RESULTS
Mesocosm conditions: Seawater conditions within the climate treatments representing future climate scenarios approximated the desired target values for temperature and pH of +1 °C and -0.1 pH units (RCP 2.6), and +2 °C and -0.2 pH units (RCP 4.5) relative to Ambient conditions (Fig. 1, Table S1). On average, all three treatments varied with natural ambient fluctuations and were warmer and more acidic during the summer trial than during the winter trial.
Silvetia biomass: The climate change treatments within the mesocosms affected final Silvetia biomass in both the summer (ANOVA: F2,27 = 11.4, p < 0.001) and winter (F2,27 = 13.4, p < 0.001, Fig. 2, Table S2). Specifically, during the summer trial, Silvetia biomass declined in all three climate treatments relative to starting biomass (Tukey’s: p < 0.001 for all), with more pronounced declines under both future climate scenarios relative to Ambient conditions (Tukey’s: p = 0.003 & p < 0.001 for RCP 2.6 & RCP 4.5, respectively). However, Silvetia biomass loss did not differ between the two future climate scenarios (Tukey’s: p = 0.710). In contrast,Silvetia biomass increased significantly under Ambient (p = 0.004) and RCP 2.6 (p < 0.001) conditions relative to its starting biomass during the winter trial but did not change under RCP 4.5 conditions (p = 0.467). Consequently, biomass under Ambient and RCP 2.6 conditions remained similar to one another (p = 0.828) in the Ambient and RCP 2.6 mesocosms but were both higher than in the RCP 4.5 mesocosms (p < 0.001 for both). Overall, final Silvetiabiomass was higher in the winter trial (74.2 ± 5.6 g, mean ± SE) relative to the summer trial (37.8 ± 11.8 g, mean ± SE). When comparing the same climate treatments across seasons (e.g., Ambient in summer to Ambient in winter), biomass of all three winter treatments were higher than those of the summer treatments (Fig. 2).
Silvetia quantum yield: Similar to biomass,Silvetia quantum yield varied among the climate treatments in both summer (ANOVA: F2,27 = 6.5, p = 0.005) and winter (F2,27 = 0.5, p = 0.635, Fig. 3, Table S3), but this appeared to differ between the two seasons. Specifically, quantum yield was generally higher in the winter (0.63 ± 0.05 ΦPSII, mean ± SE) than in the summer (0.46 ± 0.12 ΦPSII, mean ± SE). In summer, Silvetia quantum yield varied among climate change treatments and was significantly lower under RCP 4.5 conditions relative to Ambient conditions (Tukey’s: p = 0.004), but otherwise it did not differ between Ambient and RCP 2.6 conditions (p = 0.302) or between RCP 2.6 vs. RCP 4.5 conditions (p = 0.115). In contrast, quantum yield did not vary among the climate treatments in the winter trial (ANOVA: F2,27 = 0.5, p = 0.635), though the quantum yield of every winter climate treatment was higher than the summer counterpart (Fig. 3).
Mesocosm assemblages: When climate change treatments reducedSilvetia biomass in the Silvetia Present treatments, significant changes to the associated understory assemblage followed [PERMANOVA: pseudo-F2,54 = 2.8 & 1.7, p = 0.001 & 0.156 for summer and winter (differences found via a prioritesting), respectively; Fig. 4, Table S4]. Specifically, during the summer trial when biomass of the Silvetia canopy was reduced under future climate scenarios, the understory communities shifted relative to Ambient (Pairwise tests: p = 0.012 & 0.013 for comparisons of RCP 2.6 & RCP 4.5 to Ambient, respectively, Fig. 4A, Table S5). However, the understory assemblages were not different between RCP 2.6 and RCP 4.5 (p = 0.583). In contrast, no shifts in the understory assemblages occurred under either climate change scenario relative to Ambient in the absence of the Silvetia canopies (Pairwise tests: p = 0.165 & 0.420 for comparisons of RCP 2.6 & RCP 4.5 to Ambient, respectively, Table S6), and the understory assemblages were again not different between RCP 2.6 and RCP 4.5 (p = 0.460, Fig. 4B). Similarly, understory shifts during the winter trial occurred in the presence of aSilvetia canopy, but only under RCP 4.5 conditions (i.e., whenSilvetia biomass was lower than it was in Ambient & RCP 2.6 treatments, Pairwise tests: p = 0.001 for both, Fig. 4C). In the absence of a canopy, like the summer trial, there were no differences between understory assemblages of either climate change scenario relative to Ambient (Pairwise tests: p = 0.831 & 0.065 for comparisons of RCP 2.6 & RCP 4.5 to Ambient, respectively, Table S6) or between RCP 2.6 and RCP 4.5 (p = 0.655).
When future climate scenarios shifted the understory assemblage beneath a Silvetia canopy compared to the Ambient treatment (i.e., RCP 2.6 & RCP 4.5 in summer, and RCP 4.5 in winter), we observed the same ranking in taxa with respect to their contribution to this dissimilarity. From most important to least important, this ranking was the same in these three comparisons - Centroceras ,Corallina , Chondracanthus , Laurencia (Table 1).
