-
Notifications
You must be signed in to change notification settings - Fork 1
2.1 ASPIC
ASPIC fits a Schaefer, Fox, or generalized surplus-production (biomass dynamics) model to catch and relative-abundance (or standardized fishing effort) data, without making an equilibrium approximation.
The software is a suite of programs, including ASPIC for model fitting, ASPICP and AGRAPH for projections and graphics, and ASPIC 5 to 7 to convert input files from ASPIC 5 to ASPIC 7 format.
Current Entry April, 2017 (vers. 7.05)
First Entry July 2000 (vers. 3.82)
First ICCAT Reference Working paper SCRS/91/24 (ICCAT Coll. Vol. Sci. Pap. 37:218–229)
Catalogue Committee
Proposer:
Seconder:
Reviewers:
ICCAT Secretariat
Download ASPIC 7 Suite installer
Alternate link - Author's Web site
The ACPIC 7 Suite User's Guide is included in the installation as a PDF file. A Quick Start Guide for new users is also included.
Typing aspic7 -help on a command line gives brief help information. The Quick Start Guide provides a basic tutorial; the User's Guide, a detailed explanation, including tips for reliable operation. Numerous sample input files are installed with the ASPIC Suite, and these can be used as templates for the user's own analyses. Corresponding sample output files are provided, for further guidance.
Michael H. Prager
Prager Consulting
http://www.mhprager.com/
mike.prager@mhprager.com
For legal reasons, only executable code and support files can be distributed freely, along with user’s manual and sample files. However, source code is available from the author upon written request and with agreement to certain restrictions on redistribution. These restrictions are necessary because ASPIC includes a proprietary optimization routine from the book Numerical Recipes.
ASPIC does not require other software to run, except a PC operating system. The current executable runs on 32-bit and 64-bit versions of Windows. It is reported to run on Apple computers with Intel processors and suitable Windows support. Versions of ASPIC for Linux may be available on request.
The main program, ASPIC 7, fits surplus-production models to catch and relative abundance (or standardized fishing-effort) data. In doing this, the program uses a forward-projecting, observation-error estimator, similar to those used in contemporary statistical catch-age models. Production-model forms available in ASPIC include the logistic (Schaefer), exponential-yield (Fox), and generalized (Pella-Tomlinson) models. Unlike early production-model software, ASPIC does not rely upon equilibrium approximations.
For fitting, ASPIC 7, offers several objective functions, including nonlinear least squares, least absolute values, maximum likelihood, and maximum a posteriori (MAP) estimation. Added in ASPIC 7, MAP is a form of Bayesian analysis, essentially penalized likelihood estimation, i.e., MLE influenced by user-specified priors. For optimization, a Nelder-Mead (1965) “polytope” or “simplex” algorithm is used to minimize a suitable function of the residuals in catch, CPUE, or fishing effort. To increase the chances of locating a global minimum, the optimizer is repeatedly restarted until it converges to the same solution several times in a row. An optional Monte-Carlo search, which works together with the regular minimizer, can be invoked if necessary. The program uses an analytical solution of the catch and biomass projection equations of the logistic model, and corresponding numerical approximations for the Fox and generalized models. Numerous data checking features are used to promote reliable estimation.
ASPIC estimates uncertainty in parameter estimates (and related management quantities) through bootstrapping. The user may specify up to 3000 bootstrap trials.
The theory behind ASPIC is described in several publications. Basic equations and a full description of the theoretical framework in the logistic case are given in Prager (1994, 1995). The method is also described in Quinn and Deriso (1999), p. 77. The parameterization of the Pella–Tomlinson (generalized) production model is that of Fletcher (1978). A technical appendix to the ASPIC 7 User's Guide gives further details, including explanations of features not described earlier and equations of the objective functions.
Program ASPICP provides projections of up to 100 years from ASPIC bootstrap results. (Long-term projections are meant for illustrative teaching use, not for stock management. As with any population model, only short-term projections are considered reliable for management.) Projections from ASPICP include bias-corrected confidence intervals on trajectories of relative biomass and relative fishing mortality rate, among other information.
