No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Projecting species distributions using fishery‐dependent data
Abstract
Many marine species are shifting their distributions in response to changing ocean conditions, posing significant challenges and risks for fisheries management. Species distribution models (SDMs) are used to project future species distributions in the face of a changing climate. Information to fit SDMs generally comes from two main sources: fishery‐independent (scientific surveys) and fishery‐dependent (commercial catch) data. A concern with fishery‐dependent data is that fishing locations are not independent of the underlying species abundance, potentially biasing predictions of species distributions. However, resources for fishery‐independent surveys are increasingly limited; therefore, it is critical we understand the strengths and limitations of SDMs developed from fishery‐dependent data. We used a simulation approach to evaluate the potential for fishery‐dependent data to inform SDMs and abundance estimates and quantify the bias resulting from different fishery‐dependent sampling scenarios in the California Current System (CCS). We then evaluated the ability of the SDMs to project changes in the spatial distribution of species over time and compare the time scale over which model performance degrades between the different sampling scenarios and as a function of climate bias and novelty. Our results show that data generated from fishery‐dependent sampling can still result in SDMs with high predictive skill several decades into the future, given specific forms of preferential sampling which result in low climate bias and novelty. Therefore, fishery‐dependent data may be able to supplement information from surveys that are reduced or eliminated for budgetary reasons to project species distributions into the future.
... An assessment is achieved using fishery dependent (i.e., logbooks, landings) and fishery independent (i.e., standardized scientific survey) data. Fishery data is typically available throughout the fishing season through logbooks, vessel tracking systems and observer data (Karp et al. 2023). Obtaining the data is relatively cost-effective, outside of compensating fisheries observers who take biological measurements of the catch (Karp et al. 2023). ...
... Fishery data is typically available throughout the fishing season through logbooks, vessel tracking systems and observer data (Karp et al. 2023). Obtaining the data is relatively cost-effective, outside of compensating fisheries observers who take biological measurements of the catch (Karp et al. 2023). Unfortunately, fishery data is often relatively clustered due to the nature of harvesters targeting areas of highest abundances and can thus miss changes in resource abundance in fringe habitat (Karp et al. 2023). ...
... Obtaining the data is relatively cost-effective, outside of compensating fisheries observers who take biological measurements of the catch (Karp et al. 2023). Unfortunately, fishery data is often relatively clustered due to the nature of harvesters targeting areas of highest abundances and can thus miss changes in resource abundance in fringe habitat (Karp et al. 2023). In contrast, survey data can be costly; however, the surveys can be replicated annually, and a rigid statistical sampling design can provide information on fringe effects and representative data on population changes through the stocks range (Karl et al. 2023). ...
Numerous approaches to fisheries management exist, and it is paramount that the right management strategy is implemented for a given resource. One tool available in the decision-making process is a management strategy evaluation, which aims to simulate the effects of varying management strategies and harvest control rules on a stock. Management strategies are not one-size-fits-all, and the selected strategy and associated harvest control rules should align with the species' life history, including reproductive output, as well as resource use objectives. The following report provides a review of management strategy implementation in Canada (Chapter 2), fish reproductive strategies (Chapter 3), and matrix projection models for their use in management strategy evaluation (Chapter 4). Chapter 5 includes original research and details the outcome of a management strategy evaluation of the precautionary approach, co-management, and ecosystem-based fisheries management strategies on hypothetical resources with varying life history traits, ranging from extreme r-selected to extreme K-selected. The management strategy evaluation was completed for each resource and strategy under three distinct scenarios based on stock health, economic pressure, and ecosystem health. The overall finding was that the precautionary approach, when implemented correctly, was the most balanced strategy for meeting the objectives of each scenario. The most influential factor affecting simulation outcomes was economic pressure, with co-management susceptible to both under-and over-fishing at the extremes of low and high economic pressure to fish. The report continues with a comparison of the original research results with real-world fisheries (Chapter 6) and concludes with a summary chapter (Chapter 7). Fisheries management is complex, balancing social, economic, and ecosystem objectives, but a healthy resource ultimately underpins any successful management system. A fisheries management strategy that aligns with the reproductive output of a resource stands a chance at meeting these objectives and results in a sustainable, healthy fishery. ii GENERAL SUMMARY Management strategy evaluations are tools used in fisheries management to examine outcomes of various harvest control rules through closed-loop simulation. Given the variability in reproductive strategies of biological resources, it is important to assess varying management strategies to avoid under-or overexploitation of a resource. The following major report provides a comprehensive review of fisheries management strategies in Canada, reproductive traits of various species, and matrix projection modelling. Following this review, I present the results of original research consisting of a management strategy evaluation of hypothetical resources with various reproductive strategies under different management strategies and differing stock health, economic pressure, and ecosystem health scenarios. I then provide linkages between my findings and real-world examples where such management strategies have been implemented. My findings suggest that the implementation of precautionary approach principles may be the best tool for managing various biological resources and balancing social, biological, and ecological objectives.
... Surveying year-round is often impossible due to cost and weather conditions outside the summer months, especially in high-latitude systems. In situations when the seasonal timing of surveys limits the availability of fisheries-independent data, fisheries-dependent data from onboard observers, industry-reported logbooks, or fishing receipts may provide the best available information for understanding the biology of species caught in directed fisheries or as bycatch (Crear et al., 2021;Karp et al., 2023). One important consideration when using fisheries-dependent data is that sampling effort is driven by the factors regulating fishing behavior and location of fishing grounds, such as the abundance of a target species, the desire to avoid bycatch species, or distance traveled from port (Karp et al., 2023). ...
... In situations when the seasonal timing of surveys limits the availability of fisheries-independent data, fisheries-dependent data from onboard observers, industry-reported logbooks, or fishing receipts may provide the best available information for understanding the biology of species caught in directed fisheries or as bycatch (Crear et al., 2021;Karp et al., 2023). One important consideration when using fisheries-dependent data is that sampling effort is driven by the factors regulating fishing behavior and location of fishing grounds, such as the abundance of a target species, the desire to avoid bycatch species, or distance traveled from port (Karp et al., 2023). These effects typically lead to violation of the statistical assumption that sampling locations have been chosen independently of the response variable (Conn et al., 2017;Diggle et al., 2010;Pennino et al., 2019). ...
... In spite of these concerns, fisheries-dependent data have been shown to complement fisheries-independent data to fill critical knowledge gaps for species occurrence and abundance (Pecquerie et al., 2004;Pinto et al., 2019;Rufener et al., 2021). Fisheries-dependent distribution models may produce results that are comparable to models informed by random sampling (Ducharme-Barth et al., 2022), especially when fisheries-independent data are not sampling a biased subset of climatic conditions for the species of interest (Karp et al., 2023). In some cases, fisheries-dependent data may even outperform fisheries-independent data for modeling temporal changes in species distributions (Pennino et al., 2016). ...
Declining Bristol Bay red king crab (BBRKC) abundance has triggered recent closures of this iconic Bering Sea fishery and raised interest in bycatch in non-directed fisheries as a possible conservation concern. One particular concern is the effectiveness of static closed areas for bycatch fisheries in an era of climate warming and widespread distribution shifts. However, spatial data for supporting management decisions concerning bycatch is lacking, as fisheries-independent data are collected only in the summer, and the relationship to BBRKC distribution in the fall/winter/spring, when most bycatch occurs, is unknown. We filled this information gap by using fishery-dependent data to build predictive models of BBRKC bycatch distribution in non-pelagic trawl groundfish fisheries in the data-poor seasons. We trained Boosted Regression Tree models for bycatch occurrence and abundance of four BBRKC sex-size/maturity categories, and evaluation metrics indicated good to excellent predictive ability across all models. We found that flatfish directed-fishery CPUE, summer survey CPUE for BBRKC and flatfish, and depth were important predictors for bycatch occurrence and abundance. Physical variables (ice cover and temperature) were generally less important. We also found strong correlations between the mean latitude of observed bycatch and the summer survey for BBRKC, highlighting the ability of summer survey data to predict non-summer bycatch distributions. BBRKC bycatch prediction is a tractable problem, and our results are the first step towards operating models that may be used to evaluate proposed management actions. We also conclude that northward shifts in fishery-independent and-dependent data suggest the possible value of reassessing decades-old static closure areas for managing BBRKC bycatch.
... Modeling techniques are essential to deeper understanding of marine species-environment relationships (Coll et al., 2019). It is important to consider different modeling approaches in a climatic context because different models often show strong differences as environmental or range conditions become more novel (Karp et al., 2023;Stock et al., 2020). In the comparative studies on the predictive effects of models for forecasting the PFS fishery, applying new models or improving old models is crucial for its fishery prediction Han et al., 2022a;Zhang, 2021;Zhou et al., 2021). ...
... Predictions of optimal fishing grounds, abundance, and catch per unit effort for the PFS fishery in the northwestern Indian Ocean are still dominated by the generalized additive model (GAM model) and its variants Han et al., 2022a;Mohamed et al., 2018), with newer models (e.g., machine learning models) being less frequently applied. GAM models, derived from probability theory, allow for traditional statistical extrapolation of their mean response predictions and thus can continue to fit trends to new environments, outperforming machine learning models in new environments or spatial extrapolation (Karp et al., 2023;Stock et al., 2020), meanwhile its focus on coarser spatial scales for predictive analysis (Waldock et al., 2022). Unlike GAM models, the study of predicting the distribution of marine organisms is a promising but seldom exploited area for the application of machine learning models (Gladju et al., 2022;Pickens et al., 2021;Sadaiappan et al., 2023;Smoliński and Radtke, 2017), especially in the study of the Northwest Indian Ocean PFS fishery. ...
... However, this practice usually introduces bias, mainly because commercial fishing data can hardly satisfy the assumption that the data of modeling algorithms are unbiased (Melo-Merino et al., 2020). In actual production, catches are subject to fishermen's choice, weather factors and fishing gear type, etc. (Alabia et al., 2015;Karp et al., 2023;Mugo and Saitoh, 2020;Munroe et al., 2022;Scales et al., 2017;Zhou et al., 2022), which can underestimate or overestimate the abundance of the observed area. In this study, the principle of light falling gear is to take advantage of the phototropism of PFS, and thus their catch rates are strongly influenced by the lunar phase. ...
... The adult albacore SDM was trained on the fishery-dependent data Karp et al., 2023;Pennino et al., 2016), dix S1), we also included three predictors describing gear configurations (as documented in observer records): the number of hooks between floats, the number of total hooks per set, and the length of the floatline on each set ( Table 1). As with the juvenile SDM, the adult SDM showed good skill against withheld test data (AUC = 0.89). ...
... While there is a limit to the amount of complexity that can be accommodated within existing assessment models and data aggregation procedures, we suggest that recent observed and future predicted oceanographic changes across the NPTZ (and their capacity to impact albacore habitat and fishery interactions; see and selectivity within and across pelagic longline fisheries have been shown to vary by leader construction (Ward et al., 2008), hook-type (Curran & Bigelow, 2011), and bait choice (Gilman et al., 2020). Karp et al., 2023;Pennino et al., 2016). ...
... With respect to the former, we choose to use observations from the North and South Pacific in the construction of the adult albacore SDM to sample over as broad a range of predictors as possible (Karp et al., 2023). Although it is possible some behavioral differences exist between North and South Pacific albacores, previous studies based in the South Pacific show similar environmental and latitudinal associations of juveniles and adults to those in the northern hemisphere (e.g., Williams et al., 2015). ...
The management and conservation of tuna and other transboundary marine species have to date been limited by an incomplete understanding of the oceanographic, ecological and socioeconomic factors mediating fishery overlap and interactions, and how these factors vary across expansive, open ocean habitats. Despite advances in fisheries monitoring and biologging technology, few attempts have been made to conduct integrated ecological analyses at basin scales relevant to pelagic fisheries and the highly migratory species they target. Here, we use vessel tracking data, archival tags, observer records, and machine learning to examine inter‐ and intra‐annual variability in fisheries overlap (2013–2020) of five pelagic longline fishing fleets with North Pacific albacore tuna ( Thunnus alalunga , Scombridae). Although progressive declines in catch and biomass have been observed over the past several decades, the North Pacific albacore is one of the only Pacific tuna stocks primarily targeted by pelagic longlines not currently listed as overfished or experiencing overfishing. We find that fishery overlap varies significantly across time and space as mediated by (1) differences in habitat preferences between juvenile and adult albacore; (2) variation of oceanographic features known to aggregate pelagic biomass; and (3) the different spatial niches targeted by shallow‐set and deep‐set longline fishing gear. These findings may have significant implications for stock assessment in this and other transboundary fishery systems, particularly the reliance on fishery‐dependent data to index abundance. Indeed, we argue that additional consideration of how overlap, catchability, and size selectivity parameters vary over time and space may be required to ensure the development of robust, equitable, and climate‐resilient harvest control rules.
