Code and data to replicate analyses and plots for Harris et al. 2023 (in revision, Frontiers in Marine Science). This stusy tested lionfish traps in the northern Gulf of Mexico. Plots and models were built in R (version 4.1.1). The ZINB models were developed with the pscl (Jackman 2020) and glmmTMB (Brooks et al. 2017) packages. Data manipulation and plotting used the ggplot (Wickham 2016) and dplyr (Wickham et al. 2016) packages.
Deepwater traps offer a compelling market-based solution to control invasive lionfish populations on reefs too deep for SCUBA removals (generally >40 m). A scalable harvest gear must have minimal bycatch, not cause negative environmental effects, and allow for profitable fishing. We tested the efficacy of wooden slat lobster traps, wire sea bass traps, and experimental non-containment Gittings traps with 300 trap deployments on 120 mesophotic (38–78 m deep) natural reef sites in the northeastern Gulf of Mexico. Time-lapse video monitoring of the traps showed that lionfish were attracted to the vicinity of the three trap types we tested but rarely entered the lobster or sea bass traps. Lobster traps demonstrated more promise than the sea bass traps, whose high bycatch rates suggest they would be untenable for directed lionfish harvest. Gittings traps demonstrated the highest catch rates of lionfish and the lowest bycatch. However, operational issues were also observed, and lionfish catch rates were still far too low to support a commercial fishery. Continued lionfish trap research and testing will be necessary to assess the suitability of lionfish traps.
Experimental trap deployments were conducted near northeastern Gulf of Mexico (nGOM) mesophotic natural reefs in depths of 38–78 m. Traps were deployed within four distinct reef areas: the Trysler Grounds, Yellow Gravel, Pensacola Edge, and the Desoto Canyon Rim. The former three are located in the west subregion and the latter is in the east subregion. Ten deployment trips were equally divided between the two subregions (n = 5 each) with 87.2 degrees longitude demarcating east vs. west subregions. Traps were deployed at 12 reef sites during each trip. Reef sites for trap deployment were selected based on real-time acoustic sonar collected from echo sounders on board the chartered fishing vessels. Site characteristics were ground-truthed and categorized from video files collected with the ROV. Each trap deployment site was ≥0.5 km from adjacent deployment sites within the same reef area, which exceeds the typical lionfish home range estimates (Bacheler et al., 2015; McCallister et al., 2018; Dahl and Patterson III, 2020). The trap “configuration” consisted of 1, 2, 3, or 4 traps of a single type deployed at a given site, with each trap type and configuration combination deployed once among the 12 sites per trip. The trap type and deployment combination was randomly treated per reef site.
Recruitment of fishes to traps was monitored with remote time-lapse camera units attached to one trap at each trap deployment site. Each camera unit included a GoPro Hero4 camera paired with a CamDo Blink time-lapse controller housed within a GolemGear deep water housing. Cameras were deployed on subsurface buoys 2 m above the traps and programmed to record 10 seconds of video every hour during daylight hours. For each time-lapse video, the maximum number of individuals of each species (or higher-order taxa) observed in any one video frame was counted as the “minimum count”, which is the minimum number of a given taxon present in a still frame of the video. Fish counts were categorized as “inside” (i.e., recruited inside the lobster and sea bass traps or within the 2-m diameter footprint area of the Gittings traps) or "outside" of a trap (i.e., observed within the video frame but not inside the trap). Video counts were taken from one diel period per day (i.e., dawn, midday, and dusk for each day), resulting in 1714 video samples. Dawn videos occurred within two hours of sunrise and were selected as the first video of a given day where there was enough light for the video reader to identify fishes. Similarly, dusk videos were those occurring within two hours of sunset and were selected as the last video of the day with enough light to identify fishes. Midday videos were taken at noon.
A generalized linear modeling (GLM) framework was used to test how the deployment or reef site factors significantly affected catches. Models were computed separately for each fishery category: i.e., lionfish catches, fishery species catches, and non-fishery species catches (Eq. 1). Given the excess number of zeros in our catches, a zero-inflated approach was used whereby a logistic regression (i.e., the binomial component) estimated the probability zero catch and a separate regression model (i.e., the positive component) estimated the relationship between the non-zero catch per unit effort (CPUE in g) and the predictors. Quantile-quantile plots indicated that the non-zero data were overdispersed for the three response categories and best fit by a negative binomial distribution; thus, zero-inflated negative binomial (ZINB) models were used. Specifically, the ZINB positive component (Eq. 1A, log-linked negative binomial GLM) assessed the effects of trap type (lobster, Gittings, sea bass), trap configuration (1, 2, 3, or 4 traps), soak time of traps at the given site (numeric days), lionfish density at the site, site depth (m), site relief (maximum reef height in m estimated from ROV surveys), and region (east or west). The ZINB binomial component (Eq. 1B, logit-linked logistic GLM) assessed whether the deployment factors of trap type and trap configuration affected the probability of zero catches.
A similar zero-inflated GLM approach was applied to the time-lapse video data to assess the significant predictors for lionfish recruitment, i.e., the number of lionfish inside the footprint of the Gittings trap or inside the container of the lobster and sea bass traps. Similar to the catch data, the QQ plots indicated that the data was zero-inflated and overdispersed, thus a ZINB model was used. The ZINB positive component (Eq. 2A, log-linked negative binomial GLM) assessed whether trap type, lionfish density, and diel period (dawn, midday, or dusk) had significant effects. The effect of these factors specific to each experimental trap type was examined by testing the interaction of trap type × diel period and the interaction of trap type × lionfish density. The ZINB binomial component (Eq. 2B, logit-linked logistic GLM) assessed whether trap type or lionfish density affected the probability of zero lionfish counts. To account for repeated measures from videos taken during a given deployment at a reef site, reef site was included as a random effect (random intercept) in the models and assumed to be normally distributed with a mean of zero and variance σ2.
This repository is a scientific product and is not official communication of the National Oceanic and Atmospheric Administration, or the United States Department of Commerce. All NOAA GitHub project code is provided on an ‘as is’ basis and the user assumes responsibility for its use. Any claims against the Department of Commerce or Department of Commerce bureaus stemming from the use of this GitHub project will be governed by all applicable Federal law. Any reference to specific commercial products, processes, or services by service mark, trademark, manufacturer, or otherwise, does not constitute or imply their endorsement, recommendation or favoring by the Department of Commerce. The Department of Commerce seal and logo, or the seal and logo of a DOC bureau, shall not be used in any manner to imply endorsement of any commercial product or activity by DOC or the United States Government.
Methods for data collection were reviewed and approved by the University of Florida's Institutional Animal Care and Use Committee (UF IACUC Protocols #201810225 and #201810394) in accordance with relevant guidelines and regulations by American Veterinary Medicine Association. Authorization for experimental trap deployments in US federal waters and the collection of fishes for research purposes was granted by a Letter of Acknowledgment from the NOAA Fisheries Southeast Regional Office following the definitions and guidance at 50 CFR 600.10.