In the summer and relative to Ambient climate treatments, 40% of the dissimilarity with RCP 2.6 treatments was driven by a decrease inCentroceras (44.3 ± 27.5% to 33.4 ± 14.7%, mean ratio of recovered biomass to initial biomass ± SE). In addition, 34% of the dissimilarity was driven by a decrease in Corallina (from 33.4 ± 11.5% to 16.5 ± 8.9%), 23% by a decrease in Chondracanthus(20.7 ± 13.6% to 19.4 ± 12.2%), and lastly, 3% by an increase inLaurencia (from 28.3 ± 0.7% to 29.3 ± 2.0%). For summer RCP 4.5, a decrease in Centroceras (44.3 ± 27.5% to 36.0 ± 21.9%) and Corallina (33.4 ± 11.5% to 16.5 ± 7.0%) drove 40% and 30% of the dissimilarity, respectively, while an increase inChondracanthus (20.7 ± 13.6% to 28.4 ± 18.8%) andLaurencia (28.3 ± 0.7% to 30.1 ± 5.8%) drove 26% and 4% of the dissimilarity, respectively. Lastly, in winter and relative to Ambient, all genera in RCP 4.5 decreased with 31% of the dissimilarity driven by Centroceras (56.2 ± 27.8% to 29.3 ± 9.6%), 28% byCorallina (58.2 ± 11.7% to 31.5 ± 11.3%), 27% byChondracanthus (40.6 ± 25.1% to 20.1 ± 14.8%), and 14% byLaurencia (from 43.4 ± 16.9% to 28.2 ± 0.4%, Table S7).
Overall, the top two genera (Centroceras and Corallina ) consistently declined under future climate scenarios relative to Ambient climates in these three treatments (RCP 2.6 & RCP 4.5 in summer, and RCP 4.5 in winter, Fig. S1). The other two genera showed more variable responses to these treatments. For example, Chondracanthusdecreased in two of these treatments (summer, RCP 2.6 and winter, RCP 4.5) but increased in another (summer, RCP 4.5). Similarly,Laurencia decreased in one of these treatments (winter, RCP 4.5) but increased in two other treatments (summer, RCP 2.6 and summer, RCP 4.5).
Field assemblage: The effect of Silvetia removal on the understory assemblages in our field plots depended upon season and the initial state of the understory assemblages (PERMANOVA: pseudo-F1,84 = 3.6, p = 0.002, Fig. 5, Table S10). Specifically, in the fall, the Silvetia Canopy treatment influenced the assemblage in the Understory Full treatments (Pairwise tests: p = 0.009, Fig. 5A, Table S11). Under this scenario, the greatest contributors to dissimilarity (listed in order of importance, Table 2) were Centroceras (~17%), Corallina(~16%), Laurencia (~13%), Bare Rock (~10%), Gigartina (~10%), and Gelidium (~9%). Importantly, the top two species that responded to Silvetia loss were Centrocerasand Corallina were the same top two species that were impacted by the climate manipulations in our mesocosm experiment. Silvetiaabsence was associated with an increase in the average percent cover ofCentroceras (from 7.8 ± 13.7% to 25.3 ± 18.2%) while these conditions led to a decrease in Corallina (from 25.2 ± 20.3% to 11.6 ± 13.2%). Laurencia also increased in the absence of a canopy (from 10.0 ± 15.7% to 17.3 ± 19.8%), as did Gigartina(from 7.1 ± 13.1% to 10.2 ± 16.3%) and Bare Rock (from 9.2 ± 9.5% to 14.7 ± 16.5%), while Gelidium declined (from 12.3 ± 20.4% to 0.0 ± 0.0%) under these conditions (Table S13, Fig. S3).
The Silvetia Canopy effect observed in the Understory Full treatments during the fall dissipated by winter (Pairwise tests: p = 0.906, Fig. 5C). The canopy did not influence the understory assemblage when the understory was cleared at the start of the experiment in fall (p = 0.361, Fig. 5B) but did have an effect by winter (p = 0.021, Fig. 5D). Under this understory treatment, the greatest contributors to dissimilarity (listed in order of importance, Table 2) were Bare Rock (~24%), Centroceras (~18%),Corallina (~13%), Gigartina(~10%), and Gelidium (~8%). The average percent cover of Bare Rock increased in the absence ofSilvetia (from 15.9 ± 9.9% to 43.1 ± 15.8%), whileGigartina and Gelidium decreased (from 11.1 ± 18.8% to 5.0 ± 8.2% and from 8.6 ± 11.9% to 3.0 ± 5.9%, respectively). Although Centroceras and Corallina were ranked second and third in order of importance, their trends were more ambiguous. However, both slightly increased in the absence of Silvetia (from 16.8 ± 22.3% to 18.2 ± 17.0% and from 14.8 ± 16.0% to 16.6 ± 11.4%, respectively, Table S13, Fig. S3).
As a caveat to these results, prior to initiating the field manipulations in summer, a priori testing revealed that assemblages differed between Understory Full treatments (p = 0.009, Table S8, Table S9, Fig. S2). Thus, the effect of Silvetiaremoval on fall, Understory Full assemblages could be confounded with the starting state of the assemblages. However, because the starting patterns of some genera were not consistently maintained across every season (e.g., Corallina had similar starting abundances in summer but decreased in the absence of a canopy in fall, Fig. S3), it is likely that the differences found during subsequent sampling resulted from manipulating the canopy and understory rather than a holdover from the starting state of the assemblage.