... In the framework of fishery management, in the absence of highquality data coming from independent surveys and research programs, occurrence records can be obtained from fishery-dependent data. These data usually come from scientific observers measuring catches on board of commercial fishing vessels or during the landing operations, or from logbooks required by local, national, or international regulations; they may represent the only commercial species distribution data available (Karp et al., 2022), also in the context of MPAs management. These data can be of poor quality compared to fishery-independent data (Conn et al., 2017;Thorson et al., 2020) since (i) they are not taken following a sampling design; (ii) specimens can be identified to taxonomic levels higher than species; (iii) catches can come from different vessels and gears; and (iv) sampling effort is often unknown. ...
... Furthermore, the future distribution of coastal fish species requires in-depth investigation regarding predictions of how will be their habitat used under various climate change scenarios. This aspect, crucial for a comprehensive understanding of potential distribution shifts in marine ecosystems and thus to ensure sustainable fishing in the future and to assess the climate risk on the fishing communities (Bang et al., 2022;Chamberlain et al., 2023;Karp et al., 2022;Rogers et al., 2019), has not been explored in the present study and warrants further research. ...
Marine protected areas (MPAs) make an essential contribution to the spatial management of critical areas, the conservation of coastal species exploited by human activities, and the sustainable use of marine resources. Within MPAs, fishing closure areas are among the most used small‐scale fishery management tools, even though their effectiveness largely remains untested or controversial. To reduce the impact of small‐scale fisheries on marine resources, a seasonal fishing closure area (SFCA) was established beginning in 2022 in autumn–winter season inside the Capo Caccia–Isola Piana MPA (Sardinia, northwestern Mediterranean Sea). Here, we assessed a posteriori whether the areas of higher habitat suitability for eight species/taxa of relevant ecological value and economic interest to small‐scale fisheries were included in the established SFCA, adequately meeting the ecological objectives of the MPA. Thus, landing data (from 2019 to 2023) were used as occurrence records to develop MaxEnt distribution models for the eight target species/taxa. The model outputs allow us to draw important insights about the spatial adequacy of the SFCA established within the MPA aimed to protect the most exploited marine resources. Furthermore, the modeling exercises were useful for understanding the local processes influencing species' habitat selection and to identify essential areas for the target species that could remain unrevealed in larger‐scale investigations.
... In contrast, the fishery-independent models exhibited generally lower evaluation metrics but were more broadly robust in their predictive performance and ecological realism, suggesting they may more accurately represent the realized environmental niche and geographic distribution of blue sharks beyond the footprint of the fishery. This distinction regarding the relative strengths of different data types may have even greater relevance for model projections to understand how species' distributions and their interactions with fisheries may shift under climate change (Karp et al., 2023). ...
... While previous studies have suggested that fisherydependent and fishery-independent datasets can lead to consistent estimates of species' habitats (Karp et al., 2023;Pennino et al., 2016), our results suggested that models trained with heavily biased data may significantly diverge F I G U R E 6 Proportion of presences (sensitivity, a) and "true" absences from the observer data (specificity, b) correctly predicted by each selected model (Table 4) and dataset combination. Model predictions were considered "correct" when predicted suitability was greater than the 75% quantile for presence observations and less than the 25% quantile for absences in the observer data. ...
Species distribution models (SDMs) are becoming an important tool for marine conservation and management. Yet while there is an increasing diversity and volume of marine biodiversity data for training SDMs, little practical guidance is available on how to leverage distinct data types to build robust models. We explored the effect of different data types on the fit, performance and predictive ability of SDMs by comparing models trained with four data types for a heavily exploited pelagic fish, the blue shark (Prionace glauca), in the Northwest Atlantic: two fishery dependent (conventional mark‐recapture tags, fisheries observer records) and two fishery independent (satellite‐linked electronic tags, pop‐up archival tags). We found that all four data types can result in robust models, but differences among spatial predictions highlighted the need to consider ecological realism in model selection and interpretation regardless of data type. Differences among models were primarily attributed to biases in how each data type, and the associated representation of absences, sampled the environment and summarized the resulting species distributions. Outputs from model ensembles and a model trained on all pooled data both proved effective for combining inferences across data types and provided more ecologically realistic predictions than individual models. Our results provide valuable guidance for practitioners developing SDMs. With increasing access to diverse data sources, future work should further develop truly integrative modeling approaches that can explicitly leverage the strengths of individual data types while statistically accounting for limitations, such as sampling biases.
... Finally, there are continuity issues in scientific data collection in relation to a changing world. For example, in the context of climate change, scientific surveys that are standardized to allow for time-series development of relative changes in fish stock populations may miss important changes in stock dynamics (Karp et al., 2022). Here, fishers' knowledge contributions can assist evaluation of the need for potential changes in survey design. ...
... What makes 'anecdotal' information considered to be less true useful for monitoring change is not necessarily that it is less true, but that it is regarded as not 'systematic' (Wilson, 2009). For example, stock assessment science tends to be based on large spatial scale units, discrete sampling techniques, and standardized sampling protocols, whereas experiential knowledge is often more localized and is based on different and often variable temporal scales and continuous sampling practices and technologies (Perry and Ommer, 2003;Wilson, 2009;Karp et al., 2022). These are some of the reasons why experiential knowledge is often considered unusable in fish stock or ecological assessment models; particularly those that are already data-rich (Mackinson and Nøttestad, 1998). ...
For future sustainable management of fisheries, we anticipate deeper and more diverse information will be needed. Future needs include not only biological data, but also information that can only come from fishers, such as real-time ‘early warning’ indicators of changes at sea, socio-economic data and fishing strategies. The fishing industry, in our experience, shows clear willingness to voluntarily contribute data and experiential knowledge, but there is little evidence that current institutional frameworks for science and management are receptive and equipped to accommodate such contributions. Current approaches to producing knowledge in support of fisheries management need critical re-evaluation, including the contributions that industry can make. Using examples from well-developed advisory systems in Europe, United States, Canada, Australia and New Zealand, we investigate evidence for three interrelated issues inhibiting systematic integration of voluntary industry contributions to science: (1) concerns about data quality; (2) beliefs about limitations in useability of unique fishers’ knowledge; and (3) perceptions about the impact of industry contributions on the integrity of science. We show that whilst these issues are real, they can be addressed. Entrenching effective science-industry research collaboration (SIRC) calls for action in three specific areas; (i) a move towards alternative modes of knowledge production; (ii) establishing appropriate quality assurance frameworks; and (iii) transitioning to facilitating governance structures. Attention must also be paid to the science-policy-stakeholder interface. Better definition of industry’s role in contributing to science will improve credibility and legitimacy of the scientific process, and of resulting management.
... Our analyses are an historical snapshot in time, representing the present value of the ocean for the most lucrative fisheries operating off the U.S. West Coast between 2011 and 2020. However, climate change is predicted to alter the spatio-temporal distributions of marine resources, including many of the fisheries' targets described in our results [107,108]. Changes in these species' distributions will likely alter the concentrations of fishing effort in the future and the accessibility of these species to fishers [109,110]. For example, California market squid have been one of the most conspicuous species observed off the coast of Oregon in recent years owing to ocean warming [111,112], and their apparent poleward redistribution was qualitatively evident in our landings data. ...
The West Coast of the U.S. has a vast offshore wind energy (OWE) electricity generation potential with value on the order of billions of USD, and pressure is mounting to develop large OWE projects. However, this seascape has numerous existing resource extraction uses, including a multi-billion dollar commercial fishing industry, which create the potential for conflict. To date, spatially explicit comparisons of OWE and commercial fisheries value have not been done, but are essential for marine spatial planning and for investigating the tradeoffs of OWE development on existing marine uses. In this analysis, we generate maps of OWE levelized cost of energy and of total economic activity generated by the top eight commercial fishing targets that account for the vast majority (~84%) of landed revenue off the U.S. West Coast. We quantify spatial overlap between these two ocean uses and use multiobjective optimization to develop tradeoff frontiers to investigate implications for both sectors from established state goals or mandates for OWE power generation capacity. There are clear differences in the exposure of each fishery in their traditional fishing grounds as a function of differing OWE capacity goals and outcomes vary depending on whether OWE development goals are achieved at a state-by-state level or a region-wide level. Responsible siting of OWE projects includes careful consideration of existing commercial fishing activities, and responsible transition to renewable energies on the West Coast and elsewhere accounts for the socio-economic consequences of the total economic activity associated with each fishery.
... Species distribution models (SDMs) are widely used to study the relationship between species and their environment [30][31][32][33]. In fisheries, SDMs are commonly used for applications such as predicting the distribution of fished species under current and future climate conditions [34,35]. Various types of SDMs exist, including generalized additive models (GAMs), maximum entropy networks (MAXNETs), and generalized linear models (GLMs), among others [36][37][38]. ...
The impact of global warming on fish distribution is a key factor in fishery management and sustainable development. However, limited knowledge exists regarding the influence of environmental factors on the distribution of Evynnis cardinalis under climate change. This study addresses this gap by predicting the species distribution under current conditions and three future climate scenarios (SSP126, SSP370, and SSP585) using five individual models and four ensemble models. The results demonstrate that the ensemble models outperform the single models, with majority voting (EMca) achieving the highest accuracy (ROC = 0.97, TSS = 0.85). Bathymetry (BM) and the sea surface height (SSH) are the primary factors influencing the distribution. The predictions indicate that the currently suitable habitats of E. cardinalis are primarily located in the Beibu Gulf region of the northern South China Sea. Under future climate scenarios, suitable habitat areas are expected to expand to higher latitudes and deeper waters, though highly suitable habitats in the western Guangdong coastal waters, western Beibu Gulf, and southwestern offshore waters of Hainan Island will significantly decrease.
... Commercial catch data are more accessible and allow for the consideration of long-term seasonal variability; however, the data have biases due to preferential sampling of commercial fishing vessels (Karp et al. 2023). This preferential sampling could include biases from intense sampling in coastal regions because ports constrain fishing vessel distribution and, in turn, effort. ...
Chub mackerel Scomber japonicus (subsequently referred to as mackerel), a commercially important small pelagic fish in Korea, is highly sensitive to environmental changes and has shifted its spatial distribution owing to climate change in recent decades. This study examined projected changes in the seasonal potential distribution of mackerel in Korean waters in the 2050s. Three species distribution models—a maximum entropy model, a generalised additive model, and boosted regression trees—were fitted using mackerel presence and 5 skillful environmental variables (temperature, salinity, current velocity, and chlorophyll concentration at the surface, and mixed layer depth) over 18 yr (1998-2015) and projected under 3 CMIP6 future scenarios. The distribution models projected future changes in mackerel habitat with high seasonal and regional variability. Mackerel habitat was projected to increase by 13.35-42.01% throughout the year in the East Sea and decrease by up to 12.73% in the northern East China Sea and by 5.28-20.93% in the Yellow Sea in spring and summer. The habitat gains and losses of mackerel were mainly driven by the predicted temperature increases and salinity decreases. The habitat contraction in spawning areas—mainly in the Yellow and northern East China Seas—contributes to the loss of spawning habitats, which could considerably change the abundance and timing of spawning and, in turn, fisheries productivity. Our findings suggest that future changes in the seasonal potential distribution of mackerel and their potential impacts on fishing communities should be considered to effectively plan future management strategies, particularly for environmentally sensitive species such as mackerel.
... Species distribution models (SDMs) have been widely used in ecological analysis for a multitude of purposes, such as making inference of ecological niches (Guisan and Zimmermann, 2000), the assessment of climate change impacts on habitats (Karp et al., 2023), the prediction of future invader and invasive species (Fournier et al., 2019;Landis et al., 2000), the suggestion of areas for protection (Paradinas et al., 2022), and the refinement of biodiversity inventories (Staniczenko et al., 2017). ...
... Additionally, incorporating various data sources, such as historical ecology or eDNA, into actionable datasets is crucial for modeling future distributions amidst changing political and ecological seascapes. For example, combining commercial (fisheries-dependent) and survey-based (fisheries-independent) data, complemented with information from citizen science, are considered to address gaps in data coverage to improve species distribution mapping and modeling (Braun et al., 2023;Karp et al., 2023;Rufener et al., 2021). ...
The 2023 Annual Symposium of the Fisheries Society of the British Isles hosted opportunities for researchers, scientists, and policy makers to reflect on the state of art of predicting fish distributions and consider the implications to the marine and aquatic environments of a changing climate. The outcome of one special interest group at the Symposium was a collection of questions, organized under five themes, which begin to capture the state of the field and identify priorities for research and management over the coming years. The five themes were Physiology, Mechanisms, Detect and Measure, Manage, and Wider Ecosystems. The questions, 25 of them, addressed concepts which remain poorly understood, are data deficient, and/or are likely to be impacted in measurable or profound ways by climate change. Moving from the first to the last theme, the questions expanded in the scope of their considerations, from specific processes within the individual to ecosystem‐wide impacts, but no one question is bigger than any other: each is important in detecting, understanding, and predicting fish distributions, and each will be impacted by an aspect of climate change. In this way, our questions, particularly those concerning unknown mechanisms and data deficiencies, aimed to offer a guide to other researchers, managers, and policy makers in the prioritization of future work as a changing climate is expected to have complex and disperse impacts on fish populations and distributions that will require a coordinated effort to address.
