Assumptions and limitations
In addition to variability, clear patterns emerged between experimental and observational studies moving from neutral effects to facilitation of turfs on canopy species. However, it is important to note that observational studies were conducted in intertidal, shallow subtidal and/or high latitude environments, all factors that influence facilitation interactions. Thus, the different results for observational and experimental studies may be attributable instead to these confounding differences. Here, the included observational studies generally used a study design where the number of canopy individuals were counted on a given substrate (e.g., turf, crust, bare rock), scaled by the availability of that substrate (n = 68 out of 91 observational studies). For three pairwise interactions in this study, effects were estimated both observationally and experimentally, yielding similar effect size estimates (Appendix S3 Fig. S3). More work is needed to understand the relationship between experimental and observational effect sizes in non-trophic interactions, but the similarity here in response ratios is encouraging.
Another limitation of the present study is an imbalance in sample size among turf functional groups. Of the four groups, there are fewer estimates of the effect of non-coralline crustose algae on canopies (nnon-coralline crust = 28, nnon-coralline turf = 79, ncoralline crust = 114, ncoralline turf = 102). Estimates of the abundance and importance of these species vary in canopy-forming systems: in an Alaskan subtidal system, non-coralline crusts dominate the substrate relative to other species, while classic studies fail to mention non-coralline crusts as a component of the community (e.g., Reed & Foster 1984). The present study includes no estimates of the effect of non-coralline crusts in deep water (deepest is -5m relative to MLLW) and our finding that the non-coralline crust effect is more positive at shallower depths (Fig. 4) should be interpreted with caution.