The determination of data-poor catch limits in the United States: is there a better way? Open Access

Abstract

Methods for determining appropriate management actions for data-poor stocks, including annual catch limits (ACLs), have seen an explosion of research interest in the past decade. We perform an inventory of methods for determining ACLs for stocks in the United States, and find that ACLs are assigned to 371 stocks and/or stock complexes with 193 (52%) determined using methods involving catch data only. The proportion of ACLs involving these methods varies widely among fisheries management regions, with all the 67 ACLs in the Caribbean determined using recent catch when compared with 1 of 33 ACLs in the New England region (US Northeast). Given this prevalence of data-poor ACLs, we recommend additional research regarding the potential effectiveness of simple management procedures for data-poor stocks that are currently managed using ACLs. In particular, simple management procedures may allow a broader range of data types and management instruments that better suit the particulars of individual regions and stocks.

Introduction

The Magnuson-Stevens Reauthorization Act (MSRA) (DOC, 2007) requires setting an annual catch limit (ACL) for each stock, with very few exceptions. The National Marine Fisheries Service established the process of calculating an ACL in its National Standard 1 Guidelines (NMFS, 2009). The process starts with estimating maximum sustainable yield (MSY) for a given stock (NMFS, 2009). Therefore, management of marine fisheries in the exclusive economic zone (EEZ) of the United States is generally based on the concept of MSY.

Estimating MSY requires knowledge of the stock's productivity, i.e. the magnitude of production that occurs due to individual growth, compensatory recruitment, and natural mortality. This generally requires information regarding some combination of catch, fishing effort, indices of stock abundance, and/or fishing effort, either from sampling data or information from species with similar catch, effort, or index trends (e.g. Thorson et al., 2013). Fisheries science has spent over 50 years developing ecological models for population dynamics and statistical tools for fitting these models to available data (Beverton and Holt, 1957). In general, population dynamics and statistical methods require at least two of these data types—many fisheries scientists are introduced to stock assessment by learning about surplus production models, which combine catch data with an index of either stock abundance or fishing effort (Hilborn and Walters, 1992).

Many fisheries management regions in the United States have collected a time series of catch data by requiring logbooks for commercial fishers or sampling fishing vessels at port, and in some cases by sampling individuals that fish recreationally. Therefore, many US stocks have at least a partial time series of landings data. However, this time series may have a short length, and may not include fish that were caught and subsequently discarded. We refer to the sum of landings and discards as total catch, and total catch is therefore often short, missing, and/or imprecisely known.

Given the prevalence of catch-based data collection in the United States and elsewhere, many stocks have at least some information regarding recent landings or catch. For many stocks, however, additional data that may inform an assessment (e.g. length and/or age compositions, surveys) exist but are either undiscovered or not incorporated into an assessment for any number of reasons. Stocks with only catch data available for setting an ACL pose a unique challenge to stock assessment scientists, because catch will often not be informative about stock dynamics in the absence of meta-analytic or auxiliary information. It is therefore difficult to determine productivity, and hence ACLs for these stocks, and estimates will be highly imprecise as a proxy for MSY if possible to obtain at all. Yet, ACLs for each stock, individually or within a complex, are required by National Standard 1 Guidelines (NMFS, 2009).

Catch-only methods available

Ever since the passage of the MSRA requiring ACLs for every stock, stock assessment scientists have focused attention on developing methods to estimate ACLs using only catch data, or catch data in conjunction with rough estimates of species life history and current status (Berkson et al., 2011). This follows decades of research regarding management procedures, defined here as methods for setting and triggering fisheries management actions in response to changes in available data (Caddy, 2002), to which we will return in a later section.

One of the earliest attempts to develop a total allowable catch (TAC) for catch-only stocks was developed by Restrepo et al. (1998) as part of the Technical Guidance document for the 1998 National Standard 1. The Restrepo approach falls into the category of scalar approaches (Berkson et al., 2011). Scalar approaches involve selecting a specific portion of the time series of the catch, and applying a summary statistic (e.g. mean, median, 75th percentile, 0.5 multiplied by the average catch, maximum) to it. The approach is designed to identify an allowable catch that is likely to be sustainable given the catches that have been observed in the past. In some applications, the summary statistic chosen can be dependent on the perceived depletion level of the stock. The selection of period and summary statistic selected are both based on informed judgement and can be quite subjective.