... A large proportion of the occurrence data we retrieved from the global repositories was fishery-dependent in conjunction with a smaller proportion of fishery-independent data. Although fishery dependent data are inherently biased, they can still be useful to supplement other data sources, if the SDMs account properly for preferential sampling and other potential bias sources (Karp et al., 2023). We developed here a methodological approach to generate 3D fish distribution based exclusively on public species occurrence data and environmental correlates that conform with ecological niche theory, using shape-constrained GAMs. ...
Fisheries have a crucial contribution, with animal protein supply and economic income, to the subsistence and blue economy of several human societies of the Atlantic Ocean, the second largest water body in the planet. However, an accurate distribution of commercial fish across the Atlantic and through the water column is still unknown. The wide use of Species Distribution Models (SDMs) for marine fish mapping generally faces two shortcomings: (i) ignoring the vertical dimension of the ocean; and (ii) ignoring the ecological niche theory in the model fitting. Our aim is to develop 3D habitat models of the main commercial fishes across the Atlantic Ocean, accounting for 67 % of the total biomass catches, to provide an enhanced spatial representation of the environmental niche of the fish species. In particular, here we (1) explore the macroecological patterns testing if latitudinal-vertical distribution of main commercial fish species follows the isothermal distribution across the Atlantic ocean; (2) apply a novel 3D modelling approach incorporating depth dimension into the environmental data and based exclusively on public species occurrence data; (3) use Shape-Constrained Generalized Additive Models (SC-GAMs) to build SDMs in accordance with the ecological niche theory (GAM-NICHE model), avoiding potential model overfitting and hence allowing automatic model selection; and (4) estimate potential fish catch biomass in the 3D space based on the species probability of occurrence. Our results indicated that latitudinal-vertical distribution follows the prevailing isothermal distribution in the ocean, confirming that an accurate representation of stock distributions needs 3D modelling and incorporate explicitly depth dimension into the environmental data. The species response curves to 3D environmental gradients for the 30 main commercial fish species of the Atlantic yielded very good model accuracy performance (78-98 %). The developed 3D models of fish occurrence probability have the capability to be improved with the updates of new data for data-poor species, and to be projected under climate change scenarios. The obtained 3D maps conform useful and new knowledge that may help policy makers to balance the need for environmental protection with sustainable marine resource exploitation of the Atlantic Ocean.
... And this feature was also found on the European anchovy with a clearer age range (40-60days) [35], belonging to the same genus as Japanese anchovy, which showed different growth between Tyrrthenian Sea and Strait of Slicily of the Mediterranean [36,37], and the fat content also showed large monthly changes in 2004 and 2005 in the Yellow Sea of China [37]. The drastic responses to environment, in common with the most short-lived fish and r-strategists, increase the uncertainty for its fishery management [11], specifically for climate change [38]. Anchovy stock is characterized by sudden and rapid fluctuations of biomass [39], therefore requires frequent assessments and management procedure adjustments for its sustainable objectives. ...
Most fisheries in developing countries are data-limited, which creates a huge challenge to the fishery management. Therefore full utilization of the available information is essential. Japanese anchovy (Engraulis japonicus) plays a key role in the marine ecosystem, but the anchovy population in the Yellow Sea of China has declined in the recent decades, which is also a data-limited fishery. In order to implement robust management strategies to recover the anchovy stock, based on the available information as well as the associated uncertainties, here we examine its management strategy evaluation using a computer software package data-limited management toolkit (DLMtool). Results indicated that fishing pressure on the Yellow Sea anchovy should be slightly reduced and the length at first capture should exceed the length at first maturity. These are high probably conductive to anchovy population recovery. We also selected some management procedures performed well in this study, such as minlenLopt1. Paying attention to quantitative population dynamics and perfecting data collection are important for future anchovy fisheries management against uncertainties in both model and data.
... Species Distribution Models (SDMs) have become a widely used tool for the analysis of the ecology realms, allowing to infer ecological niches (Guisan and Zimmermann, 2000), the effect of climate change on suitability habitats (Karp et al., 2022), predicting new presence localities for species (Fois et al., 2018), predicting invaders and invasive species (Fournier et al., 2019;Landis et al., 2000), proposing new protected areas or refining biodiversity inventories (Staniczenko et al., 2017). ...
In ecology we may find scenarios where the same phenomenon (species occurrence, species abundance, etc.) is observed using two different types of samplers. For instance, species data can be collected from scientific surveys with a completely random sample pattern, but also from opportunistic sampling (e.g., whale or bird watching fishery commercial vessels), in which observers tend to look for a specific species in areas where they expect to find it. Species Distribution Models (SDMs) are a widely used tool for analyzing this kind of ecological data. Specifically, we have two models available for the above data: an independent model (IM) for the data coming from a complete random sampler and a dependent model (DM) for data from opportunistic sampling. In this work, we propose a sequential Bayesian procedure to connect these two models through the update of prior distributions. Implementation of the Bayesian paradigm is done through the integrated nested Laplace approximation (INLA) methodology, a good option to make inference and prediction in spatial models with high performance and low computational costs. This sequential approach has been evaluated by simulating several scenarios and comparing the results of sharing information from one model to another using different criteria. Our main results imply that, in general, it is better to share information from the independent (completely random) to the dependent model than the alternative way. However, it depends on different factors such as the spatial range or the spatial arrangement of sampling locations.
... In order to identify the social and economic factors impacting catch rates and account for them in CPUE standardization, it is necessary to assimilate the experiential knowledge of harvesters and processors (Steins et al., 2020;Mackinson, 2022;Steins et al., 2022). Novel modeling tools, such as spatiotemporal delta-generalized linear mixed models, structured additive distributional regression, and simulations further enable researchers to identify bias in and derive population trends from fishery dependent data (Mamouridis et al., 2017;Clegg et al., 2022;Ducharme-Barth et al., 2022;Karp et al., 2022). ...
Sources of fisheries information outside of fishery-independent surveys (e.g. fishery-dependent data) are especially valuable for species that support productive fisheries and lack reliable biological information, such as the northern shortfin squid (Illex illecebrosus). Fishery-dependent data streams are available for most species, however collaboration with industry members is critical to ensure that these fishery-dependent data are collected, applied, and interpreted correctly. Despite the need for collaboration and the frequency that fishery data are used in scientific research, there is limited literature on the structure of interactions and knowledge sharing that inform the analysis and application of fishery data. Between 2019 and 2022, a group of researchers collaborated with members of the northern shortfin squid fishing industry to bring together research data sets and knowledge from harvesters and processors to better describe the fishery dynamics, distribution, life history, and oceanographic drivers of the species. The collaboration focused on developing custom standardized fishery catch per unit effort (CPUE) indices to provide indicators of population trends that accounted for the impacts of technical and economic aspects of harvesting, processing and marketing on fishing effort, selectivity and landings of northern shortfin squid. We describe the methods used to inform and interpret the CPUE analyses, focusing on novel structure of interactions we had with industry members, and suggest best practices for integrating industry knowledge into CPUE standardization. The information shared and research products produced through this science-industry research collaboration advanced understanding of northern shortfin squid population and fishery dynamics, and contributed directly to the 2022 stock assessment and management process. Given the complex and stochastic nature of the northern shortfin squid population and fishery, we found it critical to maintain open communication and trust with processors and harvesters, who have unique insight into the factors that may be driving changes in catch, landings, and productivity of the valuable resource species.
... Given the multitude benefits of using fishers' knowledge to inform policy, it begs the question why it's underutilized? For catch data, there is the issue of bias in samples for density estimates, as catch logs exclusively record instances of fishing activity, neglecting areas not targeted by fishers, which biases predictions of species distributions (Karp et al., 2022). Also, given the unsystematic way much of fishers' knowledge is handled, it is often neglected (Hind, 2015). ...
There is increasing interest in utilizing fishers’ knowledge to better understand the marine environment, given the spatial extent and temporal resolution of fishing vessel operations. Furthermore, fishers’ knowledge is part of the best available information needed for sustainable harvesting of stocks, marine spatial planning and large-scale monitoring of fishing activity. However, there are difficulties with integrating such information into advisory processes. Data is often not systematically collected in a structured manner and there are issues around sharing of information within the industry, and between industry and research partners. Decision support systems for fishing planning and routing can integrate relevant information in a systematic way, which both incentivizes vessels to share information beneficial to their operations and capture time sensitive big datasets for marine research. The project Fishguider has been developing such a web-based decision support tool since 2019, together with partners in the Norwegian fishing fleet. The objectives of the project are twofold: 1) To provide a tool which provides relevant model and observation data to skippers, thus supporting sustainable fishing activity. 2) To foster bidirectional information flow between research and fishing activity by transfer of salient knowledge (both experiential and data-driven), thereby supporting knowledge creation for research and advisory processes. Here we provide a conceptual framework of the tool, along with current status and developments, while outlining specific challenges faced. We also present experiential input from fishers’ regarding what they consider important sources of information when actively fishing, and how this has guided the development of the tool. We also explore potential benefits of utilizing such experiential knowledge generally. Moreover, we detail how such collaborations between industry and research may rapidly produce extensive, structured datasets for research and input into management of stocks. Ultimately, we suggest that such decision support services will motivate fishing vessels to collect and share data, while the available data will foster increased research, improving the decision support tool itself and consequently knowledge of the oceans, its fish stocks and fishing activities.
Context
Satellite telemetry has revolutionised the study of animal movement, particularly for mobile marine animals, whose movements and habitat make consistent, long-term observation difficult.
Aims
Summarise the movements of Rio Lady, a mature female whale shark (Rhincodon typus), to characterise these movements, and to predict expected behaviour throughout the Gulf of Mexico (GOM).
Methods
Rio Lady was tracked using satellite telemetry for over 1600 days, generating over 1400 locations and travelling over 40,000 km. State–space and move persistence modelling enabled characterisation of behaviour, and machine learning (ML) enabled the development of habitat-suitability models to predict habitat utilisation, on the basis of location transmissions and their environmental covariates.
Key results
Rio Lady exhibited annually consistent patterns of movements among three regions within the GOM. Final ML models produced seasonally dynamic predictions of habitat use throughout the GOM.
Conclusions
The application of these methods to long-term location data exemplifies how long-term movement patterns and core areas can be discovered and predicted for marine animals.
Implications
Despite our limited dataset, our integrative approach advances methods to summarise and predict behaviour of mobile species and improve understanding of their ecology.
Objective
Our objective was to use sportfishing tournament data to determine whether sizes of Dolphinfish Coryphaena hippurus have been changing in the western North Atlantic (WNA) over recent decades.
Methods
We sampled North Carolina, South Carolina, and Florida marine sportfishing tournament landings for Dolphinfish lengths. Linear models were separately fitted to length data for males and females by regressing length against year. A subset of these models (analysis of covariance) considered tournament as a factor.
Result
An analysis of covariance model with a separate regression slope for each tournament provided the best fit to the data for male and female Dolphinfish. Meaningful temporal declines in the length of males and females were found for four of the five tournaments (no changes in length were observed for the fifth tournament). Median total length declines of 168, 105, 103, and 426 mm were predicted for males, and declines of 354, 133, 131, and 246 mm were predicted for females. Declines in the largest observed sizes of Dolphinfish (97.5% confidence limit) were found for most tournament‐ and sex‐specific combinations of data and could suggest excess fishing mortality on the population.
Conclusion
Declines in Dolphinfish size in the WNA region could have ramifications for conservation of the population given that these size changes translate into reduced individual fecundity of female Dolphinfish. Causes of the size decline could be fishing effects, environmental effects, or a combination of these. Reductions in individual size may be occurring simultaneously with declines in abundance identified in other recent research using fishery‐dependent data collected in the WNA.
Long-term trends in fisheries catch are useful to monitor effects of fishing on wild populations. However, fisheries catch data are often aggregated in multi-species complexes, complicating assessments of individual species. Non-target species are often grouped together in this way, but this becomes problematic when increasingly common shifts toward targeting incidental species demand closer management focus at the species level. Species distribution models (SDMs) offer an under-utilised tool to allocate aggregated catch data among species for individual assessments. Here, we present a case study of two shovel-nosed lobsters (Thenus spp.), previously caught incidentally and recorded together in logbook records, to illustrate the design and use of catch allocation SDMs to untangle multi-species data for stock assessments of individual species. We demonstrate how catch allocation SDMs reveal previously masked species-specific catch trends from aggregated data and can identify shifts in fishing behaviour, e.g., changes in target species. Finally, we review key assumptions and limitations of this approach that may arise when applied across a broad geographic or taxonomic scope. Our aim is to provide a template to assist researchers and managers seeking to assess stocks of individual species using aggregated multi-species data.
