Search

ABSTRACT Ecosystem models predict changes in productivity and status for multiple species, and are important for incorporating climate‐linked dynamics in ecosystem‐based fisheries management. However, fishery regulations are primarily informed by single‐species stock assessment models, which estimate unexplained variation in dynamics (e.g., recruitment, survival, fishery selectivity, etc) using random effects. We review the general benefits of estimating random effects in ecosystem models: (1) better representing biomass cycles and trends for focal species; (2) conditioning interactions upon observed biomass for predators and prey; (3) easier replication of model results using formal estimation rather than informal model “tuning;” and (4) attributing process errors via comparison amongst different models. We then demonstrate these by introducing a new state‐space model EcoState (and associated R‐package) that extends mass balance dynamics from Ecopath with Ecosim. This model estimates mass balance (Ecopath) and time‐dynamics (Ecosim) parameters directly via their fit to time‐series data (biomass indices and fisheries catches) while also estimating the magnitude of process errors using RTMB. A real‐world application involving Alaska pollock (Gadus chalcogrammus) in the eastern Bering Sea suggests that fluctuations in krill consumption are associated with cycles of increased and decreased pollock production. A self‐test simulation experiment confirms that estimating process errors can improve estimates of productivity (growth and mortality) rates. Overall, we show that state‐space mass‐balance models can be fitted to time‐series data (similar to surplus‐production stock assessment models), and can attribute time‐varying productivity to both bottom‐up and top‐down drivers including the contribution of individual predator and prey interactions. [ABSTRACT FROM AUTHOR]

Fisheries scientists compare processes among species to estimate species productivity, management reference points, and climate sensitivities. Ecologists have developed "phylogenetic comparative methods" (PCMs) to address these questions, but there is surprisingly little application of PCM within fisheries science. Here, I bridge this gap by introducing PCM (including Brownian motion, Ornstein–Uhlenbeck, and Pagel's kappa and lambda models for species covariance), thereby showing that PCM generalizes the nested taxonomic random effects that are commonly used in fisheries science. I next summarize phylogenetic structural equation models (PSEMs), which extend the linear models that are commonly used in fisheries. Finally, I re‐analyse a high‐quality database used to predict mortality rates from longevity and/or growth parameters. I specifically propose a PSEM that reverts to a longevity‐based prediction when longevity information is available but uses phylogenetic corrected growth parameters otherwise. Using this single PSEM replaces the common practice of fitting and predicting using separate linear models depending upon what data are available for a given species. Cross‐validation suggests that the relationship between log‐mortality rate and longevity does not vary based on phylogeny, and therefore, linear models and PSEM both explain 82% of variance when longevity is available. When longevity is unavailable, by contrast, the linear model explains only 37% of variance while the PSEM explains 52% of variance, where this gain occurs from conditioning predictions on phylogenetic similarities. I therefore conclude that PCM and PSEM provide a general and user‐friendly replacement for linear models and can improve performance for fisheries meta‐analyses that are used for fisheries management applications. [ABSTRACT FROM AUTHOR]

Fisheries scientists and managers must track rapid shifts in fish spatial distribution to mitigate stakeholder conflict and optimize survey designs, and these spatial shifts result in part from animal movement. Information regarding animal movement can be obtained from selection experiments, tagging studies, flux through movement gates (e.g. acoustic arrays), fishery catch‐per‐unit effort (CPUE), resource surveys and genetic/chemical markers. However, there are few accessible approaches to combine these data types while accounting for spatially correlated residual patterns. We therefore discuss a movement model involving diffusion (random movement), taxis (movement towards preferred habitat) and advection (passive drift following ocean currents). We specifically outline how these movement processes can be fitted to data while discretizing space and time and estimating non‐linear habitat preferences using environmental layers as well as spatial process errors. Finally, we introduce an R package, ATM, by fitting the model to bottom trawl survey, longline fishery and tagging data for Pacific cod (Gadus macrocephalus, Gadidae) in the Bering Sea during winter/summer seasons from 1982 to 2019. Combining data types predicts an increasing proportion of cod residing in the northern Bering Sea from 2013 to 2019, and estimates are informative in a recent stock assessment model. We fit sensitivity analyses by dropping tag, survey or fishery data, and this analysis shows that tagging data are necessary to identify diffusion rates, while survey data are informative about movement among biogeographic strata. This "hybrid" species distribution model can help explain poleward movement, project distributions under future climate conditions and evaluate alternative tag‐deployment scenarios to optimize tagging designs. [ABSTRACT FROM AUTHOR]

