Perennial growth of hermatypic corals at Rottnest Island, Western Australia (32°S)

Ethics statement

The Rottnest Island Authority (RIA) provided approval and permits for this research (#2012/164683 and 2014/227143). The Government of Western Australia Department of Parks and Wildlife (DPaW) provided research permits and license to take fauna for scientific purposes (#SF009096 and #SF009705) and the Government of Western Australia Department of Fisheries provided research permits and exemption from the Fish Resources Management Act 1994 (#2202 and #2410).

Site description

This study was conducted at Rottnest Island, located approximately 20 km offshore Perth, WA (32°S, 115°E, Fig. 1B). We focused on two locations within the far eastern end of Salmon Bay given that this is the only known location at Rottnest Island where A. yongei coral are currently reported to occur (P. damicornis coral can be found at multiple sites). Coral and filamentous turf algae co-inhabit this inner reef environment exhibiting mixed assemblages typical of other near-shore habitats at high-latitude (Banks & Harriott, 1995; Thomson & Frisch, 2010). Rottnest Island can be subject to strong southwesterly swells and storms with wave heights in excess of 2 m (Richardson, Mathews & Heap, 2005); however, our study site within Salmon Bay is protected from the south-westerly swell by a limestone reef platform located ∼300 m seaward from shore (Marsh, 1992).

Environmental conditions

Continuous measurements of seawater temperature in Salmon Bay, Rottnest Island were made using HOBO temperature loggers (±0.2 °C, Onset Computer Corp.). Daily down-welling planar photo-synthetically active radiation (PAR in mol m⁻² d⁻¹) for Rottnest Island was measured from December 2013 through July of 2014 using light loggers (±5%, Odyssey Data Recording) calibrated using a high precision LiCor 192A cosine PAR sensor that was factory calibrated. Additional in-water measurements of light were made on the 24th January 2014 and on the 30th July 2014 to estimate depth-dependent light attenuation coefficients (k_d) in summer and winter, respectively. During each field sampling trip, discrete samples of water were collected from the study site using a 2.2-L Van Dorn bottle and measurements of pH on the total hydrogen ion concentration scale (pH_T) were made using a Schott Handylab pH 12 m (±0.03) using both seawater (‘Tris’) and NBS buffers and calibrated against a certified reference pH buffer provided by Andrew Dickson at the Scripps Institute of Oceanography (Batch #123). A SeaFET ocean pH sensor (Satlantic, Halifax, Nova Scotia, Canada) was also deployed at a depth of 2.5 m in Salmon Bay for 5 days in January 2014 (i.e., representing summer) and 12 days in July 2014 (i.e., representing winter) to measure diurnal changes in pH_T (±0.03). Additional seawater samples were collected and filtered with 0.7-µm nominal glass fibre filters (Whatman GF/F; Sigma Aldrich, St. Louis, Missouri, USA) and taken back to UWA for the analysis of total alkalinity (TA) and dissolved inorganic nutrients. TA (±5 µeq kg⁻¹) was measured using a modified version of the spectrophotometric approach developed by Yao & Byrne (1998). Dissolved Inorganic Carbon (DIC), the partial pressure of dissolved carbon dioxide (pCO₂), and aragonite saturation state (Ω_ar) were calculated from measured pH_T, TA, and temperature using the CO2SYS program (Lewis, Wallace & Allison, 1998) and assuming seasonally dependent salinity of between 35.4 and 35.8 (Lourey, Dunn & Waring, 2006). Concentrations of dissolved ammonium (±0.1 µM), nitrate + nitrite (±0.05 µM), and phosphate (±0.03 µM) were measured using a QuikChem 8500 Series 2 Flow Injection Analysis (FIA) System (Lachat Instrument, Milwaukee, Wisconsin, USA).