Other approaches for catch-only stocks include depletion-corrected average catch (DCAC) (MacCall, 2009) and depletion-based stock reduction analysis (DB-SRA) (Dick and MacCall, 2011). Both of these methods are based on sound population dynamics theory, but require auxiliary information, i.e. an estimate of the depletion status of the population in a recent year. DB-SRA also requires having a complete time series of catch (i.e. data going back to the beginning of the fishery), data that are often not available. Assumptions about depletion in stock reduction analyses can in some cases be replaced with compositional data (Thorson and Cope, in press), although this option has not as yet been extensively tested (Punt, 2008). These methods may also be particularly sensitive to changes in biological parameters (e.g. recruitment, Vert-pre et al. (2013)), which could otherwise be identified in more data-rich models.

The “Only Reliable Catch Stocks” (ORCS) Working Group Approach (Berkson et al., 2011) was designed to provide a more solid ecological basis to the Restrepo et al. (1998) scalar approach, while allowing regional flexibility in application. It also creates a step for managers to select an allowable risk level; a step usually included in data-moderate and -rich methods. As with other scalar methods, it requires informed judgement to select an appropriate period over which to compile historical catches, and an appropriate summary statistic for interpreting historical catches.

Additional approaches continue to be developed (e.g. Martell and Froese, 2012; Cope, 2013; Thorson et al., 2013), some of which use meta-analytic and multispecies information to “steal from the data rich to give to the data poor” (Punt et al., 2011). However, new methods will continue to be limited by the lack of data for many species. In these cases, there will remain an important role for expert judgement, both for what meta-analytic information is applicable and how to interpret the sparse data that are available.

The prevalence of catch-based ACLs

If we view the need to set ACLs for stocks that only have catch data as a challenge, then clearly the magnitude of the challenge depends on the number of ACLs that are based on catch-only methods. If only a few ACLs nationwide are derived using catch-only methods, then there may be little cause for concern. If, on the other hand, a substantial number of ACLs are derived using catch-only methods, we likely need to be concerned with the actual mandate to calculate ACLs for each stock.

We conducted an inventory of all ACLs set by the National Marine Fisheries Service in the US in July of 2013. We sought to determine how many ACLs are currently set, what proportion are set using only catch data, and what methods are used for catch-based ACLs. Management in the US EEZ is a regional process, involving eight Regional Fishery Management Councils (RFMCs) (Figure 1). Each Council has a scientific advisory body, its Scientific and Statistical Committee (SSC), whose job is to assign a value for each stock's overfishing limit (OFL) and acceptable biological catch (ABC), critical steps along the path of setting an ACL (NMFS, 2009). We contacted key RFMC staff members and SSC members to obtain the information for each region.

Many stocks are not identified individually for management purposes, but instead are managed by family or even in some cases through the use of large-scale categories, such as corals or “other invertebrates”. Because of this, there are many cases where multiple stocks or categories are managed as a complex, using one ACL. For example, the South Atlantic Fishery Management Council manages 62 stocks through the use of 37 ACLs. Nationwide, hundreds of stocks are managed through the use of 371 ACLs (Table 1, Figure 1).

National overview

Of these 371 ACLs in the United States, 193 (52%) involve catch-only methods and the proportion varies widely by region (Table 1, Figure 1). In general, the northern RFMCs (New England, Mid-Atlantic, and North Pacific) have the lowest percentage of ACLs involving catch-only methods while RFMCs in the Western Pacific and the Southeast (South Atlantic, Caribbean, and Gulf of Mexico) have the greatest.

Scalar approaches are the predominant catch-only methods being used to set ACLs and they are being used to some degree by every RFMC developing ACLs involving catch-only methods (Table 1). One might expect some degree of consistency among Councils in their application of methods, but that is not the case. Councils are using different time series of catch, summary statistics, and scalars. The presence of a large number of managed stocks within an RFMC does not necessarily result in a large percentage of catch-only method based ACLs. Both the NPFMC and PFMC manage large numbers of stocks, yet less than one-quarter of their ACLs rely on these methods. Although a great deal of effort is being spent by stock assessment scientists in the development of more advanced methods (e.g. DCAC and DBSRA), these methods are currently used primarily by the Pacific Fishery Management Council (PFMC) in the development of only ten ACLs involving catch-only methods (less than 6% of the ACLs nationwide).

A brief description of how ACLs are calculated by each RFMC (as of July, 2013) is presented in Supplementary material.

Prospects and alternatives

Regional differences in the methods used to set ACLs may be caused by differences in data availability. For example, the PFMC uses DB-SRA for many of their “data-poor” stocks whereas the SAFMC would likely consider stocks that were eligible for DB-SRA “data-moderate” due to having a complete historical time series of catch data. For several regions of the United States, catch without major error is unlikely, and even occasional surveys do not exist. For other stocks, additional data that may inform an assessment (e.g. length and/or age compositions, surveys) exist but are either undiscovered or not incorporated into an assessment for any number of reasons.