Implementation of the United Nations Sustainable Development Goals requires assessments of the global state of fish populations. While we have reliable estimates of stock status for fish populations accounting for approximately half of recent global catch, our knowledge of the state of the majority of the world's "unassessed" fish stocks remains highly uncertain. Numerous publications have produced estimates of the global status of these unassessed fisheries, but limited quantity and quality of data along with methodological differences have produced counterintuitive and conflicting results. Here, we show that despite numerous efforts, our understanding of the status of global fish stocks remains incomplete, even when new sources of broadly available data are added. Estimates of fish populations based primarily on catch histories on average performed 25% better than a random guess. But, on average, these methods assigned fisheries to the wrong FAO status category 57% of the time. Within these broad summaries, the performance of models trained on our tested data sources varied widely across regions. Substantial improvements to estimates of the state of the world's exploited fish populations depend more on expanded collection of new information and efficient use of existing data than development of new modelling methods. [ABSTRACT FROM AUTHOR]

Accounting for variation in prey mortality and predator metabolic potential arising from spatial variation in consumption is an important task in ecology and resource management. However, there is no statistical method for processing stomach content data that accounts for fine‐scale spatio‐temporal structure while expanding individual stomach samples to population‐level estimates of predation. Therefore, we developed an approach that fits a spatio‐temporal model to both prey‐biomass‐per‐predator‐biomass data (i.e. the ratio of prey biomass in stomachs to predator weight) and predator biomass survey data, to predict "predator‐expanded‐stomach‐contents" (PESCs). PESC estimates can be used to visualize either the annual landscape of PESCs (spatio‐temporal variation), or can be aggregated across space to calculate annual variation in diet proportions (variation among prey items and among years). We demonstrated our approach in two contrasting scenarios: a data‐rich situation involving eastern Bering Sea (EBS) large‐size walleye pollock (Gadus chalcogrammus, Gadidae) for 1992–2015; and a data‐limited situation involving West Florida Shelf red grouper (Epinephelus morio, Epinephelidae) for 2011–2015. Large walleye pollock PESC was predicted to be higher in very warm years on the Middle Shelf of the EBS, where food is abundant. Red grouper PESC was variable in north‐western Florida waters, presumably due to spatio‐temporal variation in harmful algal bloom severity. Our approach can be employed to parameterize or validate diverse ecosystem models, and can serve to address many fundamental ecological questions, such as providing an improved understanding of how climate‐driven changes in spatial overlap between predator and prey distributions might influence predation pressure. [ABSTRACT FROM AUTHOR]

Characterizing population distribution and abundance over space and time is central to population ecology and conservation of natural populations. However, species distribution models and population dynamic models have rarely been integrated into a single modelling framework. Consequently, fine‐scale spatial heterogeneity is often ignored in resource assessments. We develop and test a novel spatiotemporal assessment framework to better address fine‐scale spatial heterogeneities based on theories of fish population dynamic and spatiotemporal statistics. The spatiotemporal model links species distribution and population dynamic models within a single statistical framework that is flexible enough to permit inference for each state variable through space and time. We illustrate the model with a simulation–estimation experiment tailored to two exploited marine species: snow crab (Chionoecetes opilio, Oregoniidae) in the Eastern Bering Sea and northern shrimp (Pandalus borealis, Pandalidae) in the Gulf of Maine. These two species have different types of life history. We compare the spatiotemporal model with a spatially aggregated model and systematically evaluate the spatiotemporal model based on simulation experiments. We show that the spatiotemporal model can recover spatial patterns in population and exploitation pressure as well as provide unbiased estimates of spatially aggregated population quantities. The spatiotemporal model also implicitly accounts for individual movement rates and can outperform spatially aggregated models by accounting for time‐and‐size varying selectivity caused by spatial heterogeneity. We conclude that spatiotemporal modelling framework is a feasible and promising approach to address the spatial structure of natural resource populations, which is a major challenge in understanding population dynamics and conducting resource assessments and management. [ABSTRACT FROM AUTHOR]

