Shifts in fisheries management: adapting to regime shifts

Abstract

For many years, fisheries management was based on optimizing yield and maintaining a target biomass, with little regard given to low-frequency environmental forcing. However, this policy was often unsuccessful. In the last two to three decades, fisheries science and management have undergone a shift towards balancing sustainable yield with conservation, with the goal of including ecosystem considerations in decision-making frameworks. Scientific understanding of low-frequency climate–ocean variability, which is manifested as ecosystem regime shifts and states, has led to attempts to incorporate these shifts and states into fisheries assessment and management. To date, operationalizing these attempts to provide tactical advice has met with limited success. We review efforts to incorporate regime shifts and states into the assessment and management of fisheries resources, propose directions for future investigation and outline a potential framework to include regime shifts and changes in ecosystem states into fisheries management.

1. Introduction

In the last 25–30 years, fisheries science and management have been evolving towards ecosystem-based stock assessments and management. Much of this evolution is attributable to the overexploitation of some global fish stocks due to the early focus of fisheries science on optimizing yield, the use of deterministic models and non-precautionary decision-making frameworks so that scientific advice was not adhered to. The short-comings of those early approaches fostered a demand from scientists, stakeholders and the general public to manage fisheries resources with a balance between yield and conservation. The foundations for this focus were supported by several global policies developed in the 1990s. The UN Conference on Environment and Development (1992) Agenda 21 proposed that the precautionary approach be adopted for the protection of marine resources. It also led to an international agreement, the UN Convention on Biological Diversity, which focused on the sustainable use, conservation and biodiversity of natural resources. In 1995, the FAO Code of Conduct for Responsible Fisheries outlined the need for research on the impacts of climate and environmental factors on fishery resources. It explicitly outlined the requirement to account for uncertainty in providing management advice by applying the precautionary approach. These international policies reinforced the migration in fisheries management from maximizing fisheries production to accounting for conservation and risk management [1].

Domestically, many nations have developed guidelines and procedures which reflect the core concept of the precautionary approach or explicitly refer to the approach within a framework of sustainability. These range from polices or frameworks that are not legally mandated to actual acts of legislation. For example, the Canadian Sustainable Fisheries Framework (2009) policy explicitly outlines the application of the precautionary approach to fisheries management and provides guidance on provisional biological reference points and corresponding management actions given a species' stock status [2]. In the USA, the Sustainable Fisheries Act (1996) amendment to Magnusson–Stevens Fishery Conservation and Management Act (1976) frames the legal responsibility for sustainable fishery resource use coupled with conservation of marine resources [3]. The Australian Fisheries Administration Act (1991) requires fisheries resource exploitation to be compliant with the precautionary approach, and to have regard for the impacts on non-target species and overall ecosystem sustainability [4]. The Fisheries Act (1996) of New Zealand requires that fisheries management decisions must be based on the best available information, must consider any uncertainty in provided advice, and should be cautious when advice is uncertain, unreliable or inadequate [5].

Additionally, public support for these types of approaches is evident in the increased popularity and consumer demand for seafood sustainability certifications, such as Marine Stewardship Certification or Friend of the Sea. These seafood sustainability certifications attempt to provide the consumer with a guarantee that the fishery resource is sustainable and is being managed with the goal of overall ecosystem conservation, e.g. low impact on bycatch, threatened, endangered or protected species and on habitats.

The current approach in fisheries science and management couples risk management of exploitation with conservation goals. Science advice provides the possible consequences and the associated risk of various harvest strategies given the best estimates of current and projected stock status and uncertainty. Those consequences can include the impacts of fishing on marine food webs (e.g. non-target species' mortality) and habitat. Conversely, balancing fishing opportunities and ecosystem sustainability requires consideration of ecosystem impacts on fish productivity. The drivers of those ecosystem impacts include climate–ocean variability and ecosystem reorganization, both of which are contained in low-frequency (decadal-scale) ecosystem regime shifts and states.