We also note that ACLs, as implemented within National Standard 1 (NMFS, 2009), are inextricably tied to the concept of MSY in their very definition. MSY cannot be estimated with any precision with only an incomplete time series of catch data, and without auxiliary information or data. Additionally, estimating MSY necessarily requires estimating the absolute scale of the population, which generally requires either information regarding total catch, or a survey with known catchability. For this reason, we believe that the process of setting ACLs has privileged catch-based methods over other data-poor assessment methods and management procedures. We now turn to what we see as prospects for future data-poor model types and management procedures.

Expanding the discussion beyond MSY and catch-based methods

As noted previously, basic stock assessment models for estimating MSY generally require information for total catch combined with either (i) an index of population abundance and/or fishing effort or (ii) compositional data (either from sampling data or strong meta-analytic information). Models and model-based procedures will also generally require explicit or implicit assumptions about fishery selectivity and the population's production function. For examples, surplus production models are a well known method for combining catch and index data. Alternatively, the recently developed “catch curve stock reduction analysis” (CC-SRA) (Thorson and Cope, in press) combines total catch and compositional data. Meanwhile, methods such as virtual population analysis or integrated statistical catch-at-age models typically combine all three data types (Quinn, 2003). We therefore refer to models with all three data types as “data-rich”, methods with two data types as “data-moderate”, and methods with only one data type as “data-poor”. By this definition, the methods discussed in this paper are considered “data-poor”. By focusing on available data, this definition departs from the FAO definition of data-poor methods, which are defined by not having sufficient biological information to determine current exploitation status (Vasconcellos and Cochrane, 2005).

Management procedures and accountability measures

The catch-based ACLs highlighted in this article have two major problems. First, the stocks on which these ACLs are based have catch as their only available data usually. Second, the very concept of the ACL is based on the theory of MSY. We suggest that US fisheries management could address to some degree both of these problems by moving away from MSY-based ACLs, and instead defining management procedures that have been shown on average to perform well in achieving management objectives (e.g. high average yield and low risk of depletion) through application of predefined target and limit reference points. As defined by Butterworth et al. (1997), a management procedure is a set of rules regarding fisheries management actions that will be initiated given predetermined realizations of available data, which can then be tested via simulation to ensure that they work well on average given plausible dynamics for the population.

Importantly, management procedures might specify that management actions occur based on model-free (sometimes called “empirical”) or model-based criteria (Rademeyer et al., 2007). Model-based management procedures generally encompass stock assessments as done in the United States, where a population dynamics model is fitted to available data, to satisfy the legal requirement of evaluating status in reference to biological standards such as MSY. However, for model-free (“empirical”) management procedures, management may entirely avoid specifying a model that approximates a population's dynamics and instead focus on control rules that use available data streams to modify existing management actions in a predefined way (and with predefined accountability measures to ensure actions are followed). Alternatively, simple management procedures may use an intentionally simplified biological model, which is intended to filter measurement error from information about abundance changes (McAllister et al., 1999). This allows for a broader research agenda for fisheries scientists, and less of a requirement to estimate MSY for stocks that lack necessary data or meta-analytic information. In essence, fisheries scientists would be asked to develop and test generic management procedures to show that they are robust to relevant uncertainties (Bentley and Stokes, 2009) rather than being asked to calculate reference points in the absence of required information for a stock (Geromont and Butterworth, 2014). Fisheries scientists could still maintain broad biological research to support the development of plausible “states-of-nature” for simulation testing of proposed management procedures. However, there could perhaps be less emphasis on estimating MSY for stocks where such estimates are imprecise or rely on dubious assumptions.

Since their development in the early 1990s, many simple management procedures have been developed for different circumstances. A model-free management procedure may, for example, use a running tally of population characteristics (i.e. average length) and trigger actions when the characteristic significantly departs from its average value (Petitgas, 2009). Or it may involve comparing species densities inside and outside a marine protected area, and use this ratio to tune existing fisheries management regulations (McGilliard et al., 2011). These and other examples share a common theme, i.e. of using available data, made up of well known fisheries indicators, to modify management actions without having to fit a population dynamics model.

As one example of a tool for simple management procedures that may be applicable for data available in the United States, the southeast region (GMFMC, SAFMC, CFMC) has been developing non-equilibrium methods for estimating fishing mortality rates given the average length of catches (Ehrhardt and Ault, 1992; Gedamke and Hoenig, 2006), and these methods have been applied to assess spawning biomass reference points for reef fish near Puerto Rico (Ault et al., 2008). Similar methods have been tested elsewhere (Klaer et al., 2012), and found to perform well in closed-loop simulation. This method therefore meets the Butterworth et al. (1997) definition of a management procedure, while only using a single data type (although the Klaer et al. (2012) simulation assumed that total catch was known during the period of active management). However, like many simple modelling approaches, this method still requires expert judgement regarding the shape of selectivity and the value of biological parameters.