Fisheries scientists use biological models to determine sustainable fishing rates and forecast future dynamics. These models require both life‐history parameters (mortality, maturity, growth) and stock‐recruit parameters (juvenile production). However, there has been little research to simultaneously predict life‐history and stock‐recruit parameters. I develop the first data‐integrated life‐history model, which extends a simple model of evolutionary dynamics to field measurements of life‐history parameters as well as historical records of spawning output and subsequent recruitment. This evolutionary model predicts recruitment productivity (steepness) and variability (variance and autocorrelation in recruitment deviations) as well as mortality, maturity, growth, and size, and uses these to predict intrinsic growth rate (r) for all described fishes. The model confirms previous analysis showing little correlation between steepness and either natural mortality or asymptotic maximum size (W∞). However, it does reveal taxonomic patterns, where family Sebastidae has lower steepness (mean=0.72) and Salmonidae has elevated steepness (mean=0.79) relative to the prediction for bony fishes (class Actinopterygii, mean=0.74). Similarly, genus Sebastes has growth rate r (0.09) approaching that of several shark families (Lamniformes: 0.02; Carcharhiniformes: 0.02). A cross‐validation experiment confirms that the model is accurate, explains a substantial portion of variance (32%–67%), but generates standard errors that are somewhat too small. Predictive intervals are tighter for species than for higher‐level organizations (e.g. families), and predictions (including intervals) are available for all fishes worldwide in R package FishLife. I conclude by outlining how multivariate predictions of life‐history and stock‐recruit parameters could be useful for stock assessment, decision theory, ensemble modelling and strategic management. [ABSTRACT FROM AUTHOR]

Multispecies models are widely used to evaluate management trade‐offs arising from species interactions. However, identifying climate impacts and sensitive habitats requires integrating spatial heterogeneity and environmental impacts into multispecies models at fine spatial scales. We therefore develop a spatio‐temporal model of intermediate complexity for ecosystem assessments (a "MICE‐in‐space"), which is fitted to survey sampling data and time series of fishing mortality using maximum‐likelihood techniques. The model is implemented in the VAST R package, and it can be configured to range from purely descriptive to including ratio‐dependent interactions among species. We demonstrate this model using data for four groundfishes in the Gulf of Alaska using data from 1982 to 2015. Model selection for this case‐study shows that models with species interactions are parsimonious, although a model specifying separate density dependence without interactions also has substantial support. The AIC‐selected model estimates a significant, negative impact of Alaska pollock (Gadus chalcogrammus, Gadidae) on productivity of other species and suggests that recent fishing mortality for Pacific cod (G. microcephalus, Gadidae) is above the biological reference point (BRP) resulting in 40% of unfished biomass; other models show similar trends but different scales due to different BRP estimates. A simulation experiment shows that fitting a model with fewer species at a coarse spatial resolution degrades estimation performance, but that interactions and biological reference points can still be estimated accurately. We conclude that MICE‐in‐space models can simultaneously estimate fishing impacts, species trade‐offs, biological reference points and habitat quality. They are therefore suitable to forecast short‐term climate impacts, optimize survey designs and designate protected habitats. [ABSTRACT FROM AUTHOR]

Analysing how fish populations and their ecological communities respond to perturbations such as fishing and environmental variation is crucial to fisheries science. Researchers often predict fish population dynamics using species‐level life‐history parameters that are treated as fixed over time, while ignoring the impact of intraspecific variation on ecosystem dynamics. However, there is increasing recognition of the need to include processes operating at ecosystem levels (changes in drivers of productivity) while also accounting for variation over space, time and among individuals. To address similar challenges, community ecologists studying plants, insects and other taxa increasingly measure phenotypic characteristics of individual animals that affect fitness or ecological function (termed "functional traits"). Here, we review the history of trait‐based methods in fish and other taxa, and argue that fisheries science could see benefits by integrating trait‐based approaches within existing fisheries analyses. We argue that measuring and modelling functional traits can improve estimates of population and community dynamics, and rapidly detect responses to fishing and environmental drivers. We support this claim using three concrete examples: how trait‐based approaches could account for time‐varying parameters in population models; improve fisheries management and harvest control rules; and inform size‐based models of marine communities. We then present a step‐by‐step primer for how trait‐based methods could be adapted to complement existing models and analyses in fisheries science. Finally, we call for the creation and expansion of publicly available trait databases to facilitate adapting trait‐based methods in fisheries science, to complement existing public databases of life‐history parameters for marine organisms. [ABSTRACT FROM AUTHOR]

