & Elphick, 2010).

Algae morphotype composition and height across sampling seasons and islands

Algae morphotype composition within sampling areas and season were compared by PCA ordination of the dive transects sampled. The PCA explained 60.5% of the variability of the algae cover data (Fig. 3, Table S1). The ANOSIM pairwise comparisons analysis verified that there were significant differences in composition of morphotypes between all island and season combinations (p < 0.001, Fig. 3). The largest differences were found between transects surveyed in summer in Menorca and all other island-season combinations (see Fig. 3 and Table S1). Algae cover was most similar between spring and summer in Mallorca (see Fig. 3). Transects sampled in spring in Menorca were similar to those sampled in Mallorca (see overlap of PCA space Fig. 3 and lower summed Euclidian squared distance Table S2). Overall, seasonal changes in algae cover for both islands were similar for most morphotypes (Table 1). In both islands, the average transects cover of 1-ET (Cystoseira spp.) were lower in summer compared to spring as was the cover of 2-SL and 3-FI (Dictyota spp. and similar). In contrast, the cover of 7-LB (P. pavonica) was higher in both islands during summer as did the occurrence of barren patches (no vegetation over rock) (Table 1). However, the magnitude of seasonal change was more pronounced in Menorca. Furthermore, transects surveyed in summer were significantly different to those sampled in Mallorca (Fig. 3 and Table S2) with higher cover in morphotype 7-LB (P. pavonica) and lower turf forming morphotypes 8-TF (Corallinaceae) compared to Mallorca (Fig. 3 and Table 1). Morphotypes 2-SL and 3-FI (Dictyota spp. and similar) formed part of many transects in Mallorca in late summer, while these morphotypes were almost absent from transects in Menorca (Fig. 1 and Table 1).

The distance-based linear model (DistLM) relating environmental variables to the PCA ordination of algae cover showed that of the 6 environmental variables considered, three (Temperature, wave stress and rugosity) proved to show significant relationships with the PCA ordination pattern (Fig. 3 and Table S3). The model with the best fit (lowest AIC) containing these three variables had an r² of 0.2 thus explaining about 20% of the variability. Mean depth, slope and herbivore density (Salpa salpa and urchins) were not significantly correlated with the ordination (Table S3). Temperature and wave stress were associated with PC1 as indicated by the horizontal orientation of the eigenvectors in Fig. 3 and thus could be related to the distinct algae cover in Menorca over the summer period. In general, average wave stress was higher (H_s 0.7) and temperature lower (25.8 °C) in Menorca compared to Mallorca during summer surveys (H_s 0.4 and 26.9 °C respectively).

Additional to the surface cover, the height of algae varied between morphotypes, seasons and islands (Fig. 4). The height of morphotypes was higher in Mallorca compared to Menorca. Seasonal changes in the height of cover were, however, relatively consistent over both areas. Overall, Cystoseira spp. (1-ET) decreased in height between spring and summer surveys (Fig. 4). In Mallorca Cystoseira spp. decreased in height from spring to late summer from an average 11.2 cm to 4.3 cm, while in Menorca it decreased from 7.6 cm to 5.1 cm. These reported changes were statistically significant (Tables S4 and S5). Other morphotypes decreased less in height (2-SL) or approximately maintained their height (3-FI, 6-BT and 8-TF). In summer, P. pavonica (7-LB) was the only algae that significantly increased in height in both areas from 4.2 to 6.9 cm in Mallorca and 3.8 to 6.8 in Menorca. During the summer period in Mallorca filamentous morphotype algae (3-FI) were found to be the highest with 10.6 cm, while in Menorca both filamentous (3-FI) and bulbus tree-like (6-BT) were the highest with 7.8 and 7.9 cm respectively.

Taxonomic composition and size of prey items found in juvenile fish stomachs

Stomach samples of the three-study species showed that all three species shared many prey taxa (see Fig. 5). D. vulgaris and S. occelatus diets were dominated harpacticoids (both 53%) while stomachs of C. julis predominantly by gastropods (50%) and to a lesser extent Harpacticoids (34%). D. vulgaris stomach samples were furthermore dominated by Ostracods (14%), Amphipods (9%) and Gastropods (9%), while S. occelatus also contained sea mites (Acari 34%) and Gastropods (15%).

For all species, significant upper quantiles (p < 0.05) were found determining the upper limit of prey sizes that were consumed at a certain size (Fig. 5). Comparing the upper limit of prey sizes consumed with increasing size of fish showed that fish below 60 mm, of the three-study species consumed very similar size prey when considering compact body prey such as Harpacticoids or Amphipods. In general prey sizes were between 1.33–1.61 mm. Both D. vulgaris and C. julis consumed elongated prey consisting mainly of Polychaetes reflected in the % contribution to the diet, 3% and 6% respectively (Fig. 5). Other elongated taxa such, Diptera larvae, Caprellidea and Tanaidacea contributed less to this prey category. Stomachs of S. occelatus had too few elongated prey items to perform an upper quantile regression analysis. The size of elongated prey found in C. julis were slightly larger for fish at a size of 60 mm compared to D. vulgaris. The size elongated taxa consumed was below 6.18 mm and 7.83 mm respectively (see Fig. 5). Overall the stomach content reflected well the dominant taxa found in the different algae morphotypes (Table S6).

