Disentangling the impacts of heat wave magnitude, duration and timing on the structure and diversity of sessile marine assemblages

Study site and sessile assemblages

Experimental assemblages were grown under natural conditions on panels deployed at Torquay Marina, southwest U.K. (50°27′31.44″N, 3°31′45.20″W). The study site is relatively unimpacted by freshening events and supports a high abundance and diversity of sessile invertebrate fauna. Submerged hard surfaces are colonised by a rich marine fauna, dominated by ascidians (e.g., Ciona intestinalis, Diplosoma listerianum, Botryllus schlosseri), bryozoans (e.g., Electra pilosa, Cryptosula pallasina) and polychaetes (e.g., Spirobranchus sp.). Filamentous algae and macroalgae are also common on vertical and upward-facing surfaces that receive sufficient light. Several non-native species have also been recorded at the study site (e.g., the bryozoans Watersipora subtorquata and Bugula neritina, the ascidians Asterocarpa humilus and Corella eumyota).

Experimental design

The experiment comprised of 3 phases; (1) a ‘colonisation’ phase in the field, (2) an ‘experimental’ phase within mesocosms of either 1 or 2 weeks in duration and (3) a final ‘recovery’ phase in the field (Fig. 1). The experiment was initiated at 3 different times throughout the summer of 2013 to investigate the effects of HW timing on ecological responses. For Phase 1, 36 settlement panels (20 × 20 cm, constructed from black double-skinned sheets of polypropylene, ‘Correx’) were attached to a fibreglass reinforced plastic (FRP) grid in 6 rows of 6 panels (Fig. S1). The FRP grid (L = 148 cm, W = 121 cm, D = 2.5 cm) was suspended horizontally at a depth of 1.5 m directly beneath a pontoon near to the seaward entrance of the marina. Panels were faced down towards the seabed in order to reduce light and sedimentation levels and therefore select for sessile fauna (rather than algae). Faunal assemblages were targeted because they are relatively well known (in terms of both taxonomy and species-specific physiology), often comprise multiple non-native species, and are relatively speciose. After a colonisation phase of ∼6 weeks, 30 panels were transported in cool boxes to the mesocosm facility at Plymouth Marine Laboratory (PML) within 1 h. Six panels were selected at random to remain on the grid throughout the experiment to serve as field-based controls. Seawater temperature at the study site was continuously monitored (every 20 min) with a ‘Hobo’ temperature logger attached to the experimental grids.

Within the mesocosm facility, 15 × 1 tonne capacity tanks were employed for Phase 2 (Fig. S1). Each tank held 0.58 m³ of freshly collected seawater which was constantly circulated using a peristaltic pump (Sci-Q 323; Walston Marlow Bredel, Falmouth, UK) and aerated by 2 air stones. Tanks were randomly assigned to one of 3 HW Magnitude treatments; 5 control tanks (hereafter ‘C’) were maintained at ambient sea temperature at the collection site (∼16.0 °C), 5 tanks were maintained at 3 °C above ambient (hereafter ‘T1’) and 5 tanks were maintained at 5 °C above ambient (hereafter ‘T2’). Temperatures in the control tanks were maintained through equilibration with air temperature within the mesocosm facility, which was controlled with a Refrigerating Heat Pump unit (Bartlett Refrigeration Limited, Exeter, UK). Temperatures in the treatment tanks were elevated and maintained using aquarium heaters (AquaVital 300W submersible heater thermostat) and electrical timers (Fig. 1). HW Magnitudes were representative of actual short-term seawater warming anomalies recorded in marine habitats (e.g., Jones, Berkelmans & Oliver, 1997; Garrabou et al., 2009; Wernberg et al., 2013). Two panels were selected at random and allocated to each tank, 1 for a 1 week HW duration and the other for a 2 week duration treatment (see below). Panels were held horizontally in the tanks, facing downwards, by attachment to a 1 m² suspended plastic mesh.

Throughout the experimental phase, ‘Hobo’ loggers were deployed in each tank to record temperature every 15 min, and salinity and temperature were manually recorded 3 times per week and adjustments made as necessary. Approximately 0.2 tonnes of seawater per tank (i.e., 1/3 of the total volume) was replaced each week. Natural light regime was approximated using daylight simulation lights within the mesocosm with an average 12-h photoperiod per day. Throughout Phase 2, panel assemblages were fed to satiation (as indicated by continuous faecal production and replete ascidian intestines) through the addition of an algal cell suspension (800 ml per tank, three times per week). The suspension comprised a mix of the unicellular algae Isochrysis galbana, Tetraselmis suecica and Rhinomonas reticulata (cultured at PML) in seawater, with an average cell density of ∼5 × 10⁶.