Forecasting distribution shifts under novel environmental conditions is a major task for ecologists and conservationists. Researchers forecast distribution shifts using several tools including: predicting from an empirical relationship between a summary of distribution (population centroid) and annual time series ("annual regression," AR); or fitting a habitat‐envelope model to historical distribution and forecasting given predictions of future environmental conditions ("habitat envelope," HE). However, surprisingly little research has estimated forecast skill by fitting to historical data, forecasting distribution shifts and comparing forecasts with subsequent observations of distribution shifts. I demonstrate the important role of retrospective skill testing by forecasting poleward movement over 1‐, 2‐ or 3‐year periods for 20 fish and crab species in the Eastern Bering Sea and comparing forecasts with observed shifts. I specifically introduce an alternative vector‐autoregressive spatio‐temporal (VAST) forecasting model, which can include species temperature responses, and compare skill for AR, HE and VAST forecasts. Results show that the HE forecast has 30%–43% greater variance than predicting that future distribution is identical to the estimated distribution in the final year (a "persistence" forecast). Meanwhile, the AR explains 2%–6% and VAST explains 8%–25% of variance in poleward movement, and both have better performance than a persistence forecast. HE and AR both generate forecast intervals that are too narrow, while VAST models with or without temperature have appropriate width for forecast intervals. Retrospective skill testing for more regions and taxa should be used as a test bed to guide future improvements in methods for forecasting distribution shifts. [ABSTRACT FROM AUTHOR]

Many of the world's fisheries are unassessed, with little information about population status or risk of overfishing. Unassessed fisheries are particularly predominant in developing countries and in small‐scale fisheries, where they are important for food security. Several catch‐only methods based on time series of fishery catch and commonly available life‐history traits have been developed to estimate stock status (defined as biomass relative to biomass at maximum sustainable yield: B/BMSY). While their stock status performance has been extensively studied, performance of catch‐only models as a management tool is unknown. We evaluated the extent to which a superensemble of three prominent catch‐only models can provide a reliable basis for fisheries management and how performance compares across management strategies that control catch or fishing effort. We used a management strategy evaluation framework to determine whether a superensemble of catch‐only models can reliably inform harvest control rules (HCRs). Across five simulated fish life histories and two harvest‐dynamic types, catch‐only models and HCR combinations reduced the risk of overfishing and increased the proportion of stocks above BMSY compared to business as usual, though often resulted in poor yields. Precautionary HCRs based on fishing effort were robust and insensitive to error in catch‐only models, while catch‐based HCRs caused high probabilities of overfishing and more overfished populations. Catch‐only methods tended to overestimate B/BMSY for our simulated data sets. The catch‐only superensemble combined with precautionary effort‐based HCRs could be part of a stepping stone approach for managing some data‐limited stocks while working towards more data‐moderate assessment methods. [ABSTRACT FROM AUTHOR]

Research has estimated associations between water temperature and the spatial distribution of marine fishes based upon correlations between temperature and the centroid of fish distribution (centre of gravity, COG). Analysts have then projected future water temperatures to forecast shifts in COG, but often neglected to demonstrate that temperature explains a substantial portion of historical distribution shifts. We argue that estimating the proportion of observed distributional shifts that can be attributed to temperature vs. other factors is a critical first step in forecasting future changes. We illustrate this approach using Gadus chalcogrammus (Walleye pollock) in the Eastern Bering Sea, and use a vector-autoregressive spatiotemporal model to attribute variation in COG from 1982 to 2015 to three factors: local or regional changes in surface and bottom temperature ('temperature effects'), fluctuations in size-structure that cause COG to be skewed towards juvenile or adult habitats ('size-structured effects') or otherwise unexplained spatiotemporal variation in distribution ('unexplained effects'). We find that the majority of variation in COG (including the north-west trend since 1982) is largely unexplained by temperature or size-structured effects. Temperature alone generates a small portion of primarily north-south variation in COG, while size-structured effects generate a small portion of east-west variation. We therefore conclude that projections of future distribution based on temperature alone are likely to miss a substantial portion of both the interannual variation and interdecadal trends in COG for this species. More generally, we suggest that decomposing variation in COG into multiple causal factors is a vital first step for projecting likely impacts of temperature change. [ABSTRACT FROM AUTHOR]

Fisheries dynamics can be thought of as the reciprocal relationship between an exploited population and the fishers and/or managers determining the exploitation patterns. Sustainable production of protein of these coupled human-natural systems requires an understanding of their dynamics. Here, we characterized the fishery dynamics for 173 fisheries from around the globe by applying general additive models to estimated fishing mortality and spawning biomass from the RAM Legacy Database. GAMs specified to mimic production models and more flexible GAMs were applied. We show observed dynamics do not always match assumptions made in management using 'classical' fisheries models, and the suitability of these assumptions varies significantly according to large marine ecosystem, habitat, variability in recruitment, maximum weight of a species and minimum observed stock biomass. These results identify circumstances in which simple models may be useful for management. However, adding flexibility to classical models often did not substantially improve performance, which suggests in many cases considering only biomass and removals will not be sufficient to model fishery dynamics. Knowledge of the suitability of common assumptions in management should be used in selecting modelling frameworks, setting management targets, testing management strategies and developing tools to manage data-limited fisheries. Effectively balancing expectations of future protein production from capture fisheries and risk of undesirable outcomes (e.g., 'fisheries collapse') depends on understanding how well we can expect to predict future dynamics of a fishery using current management paradigms. [ABSTRACT FROM AUTHOR]

