A field-based investigation of behavioural interactions between invasive green crab (Carcinus maenas), rock crab (Cancer irroratus), and American lobster (Homarus americanus) in southern Newfoundland

Marine species invasions pose a global threat to native biodiversity and commercial fisheries. The European green crab (Carcinus maenas) is one of the most successful marine invaders worldwide and has, in the last decade, invaded the southern and western coastal waters of the island of Newfoundland, Newfoundland and Labrador (NL), Canada. Impacts of green crab on the American lobster (Homarus americanus), which are native to Newfoundland, are not well understood, particularly for interactions around deployed fishing gear. Declines in lobster catch rates in invaded systems (i.e., Placentia Bay, NL), have prompted concerns among lobster fishers that green crab are interfering with lobster catch. Here, we conducted a field experiment in a recently-invaded bay (2013) in which we deployed lobster traps pre-stocked with green crab, native rock crab (Cancer irroratus) (a procedural control), or empty (control). We compared catch per unit effort across each category, and used underwater cameras to directly observe trap performance in situ. In addition, we used SCUBA surveys to determine the correlation between ambient density of lobster and green crab in the ecosystem and the catch processes of lobster in traps. We found: (1) Regardless of the species of crab stocked, crab presence reduced the total number of lobster that attempted to enter the trap, and also reduced entry success rate, (2) lobster consumed green crab, rock crab and other lobster inside traps and (3) there was a positive association between lobster catch and ambient lobster density. Our results suggest that while there was a relationship between in-trap crab density and trap catch rates, it was not linked to the non-native/native status of the crab species.

To record underwater video of lobster traps fishing in situ we designed and assembled a 115 camera system capable of recording full high-definition videos for 13 continuous hours at high 116 resolution (1080p) (see Bergshoeff et al., 2017). We mounted the camera housing to a wooden 117 frame constructed around standard commercial single-parlour wire-mesh (3.8 cm mesh size) 118 lobster traps capable of catching all types of crustaceans. We secured the housing to this frame 119 and oriented the camera 58 cm above the trap pointing downward, giving a top-down field of 120 view (FOV) of approximately 105 cm by 170 cm underwater. Using this setup we were able to 121 observe lobster entering, exiting, inside the trap, as well as those around the trap for 13 122 continuous hours. We attached a Data Storage Tag (DST) to record the depth at which each trap 123 was deployed, but due to a technological malfunction these data were not recoverable. We 124 deployed the traps at depths shallow enough for ambient light to illuminate the pot during the 125 day, therefore we did not use external lights. Traps were deployed within depths reflective of 126 where commercial lobster traps are set (i.e. < 20 m at low tide; DFO, 2016). 127 2.2 Study site and fieldwork 128 We conducted field work in nearshore lobster fishing areas around Little Harbour East 129 and Little Bay East, Fortune Bay on the southern coast of Newfoundland ( Figure 1) for five 130 weeks between July-September 2016, directly following the closure of the two-month lobster Due to logistical limitations, we were only able conduct dive surveys for 11 of 17 177 deployments (one dive per trap, n dives = 33). Within each dive, divers recorded the depth at 178 which the trap was deployed, and collected data along four transects oriented North, South, West 179 and East of each trap ( Figure 2B). 188 2.5 Video analysis: 189 We collected approximately 663 h of underwater video footage across 17 deployments. 190 Following the field study, we scored the videos manually, following protocols described in 191 previous literature (Favaro et al., 2012;Jury et al., 2001;Meintzer et al., 2017). Specifically, we 192 recorded the following quantitative parameters for lobster, rock crab and green crab: (1) the 193 number, direction and duration of entry attempts as well as the proportion of those entries that 194 were successful versus failed, (2) the number, direction and duration of exits from the trap, (3) 195 the time spent feeding on the bait, (4) the number and duration of interspecific aggression events, 196 and (5) the number and duration of predation events.
We then conducted stepwise model simplification, sequentially dropping non-significant 230 terms until all terms in the model were statistically significant (procedure outlined in Crawley, 231 2012). This procedure was used for all models presented in this paper, and we used the lme4 232 package (Bates et al., 2017) to fit models. We verified model assumptions by plotting residuals 233 against fitted values. Residuals met assumptions for normality, homogeneity and independence, 234 and there was no evidence of overdispersion. We interpreted reduced models.

