A probability-based approach to setting annual catch levels.

The requirement of setting annual catch limits to prevent overfishing has been added to the Magnuson-Stevens Fishery Conservation and Management Reauthorization Act of 2006 (MSRA). Because this requirement is new, a body of applied scientific practice for deriving annual catch limits and accompanying targets does not yet exist. This article demonstrates an approach to setting levels of catch that is intended to keep the probability of future overfishing at a preset low level. The proposed framework is based on stochastic projection with uncertainty in population dynamics. The framework extends common projection methodology by including uncertainty in the limit reference point and in management implementation, and by making explicit the risk of overfishing that managers consider acceptable. The approach is illustrated with application to gag (Mycteroperca microlepis), a grouper that inhabits the waters off the southeastern United States. Although devised to satisfy new legislation of the MSRA, the framework has potential application to any fishery where the management goal is to limit the risk of overfishing by controlling catch.

User’s Guide to ADMB2R: A Set of AD Model Builder Output Routines Compatible with the R Statistics Language

ADMB2R is a collection of AD Model Builder routines for saving complex data structures into a file that can be read in the R statistics environment with a single command.1 ADMB2R provides both the means to transfer data structures significantly more complex than simple tables, and an archive mechanism to store data for future reference. We developed this software because we write and run computationally intensive numerical models in Fortran, C++, and AD Model Builder. We then analyse results with R. We desired to automate data transfer to speed diagnostics during working-group meetings. We thus developed the ADMB2R interface to write an R data object (of type list) to a plain-text file. The master list can contain any number of matrices, values, dataframes, vectors or lists, all of which can be read into R with a single call to the dget function. This allows easy transfer of structured data from compiled models to R. Having the capacity to transfer model data, metadata, and results has sharply reduced the time spent on diagnostics, and at the same time, our diagnostic capabilities have improved tremendously. The simplicity of this interface and the capabilities of R have enabled us to automate graph and table creation for formal reports. Finally, the persistent storage in files makes it easier to treat model results in analyses or meta-analyses devised months—or even years—later. We offer ADMB2R to others in the hope that they will find it useful. (PDF contains 30 pages)

User’s Guide to C2R: A Set of C Language Output Routines Compatible with the R Statistics Language

C2R is a collection of C routines for saving complex data structures into a file that can be read in the R statistics environment with a single command.1 C2R provides both the means to transfer data structures significantly more complex than simple tables, and an archive mechanism to store data for future reference. We developed this software because we write and run computationally intensive numerical models in Fortran, C++, and AD Model Builder. We then analyse results with R. We desired to automate data transfer to speed diagnostics during working-group meetings. We thus developed the C2R interface to write an R data object (of type list) to a plain-text file. The master list can contain any number of matrices, values, dataframes, vectors or lists, all of which can be read into R with a single call to the dget function. This allows easy transfer of structured data from compiled models to R. Having the capacity to transfer model data, metadata, and results has sharply reduced the time spent on diagnostics, and at the same time, our diagnostic capabilities have improved tremendously. The simplicity of this interface and the capabilities of R have enabled us to automate graph and table creation for formal reports. Finally, the persistent storage in files makes it easier to treat model results in analyses or meta-analyses devised months—or even years—later. We offer C2R to others in the hope that they will find it useful. (PDF contains 27 pages)

User’s Guide to For2R: A Module of Fortran 95 Output Routines Compatible with the R Statistics Language

For2R is a collection of Fortran routines for saving complex data structures into a file that can be read in the R statistics environment with a single command.1 For2R provides both the means to transfer data structures significantly more complex than simple tables, and an archive mechanism to store data for future reference. We developed this software because we write and run computationally intensive numerical models in Fortran, C++, and AD Model Builder. We then analyse results with R. We desired to automate data transfer to speed diagnostics during working-group meetings. We thus developed the For2R interface to write an R data object (of type list) to a plain-text file. The master list can contain any number of matrices, values, dataframes, vectors or lists, all of which can be read into R with a single call to the dget function. This allows easy transfer of structured data from compiled models to R. Having the capacity to transfer model data, metadata, and results has sharply reduced the time spent on diagnostics, and at the same time, our diagnostic capabilities have improved tremendously. The simplicity of this interface and the capabilities of R have enabled us to automate graph and table creation for formal reports. Finally, the persistent storage in files makes it easier to treat model results in analyses or meta-analyses devised months—or even years—later. We offer For2R to others in the hope that they will find it useful. (PDF contains 31 pages)