Fishery managers must often reconcile conflicting estimates of population status and trend. Superensemble models, commonly used in climate and weather forecasting, may provide an effective solution. This approach uses predictions from multiple models as covariates in an additional 'superensemble' model fitted to known data. We evaluated the potential for ensemble averages and superensemble models (ensemble methods) to improve estimates of population status and trend for fisheries. We fit four widely applicable data-limited models that estimate stock biomass relative to equilibrium biomass at maximum sustainable yield (B/ BMSY). We combined these estimates of recent fishery status and trends in B/ BMSY with four ensemble methods: an ensemble average and three superensembles (a linear model, a random forest and a boosted regression tree). We trained our superensembles on 5,760 simulated stocks and tested them with cross-validation and against a global database of 249 stock assessments. Ensemble methods substantially improved estimates of population status and trend. Random forest and boosted regression trees performed the best at estimating population status: inaccuracy (median absolute proportional error) decreased from 0.42 - 0.56 to 0.32 - 0.33, rank-order correlation between predicted and true status improved from 0.02 - 0.32 to 0.44 - 0.48 and bias (median proportional error) declined from −0.22 - 0.31 to −0.12 - 0.03. We found similar improvements when predicting trend and when applying the simulation-trained superensembles to catch data for global fish stocks. Superensembles can optimally leverage multiple model predictions; however, they must be tested, formed from a diverse set of accurate models and built on a data set representative of the populations to which they are applied. [ABSTRACT FROM AUTHOR]

Meta-analysis has been an integral tool for fisheries researchers since the late 1990s. However, there remain few guidelines for the design, implementation or interpretation of meta-analyses in the field of fisheries. Here, we provide the necessary background for readers, authors and reviewers, including a brief history of the use of meta-analysis in fisheries, an overview of common model types and distinctions, and examples of different goals that can be achieved using meta-analysis. We outline the primary challenges in implementing meta-analyses, including difficulties in discriminating between alternative hypotheses that can explain the data with equal plausibility, the importance of validating results using multiple lines of evidence, the trade-off between complexity and sample size and problems associated with the use of model output. For each of these challenges, we also provide suggestions, such as the use of propensity scores for dealing with selection bias and the use of covariates to control for confounding effects. These challenges are then illustrated with examples from diverse subfields of fisheries, including (i) the analysis of the stock-recruit relationship, (ii) fisheries management, rebuilding and population viability, (iii) habitat-specific vital rates, (iv) life-history theory and (v) the evaluation of marine reserves. We conclude with our reasons for believing that meta-analysis will continue to grow in importance for these and many other research goals in fisheries science and argue that standards of practice are therefore essential. [ABSTRACT FROM AUTHOR]

Stakeholders increasingly expect ecosystem assessments as part of advice on fisheries management. Quantitative models to support fisheries decision-making may be either strategic ('big picture', direction-setting and contextual) or tactical (focused on management actions on short timescales), with some strategic models informing the development of tactical models. We describe and review ' Models of Intermediate Complexity for Ecosystem assessments' ( MICE) that have a tactical focus, including use as ecosystem assessment tools. MICE are context- and question-driven and limit complexity by restricting the focus to those components of the ecosystem needed to address the main effects of the management question under consideration. Stakeholder participation and dialogue is an integral part of this process. MICE estimate parameters through fitting to data, use statistical diagnostic tools to evaluate model performance and account for a broad range of uncertainties. These models therefore address many of the impediments to greater use of ecosystem models in strategic and particularly tactical decision-making for marine resource management and conservation. MICE are capable of producing outputs that could be used for tactical decision-making, but our summary of existing models suggests this has not occurred in any meaningful way to date. We use a model of the pelagic ecosystem in the Coral Sea and a linked catchment and ocean model of the Gulf of Carpentaria, Australia, to illustrate how MICE can be constructed. We summarize the major advantages of the approach, indicate opportunities for the development of further applications and identify the major challenges to broad adoption of the approach. [ABSTRACT FROM AUTHOR]