276
The distribution of our catch data was best explained by a Poisson distribution. We 277 simplified the model using stepwise removal of non-significant terms (Crawley, 2012) and fit the   Figures 1 and 3). 297 We caught a total of 326 lobsters, six green crabs, and three rock crabs across the entire 298 study. Traps pre-stocked with green crab caught between 47% fewer, and 1.97% more lobster 299 than unstocked traps (95% C.I.; β = -0.312, S.E. = 0.168, p = 0.064; Figure 3A; Table 1). Traps 300 pre-stocked with rock crab caught between 54% and 8.9% fewer lobster than unstocked traps, 301 (95% C.I.; β = -0.436, S.E. = 0.175, p = 0.013; Figure 3A; Table 1). Manuscript to be reviewed 302 We found no significant impact of the presence of the camera apparatus on lobster catch 303 (GLMM: No camera, β = -0.225, S.E. = 0.140, z = -1.610, p = 0.107; Figure 3A; Table 1). Soak 304 duration did not significantly influence this relationship and was removed from the model via  Lobster size (carapace length) ranged from 48 to 98 mm (mean ± 1 SD = 80.4 ± 6.1 mm). 309 The average size of lobster caught did not vary substantially across trap pre-stocking condition 310 and 70.9% of the lobster captured were of sub-legal size (< 82.5 mm) ( Figure 3B). Manuscript to be reviewed 324 When we accounted for lobster entry attempts through the entrance only (i.e., excluding 325 small individuals that attempted entries through the wire cells and escape slats) we found 30.1% 326 (N = 139) were successful in the unstocked trap, while 17.1% (N = 106) and 17.7% (N = 105) of 327 entry attempts were successful for green crab and rock crab pre-stocked traps, respectively 328 ( Figure 5).

329
The proportion of successful entries was higher in the unstocked traps than in either the 330 green crab or rock crab pre-stocked traps for all lobster attempts (x 2 = 190.072, df = 2, p < 0.001; 331 Figure 6A) as well as for those only through the trap entrance (x 2 = 36.049, df = 2, p < 0.001; 332 Figure 6B). Smaller lobster were able to easily crawl through the wire cells on the kitchen and 333 parlour sides of the trap, as well as through the escape slats on the top and bottom of the trap. We 334 did not observe any difference in the proportion of successful entries across trap pre-stocking 335 conditions for attempts made by small lobster through the wire cells or escape slats ( Figure 6C). 336 We did not detect any difference in the duration of successful lobster entry attempts through the 337 trap entrances according to trap pre-stocking condition (Supplementary Figure 4; Table 2).

338
There was little difference between the number of lobster inside traps after a full 339 deployment versus at the end of the video analysis period for each trap pre-stocking condition 340 (mean difference ± 1 SD = unstocked, 1.2 ± 2.5; green crab pre-stock, -0.3 ± 2.9; rock crab pre-341 stock, -0.5 ± 2.5; Supplementary Figure 5). 342 We observed very few green crab or rock crab entering traps in any stocking condition. 343 We observed 45 entry attempts made by green crab, of which 27 were successful. Rock crab 344 made 18 entry attempts, of which 17 were successful. Green crab and rock crab were observed 345 exiting the camera traps 19 and 13 times respectively. All these crabs occurred in just two 346 deployments, meaning we observed no crab entry in most videos. Manuscript to be reviewed 347 We observed 60 events in which lobster consumed all or part of another organism while 348 in the trap. There were 36 predation events against tethered rock crab, 20 against tethered green 349 crab, and four against other lobster. Lobsters that engaged in predation activities tended to have 350 larger claw sizes than those that did not predate on pre-stocked rock crab, green crab, or lobster 351 (Supplementary Figure 6).  Table 3.