Population characteristics of gulf menhaden, Brevoortia patronus

The status of the Gulf menhaden, Brevoortia patronus, fishery was assessed with purse-seine landings data from 1946 to 1997 and port sampling data from 1964 to 1997. These data were analyzed to determine growth rates, biological reference points for fi shing mortality from yield per recruit and maximum spawning potential analyses, spawner-recruit relationships, and maximum sustainable yield (MSY). The separable virtual population approach was used for the period 1976–97 (augmented by earlier analyses for 1964–75) to obtain point estimates of stock size, recruits to age 1, spawning stock size, and fishing mortality rates. Exploitation rates for age-1 fi sh ranged between 11% and 45%, for age-2 fi sh between 32% and 72%, and for age-3 fi sh between 32% and 76%. Biological reference points from yield per recruit (F0.1: 1.5–2.5/yr) and spawning potential ratio (F20: 1.3–1.9/yr and F30: 0.8–1.2/yr) were obtained for comparison with recent estimates of F (0.6–0.8/yr). Recent spawning stock estimates (as biomass or eggs) are above the long-term average, while recent recruits to age 1 are comparable to the long-term average. Parameters from Ricker-type spawner-recruit relations were estimated, although considerable unexplained variability remained. Recent survival to age-1 recruitment has generally been below that expected based on the Ricker spawner-recruit relation. Estimates of long-term MSY from PRODFIT and ASPIC estimation of production model ranged between 717,000 t and 753,000 t, respectively. Declines in landings between 1988 and 1992 raised concerns about the status of the Gulf menhaden stock. Landings have fl uctuated without trend since 1992, averaging about 571,000 t. However, Gulf menhaden are short lived and highly fecund. Thus, variation in recruitment to age 1, largely mediated by environmental conditions, infl uences fi shing success over the next two years (as age-1 and age-2 fi sh). Comparisons of recent estimates of fi shing mortality to biological reference points do not suggest overfishing. (PDF file contains 22 pages.)

Deriving acceptable biological catch from the overfishing limit: implications for assessment models

The recently revised Magnuson–Stevens Fishery Conservation and Management Act requires that U.S. fishery management councils avoid overfishing by setting annual catch limits (ACLs) not exceeding recommendations of the councils’ scientific advisers. To meet that requirement, the scientific advisers will need to know the overfishing limit (OFL) estimated in each stock assessment, with OFL being the catch available from applying the limit fishing mortality rate to current or projected stock biomass. The advisers then will derive ‘‘acceptable biological catch’’ (ABC) from OFL by reducing OFL to allow for scientific uncertainty, and ABC becomes their recommendation to the council. We suggest methodology based on simple probability theory by which scientific advisers can compute ABC from OFL and the statistical distribution of OFL as estimated by a stock assessment. Our method includes approximations to the distribution of OFL if it is not known from the assessment; however, we find it preferable to have the assessment model estimate the distribution of OFL directly. Probability-based methods such as this one provide well-defined approaches to setting ABC and may be helpful to scientific advisers as they translate the new legal requirement into concrete advice.

Severe decline in abundance of the red porgy (Pagrus pagrus) population off the southeastern United States

Status of the southeastern U.S. stock of red porgy (Pagrus pagrus) was estimated from fishery-dependent and fishery-independent data, 1972–97. Annual population numbers and fishing mortality rates at age were estimated from virtual population analysis (VPA) calibrated with fishery-independent data. For the VPA, a primary matrix of catch at age was based on age-length keys from fishery-independent samples; an alternate matrix was based on fishery-dependent keys. Additional estimates of stock status were obtained from a surplus-production model, also calibrated with fishery-independent indices of abundance. Results describe a dramatic increase in exploitation of this stock and concomitant decline in abundance. Estimated fully recruited fishing mortality rate (F) from the primary catch matrix increased from 0.10/yr in 1975 to 0.88/yr in 1997, and estimated static spawning potential ratio (SPR) declined from about 67% to about 18%. Estimated recruitment to age 1 declined from a peak of 3.0 million fish in 1973–74 to 94,000 fish in 1997, a decline of 96.9%. Estimated spawning-stock biomass declined from a peak of 3530 t in 1979 to 397 t in 1997, a decline of 88.8%. Results from the alternate catch matrix were similar. Retrospective patterns in the VPA suggest that the future estimates of this population decline will be severe, but may be less than present estimates. Long-term and marked declines in recruitment, spawning stock, and catch per unit of effort (both fishery-derived and fishery-independent)are consistent with severe overexploitation during a period of reduced recruitment. Although F prior to 1995 has generally been estimated at or below the current management criterion for overfishing (F equivalent to SPR=35%), the recent spawning-stock biomass is well below the biomass that could support maximum sustainable yield. Significant reductions in fishing mortality will be needed for rebuilding the southeastern U.S. stock.