Coral calcification rates

Branches of A. yongei and P. damicornis were collected from replicate colonies (n = 10 for A. yongei, n = 9 for P. damicornis; one branch from each replicate colony) within Salmon Bay, Rottnest Island in December 2012. Samples were buoyantly weighed, epoxied onto acrylic plates (Fig. 2; Z-Spar™ Splash Zone Compound A-788) and then buoyantly re-weighed to determine the weight of the tile and marine epoxy. Each sample tile was then secured randomly onto an aluminium tripod (Fig. 2) or a concrete block deployed at two different locations within Salmon Bay (depths of 1.5 m and 2.5 m, respectively), close to the parent colonies. Tiled coral samples were first allowed to recover for ∼4 weeks in situ before their buoyant weights were initially measured in January 2013. Some samples died at the very beginning of the study (n = 5 of A. yongei and n = 3 of P. damicornis) and could not be included in the study, thus resulting in a final sample size of n = 5 for A. yongei and n = 6 for P. damicornis at the beginning of the experiment (i.e., January–February 2013). To increase our overall sample number, additional coral samples were deployed in February 2013 (n = 11 of A. yongei and n = 3 of P. damicornis) all of which survived, giving a total sample size of n = 16 for A. yongei and n = 9 for P. damicornis. In November 2013, we deployed even more samples of each coral species to investigate potential size effects on rates of calcification for both species (n = 6 of A. yongei and n = 4 of P. damicornis); however, data from these colonies were used solely to examine relationships between calcification and colony size (Figs. 7C and 7D) and thus were not included in the analysis of seasonal growth trends.

At 1–3 monthly intervals, coral samples on tiles were detached from their support structures, transported to the shore on snorkel, weighed on shore using the buoyant weight technique (Bak, 1973; Jokiel, Maragos & Franzisket, 1978; Davies, 1989), and then returned to the water ∼2 h later. On shore, the coral colonies were kept in aerated 50-L containers and the plastic tiles were cleared of any additional growth of epilithic organisms before being weighed. The deployment of two control tiles consisting solely of a tile with epoxy and no coral colony indicated that the long-term changes in the mass of the tiles due to residual epiphytic growth after cleaning were less than 5% of the measured changes in buoyant weight over any given growth interval.

Changes in skeletal mass were calculated from the change in buoyant weight normalized to total bioactive surface area measured using both geometric estimation (Naumann et al., 2009) and X-ray computed tomography (X-ray CT; Laforsch et al., 2008). For the geometric estimation, colony surface area was calculated by summing the total area of all branches assuming cylindrical branch geometry. For the more accurate X-ray CT, 3D scans of colonies were created using the SkyScan In-vivo micro X-ray CT (UWA Centre for Microscopy, Characterization and Analysis) calibrated against two different rectanguloid sections of coral skeleton. Reconstructions of the scans were performed by NRecon software and surface area analyses were performed in CTAn software (Bruker Microct, Kontich, Belgium). To prevent sacrificing any living sample colonies growing on tiles during the course of the experiment, the surface area of each colony was calculated using regressions of total surface area versus dry skeletal weight created from a range of separately collected colonies of P. damicornis and A. yongei whose individual masses spanned a very large range from 2–370 g (A. yongei: RMSE = 1.1 cm², R² = 0.99, y = 4.1x^0.97, n = 26; P. damicornis: RMSE = 1.0 cm², y = 2.1x^1.20, R² = 0.99, n = 14). Due to size restraints, the X-ray CT scanner was only used to analyze small colonies less than 30 g (<15 cm); therefore, for larger colonies, we estimated total colony surface area using geometric estimation and then corrected these estimates based on regressions of X-ray CT surface area against geometric estimation surface area (A. yongei: y = 1.20x–7.64, RMSE = 3.8 cm², R² = 0.98, n = 10 and P. damicornis: y = 1.19x–4.13, RMSE = 5.9 cm², R² = 0.95, n = 10). Rates of calcification for each coral colony were then calculated based on the change in dry skeletal mass, the total surface area estimated from the initial and final mass, and the number of days between buoyant weighing.

