For Table and Figure references below, see the document "MinnowTrap_statistical_analysis.pdf" in the Supplemental Files section.

Study Site Selection

We conducted our study across eight discrete seagrass meadows (hereafter referred to as landscapes) located in Back Sound, North Carolina (NC), USA (3442′ N to 3439′ N, 7637′ W to 7631′ W) (Fig. S1). All of our sampled landscapes were composed of a mixture of Back Sound's dominant seagrasses: eelgrass and shoal grass, Halodule wrightii (Ascherson 1868) (Yeager et al. 2016). Landscapes were chosen based upon available aerial imagery in Google Earth Pro as of February 19, 2017, and ground-truthed for changes in seasonal seagrass growth/senescence using summer, 2017, drone photography and ImageJ 1.x (Schneider et al. 2012). No discernable differences in landscape fragmentation states (e.g. total area, number of patches) were found between the two aerial imagery sources. All landscapes were relatively shallow (1-1.5 m depth at high tide), reasonably isolated from other seagrass beds (distance to nearest seagrass meadow = 112 17 m [mean standard error]) and were appropriately sized to encompass short-term (e.g., daily, monthly) movements of common seagrass-associated fauna in this system (Yeager et al. 2016). We identified similarly sized landscapes (25882 6592 m2) available in Back Sound by defining the minimum convex polygon surrounding the seagrass meadow, regardless of the total seagrass cover within the polygon. Among eight candidate landscapes of similar size, we defined four continuous landscapes and four fragmented landscapes based on the number of patches, the perimeter-to-area ratio, and the largest patch's percent cover of the total seagrass area (Table 1). Seagrass fragmentation is often naturally coupled with habitat loss (Wilcove et al. 1986), resulting in the mean seagrass area of our fragmented landscapes being nearly half that of our continuous landscapes (Table 1). Thus, our experiment was designed to examine the effects of fragmentation (i.e., the breaking apart of habitat concomitant with habitat loss) rather than fragmentation per se (i.e., the breaking apart of habitat without habitat loss; sensu Fahrig 2003).

Seagrass-associated Fauna Sampling

Seagrass-associated fauna were sampled to explore relationships between our observed predation/depredation rates and seasonal faunal densities within each fragmentation state. We sampled seagrass-associated fauna during each crab tethering cycle with four baited (with ~ 8 pieces of dried dog food; Mahoney et al. 2018) Gee-style minnow traps (41-cm long, 22-cm wide, 0.3-cm galvanized mesh-wire cylinders, with 4-cm diameter funneled openings) haphazardly deployed in each landscape. We acknowledge that baiting traps increases catch rates in our system and could bias the captured community toward predators and scavengers (our target community), yet this increase in catch rate seems to be uniform across seagrass habitat structure variables (Mahoney et al. 2018). At 24 h, faunae were enumerated, identified to the lowest taxonomical level possible, and released.
Point measurements of water temperature (C) were taken in each landscape at the location and time of all faunal sampling using hand-held thermometers (Table S1). We chose temperature as our seasonality proxy (Fig. S2) because several other seasonally affected factors including faunal densities correlate with water temperature variability. Additionally, the measurement of temperature is easy, cheap, reliable, and comparable to previous studies.

Equipment:
* Generic brand dog food
* Memphis and twine Gee-style minnow traps (41-cm long, 22-cm wide, 0.3-cm galvanized mesh-wire cylinders, with 4-cm diameter funneled openings)