Experimental Studies
Overall, the experimental removal of an algal understory had a mean neutral effect on the canopy (random effects model, mean LNRRexp = 1.55 ± 0.94, p = 0.099), with interactions ranging from strongly competitive to strongly facilitative (Fig. 3). The mean interaction did not change over the course of an experiment, measured as the number of days since the start of the experiment (ϐdays = 0.0012, p = 0.052). The effect of turf species on the canopy varied among turf functional groups (overall Wald-type test, QM[4], exp = 17.16, p = 0.0018, pseudo-R2 = 0.35, Fig. 3). Specifically, the overall competitive effect of turfs on the canopy was driven by the negative effects of coralline turf and non-coralline crust (ϐcoralline turf = 3.59, p = 0.0009; ϐnon-coralline turf = 2.91, p = 0.034); all other turf species had no significant effect on the canopy (all p> 0.05; Appendix S3 Table S1). However, the effect of turf species differed among canopy types. Among kelps, non-coralline turf taxa also had a negative effect on the canopy (ϐnon-coralline turf = 5.04, p = 0.026). Among Fucales, all turf functional groups had an overall neutral effect on the canopy (all p > 0.10, Appendix S3 Table S1).
However, these overall effects marked geographical and latitudinal variation in the effect of turfs on canopies. In line with the widespread paradigm of kelp forest interactions, the effect of turfs was overall competitive in subtidal systems. Specifically, the effect of coralline turf on canopy taxa was competitive in the subtidal, and increasingly facilitative in the intertidal (Fig. 4, pseudo-R2 = 0.16, ϐdepth * coralline turf = -0.61, p = 0.0003). Though found in a narrower depth range than turfs, non-coralline crust also had an increasingly positive effect on the canopy at shallower depth (ϐdepth * non-coralline crust = -1.33, p = 0.009). The effect of coralline crust and non-coralline turf on the canopy did not change across depth, and this pattern held for canopy kelps (Appendix S3 Table S2). Studies of canopy Fucales, conducted primarily in the intertidal zone, demonstrated no difference in interactions across depth (all functional groups, p functional group: depth > 0.05, pseudo-R2 = 0.27; Appendix S3 Table S2).
The effect of turfs on the canopy varied across latitude (QM[8] = 26.69, p = 0.0008, pseudo-R2 = 0.39). In particular, the effect of non-coralline crust became more competitive at higher latitudes (ϐlatitude * non-coralline crust = 0.40, p = 0.027). Further, the effect of coralline turf was more facilitative at higher latitudes, with marginal significance (ϐlatitude * coralline turf = -0.35, p = 0.058; Appendix S3 Table S3).
Next, we looked only at variation in the interaction across life history stage of the canopy, and did not include variation across depth and latitude. The effect of turfs on the canopy differed among canopy life history stages, depending on the turf functional group and the identity of the canopy (QM[31] = 45.13, p < 0.0001, pseudo-R2 = 0.13; Fig. 3). Turf species had competitive effects on the canopy, primarily at early life history stages. Coralline crust and all turfs negatively affected germling canopies (overall effect, p for linear contrasts < 0.01: ϐcoralline crust = 4.77, p = 0.029; ϐcoralline turf = 9.65, p = 0.00008; ϐnon-coralline turf = 5.43, p = 0.014). Further, coralline crusts negatively affected the later canopy “recruit” stage (overall effect: ϐcoralline crust = 4.77, p = 0.004). All other effects were not modified by life history stage of the canopy (Appendix S3 Table S4). Finally, considering only studies that manipulated herbivory (n = 28), the mean effect of turfs on the canopy did not depend on the presence or absence of herbivores (linear contrast, ϐ-herbivore - +herbivore = 2.84 ± 2.10,p = 0.18; pseudo-R2 = 0.44, Appendix S3 Figure S1).