Coral linear extension rates

We measured the linear extension rates of three coral branches within five different A. yongei colonies randomly selected from across the eastern part of Salmon Bay to assess how growth (extension) rates for A. yongei varied across the total study area. Each branch was tagged 2 cm from the growing tip using coloured plastic cable ties (Shinn, 1966; Anderson, Pratchett & Baird, 2012) and measured at 1–3 month intervals. Linear extension rates were not measured for P. damicornis because the highly reticulate morphology of this coral made distal growth measurements unreliable.

Data analyses

Seasonal changes in TA, pH_T, Ω_ar and nutrients were tested using an independent t test. Repeated measures analysis of variance (rANOVA) with a Bonferroni post-hoc test was performed to test for significant differences in rates of calcification and rates of linear extension across time. After finding no significant effect of colony location on linear extension rate (F_28,70 = 1.4, p > 0.05), colony data were pooled. A priori analyses further indicated that rates of calcification at the two locations within Salmon Bay (i.e., tripod at 1.5 m and brick at 2.5 m) did not significantly differ (F_1,145 = 1.5, p > 0.05); therefore, these data were also pooled.

We tested for differences between the samples that survived the initial deployment and those new samples that were deployed ∼8 weeks later in Feb 2013 using a one-way ANOVA. The study-averaged calcification rates for corals from the two deployments differed by less than 6% for both species (A. yongei: 1.67 vs. 1.60 mg CaCO₃ cm⁻² d⁻¹; P. damicornis: 0.66 vs. 0.70 CaCO₃ cm⁻² d⁻¹) and we found that there were no significant differences between the two deployments (A. yongei: F_1,146 = 0.79, p > 0.05, P. damicornis: F_1,77 = 0.31, p > 0.05); therefore, these data were pooled. Changes in seasonally averaged calcification rates and linear extension rates were also tested using a one-way ANOVA. For all data analyses, measurements from November through April were treated as ‘summer’ and measurements from May through October as ‘winter’ based on our temperature records (Fig. 3).

Statistical significance was determined at the p <0.05 level and dependent variables were tested for normality (Shapiro–Wilk test). Before conducting rANOVA tests, the dependent variables were tested for sphericity (Mauchly’s test). When calcification rate data for A. yongei and P. damicornis deviated from the assumptions of sphericity, we adjusted the degrees of freedom using the Greenhouse-Geisser adjustment with an epsilon of 0.35 and 0.43, respectively (Quinn & Keough, 2002). Before conducting linear regression analyses we tested for independence of observations (Durbin-Watson statistic) and homoscedasticity in the variance of errors, as assessed by inspection of the regression-standardized residuals and regression standardized predicted values. All statistical analyses were performed using SPSS version 22 statistical software (IBM, Foster City, California, USA) and time series analyses were performed using Matlab version 7.2 (The Mathworks, Natick, Massachusetts, USA).

Environmental data

Weekly average water temperatures were found to range from 18.2 °C (July 2013) to 24.3 °C (February 2013) giving a seasonal temperature range of ∼6 °C (Fig. 3). These findings are consistent with earlier measurements made at another shallow nearshore site between 1978 and 1991 less than 1 km from our study site (18.5°–23.5 °C; Marsh, 1992). The daily range in hourly mean temperatures (minimum to maximum) varied seasonally but were generally minor, ranging on average from 0.4 °C in July to 1.2 °C in January. Daily average light at Rottnest Island was, however, more than two-fold higher in summer (54 mol m⁻² d⁻¹) than in winter (19 mol m⁻² d⁻¹; Fig. 4). The diffuse attenuation coefficient for seawater (k_d) in Salmon Bay measured at the depth of the tripod (1.5 m) was very low in both seasons (0.053 m⁻¹ during summer vs. 0.075 m⁻¹ during winter), corresponding to a seasonal range in light reaching the benthos of 15–48 mol m⁻² d⁻¹. Light levels were ∼80% of surface values at the tripod site (1.5 m) and ∼70% at the brick site (2.5 m). Thus, light levels these two deployment locations generally differed by only ∼10%.