Yellowfin tuna, Thunnus albacares, represents an important component of commercial and recreational fisheries in the Gulf of Mexico (GoM). We investigated the influence of environmental conditions on the spatiotemporal distribution of yellowfin tuna using fisheries’ catch data spanning 2012–2019 within Mexican waters. We implemented hierarchical Bayesian regression models with spatial and temporal random effects and fixed effects of several environmental covariates to predict habitat suitability (HS) for the species. The best model included spatial and interannual anomalies of the absolute dynamic topography of the ocean surface (ADTSA and ADTIA, respectively), bottom depth, and a seasonal cyclical random effect. High catches occurred mainly towards anticyclonic features at bottom depths > 1000 m. The spatial extent of HS was higher in years with positive ADTIA, which implies more anticyclonic activity. The highest values of HS (> 0.7) generally occurred at positive ADTSA in oceanic waters of the central and northern GoM. However, high HS values (> 0.6) were observed in the southern GoM, in waters with cyclonic activity during summer. Our results highlight the importance of mesoscale features for the spatiotemporal distribution of yellowfin tunas and could help to develop dynamic fisheries management strategies in Mexico and the U.S. for this valuable resource.
Effective and sustainable management of small-scale fisheries (SSF) is challenging. We describe a novel approach to identify important fishing grounds for SSF, by implementing a habitat modelling approach, using environmental predictors and Automatic Identification System (AIS)-B data coupled with logbook and First Sales Notes data, within the SE Bay of Biscay. Fishing activity patterns and catches of longliners and netters are used to determine the main environmental characteristics of the fishing grounds, and a habitat modelling approach is implemented to predict the zones that fulfil similar environmental characteristics across a larger geographical extent. Generalized additive mixed models (GAMMs) were built for 24 fish species, and to identify other zones that fulfil similar characteristics and, thus, could be considered relevant for the species targeted by each gear type. Most of the models showed a good prediction capacity. The models included between one and four predictor variables. ‘Depth of mixing layer’ and ‘benthic rocky habitat’ were the variables included more frequently for fish species captured by netter’s fleet. For longliners, the ‘seafloor slope’ and ‘benthic rocky habitat’ were the two most important variables. The predictive maps provide relevant information to assist in management and marine spatial planning.
Context
There have been many studies using species distribution models (SDMs) to predict shifts in species distributions due to environmental changes, but few consider effects of data quantity, data quality, or species response shape. Modeling studies using field-sampled data may be impaired to an unknown degree by lack of knowledge on species’ true relationships with environmental changes.
Objectives
Using simulations with known relationships we assess model predictions, and investigate which models are more sensitive to sample size, detection limit, or species response shape issues when different SDMs are used for predicting species distribution shifts under environmental changes.
Methods
We simulated 16 species response relationships to ecological gradients differing in response shape (skewness and kurtosis) using a generalized β-function. Populations were randomly sampled at different sample sizes and detection limits. Linear discriminant analysis (LDA), multiple logistic regression (MLR), generalized additive models (GAM), boosted regression trees (BRT), random forests (RF), artificial neural networks (ANN), and maximum entropy models (MaxEnt) were developed on sampled datasets and compared for predicting species occurrence. We used these SDMs to predict distribution patterns for virtual species with different response shapes across a real landscape of varying heterogeneity in environmental conditions, and compared them with the probability of occurrence generated by the β-function.
Results
GAM and BRT were sensitive to both sample size and detection limit changes; RF was more affected by detection limit; ANN and MaxEnt were more affected by sample size; LDA and MLR were sensitive to species response shape changes.
Conclusions
Overall, if little is known about species response to environmental changes, ANN is recommended especially for large sample size. If a focal species is likely to occur only in a narrow range of environmental conditions, GAM and BRT are preferred for large good-quality datasets, and GAM tends to perform slightly better under varied data conditions; RF is recommended for limited amounts of good-quality data. If a focal species is likely to be present in a wide range of environmental conditions, MaxEnt is preferred but caution should be taken for small sample size. If the goal is to identify potential distributions of invasive or endangered species but data quantity and quality are very limited, LDA and MLR are recommended as they generally provide reasonable model sensitivity.
The Northwest Atlantic Ocean and Gulf of Mexico are among the fastest warming ocean regions, a trend that is expected to continue through this century with far-reaching implications for marine ecosystems. We examine the distribution of 12 highly migratory top predator species using predictive models and project expected habitat changes using downscaled climate models. Our models predict widespread losses of suitable habitat for most species, concurrent with substantial northward displacement of core habitats >500 km. These changes include up to >70% loss of suitable habitat area for some commercially and ecologically important species. We also identify predicted hot spots of multi-species habitat loss focused offshore of the U.S. Southeast and Mid-Atlantic coasts. For several species, the predicted changes are already underway, which are likely to have substantial impacts on the efficacy of static regulatory frameworks used to manage highly migratory species. The ongoing and projected effects of climate change highlight the urgent need to adaptively and proactively manage dynamic marine ecosystems.
Species distribution models (SDMs) are becoming an important tool for marine conservation and management. Yet while there is an increasing diversity and volume of marine biodiversity data for training SDMs, little practical guidance is available on how to leverage distinct data types to build robust models. We explored the effect of different data types on the fit, performance and predictive ability of SDMs by comparing models trained with four data types for a heavily exploited pelagic fish, the blue shark (Prionace glauca), in the Northwest Atlantic: two fishery-dependent (conventional mark-recapture tags, fisheries observer records) and two fishery-independent (satellite-linked electronic tags, pop-up archival tags). We found that all four data types can result in robust models, but differences among spatial predictions highlighted the need to consider ecological realism in model selection and interpretation regardless of data type. Differences among models were primarily attributed to biases in how each data type, and the associated representation of absences, sampled the environment and summarized the resulting species distributions. Outputs from model ensembles and a model trained on all pooled data both proved effective for combining inferences across data types and provided more ecologically realistic predictions
than individual models. Our results provide valuable guidance for practitioners developing SDMs. With increasing access to diverse data sources, future work should further develop truly integrative modeling approaches that can explicitly leverage strengths of individual data types while statistically accounting for limitations, such as sampling biases.
Projecting the future distributions of commercially and ecologically important species has become a critical approach for ecosystem managers to strategically anticipate change, but large uncertainties in projections limit climate adaptation planning. Although distribution projections are primarily used to understand the scope of potential change ‐ rather than accurately predict specific outcomes ‐ it is nonetheless essential to understand where and why projections can give implausible results and to identify which processes contribute to uncertainty. Here, we use a series of simulated species distributions, an ensemble of 252 species distribution models, and an ensemble of three regional ocean climate projections, to isolate the influences of uncertainty from earth system model spread and from ecological modeling. The simulations encompass marine species with different functional traits and ecological preferences to more broadly address resource manager and fishery stakeholder needs, and provide a simulated true‐state with which to evaluate projections. We present our results relative to the degree of environmental extrapolation from historical conditions, which helps facilitate interpretation by ecological modelers working in diverse systems. We found uncertainty associated with species distribution models can exceed uncertainty generated from diverging earth system models (up to 70% of total uncertainty by 2100), and that this result was consistent across species traits. Species distribution model uncertainty increased through time and was primarily related to the degree to which models extrapolated into novel environmental conditions but moderated by how well models captured the underlying dynamics driving species distributions. The predictive power of simulated species distribution models remained relatively high in the first 30 years of projections, in alignment with the time period in which stakeholders make strategic decisions based on climate information. By understanding sources of uncertainty, and how they change at different forecast horizons, we provide recommendations for projecting species distribution models under global climate change.
A useful measure of general climate stress is where and when novel habitats emerge. Here we evaluate ‘climate envelope novelty’–a spatial indicator of system-level habitat change–in the California Current System (CCS), by quantifying the emergence of novel ocean conditions in multivariate physical-biogeochemical space. We use downscaled climate projections from three earth system models out to 2100 under emission scenario RCP8.5, and detect novelty at multiple spatial-temporal scales using two methods (n-dimensional hypervolumes and extrapolation detection). Under high emissions, persistent novelty doesn’t appear until around 2040 and then only in small patches of Southern California and the Pacific North West. However, novelty increases rapidly after this (especially in warmer seasons), so that by 2060 up to 50% of the CCS in an average year has shifted to a novel local climate, which increases to 100% by 2090. These results are for the average year, and the first years to experience these levels of novelty typically occur 20 years sooner. The ecosystem will increasingly experience novel combinations of warmer temperatures, lower dissolved oxygen (especially inshore), and a shallower mixed layer (especially offshore). The emergence of extensive local novelty year-round has implications for the required ubiquitous redistribution or adaptation of CCS ecology, and the emergence of extensive regional novelty in warmer months has implications for bioregional change and regionally emerging fisheries. One of our climate projections showed considerably less novelty, indicating that realistic uncertainties in climate change (especially the rate of warming) can mean the difference between a mostly novel or mostly analog future.
Developing Species Distribution Models (SDM) for marine exploited species is a major challenge in fisheries ecology. Classical modelling approaches typically rely on fish research survey data. They benefit from a standardized sampling design and a controlled catchability, but they usually occur once or twice a year and they may sample a relatively small number of spatial locations. Spatial monitoring of commercial data (based on logbooks crossed with Vessel Monitoring Systems) can provide an additional extensive data source to inform fish spatial distribution.
We propose a spatial hierarchical framework integrating both data sources while accounting for preferential sampling (PS) of commercial data.
From simulations, we demonstrate that PS should be accounted for in estimation when PS is actually strong. When commercial data far exceed scientific data, the later bring little information to spatial predictions in the areas sampled by commercial data, but bring information in areas with low fishing intensity and provide a validation dataset to assess the integrated model consistency. We applied the framework to three demersal species (hake, sole, and squids) in the Bay of Biscay that emphasize contrasted PS intensity and we demonstrate that the framework can account for several fleets with varying catchabilities and PS behaviours.
The ocean is warming, losing oxygen and being acidified, primarily as a result of anthropogenic carbon emissions. With ocean warming, acidification and deoxygenation projected to increase for decades, extreme events, such as marine heatwaves, will intensify, occur more often, persist for longer periods of time and extend over larger regions. Nevertheless, our understanding of oceanic extreme events that are associated with warming, low oxygen concentrations or high acidity, as well as their impacts on marine ecosystems, remains limited. Compound events—that is, multiple extreme events that occur simultaneously or in close sequence—are of particular concern, as their individual effects may interact synergistically. Here we assess patterns and trends in open ocean extremes based on the existing literature as well as global and regional model simulations. Furthermore, we discuss the potential impacts of individual and compound extremes on marine organisms and ecosystems. We propose a pathway to improve the understanding of extreme events and the capacity of marine life to respond to them. The conditions exhibited by present extreme events may be a harbinger of what may become normal in the future. As a consequence, pursuing this research effort may also help us to better understand the responses of marine organisms and ecosystems to future climate change.
Abundance indices derived from fisheries-dependent data (catch-per-unit-effort or CPUE) are known to have potential for bias, in part because of the usual non-random nature of fisheries spatial distributions. However, given the cost and lack of availability of fisheries-independent surveys, fisheries-dependent CPUE remains a common and informative input to fisheries stock assessments. Recent research efforts have focused on the development of spatiotemporal delta-generalized linear mixed models (GLMMs) which simultaneously standardize the CPUE and predict abundance in unfished areas when estimating the abundance index. These models can include local seasonal environmental covariates (e.g. sea surface temperature) and a spatially varying response to regional annual indices (e.g. the El Niño Southern Oscillation) to interpolate into unfished areas. Spatiotemporal delta-GLMMs have been demonstrated in simulation studies to perform better than conventional, non-spatial delta-generalized linear models (GLMs). However, spatiotemporal delta-GLMMs have rarely been evaluated in situations where fisheries spatial sampling patterns change over time (e.g. fisheries expansion or spatial closures). This study develops a simulation framework to evaluate 1) how the nature of fisheries-dependent spatial sampling patterns may bias estimated abundance indices, 2) how shifts in spatial sampling over time impact our ability to estimate temporal changes in catchability, and 3) how including seasonal environmental covariates and/or regional annual indices in spatiotemporal delta-GLMMs can improve the estimation of abundance indices given shifts in spatial sampling. Spatiotemporal delta-GLMMs are then applied to a case study example where the spatial sampling pattern changed dramatically over time (contraction of the Japanese pole-and-line fishery for skipjack tuna Katsuwonus pelamis in the western and central Pacific Ocean). Results from simulations indicate that spatial sampling in proportion to the underlying biomass can produce similar abundance indices to those produced under random sampling. Though estimated abundance indices were not perfect, spatiotemporal GLMMs were generally able to disentangle shifts in spatial sampling from temporal changes in catchability when shifts in spatial sampling were not too extreme. Lastly, the inclusion of seasonal environmental covariates and/or regional oceanographic indices in spatiotemporal GLMMs did not improve abundance index estimation and in some cases resulted in degraded model performance.