Following a 1 week HW duration, 15 panels (i.e., 1 panel per tank) were returned to their original grid in Torquay Marina for a recovery period (Phase 3). After a further week, the remaining 15 panels were also returned to their original grid (Fig. 1). Following a recovery period of ∼3 weeks for the 1-week HW duration and ∼2 weeks for the 2-week HW duration panels, all panels were harvested and preserved in 80% IMS for subsequent scoring and analysis (see below). The entire 10-week experiment (Phase 1–3) was repeated 3 times, with each experimental HW ‘Timing’ commencing ∼2 weeks apart (from May 25th until June 21st), so that the timing of the simulated warming events was staggered throughout the summer (Fig. 1).

Panel analysis

Panel assemblages were quantified at the end of Phase 1, 2 and 3. Panels were removed from the mesocosm tanks (Phase 1 and 2) or IMS containers (Phase 3) and submerged in a shallow Pyrex dish of seawater/IMS for examination under a dissecting light microscope. A wire mesh overlay, which projected 121 small squares (∼1.3 × 1.3 cm) for scoring, was used to analyse the inner 15 × 15 cm portion of each panel (the peripheral 2.5 cm was not scored to account for edge effects). Sessile organisms were identified to as high a taxonomic level as possible, always to genus and predominantly to species, and quantified by scoring every individual (solitary taxa) or colony (colonial taxa) occurring in each sub-sampled square of the scoring area. Any individual or colony occupying more than one square was scored multiple times accordingly, thereby generating an abundance score weighted by spatial coverage. Colonies or individuals that were heavily overgrown and clearly smothered by other encrusting taxa were assumed to be non-functional and were not counted. After the final analysis (Phase 3), all sessile fauna was scraped from the panel, dried at >60 °C for 48 h and then weighed to determine the total dry biomass of each panel.

Statistical analysis

Data were analysed as 2 independent experiments based on HW duration (i.e., 1 and 2 week exposures), as direct comparisons were not appropriate due to potential confounding effects of total time spent in the mesocosms and in recovery (Phase 3). Instead, the effects of HW Magnitude were examined for each dataset separately before making qualitative comparisons of their relative effects between the 1 and 2 week HW durations. Initially, patterns in assemblage structure at the end of the experiment (i.e., after the Phase 3) were compared between the HW Magnitude treatments (i.e., control, T1, T2 and field controls) for each HW Timing (i.e., Timing 1, 2 and 3). A Bray-Curtis similarity matrix based on square-root transformed abundance data was used to construct Principal Coordinate Analysis (PCO) plots. Permutational multivariate analysis of variance (PERMANOVA) was used to examine differences between treatments, by employing a 2-way crossed design with HW Magnitude and Timing as fixed factors. Tests used 4,999 permutations under a reduced model, with significance accepted at P < 0.05. In addition, variability in assemblage level metrics (i.e., total abundance, species richness, total biomass) at the end of the experiment was examined with univariate permutational ANOVA (using the same model described above but with similarity matrices based on Euclidean distances).

As panel assemblage structure was highly variable after the colonisation stage (Phase 1), further analyses were conducted on relative changes in assemblage structure, in addition to absolute assemblage structure at the end of the experiment. Here, the percent cover of each taxa at the end of the colonisation phase (Phase 1) was subtracted from the final percentage cover recorded at the end of the experiment (Phase 3). Thus, a positive value indicated an increase in spatial cover from the initial colonisation phase through the experimental and grow-out phase whereas a negative value indicated a decrease. Multivariate assemblage-level analysis were then conducted on relative changes, first by adding 1,000 to each score to generate positive values, then by constructing similarity matrices based on Euclidean distances of untransformed data. For the 1 week and 2 week HW duration experiments independently, differences in assemblage structure between HW Magnitude and Timing were then visually and statistically examined using the same approach described above for patterns of absolute assemblage structure. Where a significant interaction between HW Magnitude and Timing was detected, pairwise comparisons within each level of the interaction term were conducted. Patterns of relative change in the abundance of dominant species, both native and non-native, in response to the HW Magnitude treatments were also examined, using one-way PERMANOVA to test for differences between HW Magnitude within each Timing. All statistical procedures were conducted with the PERMANOVA add-on (Anderson, Gorley & Clarke, 2008) for PRIMER v.6 software (Clarke & Warwick, 2001).

Heat wave simulations

Temperatures in the treatment tanks were stable throughout the experimental period (Phase 2), with mean values (±SD) of 16.9 ± 0.5 °C, 19.8 ± 0.3 °C and 21.8 ± 0.4 °C for C, T1 and T2, respectively. Control tank temperatures were well-aligned with monthly mean temperature for the study site (based on in situ logger data for 2010–2013). Daily mean temperatures for the field site indicated that a short-term warming event occurred through mid-July 2013 (Fig. 1). Even so, daily mean temperatures for T1 and T2 were greater than the maximum daily mean recorded in the field (Fig. 1).

