A model-based approach to standardizing American lobster (Homarus americanus) ventless trap abundance indices
Introduction
American lobster (Homarus americanus) supports the most valuable single-species fishery in North America, with an ex-vessel value of $US 1.7 billion in 2017 (DFO, 2020; NOAA, 2020). Despite this economic significance, the species’ dynamic life history can make it difficult to infer population trends across life stages. Canadian lobster stock assessments historically have relied on fishery-dependent data to derive catch per unit effort (CPUE) indices for inferring population trajectories (DFO, 2013). However, using industry effort data may provide misleading population trajectory signals given such CPUE data are derived primarily from densely populated lobster areas where effort is concentrated, and do not necessarily include the entirety of the population or stock bounds. Such dynamics have often led to hyperstability when using CPUE to infer population abundance (Harley et al., 2001). Further, improvements in fishing strategies and gear are constantly evolving and can influence CPUE (Smith and Tremblay., 2003) yet are usually undocumented. While commercial CPUE can be useful in assessing the performance of the fishery, these uncertainties have led to their exclusion as abundance indices in U.S. American lobster stock assessments.
Historically, bottom otter trawl survey data from multi-species surveys have been used in U.S. American lobster stock assessments as indicators of population trends (Atlantic States Marine Fisheries Commission (ASMFC, 2009, 2015). Two of the greatest drawbacks for using otter trawl data to represent lobster stock abundance are that (1) they typically do not sample in the preferable cobble or boulder habitat of lobsters (Wahle and Steneck., 1991) in fear of entangling the trawl, and that (2) they cannot fish in areas or paths where fixed fishing gear exists (Smith and Tremblay., 2003). Trap surveys have often been considered a better method for assessing the structure-oriented lobster population (Tremblay et al., 2009). To address this concern for the U.S. lobster stocks, states within the northeast U.S. initiated complementary lobster ventless trap surveys designed to target sub-legal lobsters (or recruits) across various habitat types in state waters (0–3 nautical miles from shore). The Coastwide Ventless Trap Survey (VTS) employs a random stratified sampling design and gear modeled after the commercial fishing configuration of traps strung together over trawl lines, but with the inclusion of ventless traps to retain sub-legal and legal sized lobsters. The catch of lobsters per ventless trap (CPVT) is used to provide an estimate of the relative abundance of lobsters through space and time. The VTS was designed with depth and National Marine Fisheries Service (NMFS) statistical areas defining the stratification based on pilot surveys and analyses (Pugh and Glenn, 2020), which found that a stratification scheme based on depth provided the most accurate estimates of relative abundance, as represented by reduced variability in CPVT. The NMFS statistical area inclusion in survey stratification is intended to account for the latitudinal gradient of environmental drivers on lobsters, such as temperature. Mean CPVT estimates in concert with strata-specific weights based on the survey area domain are then used to derive annual abundance indices.
Male and female CPVT indices were first used in 2015 as fisheries-independent assessment model inputs for the two U.S. stocks, Southern New England and the Gulf of Maine/Georges Bank (Atlantic States Marine Fisheries Commission (ASMFC, 2015). This inclusion served as a major advancement for better understanding the sub-legal lobster population and population trends in the summer, a season that previously lacked relative abundance information. While the VTS abides by a standard set of methods, unforeseen changes in sampling design can occur due to inclement weather and gear loss. Many factors have been found to influence trap effectiveness and catchability for crabs and lobsters (Miller, 1990; Tremblay and Smith, 2001; Smith and Tremblay., 2003; Geraldi et al., 2009), but have been unaccounted for when constructing relative abundance estimates. Within stock assessment models, survey indices are often scaled to the stock or population size using non-linear, time-invariant relationships (Hilborn and Walters, 1992). While the estimated parameters of these functions (i.e., catchability coefficients) capture the average magnitude of survey catchability, process error in terms of time-varying catchability is often not considered.
Model-based approaches to standardizing fisheries-independent catch data have been widely viewed as a useful tool in providing more accurate estimates of abundance indices (Maunder and Punt, 2004). Utilizing these approaches provide several benefits, including an understanding of the effects variables have on species observed abundance, reducing the variance on the abundance predictions, corrected-abundance estimates for when these variables deviate, a tool to help account for years with missed sampling, and a method to account for random effects on observed abundances. This work aimed to understand the impact of several covariates on lobster abundance trends for the inshore components of the Gulf of Maine (GOM) and Southern New England (SNE) lobster stocks, and account for these sources of variability when deriving abundance indices. Model-based approaches were utilized to derive sex- and stock-specific annual abundance estimates from the lobster ventless trap surveys. Model-derived male and female GOM and SNE CPVT indices were constructed and compared to those based on the design-based approach to ascertain their differences and advantages, and ultimately the significance of including factors documented to influence trap catchability in deriving abundance estimates for the VTS.