Monitoring and assessment of natural resources often require inputs from multiple data sources. In fisheries science, for example, the inference of a species’ abundance distribution relies on two main data sources, namely commercial fisheries and scientific survey data. Despite efforts to combine these data into an integrated statistical model, their coupling is frequently hampered due to differences in their sampling designs, which imposes distinct bias sources in the estimator of the abundance distribution. We developed a flexible species distribution model (SDM) that can integrate both data sources while filtering out their relative bias contributions. We applied the model on three different age groups of the western Baltic cod stock. For each age group, we tested the model on (1) survey data and (2) integrated data (survey + commercial) as a means to compare their differences and investigate how the inclusion of commercial fisheries data improved the spatiotemporal abundance estimator and parameter estimates. Moreover, we proposed a novel validation approach to evaluate whether the inclusion of commercial fisheries data in the integrated model is not in direct contradiction with the survey data. Following our approach, the results indicated that the use of commercial fisheries data is suitable for the integrated model. Across all age groups, our results demonstrated how commercial fisheries supplied additional information on cod’s spatiotemporal abundance dynamics, highlighting sometimes abundance hot spots that were not detected by the survey model alone. Additionally, the integrated model provided a reduction of up to 20% and 10% in the uncertainty (SE) of the predicted abundance fields and fixed‐effect parameters, respectively. The proposed model represents thus a valuable benchmark for evaluating spatiotemporal dynamics of fish, and strengthens the science‐based advice for marine policymakers.
Statistical models built using different data sources and methods can exhibit conflicting patterns. We used the northern stock of black sea bass (Centropristis striata) as a case study to assess the impacts of using different fisheries data sources and laboratory-derived physiological metrics in the development of thermal habitat models for marine fishes. We constructed thermal habitat models using general-
ized additive models (GAMs) based on various fisheries datasets as input, including the NOAA Northeast Fisheries Science Center (NEFSC) bottom trawl surveys, various inshore fisheries-independent trawl surveys (state waters), NEFSC fisheries-dependent observer data, and laboratory- based physiological metrics. We compared each model's GAM response curve and coupled them to historical ocean conditions in the U.S. Northeast Shelf using bias- corrected ocean temperature output from a
regional ocean model. Thermal habitat models based on shelf- wide data (NEFSC fisheries- dependent observer data and fisheries- independent spring and fall surveys) explained the most variation in black sea bass presence/absence data at ~15% deviance explained. Models based on a narrower range of sampled thermal habitat from inshore survey data in the Northeast Area Monitoring and Assessment Program (NEAMAP) and the geographically isolated Long Island Sound data performed poorly.
All models had similar lower thermal limits around 8.5°C, but thermal optima, when present, ranged from 16.7 to 24.8°C. The GAMs could reliably predict habitat from years excluded from model training, but due to strong seasonal temperature fluctuations in the region, could not be used to predict habitat in seasons excluded from training. We conclude that survey data source can greatly impact development and interpretation of thermal habitat models for marine fishes. We suggest that model development be based on data sources that sample the widest range of ocean temperature and physical habitat throughout multiple seasons when possible, and encourage thorough consideration of how data gaps may influence model uncertainty.
Spatial management for highly migratory species (HMS) is difficult due to many species’ mobile habits and the dynamic nature of oceanic habitats. Current static spatial management areas for fisheries in the United States have been in place for extended periods of time with limited data collection inside the areas, making any analysis of their efficacy challenging. Spatial modeling approaches can be specifically designed to integrate species data from outside of closed areas to project species distributions inside and outside closed areas relative to the fishery. We developed HMS-PRedictive Spatial Modeling (PRiSM), which uses fishery-dependent observer data of species’ presence–absence, oceanographic covariates, and gear covariates in a generalized additive model (GAM) framework to produce fishery interaction spatial models. Species fishery interaction distributions were generated monthly within the domain of two HMS longline fisheries and used to produce a series of performance metrics for HMS closed areas. PRiSM was tested on bycatch species, including shortfin mako shark (Isurus oxyrinchus), billfish (Istiophoridae), and leatherback sea turtle (Dermochelys coriacea) in a pelagic longline fishery, and sandbar shark (Carcharhinus plumbeus), dusky shark (C. obscurus), and scalloped hammerhead shark (Sphyrna lewini) in a bottom longline fishery. Model validation procedures suggest PRiSM performed well for these species. The closed area performance metrics provided an objective and flexible framework to compare distributions between closed and open areas under recent environmental conditions. Fisheries managers can use the metrics generated by PRiSM to supplement other streams of information and guide spatial management decisions to support sustainable fisheries.
Management strategy evaluation (MSE) is a simulation approach that serves as a “light on the hill” (Smith, 1994) to test options for marine management, monitoring, and assessment against simulated ecosystem and fishery dynamics, including uncertainty in ecological and fishery processes and observations. MSE has become a key method to evaluate trade-offs between management objectives and to communicate with decision makers. Here we describe how and why MSE is continuing to grow from a single species approach to one relevant to multi-species and ecosystem-based management. In particular, different ecosystem modeling approaches can fit within the MSE process to meet particular natural resource management needs. We present four case studies that illustrate how MSE is expanding to include ecosystem considerations and ecosystem models as ‘operating models’ (i.e., virtual test worlds), to simulate monitoring, assessment, and harvest control rules, and to evaluate tradeoffs via performance metrics. We highlight United States case studies related to fisheries regulations and climate, which support NOAA’s policy goals related to the Ecosystem Based Fishery Roadmap and Climate Science Strategy but vary in the complexity of population, ecosystem, and assessment representation. We emphasize methods, tool development, and lessons learned that are relevant beyond the United States, and the additional benefits relative to single-species MSE approaches.
Machine learning algorithms have become very popular for spatial mapping of the environment due to their ability to fit nonlinear and complex relationships. However, this ability comes with the disadvantage that they can only be applied to new data if these are similar to the training data. Since spatial mapping requires predictions to new geographic space which in many cases goes along with new predictor properties, a method to assess the area to which a prediction model can be reliably applied is required.
Here, we suggest a methodology that delineates the ‘area of applicability’ (AOA) that we define as the area where we enabled the model to learn about relationships based on the training data, and where the estimated cross‐validation performance holds. We first propose a ‘dissimilarity index’ (DI) that is based on the minimum distance to the training data in the multidimensional predictor space, with predictors being weighted by their respective importance in the model. The AOA is then derived by applying a threshold which is the (outlier‐removed) maximum DI of the training data derived via cross‐validation. We further use the relationship between the DI and the cross‐validation performance to map the estimated performance of predictions. We illustrate the approach in a simulated case study chosen to mimic ecological realities and test the credibility by using a large set of simulated data.
The simulation studies showed that the prediction error within the AOA is comparable to the cross‐validation error of the trained model, while the cross‐validation error does not apply outside the AOA. This applies to models being trained with randomly distributed training data, as well as when training data are clustered in space and where spatial cross‐validation is applied. Using the relationship between DI and cross‐validation performance showed potential to limit predictions to the area where a user‐defined performance applies.
We suggest to add the AOA computation to the modeller's standard toolkit and to present predictions for the AOA only. We further suggest to report a map of DI‐dependent performance estimates alongside prediction maps and complementary to (cross‐)validation performance measures and the common uncertainty estimates.
Given the ecological and economic importance of eastern boundary upwelling systems like the California Current System (CCS), their evolution under climate change is of considerable interest for resource management. However, the spatial resolution of global earth system models (ESMs) is typically too coarse to properly resolve coastal winds and upwelling dynamics that are key to structuring these ecosystems. Here we use a high-resolution (0.1°) regional ocean circulation model coupled with a biogeochemical model to dynamically downscale ESMs and produce climate projections for the CCS under the high emission scenario, Representative Concentration Pathway 8.5. To capture model uncertainty in the projections, we downscale three ESMs: GFDL-ESM2M, HadGEM2-ES, and IPSL-CM5A-MR, which span the CMIP5 range for future changes in both the mean and variance of physical and biogeochemical CCS properties. The forcing of the regional ocean model is constructed with a “time-varying delta” method, which removes the mean bias of the ESM forcing and resolves the full transient ocean response from 1980 to 2100. We found that all models agree in the direction of the future change in offshore waters: an intensification of upwelling favorable winds in the northern CCS, an overall surface warming, and an enrichment of nitrate and corresponding decrease in dissolved oxygen below the surface mixed layer. However, differences in projections of these properties arise in the coastal region, producing different responses of the future biogeochemical variables. Two of the models display an increase of surface chlorophyll in the northern CCS, consistent with a combination of higher nitrate content in source waters and an intensification of upwelling favorable winds. All three models display a decrease of chlorophyll in the southern CCS, which appears to be driven by decreased upwelling favorable winds and enhanced stratification, and, for the HadGEM2-ES forced run, decreased nitrate content in upwelling source waters in nearshore regions. While trends in the downscaled models reflect those in the ESMs that force them, the ESM and downscaled solutions differ more for biogeochemical than for physical variables.
Time-area closures are a valuable tool for mitigating fisheries bycatch. There is increasing recognition that dynamic closures, which have boundaries that vary across space and time, can be more effective than static closures at protecting mobile species in dynamic environments. We created a management strategy evaluation to compare static and dynamic closures in a simulated fishery based on the California drift gillnet swordfish fishery, with closures aimed at reducing bycatch of leatherback turtles. We tested eight operating models that varied swordfish and leatherback distributions, and within each evaluated the performance of three static and five dynamic closure strategies. We repeated this under 20 and 50% simulated observer coverage to alter the data available for closure creation. We found that static closures can be effective for reducing bycatch of species with more geographically associated distributions, but to avoid redistributing bycatch the static areas closed should be based on potential (not just observed) bycatch. Only dynamic closures were effective at reducing bycatch for more dynamic leatherback distributions, and they generally reduced bycatch risk more than they reduced target catch. Dynamic closures were less likely to redistribute fishing into rarely fished areas, by leaving open pockets of lower risk habitat, but these closures were often fragmented which would create practical challenges for fishers and managers and require a mobile fleet. Given our simulation’s catch rates, 20% observer coverage was sufficient to create useful closures and increasing coverage to 50% added only minor improvement in closure performance. Even strict static or dynamic closures reduced leatherback bycatch by only 30–50% per season, because the simulated leatherback distributions were broad and open areas contained considerable bycatch risk. Perfect knowledge of the leatherback distribution provided an additional 5–15% bycatch reduction over a dynamic closure with realistic predictive accuracy. This moderate level of bycatch reduction highlights the limitations of redistributing fishing effort to reduce bycatch of broadly distributed and rarely encountered species, and indicates that, for these species, spatial management may work best when used with other bycatch mitigation approaches. We recommend future research explores methods for considering model uncertainty in the spatial and temporal resolution of dynamic closures.
Many fish species are shifting spatial distributions in response to climate change, but projecting these shifts and measuring their impact at fine scales are challenging. We present a simulation that projects change in fishery landings due to spatial distribution shifts, by combining regional ocean and biogeochemical models (forced by three earth system models, ESMs: GFDL‐ESM2M, HadGEM2‐ES, IPSL‐CM5A‐MR), correlative models for species distribution and port‐level landings, and a simulation framework which provides realistic values for species abundance and fishery conditions using an historical “reference period”. We demonstrate this approach for the northern subpopulation of Pacific sardine, an iconic commercial species for the U.S. West Coast. We found a northward shift in sardine landings (based on the northern subpopulation's habitat suitability), with projected declines at southern ports (20%–50% decline by 2080) and an increase (up to 50%) or no change at northern ports, and this was consistent across the three ESMs. Total sardine landings were more uncertain, with HadGEM2 indicating a 20% decline from 2000 to 15 levels by 2070 (a rate of 170 mt/y), IPSL a 10% increase (115 mt/y), and GFDL an 15% increase by the year ~2050 followed by a sharp decrease. The ESMs also differed in their projected change to the timing of the fishing season and frequency of fishery closures. Our simulation also identified key constraints on future landings that can be targeted by more tactical assessment; these included the seasonality of quota allocation and the abundance of other species in the catch portfolio.
We apply the concept of threat points in game theory to explore the stability of current joint management arrangements for shared transboundary fish stocks between Canada and the United States. We use three examples to explore the effects of projected impacts of climate change on the productivity and distribution of these stocks between the exclusive economic zones of the two countries. The three stocks that we study are: Atlantic cod (Gadus morhua) and yellowtail flounder (Limanda ferruginea) within the Gulf of Maine, and Pacific halibut (Hippoglossus stenolepis) in the Pacific Coast. We define a threat point as the payoffs that the fisheries in Canada and the United States take home given the current management agreement between the two countries. This is an application of John Nash's threat point, defined as the minimum payoffs that each player in a game theoretic model must receive for the solution to a cooperative game to be stable, which is usually the outcome of a noncooperative game. First, we compute the threat points, that is, the current profits that Canada and the United States derive from the three shared stocks, respectively. Next, we build an ensemble of climate-marine ecosystem and economic models and use them to determine how climate change is likely to change current profits received by each country relative to the shifts in their threat points. We find that in some cases the profits obtained by fisheries in Canada and the United States would change under climate change both in absolute and relative terms resulting in relative changes in threat points. These relative changes in threat points serve as the basis for our discussion of the stability of current transboundary management agreements between Canada and the United States for these important shared stocks in the face of climate change.