Assemblage level responses

A total of 34 taxa (24 species and 10 distinct genera), 7 of which were classified as non-native, were recorded on the 108 experimental panels (Table S1). Overall, panel assemblages were well developed by the commencement of the experimental stage (Phase 2), with an average richness of 14.5 ± 0.5 species per panel and an average total faunal cover of 67.4 ± 3.5% (Fig. S2). At the end of the experimental runs (Phase 3) average panel richness and total cover (%) was 15.4 ± 0.3 and 80.3 ± 6.7%, respectively (Fig. S2). Following colonisation (Phase 1), panel assemblages were fairly variable in structure, most likely due to high spatial variability in settlement and recruitment (Fig. S3). Even so, panel assemblages within each experimental event were generally clustered together whereas no partitioning between the randomly assigned treatments was observable (Fig. S3).

Partitioning in absolute multivariate assemblage structure at the end of the experiment was principally related to the timing of the experimental run, in that panel assemblages were clustered by the ‘HW ‘Timing’ rather than by HW ‘Magnitude’ for both the 1 week and 2 week durations (Fig. 2). Within each Timing, there was no clear partitioning in multivariate assemblage structure between the Magnitude treatments (i.e., C, T1, T2), indicating that the HW simulations had little effect on overall assemblage structure based on absolute abundances. However, the assemblages on the field controls (FC) were generally dissimilar to those on the experimental panels (Table 1 and Fig. S4), suggesting that the 1 or 2 weeks spent in the mesocosm facility influenced community succession to some degree, as could be expected. Even so, when compared to variability between Timings, differences in assemblage structure between field controls and experimental controls were minimal (Fig. 2). For the 1 week HW duration, a significant effect of Magnitude was detected, and post-hoc tests indicated that the C panels supported distinct assemblages from the T2 panels (Table 1). For the 2 week HW duration, significant differences between assemblages on C panels and those on T1 and T2 panels was recorded following HW Timing 2 (Table 1). Univariate permutational ANOVA indicated that total abundance and species richness differed between treatments, but only following the 2 week duration experiments (Table S2). Pairwise tests showed that the Field Controls differed from the experimental treatments (which were all statistically similar), again suggesting some effect of the experimental procedure. Plots of species richness, total abundance and total biomass (dry weight) at the end of the experiment (after Phase 3) showed no clear differences between the HW Magnitude treatments (Fig. 3).

Analysis of variability in assemblage structure based on relative changes in abundance (i.e., the difference between initial abundances after Phase 1 and final abundances after Phase 3) showed that responses to the HW Magnitude treatments varied considerably between HW Timings (Fig. 4). For example, variability in shifts in assemblage structure was much greater for Timing 1 and 2 compared with Timing 3, for both the 1 and 2 week HW durations (Fig. 4). PERMANOVA detected significant differences in assemblage structure (based on relative changes in abundance) between Timings for both HW durations (Table 2). For the 1 week HW duration experiments, the main and interactive effects of Magnitude were non-significant (Table 2). However, for the 2 week HW duration experiment, a significant Magnitude × Timing interaction was detected (Table 2). Post-hoc pairwise comparisons between HW Magnitudes for each HW Timing showed that assemblage structure for the T2 panels was significantly different to those observed for the C and T1 panels following Timing 1 (Table 2). This was also evident in the PCO plots, with clear partitioning between T2 assemblages and C/T1 assemblages following Timing 1 (Fig. 5). For Timing 2 and 3, however, no significant differences were detected between HW Magnitudes (Table 2), although some partitioning between T2 and C/T1 assemblages was again evident following Timing 2 (Fig. 5).

Species level responses

There were no consistent trends in the responses of native dominant species to the HW Magnitude treatments (Fig. 6). Relative changes in abundances were variable, both between species and between Timings for the same species (Fig. 6). The only significant differences between HW Magnitudes were observed for the colonial ascidian Botryllus schlosseri, after 2 of the 1-week HW duration events, and the bryozoan Electra pilosa, after a 2-week HW event (Fig. 6 and Table S3). For B. schlosseri, the rate of increase in abundance was significantly lower on the T2 panels compared with the control panels (Fig. 6 and Table S3). For E. pilosa, which decreased in abundance during the HW Timing 1, the rate of decline was greater on the T2 panels compared with the control panels (Fig. 6 and Table S3).

The responses of non-native species were similarly variable and no clear trends were observed (Fig. 7 and Table S4). The ascidian Corella eumyota and the bryozoans Bugula neritina and Tricellaria inopinata generally increased in abundance between Phase 1 and Phase 2 and the magnitude of increase did not vary between the HW Magnitude treatments (Fig. 7 and Table S4). The compass ascidian, Asterocarpa humilis, decreased in abundance in HW Timing 2, and the magnitude of decline was significantly greater on the HW Magnitude treatments T1 and T2 compared with the control panels (Fig. 7 and Table S4).