In this paper, we review efforts to incorporate regime shifts and states into the assessment and management of fisheries resources. To date, efforts to move beyond identifying environmental forcing impacts on fish productivity to providing tactical fisheries management advice based on enviromental forcing have not been widespread. There are only a few examples where the relationship between environmental forcing and fish productivity have been operationalized and are used in fisheries management, and even one of those has had difficulties in implementation. We outline the common obstacles and pitfalls to full implementation of incorporating regime shifts and states into fisheries assessment and management. We then propose directions for future avenues of investigation and outline a potential framework to include regime shifts and states into fisheries management.

2. Incorporating environmental forcing in fisheries assessment and management

Fisheries stock assessment involves synthesizing biological and fishery information to estimate historical and current abundance and to forecast abundance under alternative harvest policies and the application of harvest control rules. Harvest control rules are sets of predefined management decisions (e.g. harvest rate, total allowable catch (TAC) or time–area closures) that are a function of the status of the stock (as indicated by a stock assessment) in relation to biological reference points. It is common to account for environmental variability in stock assessments by, for example, estimating annual deviations about the stock–recruitment relationship which reflect environmental influences on reproductive success. However, inclusion of low-frequency environmental effects is much less common.

Methods to incorporate regime shifts and states into fisheries stock assessment are encompassed by the Ecosystem Approach to Fisheries, which broadly accounts for ‘ecosystem considerations’ in fisheries management advice [6]. Those methods provide both tactical and strategic advice. Here, tactical advice is the direct provision of management advice through the inclusion of ecosystem-state information either into the assessment model or in the harvest control rule. The inclusion in the assessment can be relatively straightforward, such as the inclusion of environmental variables when fitting stock–recruitment relationships [7], or complicated, such as mechanistic bottom-up modelling [8]. The inclusion of environmental data in harvest control rules is accomplished through the estimation of biological or fishery reference points which change over time as a function of such data [9], or inclusion in determination of allowable harvest rates [10]. Strategic advice incorporates environmental variability into the testing of management strategies (e.g. Management Strategy Evaluation (MSE) [11]), to determine whether the inclusion of regime shifts or states improves the balance between yield and conservation [12], and the consequences of ignoring regime shifts or states when providing management advice [13].

(a). Operationalized examples

These approaches have been applied only infrequently. We are aware of only a few examples that have operationalized inclusion of low-frequency environmental influences into the provision of tactical advice and outline two of these in detail. The first example is the northern subpopulation of Pacific sardine (Sagax sardinops) in the California Current. Management advice, until 2012, implicitly used the relationship between recruits-per-spawner and the environment, as quantified by sea surface temperature (SST) at the pier of the Scripps Institute of Oceanography in La Jolla, CA, USA [7] to establish the fishing mortality corresponding to maximum sustainable yield (F_MSY; [10]). This was achieved through a nonlinear relationship between F_MSY and SST which was bounded by 0.05 and 0.15 for F [14]. However, this use of SST was removed from the harvest control rule in 2012 [14] after analysis of updated data through 2009 suggested that the relationship was no longer significant [15]. Since then, the updated data have been reassessed using the same methodology originally employed and a considerable, significant part of the fish recruitment and recruits per spawner variability is indeed explained by the relationship with temperature, albeit SST averaged over the main spawning area and not the pier temperature [16,17]. This debate highlights the limitations of depending on a single measure of environmental variability to reflect overall changes in ecosystem state, and the need to update and revise relationships when necessary [16]. It is not hard to imagine the difficulty in re-introducing a temperature–recruitment relationship into the harvest control rule for Pacific sardine now that the credibility of the ecological link to stock assessment has been tarnished. However, given the importance of environmental forcing to Pacific sardine abundance and distribution, such a reintroduction is currently underway.