Marine fisheries around the globe are increasingly exposed to external drivers of social and ecological change. Though diversification and flexibility have historically helped marine resource users negotiate risk and adversity, much of modern fisheries management treats fishermen as specialists using specific gear types to target specific species. Here, we describe the evolution of harvest portfolios amongst Pacific Northwest fishermen over 35+ years with explicit attention to changes in the structure and function of the albacore (Thunnus alalunga, Scombridae) troll and pole-and-line fishery. Our analysis indicates that recent social-ecological changes have had heterogenous impacts upon the livelihood strategies favoured by different segments of regional fishing fleets. As ecological change and regulatory reform have restricted access to a number of fisheries, many of the regional small (<45 ft) and medium (45-60 ft) boat fishermen who continue to pursue diverse livelihood strategies have increasingly relied upon the ability to opportunistically target albacore in coastal waters while retaining more of the value generated by such catch. In contrast, large vessels (>60 ft) targeting albacore are more specialized now than previously observed, even as participation in multiple fisheries has become increasingly common for this size class. In describing divergent trajectories associated with the albacore fishery, one of the US West Coast's last open-access fisheries, we highlight the diverse strategies and mechanisms utilized to sustain fisheries livelihoods in the modern era while arguing that alternative approaches to management and licensing may be required to maintain the viability of small-scale fishing operations worldwide moving forward.
Spatial distributions of marine fauna are determined by complex interactions between environmental conditions and animal behaviors. As climate change leads to warmer, more acidic, and less oxygenated oceans, species are shifting away from their historical distribution ranges, and these trends are expected to continue into the future. Correlative Species Distribution Models (SDMs) can be used to project future habitat extent for marine species, with many different statistical methods available. However, it is vital to assess how different statistical methods behave under novel environmental conditions before using these models for management advice, and to consider whether future projections based on these techniques are biologically reasonable. In this study, we built SDMs for adults and larvae of two ecologically important pelagic fishes in the California Current System (CCS): Pacific sardine (Sardinops sagax) and northern anchovy (Engraulis mordax). We used five different SDM methods, ranging from simple [thermal niche model (TNM)] to complex (artificial neural networks). Our results show that some SDMs trained on data collected between 2003 and 2013 lost substantial predictive skill when applied to observations from more recent years, when ocean temperatures associated with a marine heatwave were outside the range of historical measurements. This decrease in skill was particularly apparent for adult sardine, which showed non-stationary relationships between catch locations and sea surface temperature (SST) through time. While sardine adults and larvae shifted their distributions markedly during the marine heatwave, anchovy largely maintained their historical spatiotemporal distributions. Our results suggest that correlative relationships between species and their environment can become unreliable during anomalous conditions. Understanding the underlying physiology of marine species is therefore essential for the construction of SDMs that are robust to rapidly changing environments. Developing distribution models that offer skillful predictions into the future for species such as sardine and anchovy, which are migratory and include separate sub-stocks, may be particularly challenging.
Invited Discussion : Bertrand Clarke - Meng Li - Peter Grunwald and Rianne de Heide Contributed Discussion : A. Philip Dawid - William Weimin Yoo - Robert L. Winkler, Victor Richmond R. Jose, Kenneth C. Lichtendahl Jr., and Yael Grushka-Cockayne - Kenichiro McAlinn, Knut Are Aastveit, and Mike West - Minsuk Shin - Tianjian Zhou - Lennart Hoogerheide and Herman K. van Dijk - Haakon C. Bakka, Daniela Castro-Camilo, Maria Franco-Villoria, Anna Freni-Sterrantino, Rapha¨el Huser, Thomas Opitz, and Havard Rue - Marco A. R. Ferreira - Luis Pericchi - Christopher T. Franck - Eduard Belitser and Nurzhan Nurushev - Matteo Iacopini, and Stefano Tonellato - Merlise Clyde
Species distribution models (SDMs) are a common approach to describing species’ space‐use and spatially‐explicit abundance. With a myriad of model types, methods and parameterization options available, it is challenging to make informed decisions about how to build robust SDMs appropriate for a given purpose. One key component of SDM development is the appropriate parameterization of covariates, such as the inclusion of covariates that reflect underlying processes (e.g. abiotic and biotic covariates) and covariates that act as proxies for unobserved processes (e.g. space and time covariates). It is unclear how different SDMs apportion variance among a suite of covariates, and how parameterization decisions influence model accuracy and performance. To examine trade‐offs in covariation parameterization in SDMs, we explore the attribution of spatiotemporal and environmental variation across a suite of SDMs. We first used simulated species distributions with known environmental preferences to compare three types of SDM: a machine learning model (boosted regression tree), a semi‐parametric model (generalized additive model) and a spatiotemporal mixed‐effects model (vector autoregressive spatiotemporal model, VAST). We then applied the same comparative framework to a case study with three fish species (arrowtooth flounder, pacific cod and walleye pollock) in the eastern Bering Sea, USA. Model type and covariate parameterization both had significant effects on model accuracy and performance. We found that including either spatiotemporal or environmental covariates typically reproduced patterns of species distribution and abundance across the three models tested, but model accuracy and performance was maximized when including both spatiotemporal and environmental covariates in the same model framework. Our results reveal trade‐offs in the current generation of SDM tools between accurately estimating species abundance, accurately estimating spatial patterns, and accurately quantifying underlying species–environment relationships. These comparisons between model types and parameterization options can help SDM users better understand sources of model bias and estimate error.
Climate change is expected to have a profound impact on the distribution, abundance and diversity of marine species globally1,2. These ecological impacts of climate change will affect human communities dependent on fisheries for livelihoods and well-being³. While methods for assessing the vulnerability of species to climate change are rapidly developing⁴ and socio-ecological vulnerability assessments for fisheries are becoming available⁵, there has been less work devoted to understanding how impacts differ across fishing communities. We developed a linked socio-ecological approach to assess the exposure of fishing communities to risk from climate change, and present a case study of New England and Mid-Atlantic (USA) fishing communities. We found that the northern part of the study region was projected to gain suitable habitat and the southern part projected to lose suitable habitat for many species, but the exposure of fishing communities to risk was strongly dependent on both their spatial use of the ocean and their portfolio of species caught. A majority of fishing communities were projected to face declining future fishing opportunities unless they adapt, either through catching new species or fishing in new locations. By integrating climatic, ecological and socio-economic data at a scale relevant to fishing communities, this analysis identifies where strategies for adapting to the ecological impacts of climate change will be most needed.
Understanding which species and ecosystems will be most severely affected by warming as climate change advances is important for guiding conservation and management. Both marine and terrestrial fauna have been affected by warming1,2 but an explicit comparison of physiological sensitivity between the marine and terrestrial realms has been lacking. Assessing how close populations live to their upper thermal limits has been challenging, in part because extreme temperatures frequently drive demographic responses3,4 and yet fauna can use local thermal refugia to avoid extremes5–7. Here we show that marine ectotherms experience hourly body temperatures that are closer to their upper thermal limits than do terrestrial ectotherms across all latitudes—but that this is the case only if terrestrial species can access thermal refugia. Although not a direct prediction of population decline, this thermal safety margin provides an index of the physiological stress caused by warming. On land, the smallest thermal safety margins were found for species at mid-latitudes where the hottest hourly body temperatures occurred; by contrast, the marine species with the smallest thermal safety margins were found near the equator. We also found that local extirpations related to warming have been twice as common in the ocean as on land, which is consistent with the smaller thermal safety margins at sea. Our results suggest that different processes will exacerbate thermal vulnerability across these two realms. Higher sensitivities to warming and faster rates of colonization in the marine realm suggest that extirpations will be more frequent and species turnover faster in the ocean. By contrast, terrestrial species appear to be more vulnerable to loss of access to thermal refugia, which would make habitat fragmentation and changes in land use critical drivers of species loss on land.
Aim
Changes in abundance and shifts in distribution as a result of a warming climate have been documented for many marine species, but opportunities to test our ability to forecast such changes have been limited. This study evaluates the ability of habitat‐based density models to accurately forecast cetacean abundance and distribution during a novel year with unprecedented warm ocean temperatures caused by a sustained marine heatwave.
Location
California Current Ecosystem, USA.
Methods
We constructed generalized additive models based on cetacean sighting and environmental data from 1991 to 2009 for eight species with a diverse range of habitat associations. Models were built with three different sets of predictor variables to compare performance. Models were then used to forecast species abundance and distribution patterns during 2014, a year with anomalously warm ocean temperatures. Cetacean sighting data collected during 2014 were used to assess model forecasts.
Results
Ratios of model‐predicted abundance to observed abundance were close to 1:1 for all but one species and accurately captured changes in the number of animals in the study area during the anomalous year. Predicted distribution patterns also showed good concordance with the 2014 survey observations. Our results indicate that habitat relationships were captured sufficiently to predict both changes in abundance and shifts in distribution when conditions warmed, for both cool‐ and warm‐temperate species.
Main conclusions
Models built with multidecadal datasets were able to forecast abundance and distribution in a novel warm year for a diverse set of cetacean species. Models with the best explanatory power did not necessarily have the best predictive power. Also, they revealed species‐specific responses to warming ocean waters. Results have implications for modelling effects of climate change on cetaceans and other marine predators.
Fisheries scientists provide stock, ecosystem, habitat, and climate assessments to support interdisplinary fisheries management in the US and worldwide. These assessment activities have evolved different models, using different review standards, and are communicated using different vocabulary. Recent research shows that spatio-temporal models can estimate population density for multiple locations, times, and species, and that this is a “common currency” for addressing core goals in stock, ecosystem, habitat, and climate assessments. I therefore review the history and “design principles” for one spatio-temporal modelling package, the Vector Autoregressive Spatio-Temporal (VAST) package. I then provide guidance on fifteen major decisions that must be made by users of VAST, including: whether to use a univariate or multivariate model; when to include spatial and/or spatio-temporal variation; how many factors to use within a multivariate model; whether to include density or catchability covariates; and when to include a temporal correlation on model components. I finally demonstrate these decisions using three case studies. The first develops indices of abundance, distribution shift, and range expansion for arrowtooth flounder (Atheresthes stomias) in the Eastern Bering Sea, showing the range expansion for this species. The second involves “species ordination” of eight groundfishes in the Gulf of Alaska bottom trawl survey, which highlights the different spatial distribution of flathead sole (Hippoglossoides elassodon) relative to sablefish (Anoplopoma fimbria) and dover sole (Microstomus pacificus). The third involves a short-term forecast of the proportion of coastwide abundance for five groundfishes within three spatial strata in the US West Coast groundfish bottom trawl survey, and predicts large interannual variability (and high uncertainty) in the distribution of lingcod (Ophiodon elongatus). I conclude by recommending further research exploring the benefits and limitations of a “common currency” approach to stock, ecosystem, habitat, and climate assessments, and discuss extending this approach to optimal survey design and economic assessments.
This paper provides a theoretical understanding of sampling bias in presence-only data in the context of species distribution modelling. This understanding forms the basis for two integrated frameworks, one for detecting sampling bias of different kinds in presence-only data ( the bias assessment framework ) and one for assessing potential effects of sampling bias on species distribution models ( the bias effects framework ). We exemplify the use of these frameworks to museum data for nine insect species in Norway, for which the distribution along the two main bioclimatic gradients (related to oceanicity and temperatures) are modelled using the MaxEnt method. Models of different complexity (achieved by use of two different model selection procedures that represent spatial prediction or ecological response modelling purposes, respectively) were generated with different types of background data (uninformed and background-target-group [BTG]). The bias assessment framework made use of comparisons between observed and theoretical frequency-of-presence (FoP) curves, obtained separately for each combination of species and bioclimatic predictor, to identify potential sampling bias. The bias effects framework made use of comparisons between modelled response curves (predicted relative FoP curves) and the corresponding observed FoP curves for each combination of species and predictor. The extent to which the observed FoP curves deviated from the expected, smooth and unimodal theoretical FoP curve, varied considerably among the nine insect species. Among-curve differences were, in most cases, interpreted as indications of sampling bias. Using BTG-type background data in many cases introduced strong sampling bias. The predicted relative FoP curves from MaxEnt were, in general, similar to the corresponding observed FoP curves. This indicates that the main structure of the data-sets were adequately summarised by the MaxEnt models (with the options and settings used), in turn suggesting that shortcomings of input data such as sampling bias or omission of important predictors may overshadow the effect of modelling method on the predictive performance of distribution models. The examples indicate that the two proposed frameworks are useful for identification of sampling bias in presence-only data and for choosing settings for distribution modelling options such as the method for extraction of background data points and determining the appropriate level of model complexity.