Section snippets
Survey design
Beginning in 2006, the VTS has employed a random stratified survey design, using NMFS Statistical Area (SA) and depth as the primary strata classifications. The SAs included in the survey are 511, 512, 513, and 514 in the GOM region (no sampling is conducted in the Georges Bank region of the stock), and 538, 539, and 611 in SNE. The survey is a cooperative effort between state fisheries agencies and commercial lobstermen, in which lobstermen are contracted to deploy and retrieve survey gear
Model results
When testing model variants, the models with the lowest AIC scores across sexes and stocks were those excluding the number of traps offset (Table 2), of which all converged successfully. Excluding the site effect resulted in the greatest difference in AIC scores and weakened model fitness, suggesting that including site effect made the greatest improvement to the model. Including the unique site effect improved model performance more for SNE models than for GOM models, and for SNE females more
CPVT sources of variability
We have quantified several factors that influence catch rates of lobsters within the Coastwide Ventless Trap Survey. Such confounding factors have long been suspected to influence trap catch rates for crustacean species (Miller, 1990), and this work begins to assess and account for these factors when deriving abundance indices by removing the variability in catch associated with them. The day of year sampling occurred, unique site sampled, and the soak time of the traps all influenced lobster
Funding
The Coastwide Ventless Trap Survey is supported by the Atlantic States Marine Fisheries Commission (ASMFC). Additional funding by state agencies’ commercial and recreational lobster permit fees have supported the survey at various points through time.
Author contributions
MCM designed the research study, and MCM and JK and executed the research. MCM led the writing of the manuscript. All authors provided guidance throughout the study and revised drafts of the manuscript.
Declaration of Competing Interest
The authors declare they have no conflict of interest in these studies.
Acknowledgements
We are grateful for the many scientists and commercial fishermen that have made this survey possible. Reviews and meeting facilitation from Megan Ware and Caitlin Starks helped progress this work. These analyses are the product of the ASMFC American Lobster 2020 Stock Assessment process, with comments from the Peer-Review Panel on this work improving the manuscript. The views expressed herein are those of the authors and do not necessarily reflect the views of their agencies.
References (48)
- et al.
Generalized linear mixed models: a practical guide for ecology and evolution
Trends Ecol. Evol.
(2009) - et al.
A comparison of Americould lobster size structure and abundance using standard and ventless traps
Fish. Res.
(2015) - et al.
Changes in higher trophic level productivity, diversity and niche space in a rapidly warming continental shelf ecosystem
Sci. Total Environ.
(2020) - et al.
Whole truths vs. Half truths—and a search for clarity in long-term water temperature records
Estuar. Coast. Shelf Sci.
(2015) - et al.
Standardizing catch and effort data: a review of recent approaches
Fish. Res.
(2004) - et al.
Fishery-independent trap surveys of lobster (Homarus americanus): design considerations
Fish. Res.
(2003) - et al.
Effects of temperature on phagocytosis in Homarus americanus
Aquaculture
(1978) - et al.
Modeling spatiotemporal variability of the bioclimate envelope of Homarus americanus in the coastal waters of Maine and New Hampshire
Fish. Res.
(2016) - et al.
GLMs, GAMs, and GLMMs: an overview of theory for applications in fisheries research
Fish. Res.
(2004) Influence of behavioral interactions on lobster distribution and abundance as inferred from pot-caught samples
ICES Marine Science Symposium
(1995)
Information theory as an extension of the maximum likelihood principle
American Lobster Stock Assessment for Peer Review
Recruitment Failure in the Southern New England Lobster Stock
American Lobster Benchmark Stock Assessment and Peer Review Report
American Lobster Benchmark Stock Assessment and Peer Review Report
Influence of soak time and fish accumulation on catches of reef fishes in a multispecies trap survey
Fish. Bull.
glmmTMB balances speed and flexibility among packages for zero-inflated generalized linear mixed modeling
R Journal
Principles of Geographical Information Systems
Estimating spatial distribution of American lobster Homarus americanus using habitat variables
Mar. Ecol. Prog. Ser.
2018 framework assessment of the american lobster (Homarus americanus) in LFA 27-33
DFO Can. Sci. Advis. Sec. Res. Doc
Sources of variation in stomach contents of predators of Atlantic herring in the Northwest Atlantic during 1973–2014
Ices J. Mar. Sci.
Assessment of Lobster (Homarus americanus) in Lobster Fishing Area (LFA) 34. DFO Can. Sci. Advis. Sec. Sci. Advis. Rep
Seafisheries Landed Value by Province, 2017
A prolonged thermal stress experiment on the American lobster Homarus americanus
J. Shellfish Res.
Cited by (2)
The impact of ocean warming on juvenile American lobster recruitment off southeastern Massachusetts
2023, Fisheries Oceanography