364
The average number of ambient lobster did not vary significantly across trap pre-stocking 365 conditions (mean ± 1 SD = unstocked, 23.0 ± 12.57; green crab pre-stock, 21.18 ± 12.80; rock 366 crab pre-stock 21.73 ± 8.97; Figure 7).  We found traps pre-stocked with both species of crabs caught fewer lobster than 378 unstocked traps, though the reduction associated with green crab was not technically statistically 379 significant (p = 0.06).

380
The number of green crab observed around our traps via SCUBA surveys were relatively 381 low. As a result, we were not able to robustly assess the relationship between ambient green crab 382 density and lobster catch rates. Since completing this study, the density of green crab in Fortune 383 Bay has increased (C. McKenzie, Pers. Obs.). Repeating this study in an area with a higher 384 density of invaders may yield additional insight.

385
The sizes of captured lobster in our study were similar across all trap pre-stocking 386 conditions and the majority were of sub-legal size (< 82.5 mm). Given that our fieldwork was 387 conducted after closure of the commercial fishery, it is possible that fewer legal sized (> 82.5 388 mm) lobster were available to be captured. Previous studies have found that the presence of 389 larger individuals inside traps can reduce the catch of smaller individuals for lobster (Watson and 390 Jury, 2013) as well as for green crab and rock crab (Miller and Addison, 1995). Watson and Jury, The use of a procedural control (rock crab) was critical to this study. Since lobsters' 501 response to both green crab and rock crab was similar, it demonstrated impacts on lobster catch 502 were not specific to green crab but were rather due to the presence of either crab species in the 503 traps. This finding aligns with those of Howard et al., (2017) which found non-native crabs did 504 not reduce prey abundance via direct consumption any more than native crabs. Howard et al., 505 (2017) also reported on the paucity of studies that directly compared impacts of native versus 506 non-native species. Had we not incorporated a procedural control, our conclusion would likely 507 have been that green crab caused declines in trap effectiveness, with the implication that 508 something about their identity as an invader was the root cause. We recommend future studies 509 investigating green crab impacts incorporate a direct comparison with the impacts of native 510 species. In some cases, it may be the nature of the invasion (e.g. rapid population growth) rather 511 than the identity of the species that causes impacts.

512
Nevertheless, our findings do not imply that rapid increases in the green crab population 513 will be benign to the lobster fishery, or to lobster populations. In heavily invaded systems green 514 crab have been observed readily accessing and being captured by lobster traps (Goldstein et al., 515 2017). In the Great Bay Estuary, New Hampshire, Goldstein et al., (2017), captured ~8.5 times 516 more green crab than lobster. This implies that reductions in catch rates may not be due to a 517 specific unique quality of the invader but rather the sheer abundance of green crab relative to 518 native species. Manuscript to be reviewed 521 This study has shown the presence of crabs inside lobster traps can reduce the effectiveness of 522 lobster traps. Crabs in traps cause fewer lobster to attempt entry, and reduce the success rate of 523 entries. This effect was observable with both native and invasive crabs. Using SCUBA survey 524 data we determined the difference could not be explained by differences in ambient lobster 525 density across stocking conditions. 526 As green crab spread in extent and grow in density around Newfoundland (DFO, 2016), it will 527 become increasingly important to understand how the invasion is specifically affecting fishery 528 performance. As the density of green crab increases, there will be more potential for fishery 529 interactions. While the collapse of lobster catch rates in neighbouring Placentia Bay pre-date the 530 green crab invasion (and so cannot be blamed on green crab alone (Best et al., 2017; DFO, 531 2018)), the lack of recovery in this heavily-invaded system may signal a warning to fisheries in 532 the Fortune Bay region -that if green crab continue to grow in abundance, it is likely traps will 533 continue to perform worse. While ecosystem responses to invasion must continue to be studied, 534 future research should also examine fishing gear performance across gradients of invader density 535 to better understand how invaders can impact the capture process itself.