Faunal Communities Are Invariant to Fragmentation in Experimental Seagrass Landscapes

Abstract

Human-driven habitat fragmentation is cited as one of the most pressing threats facing many coastal ecosystems today. Many experiments have explored the consequences of fragmentation on fauna in one foundational habitat, seagrass beds, but have either surveyed along a gradient of existing patchiness, used artificial materials to mimic a natural bed, or sampled over short timescales. Here, we describe faunal responses to constructed fragmented landscapes varying from 4-400 m2 in two transplant garden experiments incorporating live eelgrass (Zostera marina L.). In experiments replicated within two subestuaries of the Chesapeake Bay, USA across multiple seasons and non-consecutive years, we comprehensively censused mesopredators and epifaunal communities using complementary quantitative methods. We found that community properties, including abundance, species richness, Simpson and functional diversity, and composition were generally unaffected by the number of patches and the size of the landscape, or the intensity of sampling. Additionally, an index of competition based on species co-occurrences revealed no trends with increasing patch size, contrary to theoretical predictions. We extend conclusions concerning the invariance of animal communities to habitat fragmentation from small-scale observational surveys and artificial experiments to experiments conducted with actual living plants and at more realistic scales. Our findings are likely a consequence of the rapid life histories and high mobility of the organisms common to eelgrass beds, and have implications for both conservation and restoration, suggesting that even small patches can rapidly promote abundant and diverse faunal communities.

Fig 1

(a) A schematic of the experimental design employed in Experiment 1. Lower middle panels depict aerial photographs of transplants from Experiment 1 in the (b) James River Estuary, and (c) York River Estuary, taken June 1997 (approximately nine months after the initial planting) (from [35]). (d) A schematic of the experimental design employed in Experiment 1. Bottom panel depicts aerial photograph of transplants from Experiment 2 in the York River Estuary taken in June 1999 (approximately nine months after the initial planting). The arrow in panel (e) denotes a separate set of transplants conducted one-year prior for unrelated purposes, and is not reported on here.

Fig 2. Mean values ± 1 SE m^-2 for community properties obtained from suction sampling of nekton in Experiment 1.

Colors and shapes correspond to small (4 m²), medium (100 m²), and large (400 m²) fragmented experimental landscapes of transplanted eelgrass. Small points correspond to replicate values. Large points are marginal means ± 1 SE estimated from generalized linear mixed effects models that account for variable sampling effort.

Fig 3. Standardized effect size (SES) from checkerboard scores, an index of nekton species co-occurrences in suction samples from Experiment 1.

Values >2 indicate significantly fewer associations than would be expected from chance, whereas values < -2 indicate the opposite. Values in the range [-2, 2] indicate random segregation of species in experimental replicates. Large points are marginal means ± 1 SE estimated from generalized linear mixed effects models that account for variable sampling effort. Colors correspond to the size of the fragmented experimental landscape: small (4 m²), medium (100 m²), and large (400 m²).

Fig 4. Non-metric multidimensional scaling of multivariate nekton community abundances from suction samples in Experiment 1.

Colors correspond to the size of the fragmented experimental landscape: small (4 m²), medium (100 m²), and large (400 m²), and shapes to the month of sampling. Ovals are 95% confidence ellipses; overlap indicates that composition is not significantly different among the set of points based on α = 0.05.

Fig 5. Mean values ± 1 SE m^-2 for community properties obtained from suction sampling of nekton in Experiment 2.

Colors and shapes correspond to small (4 m²), medium (100 m²), both fragmented and unfragmented, and large fragmented (400 m²) experimental landscapes of transplanted eelgrass. Small points correspond to replicate values. Large points are marginal means ± 1 SE estimated from generalized linear mixed effects models that account for variable sampling effort.

Fig 6. Standardized effect size (SES) from checkboard scores, an index of nekton species co-occurrences in suction samples from Experiment 2 across two seasons.

Values >2 indicate significantly fewer associations than would be expected from chance, whereas values < -2 indicate the opposite. Values in the range [-2, 2] indicate random segregation of species in experimental replicates. Large points are marginal means ± 1 SE estimated from generalized linear mixed effects models that account for variable sampling effort. Colors correspond to the size of the experimental landscape: small (4 m²), medium (100 m²), both fragmented and unfragmented, and large fragmented (400 m²).

Fig 7. Non-metric multidimensional scaling of multivariate nekton community abundances from suction samples in Experiment 2.

Colors correspond to the size of the experimental landscape: small (4 m²), medium (100 m²), both fragmented and unfragmented, and large fragmented (400 m²), and shapes to the season of sampling. Ovals are 95% confidence ellipses.

Fig 8. Mean values ± 1 SE m^-2 for community properties obtained from core sampling of epifauna in Experiment 2.

Colors and shapes correspond to small (4 m²), medium (100 m²), both fragmented and unfragmented, and large fragmented (400 m²) experimental landscapes of transplanted eelgrass. Small points correspond to replicate values. Large points are marginal means ± 1 SE estimated from generalized linear mixed effects models that account for variable sampling effort.

Fig 9. Standardized effect size (SES) from checkerboard scores, an index of epifaunal species co-occurrences in core samples from Experiment 2 across two seasons.

Values >2 indicate significantly fewer associations than would be expected from chance, whereas values < -2 indicate the opposite. Values in the range [-2, 2] indicate random segregation of species in experimental replicates. Large points are marginal means ± 1 SE estimated from generalized linear mixed effects models that account for variable sampling effort. Colors correspond to the size of the experimental landscape: small (4 m²), medium (100 m²), both fragmented and unfragmented, and large fragmented (400 m²).

Fig 10. Non-metric multidimensional scaling of multivariate epifaunal community abundances from core samples in Experiment 2.

Colors correspond to the size of the experimental landscape: small (4 m²), medium (100 m²), both fragmented and unfragmented, and large fragmented (400 m²), and shapes to the season of sampling. Ovals are 95% confidence ellipses.