Species Distribution Models (SDMs) are now being widely used in ecology for management and conservation purposes across terrestrial, freshwater, and marine realms. The increasing interest in SDMs has drawn the attention of ecologists to spatial models and, in particular, to geostatistical models, which are used to associate observations of species occurrence or abundance with environmental covariates in a finite number of locations in order to predict where (and how much of) a species is likely to be present in unsampled locations. Standard geostatistical methodology assumes that the choice of sampling locations is independent of the values of the variable of interest. However, in natural environments, due to practical limitations related to time and financial constraints, this theoretical assumption is often violated. In fact, data commonly derive from opportunistic sampling (e.g. whale or bird watching, etc.), in which observers tend to look for a specific species in areas where they expect to find it. These are examples of a what is referred to as preferential sampling, which can lead to biased predictions of the distribution of the species. The aim of this study is to discuss a SDM that addresses this problem and that it is more computationally efficient than existing MCMC methods. From a statistical point of view, we interpret the data as a marked point pattern, where the sampling locations form a point pattern and the measurements taken in those locations (i.e. species abundance or occurrence) are the associated marks. Inference and prediction of species distribution is performed using a Bayesian approach and integrated nested Laplace approximation (INLA) methodology and software are used for model fitting to minimize the computational burden. We show that abundance is highly overestimated at low abundance locations when preferential sampling effects not accounted for, in both a simulated example and a practical application using fishery data. This highlights that ecologists should be aware of the potential bias resulting from preferential sampling and account for it in a model when a survey is based on non-randomized and/or non-systematic sampling.
Predictive models are central to many scientific disciplines and vital for informing management in a rapidly changing world. However, limited understanding of the accuracy and precision of models transferred to novel conditions (their ‘transferability’) undermines confidence in their predictions. Here, 50 experts identified priority knowledge gaps which, if filled, will most improve model transfers. These are summarized into six technical and six fundamental challenges, which underlie the combined need to intensify research on the determinants of ecological predictability, including species traits and data quality, and develop best practices for transferring models. Of high importance is the identification of a widely applicable set of transferability metrics, with appropriate tools to quantify the sources and impacts of prediction uncertainty under novel conditions.
Aim
Accurate predictions of cetacean distributions are essential to their conservation but are limited by statistical challenges and a paucity of data. This study aimed at comparing the capacity of various statistical algorithms to deal with biases commonly found in nonsystematic cetacean surveys and to evaluate the potential for citizen science data to improve habitat modelling and predictions. An endangered population of humpback whales (Megaptera novaeangliae) in their breeding ground was used as a case study.
Location
New Caledonia, Oceania.
Methods
Five statistical algorithms were used to model the habitat preferences of humpback whales from 1,360 sightings collected over 14 years of nonsystematic research surveys. Three different background sampling approaches were tested when developing models from 625 crowdsourced sightings to assess methods accounting for citizen science spatial sampling bias. Model evaluation was conducted through cross‐validation and prediction to an independent satellite tracking dataset.
Results
Algorithms differed in complexity of the environmental relationships modelled, ecological interpretability and transferability. While parameter tuning had a great effect on model performances, GLMs generally had low predictive performance, SVMs were particularly hard to interpret, and BRTs had high descriptive power but showed signs of overfitting. MAXENT and especially GAMs provided a valuable complexity trade‐off, accurate predictions and were ecologically intelligible. Models showed that humpback whales favoured cool (22–23°C) and shallow waters (0–100 m deep) in coastal as well as offshore areas. Citizen science models converged with research survey models, specifically when accounting for spatial sampling bias.
Main conclusions
Marine megafauna distribution models present specific challenges that may be addressed through integrative evaluation, independent testing and appropriately tuned statistical algorithms. Specifically, controlling overfitting is a priority when predicting cetacean distributions for large‐scale conservation perspectives. Citizen science data appear to be a powerful tool to describe cetacean habitat.
Seafood is an essential source of protein for more than 3 billion people worldwide, yet bycatch of threatened species in capture fisheries remains a major impediment to fisheries sustainability. Management measures designed to reduce bycatch often result in significant economic losses and even fisheries closures. Static spatial management approaches can also be rendered ineffective by environmental variability and climate change, as productive habitats shift and introduce new interactions between human activities and protected species. We introduce a new multispecies and dynamic approach that uses daily satellite data to track ocean features and aligns scales of management, species movement, and fisheries. To accomplish this, we create species distribution models for one target species and three bycatch-sensitive species using both satellite telemetry and fisheries observer data. We then integrate species-specific probabilities of occurrence into a single predictive surface, weighing the contribution of each species by management concern. We find that dynamic closures could be 2 to 10 times smaller than existing static closures while still providing adequate protection of endangered nontarget species. Our results highlight the opportunity to implement near real-time management strategies that would both support economically viable fisheries and meet mandated conservation objectives in the face of changing ocean conditions. With recent advances in eco-informatics, dynamic management provides a new climate-ready approach to support sustainable fisheries.
Recent shifts in the geographic distribution of marine species have been linked to shifts in preferred thermal habitats. These shifts in distribution have already posed challenges for living marine resource management, and there is a strong need for projections of how species might be impacted by future changes in ocean temperatures during the 21st century. We modeled thermal habitat for 686 marine species in the Atlantic and Pacific oceans using long-term ecological survey data from the North American continental shelves. These habitat models were coupled to output from sixteen general circulation models that were run under high (RCP 8.5) and low (RCP 2.6) future greenhouse gas emission scenarios over the 21st century to produce 32 possible future outcomes for each species. The models generally agreed on the magnitude and direction of future shifts for some species (448 or 429 under RCP 8.5 and RCP 2.6, respectively), but strongly disagreed for other species (116 or 120 respectively). This allowed us to identify species with more or less robust predictions. Future shifts in species distributions were generally poleward and followed the coastline, but also varied among regions and species. Species from the U.S. and Canadian west coast including the Gulf of Alaska had the highest projected magnitude shifts in distribution, and many species shifted more than 1000 km under the high greenhouse gas emissions scenario. Following a strong mitigation scenario consistent with the Paris Agreement would likely produce substantially smaller shifts and less disruption to marine management efforts. Our projections offer an important tool for identifying species, fisheries, and management efforts that are particularly vulnerable to climate change impacts.
The California drift gillnet fleet has operated off the U.S. west coast since it developed in the late 1970's. At its full extent, fishing effort ranged from the Southern California Bight north to Oregon. The fishery initially targeted pelagic sharks (family Alopiidae), but shifted targets in the mid-1980's to swordfish, Xiphias gladius, due to the species' higher value and to the regulatory changes which created an economic advantage for targeting swordfish compared to thresher shark, Alopias vulpinas. Consequently the number of participants in the once-small fleet substantially increased. Conservation concerns over the entanglement of nontarget species, including sharks, sea turtles (family Dermochelyidae and Cheloniidae), and marine mammals, led to the enactment of a series of time and area closures over a period of 19 years to reduce the bycatch of those species. As a result, the fishery declined dramatically. This paper analyzes catch and effort data for the fishery to determine the effects of these management measures on targeted pelagic sharks and swordfish catch per unit of effort (CPUE). Unlike traditional catch-rate analysis, which seeks to standardize fishing effort heterogeneity to produce a measure of relative stock abundance, our analysis uses a linear regression of set-level catch counts on time and area combinations where fishing occurred over the full period for which data is available to estimate the effects of time and area closures on fishery catch rates while controlling for population abundance trends. Resampling methods are used to quantify uncertainty in the estimates. The analysis documents reduced common thresher and shortfin mako shark, Isurus oxyrinchus, CPUE resulting from effort shifting out of higher to lower catch rate areas. Before the 2001 establishment of the Pacific Leatherback Conservation Area (PLCA) closure which eliminated most effort in the fishery north of Pt. Conception, the fishery operated from the Mexican territorial water border northward to Oregon. The cumulative effect of all enacted time and area closures was to concentrate fishing effort within the Southern California Bight, likely contributing to a decline in vessel participation. The socioeconomic impacts on the operation of the fishery and on the usefulness of drift gillnet CPUE data for stock assessments are discussed.
Consequences of shifting species distributions
Climate change is causing geographical redistribution of plant and animal species globally. These distributional shifts are leading to new ecosystems and ecological communities, changes that will affect human society. Pecl et al. review these current and future impacts and assess their implications for sustainable development goals.
Science , this issue p. eaai9214
Climate change is driving changes in the physical and chemical properties of the ocean that have consequences for marine ecosystems. Here, we review evidence for the responses of marine life to recent climate change across ocean regions, from tropical seas to polar oceans. We consider observed changes in calcification rates, demography, abundance, distribution, and phenology of marine species. We draw on a database of observed climate change impacts on marine species, supplemented with evidence in the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. We discuss factors that limit or facilitate species' responses, such as fishing pressure, the availability of prey, habitat, light and other resources, and dispersal by ocean currents. We find that general trends in species' responses are consistent with expectations from climate change, including shifts in distribution to higher latitudes and to deeper locations, advances in spring phenology, declines in calcification, and increases in the abundance of warm-water species. The volume and type of evidence associated with species responses to climate change is variable across ocean regions and taxonomic groups, with predominance of evidence derived from the heavily-studied north Atlantic Ocean. Most investigations of the impact of climate change being associated with the impacts of changing temperature, with few observations of effects of changing oxygen, wave climate, precipitation (coastal waters), or ocean acidification. Observations of species responses that have been linked to anthropogenic climate change are widespread, but are still lacking for some taxonomic groups (e.g., phytoplankton, benthic invertebrates, marine mammals).
Species mapping is an essential tool for conservation programmes as it provides clear pictures
of the distribution of marine resources. However, in fishery ecology, the amount of objective
scientific information is limited and data may not always be directly comparable. Information
about the distribution of marine species can be derived from two main sources: fishery-
independent data (scientific surveys at sea) and fishery-dependent data (collection and
sampling by observers in commercial vessels). The aim of this paper is to compare whether
these two different sources produce similar, complementary or different results. We compare
them in the specific context of identifying the Essential Fish Habitats of three elasmobranch
species (
S. canicula, G. melastomus
and
E. spinax
). Similarity and prediction statistics are
used to compare the two different spatial patterns obtained by applying the same Bayesian
spatio-temporal modelling approach in the two sources. Results showed that the spatial
patterns obtained are similar, although differences are present. In particular, models based on
fishery-dependent data are better able to identify temporal relationships between the
probability of presence of the species and seasonal environmental variables. In contrast,
fishery-independent data better discriminate spatial locations where a species is present or
absent. Besides the spatial and temporal differences of the two datasets, the consistency of
habitat results highlights the inclusion in each dataset of most of the environmental envelope
of each species, both in time and space. Consequently, sampling data should be adapted to
each species in order to reasonably cover their environmental envelope, and a combination of
datasets will likely provide a better habitat estimation than using each dataset independently.
These findings can be useful in helping fishery managers improve definition of survey design
and analyses.
virtualspecies is a freely available package for R designed to generate virtual species distributions, a procedure increasingly used in ecology to improve species distribution models. This package combines the existing methodological approaches with the objective of generating virtual species distributions with increased ecological realism. The package includes (1) generating the probability of occurrence of a virtual species from a spatial set of environmental conditions (i.e., environmental suitability), with two different approaches; (2) converting the environmental suitability into presence-absence with a probabilistic approach; (3) introducing dispersal limitations in the realised virtual species distributions and (4) sampling occurrences with different biases in the sampling procedure. The package was designed to be extremely flexible, to allow users to simulate their own defined species-environment relationships, as well as to provide a fine control over every simulation parameter. The package also includes a function to generate random virtual species distributions. We provide a simple example in this paper showing how increasing ecological realism of the virtual species impacts the predictive performance of species distribution models. We expect that this new package will be valuable to researchers willing to test techniques and protocols of species distribution models as well as various biogeographical hypotheses.This article is protected by copyright. All rights reserved.
Aim
Spatial sampling biases in biodiversity data arise because of complex interactions between geography, species characteristics and human behaviour, including preferences for or against particular species or habitats; biases are therefore not necessarily independent of the environmental niches of species. We evaluate when correlations between spatial sampling biases and environmental niches are likely to affect species distribution models (SDMs) developed both with and without attempts to correct these biases.
Innovation
A virtual species and virtual ecologist framework was used to simulate biodiversity data with either no spatial sampling bias or biases that were correlated (positively or negatively) with one of the environmental variables used to define the environmental niches of the species. The environmental variables used to define the species niche were simulated with spatial autocorrelation operating at multiple spatial scales. Virtual samples were then used to model species distributions, with models evaluated based on their ability to rank the suitability of sites correctly.
