Literature search and meta-analysis inclusion criteria
We conducted a search using the Institute of Science Information's (ISI) Web of Science (last accessed on May 13, 2021) to gather peer-reviewed literature examining edge effects and fragmentation effects on biogenic complexity, faunal densities, and predation in seagrass ecosystems. Search terms included 1) seagrass AND 2) edge effects OR fragmentation effects AND 3) density OR predation OR survival OR mortality OR trophic interactions. We supplemented this database with additional articles known to us. All candidate studies were judged for inclusion in our meta-analysis based on the following criteria: 1) The study was an original experiment in a mesocosm or natural setting providing edge effect data (i.e., responses in patch edges vs. interiors) or fragmentation effect data (i.e., responses in fragmented vs. continuous landscapes) for one or more of our response metrics of interest in extractable form (i.e., table, figure, or text). Response metrics were natural seagrass shoot density, faunal density, and predation survival. Initially, we considered several metrics of biogenic complexity because they may respond to habitat configuration differently, yet shoot density was ultimately chosen as it was the most common metric reported. Shoot density data were only extracted from studies also examining faunal response metrics, because we were interested in examining faunal-habitat relationships in the context of proximate (e.g., shoot density) and ultimate (e.g., edge, fragmentation) drivers. For faunal density responses, if data for “nested” taxonomic levels were provided (e.g., fish, flounder), we extracted data for the lowest taxonomic level available. Prey survival responses included data expressed as, or converted to, proportion survival or survival time (e.g., h to consumption) of sessile or tethered prey. Only survival from uninhibited predator exposure was considered. 2) The response metric(s) included the mean, sample size, and either standard error (SE), standard deviation (SD), or confidence interval (CI). 3) Levels of edge effects (e.g., edge, interior) and fragmentation (e.g., fragmented, continuous) were typically expressed as discrete categories. Therefore, we accepted the operational definitions used by these studies, but also included meta-data such as edge/interior widths and distances, and fragmentation degree in our database to illustrate the range of definitions used across studies. All included studies examined fragmentation as a state (i.e., configuration), rather than an active process (i.e., changing configuration through time). For studies that included more than two discrete levels of edge (e.g., integer distances) or fragmentation (e.g., continuous, patchy, very patchy), only the most extreme levels were included in effect size calculations (e.g., the distances closest to the patch edge and center; the most continuous and fragmented landscape classifications). Figure data was extracted using DataThief III software (Tummers, 2006).
Calculating Log Response Ratios
Refer to the attached Supplemental File, "864783_Calculating_Log_Response_Ratios.pdf" for a description of how the log response ratios were calculated.