Alternatively, many US management regions routinely collect catch rate data from fishery-independent surveys. Fishery and survey catch rates have previously been used in an empirical management procedure for a two-species South African hake fishery, a sardine and anchovy fishery, and a South African rock lobster fishery (Plagányi et al., 2007). In these cases, the fishery was eventually changed to model-based management procedures, with accompanying decrease in variance for management advice (McAllister et al., 1999). Although these model-free procedures generally managed output-controls (i.e. total catch), future research could explore cpue-based empirical procedures given input-control management (i.e. gear restrictions and/or effort limits).

Because there is more inherent flexibility when designing and applying either a model-free or a model-based management procedure when compared with an MSY-based policy, there is more flexibility in the types of data that can be used. However, this flexibility should not be confused with being vague when defining management objectives. Instead, each proposed management procedure must be thoroughly tested to ensure that it performs well on average across multiple different hypotheses for the population's dynamics, where performance could be flexibly defined by stakeholders and managers in accordance with management objectives, funding, and available data. Moving to a management procedure-based paradigm could allow fisheries scientists to change their focus from developing and testing population dynamics models used for tactical decision making, to developing and testing strategic management procedures that are intended to perform well on average across stocks (Bentley and Stokes, 2009). We note that, in the latter case, developing and fitting population dynamics models would still be important for parameterizing the operating models used to test any proposed management procedure (Punt, 2008).

This change in focus from tactical population dynamics modelling fitting to strategic testing of management procedures would not by itself solve the problem of catch-only stocks, because even a management procedure-based paradigm requires additional data beyond catch. The critical point here is that the range of additional data that could be used is much broader than is the case now. In many data-poor fisheries, the types of data required to estimate MSY parameters, such as indices of abundance, may not be accessible due to limited resources (funding, personnel, equipment) or required survey complexity. While indices of abundance may not be possible to obtain, it may be possible to obtain lengths of fish caught, or a comparison of densities inside and outside an MPA. These kinds of studies may be far more feasible for multispecies fisheries that involve data-poor stocks in the US Caribbean or Western Pacific, for example. It also may be more efficient for limited stock assessment personnel to develop model-free management procedures, than try customizing a population dynamics model to the data available for every managed species.

When US marine fisheries policy was revised in 2007 as the MSRA (DOC, 2007), it required accountability measures whenever fisheries management councils were not implementing the federally legislated mandate to end overfishing (Methot et al., 2013). These accountability measures are part of a process that requests a scientific determination of MSY for every stock and, in doing so, further strengthened the central role of MSY estimation in US fisheries management. However, we note that empirical management procedures were also developed with a concept of limit reference points, wherein a strict management response would be triggered by a predefined threshold (Caddy, 2002). The specific management response is likely to vary based on the management instruments that are available in a given region (e.g. output-controls like TAC, or input-controls such as effort restrictions or gear regulations), so that the catch limits defined by data-poor harvest control rules in Australia (Little et al., 2011; Klaer et al., 2012) might not be appropriate for regions without accurate catch reporting. Regardless of the management instrument that is available for a given region, and hence the choice of what type of control rule is used during simulation testing of a management procedure, the present emphasis on MSY for defining limit reference points (via overfishing limits) seems to use to be only one of many possible methods for defining accountability measures.

Critically, stock assessment modelers in the United States are increasingly focused on population dynamics models as the chief decision-support tool, presumably due to the requirement to estimate biological reference points such as MSY. However, the development of simpler (potentially model-free) decision-support tools remains a critical priority for fisheries management. Combined with flexible management procedures, generic and simple decision-support tools using a broader range of data types may prove to be central for managing the hundreds of stocks that are currently managed primarily via ACLs determined by statistics of recent catches.

Supplementary data

Supplementary material is available at the ICESJMS online version of the manuscript.

Acknowledgements

This work was originally developed and presented at the World Conference on Stock Assessment Methods organized by the International Council for the Exploration of the Seas (ICES) in July of 2013. Dr Luiz Barbieri provided very helpful guidance on the development of the presentation, as an original co-author, before asking to be removed due to time limitations. Mark Maunder, Nokome Bentley, Mark Dickey-Collas, Rick Methot, David Newman, and Alex Chester provided very helpful comments on earlier drafts of the manuscript. The authors thank the SSC members and RFMC staff who provided the critical, regional information required for this work. The original map of the Regional Fishery Management Councils was created by Tim Haverland and provided to us by Laura Oremland, both of the NMFS Office of Science and Technology. Katyana Vert-pre adapted the map.

References

Author notes