Main conclusions
Correlations between spatial sampling bias and environmental niches frequently reduced the rank correlation of model predictions, but the relative importance of these effects varied with species type (greater decline in rank correlation as the environmental niche broadens) and data type (models built using detection/non‐detection data were less affected than those using detection‐only data). Bias‐correction effectiveness varied depending on the structure of the spatial bias but was also highly variable across methods and dependent on data type. The implications of these results are that spatial sampling bias is a greater concern for SDMs where: (1) the distribution of effort is non‐random with respect to an environmental gradient thought to be correlated with a species’ distribution; (2) the species being modelled has a broad environmental niche; and (3) the data for modelling contain only information on detections (i.e., presence only).
Climate change is altering the biogeochemical conditions of the ocean, leading to the emergence of novel environmental conditions that may drastically affect the performance of very large marine protected areas (VLMPAs) (area > 100,000 km²). Given the prominent role that VLMPAs play in ocean conservation, determining when and where novel conditions will emerge within VLMPAs is vital for ensuring a healthy ocean in the future. Here, using a non-parametric approach to detect novelty, we show that 60%–87% of the ocean and 76%–97% of VLMPAs are expected to contain novel conditions across multiple biogeochemical variables by 2100, with novel conditions in pH emerging by 2030. With most VLMPAs expected to contain environmental conditions unlike those currently within their boundaries, and given the likelihood of any of these climate futures unfolding, present-day management will need to consider alterations to current and future VLMPA design and use.
Fishing communities are increasingly required to adapt to environmentally driven changes in the availability of fish stocks. Here, we examined trends in the distribution and biomass of five commercial target species (dover sole, thornyheads, sablefish, lingcod, and petrale sole) on the US west coast to determine how their availability to fishing ports changed over 40 years. We show that the timing and magnitude of stock declines and recoveries are not experienced uniformly along the coast when they coincide with shifts in species distributions. For example, overall stock availability of sablefish was more stable in southern latitudes where a 40% regional decline in biomass was counterbalanced by a southward shift in distribution of >200 km since 2003. Greater vessel mobility and larger areal extent of fish habitat along the continental shelf buffered northerly ports from latitudinal changes in stock availability. Landings were not consistently related to stock availability, suggesting that social, economic, and regulatory factors likely constrain or facilitate the capacity for fishers to adapt to changes in fish availability. Coupled social–ecological analyses such as the one presented here are important for defining community vulnerability to current and future changes in the availability of important marine species.
Population surveys are often used to estimate the density, abundance, or distribution of natural populations. Recently, model‐based approaches to analyzing survey data have become popular because one can more readily accommodate departures from pre‐planned survey routes and construct more detailed maps than one can with design‐based procedures.
Spatial models for population distributions (SMPDs) often make the implicit assumption that locations chosen for sampling and animal abundance at those locations are conditionally independent given modelled covariates. However, this assumption may be violated when survey effort is non‐randomized, leading to preferential sampling.
We develop a hierarchical statistical modelling framework for detecting and alleviating the biasing effects of preferential sampling in spatial distribution models fitted to count data. The approach works by specifying a joint model for population density and the locations selected for sampling, and specifying a dependent correlation structure between the two processes.
Using simulation, we show that moderate levels of preferential sampling can lead to large (e.g. 40%) bias in estimates of animal density and that our modelling approach can considerably reduce this bias. In contrast, preferential sampling did not appear to bias inferences about parameters informing species–habitat relationships (i.e. slope parameters).
We apply our approach to aerial survey counts of bearded seals ( Erignathus barbatus ) in the eastern Bering Sea. As expected, models with a preferential sampling effect led to lower abundance than those without. However, several lines of reasoning (better predictive performance, higher biological realism) led us to prefer models without a preferential sampling effect for this dataset.
When population surveys break from traditional scientific survey design principles, ecologists should recognize the potentially biasing effects of preferential sampling when estimating population density or occurrence. Joint models, such as those described in this paper, can be used to test and correct for such biases. However, such models can be unstable; ultimately the best way to avoid preferential sampling bias is to incorporate design‐based principles such as randomization and/or systematic sampling into survey design.
Spatio-temporal models are increasingly used to develop indices of population abundance from fishery-dependent and –independent data. Developments in spatio-temporal index standardization were discussed at a workshop hosted by The Center for the Advancement of Population Assessment Methodology (CAPAM) titled “Development and application of spatio-temporal models to derive indices of relative abundance” (La Jolla, CA, USA; Feb. 26-March 2, 2018). This special issue includes ten submissions that highlight potential benefits arising from spatio-temporal index standardization including: (1) improved prediction for areas with little-to-no data; (2) capacity to weight densities by area to account for sampling rates that vary spatially; (3) calculation of the composition (e.g. age or length) of the indices and to weight predictions by catch to inform the composition of fishery removals, (4) estimation of distribution shifts and range expansion/contraction; and (5) capacity to combine data sources. The articles in this Special Issue highlight the need for continued development and testing for spatio-temporal index standardization methods, so that the relative strengths and weaknesses of different approaches are fully understood. The articles also identify topics that warrant further research including improved model diagnostics, accounting for biased rates of sampling (“preferential sampling”), and improvements in computational efficiency.
We describe and illustrate a spatio-temporal modelling approach for analyzing age-or size-specific catch-per-unit-effort (CPUE) data to develop indices of relative abundance and associated composition data. The approach is based on three concepts: 1) composition data that are used to determine the component of the population represented by the index should be weighted by CPUE (abundance) while the composition data used to represent the fish removed from the stock should be weighted by catch; 2) due to spatial non-randomness in fishing effort and fish distribution, the index, index composition, and catch composition, should be calculated at a fine spatial scale (e.g., 1°x1°) and summed using area weighting; and 3) fine-scale spatial stratification will likely result in under-sampled and unsampled cells and some form of smoothing method needs to be applied to inform these cells. We illustrate the concepts by applying them to yellowfin tuna (Thunnus albacares) in the eastern Pacific Ocean.
This paper investigates environmental drivers of US west coast petrale sole (Eopsetta jordani) recruitment as an initial step towards developing an environmental recruitment index that can inform the stock assessment in the absence of survey observations of age‐0 and age 1‐fish. First, a conceptual life‐history approach is used to generate life stage‐ and spatio‐temporally‐specific mechanistic hypotheses regarding oceanographic variables that likely influence survival at each life stage. Seven life history stages are considered, from female‐spawner condition through benthic recruitment as observed in the Northwest Fisheries Science Center West Coast Groundfish Bottom Trawl Survey (age‐2 fish). The study area encompasses the region from 40‐48 °N in the California Current Ecosystem. Hypotheses are tested using output from a regional ocean reanalysis model outputs and model selection techniques. Four oceanographic variables explained 73% of the variation in recruitment not accounted for by estimates based exclusively on the spawning stock size. Recruitment deviations were (1) positively correlated with degree days during the female precondition period, (2) positively correlated with mixed‐layer depth during the egg stage, (3) negatively correlated with cross‐shelf transport during the larval stage, and (4) negatively correlated with cross‐shelf transport during the benthic juvenile stage. While multiple mechanisms likely affect petrale sole recruitment at different points during their life history, the strength of the relationship is promising for stock assessment and integrated ecosystem assessment applications.
A physical-biogeochemical model is used to produce a retrospective analysis at 3-km resolution of alongshore phytoplankton variability in the California Current during 1988–2010. The simulation benefits from downscaling a regional circulation reanalysis, which provides improved physical ocean state estimates in the high-resolution domain. The emerging pattern is one of local upwelling intensification in response to increased alongshore wind stress in the lee of capes, modulated by alongshore meanders in the geostrophic circulation. While stronger upwelling occurs near most major topographic features, substantial increases in phytoplankton biomass only ensue where local circulation patterns are conducive to on-shelf retention of upwelled nutrients. Locations of peak nutrient delivery and chlorophyll accumulation also exhibit interannual variability and trends noticeably larger than the surrounding shelf regions, thereby suggesting that long-term planktonic ecosystem response in the California Current exhibits a significant local scale (O(100 km)) alongshore component.
Relative abundance trends of highly migratory species (HMS) have played a central role in debates over the health of global fisheries. However, such trends have mostly been inferred from fishery catch rates, which can provide misleading signals of relative abundance. While many biases are accounted for through traditional catch rate standardization, pelagic habitat fished is rarely directly considered. Using a method that explicitly accounts for temperature regimes, we analysed data from the US pelagic longline fishery to estimate relative abundance trends for 34 HMS in the Atlantic Ocean from 1987 through 2013. This represents one of the largest studies of HMS abundance trends. Model selection emphasized the importance of accounting for pelagic habitat fished with water column temperature being included in nearly every species' model, and in extreme cases, a temperature variable explained 50-60% of the total deviance. Our estimated trends represent observations from one fishery only, and a more integrated stock assessment should form the basis for conclusions about stock status overall. Nonetheless, our trends serve as indicators of stock abundance and they suggest that a majority of HMS (71% of analysed species) are either declining in relative abundance or declined initially with no evidence of rebuilding. Conversely, 29% of the species exhibited stable, increasing, or recovering trends; however, these trends were more prevalent among tunas than either billfishes or sharks. By estimating the effects of pelagic habitat on fishery catch rates, our results can be used in combination with ocean temperature trends and forecasts to support bycatch avoidance and other time-area management decisions.
After decades of extensive surveying, knowledge of the global distribution of species still remains inadequate for many purposes. In the short to medium term, such knowledge is unlikely to improve greatly given the often prohibitive costs of surveying and the typically limited resources available.
By forecasting biodiversity patterns in time and space, predictive models can help fill critical knowledge gaps and prioritise research to support better conservation and management.
The ability of a model to predict biodiversity metrics in novel environments is termed “transferability,” and models with high transferability will be the most useful in this context.
Despite their potentially broad utility, little guidance exists on what confers high transferability to biodiversity models.
We synthesise recent advances in biodiversity model transfers to facilitate increased understanding of what underpins successful model transferability, demonstrating that a consistent approach has so far been lacking but is essential for achieving high levels of repeatability, transparency and accountability of model transfers.
We provide a set of guidelines to support efficient learning and the improvement of model transferability.
The widely recommended procedure of Bayesian model averaging is flawed in the M-open setting in which the true data-generating process is not one of the candidate models being fit. We take the idea of stacking from the point estimation literature and generalize to the combination of predictive distributions, extending the utility function to any proper scoring rule, using Pareto smoothed importance sampling to efficiently compute the required leave-one-out posterior distributions and regularization to get more stability. We compare stacking of predictive distributions to several alternatives: stacking of means, Bayesian model averaging (BMA), pseudo-BMA using AIC-type weighting, and a variant of pseudo-BMA that is stabilized using the Bayesian bootstrap. Based on simulations and real-data applications, we recommend stacking of predictive distributions, with BB-pseudo-BMA as an approximate alternative when computation cost is an issue.
The first edition of this book has established itself as one of the leading references on generalized additive models (GAMs), and the only book on the topic to be introductory in nature with a wealth of practical examples and software implementation. It is self-contained, providing the necessary background in linear models, linear mixed models, and generalized linear models (GLMs), before presenting a balanced treatment of the theory and applications of GAMs and related models. The author bases his approach on a framework of penalized regression splines, and while firmly focused on the practical aspects of GAMs, discussions include fairly full explanations of the theory underlying the methods. Use of R software helps explain the theory and illustrates the practical application of the methodology. Each chapter contains an extensive set of exercises, with solutions in an appendix or in the book’s R data package gamair, to enable use as a course text or for self-study.
A major limitation to fully integrated ecosystem based fishery management approaches is a lack of information on the spatial distribution of marine species and the environmental conditions shaping these distributions. This is particularly problematic for deep-water species that are hard to sample and are data poor. The past decade has seen the rapid development of a suite of advanced species distribution, or ecological niche, modelling approaches developed specifically to support efficient and targeted management. However, model performance can vary significantly and the appropriateness of which methods are best for a given application remains questionable. Species distribution models were developed for three commercially valuable Hawaiian deep-water eteline snappers: Etelis coruscans (Onaga), Etelis carbunculus (Ehu) and Pristipomoides filamentosus (Opakapaka). Distributional data for these species was relatively sparse. To identify the best method, model performance and distributional accuracy was assessed and compared using three approaches: Generalised Additive Models (GAM), Boosted Regression Trees (BRT) and Maximum Entropy (MaxEnt). Independent spatial validation data found MaxEnt consistently provided better model performance with ‘good’ model predictions (AUC =>0.8). Each species was influenced by a unique combination of environmental conditions, with depth, terrain (slope) and substrate (low lying unconsolidated sediments), being the three most important in shaping their distributions. Sustainable fisheries management, marine spatial planning and environmental decision support systems rely on an understanding species distribution patterns and habitat linkages. This study demonstrates that predictive species distribution modelling approaches can be used to accurately model and map sparse species distribution data across marine landscapes. The approach used herein was found to be an accurate tool to delineate species distributions and associated habitat linkages, account for species-specific differences and support sustainable ecosystem-based management.