It is widely acknowledged that the 1977 regime shift [18] led to a marked reorganization of the ecosystem of the Gulf of Alaska and the Bering Sea from one dominated by large invertebrates, including crabs, to one dominated by groundfishes, in particular walleye pollock (Theragra chalcogramma) [19,20], which now represents the third largest single-species fishery in the world [21]. Management advice for several crab stocks in the Bering Sea and Aleutians Island region of Alaska involves recommendations for an Overfishing Level (OFL), which is the catch corresponding to F_MSY, and an Acceptable Biological Catch (ABC), which is less than the OFL to account for scientific uncertainty. The OFL is based on multiplying the spawner biomass-per-recruit by average recruitment (R_MSY) when the stock biomass was at maximum sustainable yield (B_MSY). For these stocks, B_MSY is biomass of mature males corresponding to maximum sustainable yield. Prior to 2011, the OFL (and ABC) for the Eastern Bering Sea Tanner crab (Chionoecetes bairdi) stock was based on setting R_MSY to the average recruitment from 1974 to 1980, which are animals spawned before the 1977 regime shift given the time between spawning and recruitment for this stock. Under the US Magnusson–Stevens Fishery Management and Conservation Act, B_MSY should be based on ‘prevailing environmental conditions'. As such, since 2012, the impacts of regime shifts on crab recruitment were included in management advice for Eastern Bering Sea Tanner crab by redefining the period for R_MSY to 1982 to the present to account for the marked reduction in productivity in 1977 based on fits of stock–recruitment relationships [22]. The first year of the period selected to define R_MSY was based on considering all possible years and selecting a year such that the fit of the stock–recruitment relationship is best. The subsequent assessment [23] confirmed that the results of the analyses in Punt [22] are robust to the form of the stock–recruitment relationship and additional data. In contrast to crab stocks in the Gulf of Alaska and Bering Sea, recruitment of most groundfish stocks increased after the 1997 regime shift [19,20]. Consequently, OFLs and ABCs for these stocks are based on recruitment estimates for the post-1977 period [24].

(b). Environmental forcing in single-species assessments

Aside from these two operationalized examples, a number of studies have attempted to include some form of environmental forcing in stock assessments [25–27]. More typical is the identification of a stock–recruitment relationship with environmental forcing [7,28–30]. However, these relationships can be spurious correlations that eventually break down [31], in part, due to fishing-induced impacts on spawning biomass and recruitment coincident with a directional change in the environment [32]. This topic of research remains an active area of debate and investigation, and new analyses may provide approaches to illustrate that environmental forcing is indeed a stronger influence on recruitment than spawning stock biomass [33,34]. When synchronous shifts in recruitment are observed across large marine ecosystems and coincide with regime shifts, there is obvious and strong evidence for regime-shift forcing on recruitment [34]. In fact, since recruitment has a strong influence on subsequent spawning stock biomass, a spurious relationship of spawning stock biomass on recruitment is the result, particularly if recruitment shifts abruptly [34].

Incorrectly identifying environmental forcing impacts on fish productivity can have serious consequences for fisheries management. The value of environmental information in stock assessments may not offer improved outcomes unless there is a reduction in the uncertainty of the new advice and the gains (i.e. yield, sustainable stocks) based on that advice outweigh the risks of being wrong [35]. An example of serious consequences is the Bay of Biscay anchovy (Engraulis encrasicolus) population which is assessed and managed as a single stock [36]. In 1999, the relationship between wind-driven upwelling and recruitment [37] was used to predict a failed recruitment year in 2000, and consequently, the TAC for that year was initially set at approximately half of what it should have been [1]. Halfway through the season, the results from surveys indicated the reduction was unnecessary and the TAC was doubled [1]. Aside from the instability for the fishing industry, the events resulted in intensive debates among the management bodies and scientific advisors about the utility of including environmental forcing impacts in assessment and management [1]. Subsequent re-analysis suggested an alternate model based on upwelling and stratification breakdown events could more successfully predict recruitment [38,39]. However, simulation analyses concluded that successful management would result from implementing precautionary approaches rather than incorporating the environmental forcing in recruitment estimates, principally due to uncertainty in process understanding [40]. The utility of environmental variables in the assessment and management of Bay of Biscay anchovy remains unresolved.

This failure highlights that a lack of understanding of the underlying processes, or misidentifying processes, can be a major obstacle to incorporating environmental regime shifts and states into stock–recruitment relationships. In addition, the confounding effects of fishing on spawning biomass can mask environmental signals in recruitment unless: (i) less than 50% of the total recruitment variability is due to random variability [41] and (ii) recruitment observations span at least a whole regime shift and state [32].