Potential prey availability in different algae morphotypes for juvenile fish

Comparing potential prey abundance and biomass by algae morphotype using one-way ANOVA analysis showed that there were significant differences considering the three-study species (Table S7 and Fig. 6). Post-hoc Tukey tests showed that in general 1-ET and 6-BT held significantly higher prey densities and biomass compared to other algae morphotypes (S15 and Fig. 6). Overall, the highest prey densities or biomasses in absolute terms were, irrespective of season and fish species, found in the three structurally complex algae 1-ET, 6-BT and 8-TF, compared to the more structurally simple algae, such as 2-SL, 3-FI and 7-LB. The general pattern between islands and seasons observed in prey abundances and biomass were reflective of patterns when considering all associated fauna (Fig. S1). Note that overall abundances and biomass of associated fauna were considerably higher within all algae morphotypes in spring compared to summer (Fig.  S1).

Relationship between different algae morphotypes and juvenile fish abundances

The total number of juvenile fish observed over the visual transects surveyed varied considerably between islands and species. For D. vulgaris we only encountered 39 fish in 96 transects during the spring period of which 35 were recorded in Mallorca (mean size 43.8 mm Mallorca and 46.1 mm Menorca). In contrast, similar numbers of C. julis were recorded at both islands during the summer sampling period see Table 2 (mean size 48.7 mm Mallorca and 32.7 mm Menorca). However, for S. ocellatus the number of juveniles varied considerably between islands. In 48 transects we observed 486 fish in Mallorca as opposed to only 63 in Mallorca (mean size 28.1 mm Mallorca and 33.3 mm in Menorca). This island scale difference was also reflected in other Symphodus species (see Table 2).

Due to the low abundance of D. vulgaris juveniles in visual census transects we did not investigate their relationship with algae cover and total prey biomass within transects using a Generalized Linear Mixed Modelling (GLMM) approach. Furthermore, we restricted this type of analysis to Mallorca for S. ocellatus where we had a sufficiently large sample size. C. julis was analyzed considering data from both islands.

Prior to the GLMM analysis we correlated the explanatory variables to detect any collinearity. As we found a positive correlation between 2-SL and 3-FI (Pearson correlation coefficient 0.68) and as the two algae morphotypes were structurally and taxonomically similar, their cover was pooled to avoid collinearity.

For C. julis neither algae cover, nor the total prey biomass had a significant effect on juvenile abundances. For S. ocellatus the best model including only the significant variables contained the algae morphotypes 2-SL+3-FI and 7-LB. While 2-SL+3-FI had a positive effect on fish abundances 7-LB had a negative effect (Table 3 and Fig. 7).

Conclusion and recommendations for future studies

The results of this study show that in many ways we still have a relatively rudimentary understanding of what represents a good nursery habitat for juvenile fish. Algae species such as Cystoseira that appear to have high importance in one area and season for a specific species may not necessarily have the same importance in another area or season. This may be due to the context dependent nature of juvenile occurrence and the habitat composition at the time of the study. The results suggest that fish can adapt to local situations and use algae morphotypes/species that offer the required function at the time. While at the time of our study we detected an association of S. ocellatus juveniles (3–4 cm) with Dictoytales these results do not implicate that Cystoseira habitats are not important for juvenile fish. Settlers that arrived in July-early August may well have used Cystoseira to shelter within the first two months of their lives. Equally likely, those new settlers of other species arriving in the area during the spring, when the Cystoseira forest are better developed with higher cover and canopy height may have a preference for this algae over others. While providing new insights, the present study is not able to evaluate the importance of Cystoseira habitats as a nursery habitat with certainty and as such the effect of the continuing loss of this habitat. Considerably more focused research will be required to address this question.

Abundance of juvenile fish may vary considerably in time, at any one location, due to natural stochastic processes, causing a high degree of variability. Scenarios are imaginable in which, due to a strong settlement pulse and fishery-depleted predators, there are many juveniles in a habitat. However, because of the lack of food provided by the habitat, fish may have low long-term survival potential (e.g., Macpherson et al., 1997; Planes et al., 1999; Cuadros et al., 2018). Using abundance indices alone as a tool to assess habitat quality has limitations, as these do not consider the effect of different habitat conditions on the individual fish, i.e., how a habitat affects parameters such e.g., growth and survival potential. Thus, when combined with abundance estimates, methodologies that assess fish condition, growth performance and health status may provide a more accurate picture of the prevailing conditions for fish living at a certain location. Habitat quality for juveniles should not be measured on the level of “settlement success” alone i.e., the maximum number of recently settled individuals, which may be largely a function of stochastic hydrodynamic processes (Beck et al., 2001). A better measure of habitat quality would be to estimate the survival rate of newly settled fish to reach a certain arbitrary size in good condition. However, measuring this type of “recruitment success” would be time consuming and financially costly as it would require intensive temporal sampling of juveniles from their settlement phase to their recruitment to the adult population. To increase our understanding of what constitutes high quality nursery habitats, it is important to have viable and cost-effective sampling solutions. Therefore, future studies should aim to incorporate methods and develop indicators that link both abundance estimates and population fitness parameters of juvenile fish at an agreed, predefined size range (by species)to evaluate habitat quality.