TA did not differ between seasons (2,282 ± 18 in summer vs. 2,297 ± 4 in winter, t test, t(26) = − 1.19, p > 0.05, Table 1), and while daytime seawater pH_T ranged from 7.9–8.19, on average summer pH_T was only 0.03 pH units higher than winter (mean ± SE = 8.13 ± 0.02 vs. 8.10 ± 0.02, t test, t(33) = 1.04, p > 0.05, Table 1). Continuous measurements of water column pH_T over diurnal cycles indicate daily average values that were close to equilibrium with the atmosphere (8.03 in summer and 8.07 in winter vs. 8.047 and 8.048; T = 22.1 °C and 20.7 °C, TA = 2,301, and pCO₂ = 395 µatm, Fig. S3). The higher daytime discretely measured pH was not unexpected given the observed diurnal pH amplitudes of around ±0.1 for both seasons and uncertainties in their measurement (7.9–8.15 ± 0.03 in summer and 8.0–8.2 ± 0.03 in winter, Fig. S3). The magnitude of these diurnal changes likely varied over time in accordance with changes in both offshore wave climate and local rates of community metabolism (Zhang et al., 2012; Falter et al., 2013). Consequently, Ω_ar was on average ∼0.6 lower in winter than in summer based on discrete daytime samples (3.28 ± 0.12 vs. 3.89 ± 0.14, respectively, t test, t(33) = 2.54, p < 0.05, Table 1), but only ∼0.4 lower in winter than in summer based on continuously measured pH averaged over many diurnal cycles (2.67 ± 0.15 vs. 3.08 ± 0.17). Ammonium concentrations were comparable across summer and winter (0.4–0.5 mmol m⁻³, t test, t(12) = 1.12, p > 0.05, Table 1), while nitrate concentrations were three-fold higher in winter than in summer (0.77 vs. 0.25 mmol m⁻³; t test, t(13) = − 6.22, p < 0.05, Table 1), and phosphate concentrations were six-fold higher in winter than in summer (0.19 vs. 0.03 mmol m^−3,, t-test, t(13) = − 4.12, p < 0.05, Table 1).

Rates of calcification and linear extension

Colony surface area estimated from geometric estimation was linearly correlated with the more accurate surface areas calculated from X-ray CT (A. yongei: y = 0.82x + 6.99, root mean squared error (RMSE) = 3.2 cm², R² = 0.98; P. damicornis: y = 0.8 × + 5.0, RMSE = 4.8 cm², R² = 0.95, Figs. 5A and 5B). Visual comparisons between virtual 3D reconstructions of a colony with a photograph of the same colony verified the high degree of similarity between actual fragments and the 3D reconstructions based on X-ray CT scans (Fig. 6). We further found strong linear relationships between the log of the total colony surface area and the log of dry weight for both coral species (A. yongei: RMSE = 1.1, R² = 0.99, y = 4.1x^0.97; P. damicornis: RMSE = 1.0, y = 2.1x^1.20, R² = 0.99, Figs. 7A and 7B). Rates of coral calcification were predominantly independent of size as represented by total surface area in A. yongei (b ≈ 1 ± 0.03 where ΔM ∝ SA^b, R² = 0.92, Fig. 7C). The relationship between calcification rate and size was, however, less clear for P. damicornis, which exhibited a modest allometric decrease with total surface area (b ≈ 0.7 ± 0.26 where ΔM ∝ SA^b, R² = 0.18, Fig. 7D). Thus, rates of coral calcification were largely insensitive to size; results consistent with prior studies of metabolic allometry in coral (Jokiel & Morrissey, 1986; Sebens, 1987; Elahi & Edmunds, 2007).

