Abundance of non-native crabs in intertidal habitats of New England with natural and artificial structure

Location of study site

Experiments were conducted in Kingston Bay, Massachusetts (41°59′09.9″N, 70°41′49.7″W) (Fig. 1), a shallow, protected embayment of Cape Cod Bay with both rocky intertidal shoreline and muddy sand flats and a tidal range of approximately 2.4 m. An area permitted for oyster aquaculture was located approximately 1.6 km from the study site. Two experimental sites were chosen within the bay: (1) the high intertidal zone (approx. 1.7 m above mean low water, MLW), with a mix of coarse sand and pebble substrate, as well as patches of cobble rocks; and (2) the low intertidal zone (approx. 26 cm above MLW) with a muddy sand substrate devoid of complex natural structure to separate the effects of natural vs. artificial structure. Experiments were conducted from late June to September, 2012, during which time juvenile crabs were expected to be present (Williams, 1984). Air temperature ranged from 12 to 27 °C and no major storms occurred.

General experimental procedures

Black plastic (high density polyethylene, HDPE) bags used for oyster grow-out, measuring 96 cm × 51 cm (∼0.5 m² area) with 62 mm² square mesh openings (1.1 cm maximum diagonal length), were borrowed from local oyster farmers. The bags were pouch-like in shape with a sealable opening at one end. Bags were secured with cable ties to 1 m² frames of 21.3 mm PVC pipe, (holes drilled to sink in seawater and anchored to the substrate using plastic stakes), and placed in the intertidal zone for one week. At the end of each experiment, bags were retrieved at low tide during daylight, and all visible crabs in and under the bags were collected. The bags were placed into clean plastic bins and thoroughly rinsed in fresh water to remove crabs. The rinse water was collected and strained through a 500 µm mesh sieve and crabs were frozen until analysis. All crabs collected were identified to species using morphological descriptions in Shen (1935), Gosner (1978), Hogarth (1978), and Williams (1984), and the carapace width (CW, measured at the two widest points of the carapace) was determined using a ruler for large crabs and an ocular micrometer for small crabs. The research was conducted under scientific permit number 054542 issued by the Commonwealth of Massachusetts Division of Marine Fisheries.

Experiment 1: grow-out bags with natural substrate

A three-factor experimental design tested the effects of natural substrate (sand or rocks), mesh grow-out bags (present or absent) and sampling period on crab abundance. Study plots with all treatments (sand, rocks, sand covered by mesh bag, and rocks covered by mesh bag) were established in the high intertidal zone on 6 July 2012. Each treatment, replicated twice, was deployed for 6 days, and the experiment was repeated for 3 consecutive weeks. At the end of each 6-d deployment, all materials (mesh bags, rocks) were removed to collect crabs. In all ∼0.5 m² plots, approximately 2.5 cm of sand to a depth of approx. 2.5 cm was examined to collect buried crabs.

Experiment 2: grow-out bags with bivalves

The second experiment tested the effect of grow-out bags containing bivalves on crab abundance on muddy sand substrate. The experimental design comprised 3 levels of bivalve shell status (mesh bag without oyster shells, mesh bag containing aged shells without oyster flesh, and mesh bag with living oysters (cultured Crassostrea virginica)) and sampling period. To test for effects of natural, non-living structure on the presence and abundance of crabs, approximately 80 oyster valves in 40 pairs were matched to approximate the surface of a complete oyster. Equal proportions of three size classes of shells (shell length 4–6 cm, 7–8 cm and 9–11 cm) were placed in oyster bags. To test for effects of natural, living structure on the presence and abundance of crabs, approximately 40 oysters in three size classes (shell length 4–6 cm, 7–8 cm and 9–11 cm) in equal proportion were placed in oyster bags. Bags were placed in a line parallel to the water’s edge at low intertidal height. Treatments were arranged randomly in 4 blocks, with each block containing one replicate of each treatment (n = 4). Patches of cobble-size rocks occurred 16.2 m towards the north and 9.5 m towards the south.

The experiment was conducted weekly for 7 consecutive weeks, with replicates deployed in a randomized block design over a two-day staggered period beginning 30–31 July 2012. Treatments were left undisturbed for 6 d (as in the previous experiment), then were retrieved on the seventh day and processed as described above. Following processing, bags containing oysters were held in seawater overnight to keep oysters alive before the next deployment. Bags with shell valves remained out of water for 8–9 h between deployments. Retrieved materials were deployed the following day for the next week’s experiment.