Ecosystem regime shifts and states also impact other biological parameters, such as growth. Unlike recruitment time series, there are probably several size-at-age time series that are long enough to span at least one regime shift and state. In fact, several stock assessments now include time-varying growth [42–44]. Expanding time-varying growth to include environmental forcing could include regime-shift impacts since size-at-age can dramatically change from one regime state to another [45]. However, environmental forcing of growth or other biological parameters will result from indirect impacts (e.g. prey availability) and direct impacts (i.e. physiological impacts such as mortality) which can have confounding consequences. Therefore, it is unlikely that these avenues of investigation will provide better results than those produced with environment–recruitment studies, unless a large portion of the variability is due to direct impacts. Since biological parameters, such as growth, are used directly in estimating biomass, including regime shifts in growth in stock assessments will impact estimates of biomass-based biological references points such as unfished biomass (B₀). For example, B₀ could be calculated using current values for biological parameters such as growth or using projected values for these parameters given regime impacts. Similarly, it is necessary to forecast size-at-age to apply harvest control rules.

Spatial distributions of fish stocks change with regime shifts and states [46–48]. Key assumptions underlying stock assessments may be linked to current distribution, and changes in distribution may violate these assumptions [49]. Stock assessments often rely on fishery-dependent or -independent indices of abundance, but catchability in either can be time-varying due to a number of factors [50], including low-frequency environmental forcing. Environmental forcing can impact catchability through behavioural changes such as feeding habits, densities or spatial distribution [51–53]. Time-varying catchability is seldom included in stock assessments despite proposed methods to model environmental forcing impacts on catchability [54,55]. Nevertheless, bottom temperature as a covariate for survey catchability has been successfully incorporated into the stock assessment of flathead sole (Hippoglossoides elassodon) [56]. Conversely, the inclusion of bottom temperature in time-varying survey catchability for Atlantic cod (Gadus morhua) in the Gulf of St Lawrence did not improve model fit [57]. Avenues for future investigations could focus on accounting for regime-shift impacts on fish distribution in the management decision-making process, for example, in the determination of regime-specific migration rates of trans-boundary stocks.

(c). Bottom-up mechanistic models linking climate to environmental forcing to fish productivity

This integrated approach links trophic interactions to fish recruitment and population dynamics models. Full implementation of this approach, at the one end, requires complex Regional Oceanographic Models (ROMs) that relate climate forcing to physical oceanography and lower trophic level productivity. At the other end, multi-species fisheries assessments would be required for full ecosystem-based implementation. To date, development has focused on using the outputs of Global Climate Models (GCMs) in ROMs and end-to-end biological models to forecast the impacts of climate change [8]. This approach can account for the uncertainty inherent in each data series and modelling step, and project the impact of candidate management actions given climate change. It could be applied to low-frequency climate variability (i.e. regime shifts and states) to project the impact of candidate management actions given regime shifts and long-term climate change. However, the current GCMs exclude low-frequency climate variability, and focus on projections that are at least 30 years (or more) from current. Even so, this approach would not provide tactical advice for short-term (even up to 10-year forecasting) fisheries management, but could be used for strategic advice in adaptive and mitigation planning for long-term change. Intermediate to single-species assessments and whole integrated ecosystem models with climate–ocean forcing are ‘Models of Intermediate Complexity for Ecosystem assessments’ [58]. These models limit complexity by selecting only the ecosystem components that are required to address the main impacts of the management question under consideration and are capable of providing tactical outputs for fisheries management decision-making [58]. The models include at least one ecological process and can be linked to physical models to quantify the impact of environmental forcing. To date, however, these models have not been applied for actual tactical decision-making.

