Dataset: Landscape parameters of seagrass, fish and macroinvertebrate communities within Artificial Seagrass Units (ASU) in Back Sound, NC from July to September 2018

To examine fragmentation components independently and interactively, we conducted this study on Oscar Shoal and an adjacent unnamed shoal in Back Sound, NC, USA (34°42′20" N to 34°41′60" N, 76°36′ 15" W to 76°35′17" W) during the summer of 2018. Both shoals were shallow (less than 0.5-meter depth at low tide) and historically supported expansive, ephemeral seagrass meadows (Peterson et al., 2001) that have been absent over the last decade. During 2018, these shoals had large expanses of sandy area speckled with small patches of seagrass (which were avoided during landscape siting) composed of a mixture of eelgrass, Zostera marina (Linnaeus 1753), and shoal grass, Halodule wrightii (Ascherson 1868) (Yeager et al., 2016). Both shoals were adjacent to deep boating channels between two large salt marsh complexes to the north (North River Marsh) and south (Middle Marsh).

We generated 25 unique 18-meters × 13-meters (234 square meter) landscapes along orthogonal axes of percent cover of seagrass within the landscape footprint (10 percent, 22.5 percent, 35 percent, 47.5 percent, and 60 percent) and fragmentation per se (i.e., percolation probability: 0.1, 0.225, 0.35, 0.475, 0.59). Blueprints for landscape construction in the field were generated by a modified random cluster (MRC) method (Saura and Martinez-Millan, 2000) using the randomHabitat function in the secr package in R (Efford, 2016). In using the MRC method, landscapes of low percolation probability exhibit low connectivity from one ASU to the next (i.e., are highly fragmented), which approaches a lower limit of realistic landscape fragmentation at percolation probability = 0.1 (Saura and Martinez-Millan, 2000). Additionally, patch cohesion is maximized (i.e., landscapes are contiguous) when percent cover reaches approximately 60 percent and percolation probability approaches approximately 0.59 (Saura and Martinez-Millan, 2000). Therefore, we chose 60 percent cover and percolation probability = 0.59 as the upper limits of our landscape parameters, and 10 percent cover and percolation probability = 0.1 as the lower limits. We also note that the generated landscapes fell within the range of the number of patches (approximately 1-15 patches) observed within natural, low seagrass area landscapes (approximately 250-375 square meters) in our system (Yeager et al., 2016). For simplicity, landscape parameters will be hereafter described in the context of percent cover and fragmentation per se (i.e., low percolation probability is high fragmentation per se). Landscapes were constrained to fall within 2 percent of the area input parameter and maintain consistent edge-to-area ratios within individual fragmentation per se treatment levels (secr package in R; Efford, 2016). Within an individual landscape defined by more than 1 seagrass patch, discrete patches of artificial seagrass were separated from one another by a minimum of 0.86-meter (short-side length of an ASU) of sandy matrix in all directions.

Each landscape contained 22 to 135 ASUs (total ASUs constructed = 2059) constructed to mimic shoot density, shoot width, and canopy height of Zostera marina meadows in this system (Yeager et al., 2016). For each ASU, 30-centimeter (cm) lengths of green splendorette curling ribbon (0.5-cm width) were tied to approximately 1-square meter bases of rigid black plastic VEXAR (0.86-meter x 1.2-meter, 2.5-centimeter) mesh so that each ASU had uniformly spaced "shoots" (450 square meters) with two 15-centimeter length "leaves". Each ASU required approximately 3 hours to construct, with the entire manufacturing effort requiring the equivalent of three full-time person work years (with work divided among a number of technicians and community volunteers). 

Twenty-five ASU landscapes were deployed over the course of 10 days (21-May to 31-May 2018) in haphazard order and placement: 19 were located on Oscar Shoal and the remaining 6 were located on the adjacent, unnamed shoal. All landscapes were more than 50 meters apart from each other and greater than 30 meters from natural seagrass patches. Individual ASUs were secured to the sediment surface and each other with lawn staples (16 per ASU) and zip ties (0-4 per ASU, based on the presence/absence of adjacent ASUs), respectively. The site with 60 percent cover and 0.59 percolation probability (i.e., "60 percent-0.59"; hereafter sites will be named by this convention), was replicated (on 8-Jun 2018) as a 26^th landscape to evaluate potential differences in ecological response metrics influenced by the identity of the two shoal environments.

Hurricane Florence
The study area and artificial landscapes were directly impacted by Hurricane Florence during 13-16 Sept 2018. Despite ASU re-enforcements made prior to Florence's landfall (i.e., additional lawn staples and cable ties), our landscapes experienced substantial disturbance akin to natural seagrasses in the vicinity, in many cases completely removing or burying ASUs which altered the landscape percent cover and fragmentation per se parameters. Because of this substantial disturbance, we opportunistically examined how our landscapes were altered by Hurricane Florence as a proxy for how extreme storm events may alter natural seagrass meadows. We recalculated landscape parameters based on ASU-by-ASU checks made after Hurricane Florence. Post-Florence landscape percent cover and percolation probabilities were recalculated both including and excluding ASUs that were fully buried under sediment. Within natural habitats, depending on sedimentation intensity, seagrass burial may represent temporary or permanent habitat loss (Cabaço et al., 2008). Holding the original landscape 234-square meter footprint constant, the percent cover of each landscape was recalculated from the remaining number of ASUs (including and excluding buried ASUs). Percolation probability is a fragmentation per se input parameter for landscape generation (Saura and Martinez-Millan, 2000) but is not generally used to describe existing or natural landscapes. Therefore, to determine how landscape fragmentation per se was altered, we first determined the linear relationship between the initial number of seagrass patches and our percolation probability treatments (y = 0.61 – 0.06x, r² = 0.912). We then used this relationship to predict post-Florence landscape percolation probabilities from the remaining number of patches in each landscape (including and excluding buried ASUs).

Known Issues:
See Hurricane Florence section in methods. Some landscapes were completely removed or buried and landscape parameters are denoted with 0 or NA.