Statistical analysis

In both experiments, crab abundances were analyzed in the R statistical program (R Core Team, 2015) using a generalized linear model assuming a Poisson distribution (with log-link function). GLMER (from the lme4 package) was utilized to test a mixed effect model with experimental treatment as a fixed effect and weekly observation as a random effect. P-values of fixed effects were estimated using drop1 with a likelihood ratio test. Experiment 1 (Grow-out Bags with Natural Substrate) had two fixed factors (substrate (sand/rocks), mesh bag (present/absent)) and weekly observations (3 sampling dates). Because two crab species were collected simultaneously, each crab species was analyzed separately. Experiment 2 (Grow-out Bags with Bivalves) had one fixed factor (bivalve shell status (none/non-living, aged shells/living oysters) plus the random block sampling and weekly observations (7 sampling dates)). Frequency distributions of crab body sizes were analyzed by Kolmogorov–Smirnov two-sample test (K–S) tests to determine if carapace widths differed by crab species (Experiment 1) or bivalve shell status (Experiment 2) (Sokal & Rohlf, 1981). Effect sizes were calculated for treatment comparisons: unstandardized (mean and 95% confidence interval) and standardized (e.g., Cohen’s d with 95% confidence interval, and Pearson r or omega-squared ω²).

Experiment 1: grow-out bags with natural substrate

In all 3 weekly trials, no crabs were observed in either the bare sand or mesh bag over sand treatments and no crabs were found inside any bags. The green crab, C. maenas, and Asian shore crab, H. sanguineus, were only observed in the rocks and mesh over rocks treatments (Fig. 2). After removing the sand treatment (both with and without mesh) from the model, crab densities were not measurably affected by the presence of mesh for either H. sanguineus (with mesh, mean = 25.33 m⁻², CI [15.98–34.69] m⁻²; without mesh, mean = 23.33, CI [14.97–31.69]) or C. maenas (with mesh, mean = 15.00, CI [1.42–27.58]; without mesh, mean = 14.67, CI [5.03–24.30]). Moreover, both pairwise comparisons possessed small effect sizes: H. sanguineus (Cohen’s d = 0.237, 95% CI [−0.899–1.372]); C. maenas (d = 0.031, 95% CI [−1.100–1.163]). Furthermore, each generalized linear model for each species indicated no significant differences in abundance of H. sanguineus among mesh treatments (F_1,6 = 0.246, p = 0.619), and likewise no significant differences in abundance of C. maenas among mesh treatments (F_1,6 = 0.011, p = 0.916).

H. sanguineus individuals observed in all treatments during all weeks combined ranged in size from 4 to 27 mm, and C. maenas individuals ranged in size from 2 to 30 mm (Fig. 3). The K–S test indicated a significant difference in the size distributions of C. maenas and H. sanguineus (D = 0.870, p < 0.001) and a large effect size (Cohen’s d = 1.765, 95% CI [1.455–2.075]). Green crabs (mean = 5.61 mm, 95% CI [4.52–6.70 mm]) were smaller than H. sanguineus (mean = 14.62, 95% CI [13.82–15.43]) with 80% of C. maenas individuals having a carapace width of 2–5.9 mm (Fig. 3).

Experiment 2: grow-out bags with bivalves

The only crab species found inside the mesh bags in this experiment was C. maenas. The generalized linear model indicated significant differences existed among shell status treatments (F_2,59 = 106.699, p ≪ 0.001) with a large effect size (ω² = 0.666). Crab densities in the mesh-with-living-oysters treatment (mean = 29.50 m⁻², 95% CI [25.50–33.50] m⁻²) were nearly double that of the mesh-with-shells treatment (mean = 15.21, 95% CI [12.33–18.10]) and six-fold more than the mesh-only treatment (mean = 4.93, 95% CI [3.51–6.35]) (i.e., living oysters > shells > mesh, Fig. 4). Moreover, all three pairwise comparisons possessed large effect sizes: living oysters vs. non-living shells (Cohen’s d = 1.561, 95% CI [0.964–2.161], Pearson r = 0.615); living oysters vs. mesh (d = 3.099, 95% CI [2.327–3.883], r = 0.840); non-living shells vs. mesh (d = 1.750, 95% CI [1.135–2.368], r = 0.658). The block effect was a significant source of variation (F_3,59 = 2.941, p = 0.029).

Individual green crab sizes varied among treatments (Fig. 5). Crabs in the mesh-with-living-oysters (mean = 7.00 mm, 95% CI [6.74–7.25] mm) were significantly larger than those in either the mesh-plus-shells (mean = 6.45, 95% CI [6.09–6.80]; D = 0.161, p = 0.001) or mesh-alone (mean = 5.66, 95% CI [5.15–6.16]; D = 0.264, p = 0.001) with small (d = 0.218, 95% CI [0.052–0.384]) and medium effect sizes (d = 0.538, 95% CI [0.279–0.797]), respectively. There was no significant difference between the sizes of crabs in the mesh-only and mesh-plus-shells treatments (D = 0.152, p = 0.189). There also was no significant difference in the size distribution of green crabs between the first and last week of the experiment (D = 0.117, p = 0.512).