Irrespective of whether a complex suite of models or models of intermediate complexity are used, one of the steps is identifying the underlying mechanisms that link climate–ocean processes to fish productivity, such as recruitment success or spatial distribution. As with the previous section, a potential pitfall may be identifying spurious relationships which will lead to misidentification of processes. This integrative approach is being developed in quantitative ecosystem assessments, i.e. assessment of the ecosystem of which a fishery species is one component and sustainable fisheries is one in a suite of human-driven stressors [8], the success of which is yet to be measured. Given the recent results of several strategic studies (reviewed in the next section) that investigated the direct inclusion of regime shifts or states into tactical management advice, the success of these quantitative ecosystem assessments for fisheries management are likely to be difficult to achieve in the near future.

(d). Environmental forcing in the determination of harvest control rules

This is a very active field of investigation using MSE that employs simulation testing of the performance of management strategies given uncertainty in stock assessment model estimates, data collection, reference points and management actions to achieve specified management objectives [12]. MSE can use strategic models to direct the development of tactical models and harvest control rules. Within an Ecosystem Approach to Fisheries, the overall goal of MSE is to select strategies and harvest control rules that optimize the balance between fishing opportunities and ecosystem sustainability or conservation [59–61]. MSE usually considers a much broader range of sources of uncertainty than conventional stock assessments, reflecting in part that standard methods for quantifying uncertainty in conventional stock assessments underestimate uncertainty because, for example, the uncertainty associated with pre-specifying parameters, choosing functional forms for key biological processes and dataset choices, cannot be quantified using these standard methods. MSE is currently the most relevant field of investigation for incorporating regime shifts into fisheries management since the strategic models can ask: (i) Does ignoring environmental impacts lead to poor performance? and (ii) Can we do better using an environmental control? Unfortunately, the answer for these two questions can be opposing.

Given the substantial amount of published literature on impacts of regime shifts on fish productivity [33,62–66], it seems intuitive that accounting for ecosystem regime shifts and states in harvest control rules should improve the ability to optimize the balance between yield and conservation [67,68]. However, research to date using MSE suggests that this is generally not the case [9,69,70]. Fundamental to the precautionary approach is the inclusion of uncertainty in estimating stock status relative to biological reference points. In particular, the determination of biological reference points given environmental forcing and regime states has been well investigated in an MSE framework and several approaches have been explored and are outlined below [71].

The Dynamic B₀ approach takes the parameters that are estimated by the stock assessment model and then projects the population forward from the first year that there are catches, but with no fishing to estimate B₀ [72]. This projection of B₀ has time-varying recruitment, growth and natural mortality which can be simulated with regime-like characteristics. The Moving window approach [71] estimates B₀ and B_MSY using stock assessment estimates of recruitment for a specified number of years. This approach requires 20–25 years of data for reasonable precision and is not useful if the current regime is less than 20 years [71]. The STARS approach [73,74] detects regime shifts in a time series based on a user-defined minimum duration of a regime state and t-distributions to compare an additional year in a time series for the currently selected regime. If the deviation is significantly different from the number of years of the minimum duration required, then a regime shift is confirmed. The time series can then be separated into regime states and used to define regime-specific mean recruitment for application of harvest control rules or estimation of biological reference points [71]. Given investigation of these approaches, it remains that biological reference points should not be regime specific, but should be based on the fit of the stock–recruitment relationship if catch and survey data do not span multiple regime states and shifts thereby capturing a full suite of environmental signals [9].

3. Obstacles to including regime shifts and states into fisheries management

As outlined above, a shift in fisheries science and management philosophy has already occurred that is amenable to including regime shifts and states into fisheries management. For example, there is international and domestic legislation in place to support a precautionary approach, and this approach has been embraced by the fisheries science community. It is now standard to consider uncertainty and risk in tactical advice to balance yield and conservation. There is a considerable amount of research on environmental forcing, including low-frequency regime shifts and states, that impact fish production and overall ecosystem structure and productivity. The consideration of uncertainty in fisheries science would logically be expected to now include the impacts on fish productivity due to environmental forcing and ecosystem changes. So given all of these components, why is consideration of regime shifts and states not operationalized in fisheries management, except in a very few cases?