Mean rates of calcification over the entire study ranged from 1.31–2.02 mg CaCO₃ cm⁻² d⁻¹ for A. yongei and 0.34–0.90 mg CaCO₃ cm⁻² d⁻¹ for P. damicornis (Figs. 8A and 8B). Rates of calcification were found to differ significantly for the factor of time for A. yongei (rANOVA, F_(2.8,39.6) = 7.0, p < 0.05) and P. damicornis (rANOVA, F_(3.0,17.8) = 3.9, p < 0.05). Post-hoc tests for A. yongei showed that the significant changes over time were attributed to the abrupt calcification rate declines of 25% during February 2014, which was followed by consistently low rates throughout autumn 2014; both time points were found to be significantly different from five other previous time points (Bonferroni post-hoc test, p < 0.05, Fig. 8). Similarly, post-hoc tests for P. damicornis showed that the significant changes over time were attributed to the abrupt calcification rate declines of 40% during February 2014, which was significantly different from one other time point (Bonferroni post-hoc test, p < 0.05, Fig. 8). However, following the late 2013/early 2014 summer declines, mean rates of calcification had partially recovered for P. damicornis (May 2014, Fig. 8). Average rates of calcification for A. yongei did not significantly differ between seasons (summer 2013: 1.76 ± 0.10 vs. winter 2014: 1.67 ± 0.05 vs. summer 2014: 1.60 ± 0.05 mg CaCO₃ cm⁻² d⁻¹; mean ± SE, one-way ANOVA, F_(2,129) = 1.4, p > 0.05). Similarly, average rates of calcification for P. damicornis did not significantly differ between seasons (summer 2013: 0.64 ± 0.07 vs. winter 2014: 0.82 ± 0.05 vs. summer 2014: 0.62 ± 0.08 mg CaCO₃ cm⁻² d⁻¹; mean ± SE, one-way ANOVA, F_(2,69) = 3.0, p > 0.05).

Calcification rates for P. damicornis showed a negative linear relationship with seasonal changes in temperature (y = − 0.08x + 2.48, RMSE = 0.15 mg CaCO₃ cm⁻² d⁻¹, R² = 0.45, p < 0.05, Fig. 9B). Below temperatures of 23 °C, however, rates of calcification for P. damicornis plateaued with declining temperature; thus, the relationship at lower temperatures was not significant (b ≈ 0 ± 0.07, p > 0.05, Fig. 9B). No significant correlation was found between mean rates of calcification and seasonal changes in temperature for A. yongei (y = − 0.017x + 2.0, R² = 0.003, RMSE = 0.23 mg CaCO₃ cm⁻² d⁻¹, p > 0.05, Fig. 9A). Similarly, no significant correlation was found between seasonal changes in daily light and calcification rates for either A. yongei (R² = 0.001, p > 0.05) or P. damicornis (R² = 0.17, p > 0.05).

Relatively high linear extension rates (32–74 mm y⁻¹, mean of 51 ± 4 mm y⁻¹, Fig. 8A) were found for A. yongei coral at Rottnest Island, similar to those reported more than 20 years ago (69 mm y⁻¹; Marsh, 1992). Average rates of linear extension for A. yongei significantly differed across time (rANOVA, F_(7,98) = 4.4, p < 0.05); however, post hoc tests revealed that this was a result of the significantly higher rates of extension in the month of April 2013 (Bonferroni post-hoc test, p < 0.05), which was significantly higher than four of the other time points (i.e., February 2013, May 2013, July 2013 and January 2014). Average rates of linear extension for A. yongei did not significantly differ between seasons (one-way ANOVA, F_(2,119) = 1.7, p > 0.05). No significant correlation was found between mean rates of linear extension and rates of calcification (R² = 0.12, p > 0.05).

Shallow-water coral bleaching event

In December 2013, colonies of P. damicornis and A. yongei in Salmon Bay suffered high mortality following a bleaching event. Approximately 95% of colonies living in an extensive shallow (<0.5 m) reef flat were visibly bleached (photographs of bleaching are in Figs. S2A, S2B). However, deeper colonies (>0.5 m), including the sample coral colonies growing on tiles attached to the tripod and brick, showed no visible signs of bleaching. Subsequent monitoring within four weeks following the bleaching event showed that all of the previously bleached coral had died and were overgrown by filamentous algae.