Probably the most common obstacles to operationalization include:

• — linkages between environmental variables and recruitment time series eventually break down, either due to spurious correlations or changes in the nature of the relationship;

• — typically, the length of the time series for recruitment data is shorter than the span of at least one regime shift and state;

• — the environmental and recruitment time series typically have high within-regime variability which makes it difficult for stock assessments to detect regime shifts; and

• — without a reliable way to anticipate a regime shift, predictions (even short-term) are not possible.

Error inherent in environmental, recruitment or catch measurements and in stock assessments models are currently such that little, or nothing, is gained by including regime states in fisheries advice and management [32,69–71], even if regime shifts can be anticipated [75]. The intuitive advantage of including regime states and shifts in the long-term balance in yield and conservation is observed only when it is assumed that measurements and stock assessment estimates are known with certainty, either in a single-species assessment [67] or in a multi-species ecosystem model [68]. In this situation, regime-specific harvest rates and control rules perform better in balancing long-term yield and avoidance of conservation-defined depletion than a suite of single-species control rules. However, in reality, observations and estimations are made with often very considerable (and perhaps even unresolvable) errors and uncertainty, which in general undermines the success of the inclusion of regime shifts and states.

4. A way forward

MSE provides an avenue for testing tactical models given decisions made based on knowledge of regime shifts and states. However, we propose that current attempts should not focus on directly integrating regime shifts and states in the stock assessments or in estimating biological reference points, but rather are used as supporting information to stock assessment advice. In this regard, the nature of the regime's productivity could be useful when future projections or assumptions about recruitment are needed for management decisions [71]. Historical data on environmental factors and fish productivity could be synthesized and classified into regimes to give an indication of previous regime-state attributes and corresponding relative fish productivity [76]. Biomass estimates from stock assessments (without inclusion of environmental forcing) would be produced for recruitment assumptions for below average, average and above average recruitment-state scenarios and tactical advice based on these scenarios would be summarized in decision tables with associated risk [77]. Current information on ecosystem attributes to compare with synthesized historical regime-state characteristics could be used to assign plausibility weights to each fish productivity scenario presented in the decision table, and the corresponding tactical advice. Knowing the consequences of each management action given each alternative state of nature, along with their relative plausibility, will provide decision-makers with the information to allow them to select management actions given the available uncertainties. Use of decision tables in this way is common in fisheries management advice, although the factors considered uncertain differ (e.g. the resilience of recruitment to reductions in spawning biomass, the sizes of historical catch, etc.) [44,78,79]. The selection of the most likely fish productivity scenario need not be ad hoc, and plausibility rankings could be assigned following a four-level scheme [80]:

• — how strong is the basis for the hypothesis in the data for the species or region under consideration;

• — how strong is the basis for the hypothesis in the data for a similar species or in another region;

• — how strong is the basis for the hypothesis for any species; and

• — how strong or appropriate is the theoretical basis for the hypothesis?

As a starting point, these four criteria can be used to place scenarios into relative plausibility categories such as ‘high’, ‘medium’ or ‘low’, where a higher rank is assigned to scenarios with support for criteria near the top of the list [80]. Taken together, this approach would supplement, and not replace, current stock assessments and reference point estimations. This approach is consistent with the view that moving to an ecosystem approach to fisheries should be evolutionary and not revolutionary [81]. The above approach is, however, untested and while it may provide an alternative way to move forward, it needs to be formulated as a management strategy and evaluated using MSE. As with all MSEs, this will require operationalizing the proposed steps in the algorithm to the extent possible.

The expectation of stakeholders and fishery managers is that stock assessment advice be based on the precautionary approach, account for all uncertainty and consider ecosystem impacts on fish productivity. Fisheries agencies can begin to incorporate regime-shift impacts on fish productivity into the way they assess and manage marine resources to move forward with their mandates to implement ecosystem-based fisheries management. To do so requires the framing of fish productivity into regime-specific states to inform the projection of recruitment and the definition of stock status or ‘health’ of fish stocks. However, this needs to be done pragmatically within a decision analysis framework, rather than as ‘best case’ forecasts which have been shown to be unreliable and may in the longer term lead to a loss in confidence that accounting for ecosystem factors is worthwhile.