Program AGRAPH provides quick, presentation-quality graphics of certain ASPIC and ASPICP results. Graphs are visible on screen and may be sent to a Windows printer or graphics file in one of several formats.
Program ASPIC5to7 converts an ASPIC input file from ASPIC 5 to ASPIC 7 format. Because the new format is slightly more complex, default values are used where information must be added. Thus, to ensure that the conversion was done appropriately, the analyst must inspect the converted file and edit it if necessary.
- One to twelve series of data on (1) relative abundance or fishing effort and (2) removals. Each series may be up to 150 years long.
- Starting guesses and bounds on the leading parameters (MSY, Fmsy, q, B1/K).
- When using MAP estimation, names of chosen prior distributions and their parameters.
- Control parameters (e.g., convergence criteria). The sample input files provide values that typically work well; changes generally cause worse results. Suitable values are also given in the User's Guide.
- Items for user convenience, such as run title and a text description of each data series.
ASCII files with a variety of outputs, including a special file for loading into R.
- Goodness-of-fit information (ANOVA table and R-squared).
- Likelihood values or other measures of minimization.
- Time plots of estimated and observed relative abundance.
- Time plots of residuals for each series.
- Checks that parameters are not at constraints; warning messages if they are.
- Plots of population trajectories.
- Values of Prager's coverage and nearness indices (experimental).
- Under MLE estimation, the Akaike Information Criterion values for logistic and generalized fits; also, a statistical test comparing the two models.
- Limits on number of iterations allowed for convergence.
- Error codes and error messages on screen, on the output, and in the summary files whenever a recognized error condition is encountered.
The basic model underlying ASPIC was described by Lotka (1924). It was introduced to fishery science in a quantitative way by Schaefer (1954, 1957). The fitting algorithm (forward projection) was applied to this problem by Pella (1967), who also derived an analytical solution conditioned on effort, and by Pella and Tomlinson (1969). Both of the preceding authors described applications to yellowfin tuna. A system of equations similar to that used in ASPIC was described by Schnute (1977). The polytope optimization mehod is described in Nelder and Mead (1965).
The specific combination of theory, fitting algorithm, and optimization technique, along with several characteristic extensions, are detailed in Prager (1994). Some aspects of application to Atlantic swordfish are described in Prager et al. (1995). Prager et al. (1996) examined performance under changing gear selectivity. Prager and Goodyear (2001) examined performance with inconsistently measured data. Prager (2002) compared logistic and generalized estimates from ASPIC; in the course of doing so, correctness of the generalized estimation procedure was examined. Williams and Prager (2002) compared ASPIC to PRODFIT in fitting the generalized model without and with (respectively) an equilibrium approximation, and concluded that, with the availability of contemporary computers, problems with the equilibrium approximation outweigh its usefulness. Shertzer and Prager (2002) compared two objective functions (SSE and LAV) available in ASPIC to a third objective function (least median of squares) that was temporarily implemented, but found unstable. ASPIC has been used in numerous peer-reviewed assessments in the U.S. and internationally, and by several international fish conservation and management bodies, including NAFO, ICES, and ICCAT.
Great care has been taken to write the program in standard-compliant Fortran 95 and to use the error-checking facilities of that language. The program has been compiled under several compilers with all error-checking options enabled to detect any errors in array bounds, argument matching, undefined or nonstandard language (Fortran) features, or other such errors. Collaborators have reviewed many sections of the source code.
The author has written a simulation program to generate numerous simulated data sets. These have been analyzed by ASPIC to verify that correct answers are obtained. As well, the author has offered to correct any bug encountered or suspected by others, and has analyzed several such incidents. ASPIC was used in a simulation study of the generalized production model in which nearly 50,000 simulated data sets were fit (Williams and Prager 2002). Any unexpected behavior of the program was analyzed and corrected.
Several simulation studies have been published using ASPIC. One that comes to mind is Cadrin et al. (1999). Also, Polacheck (1993) compared statistical assumptions and preferred observation-error estimators such as ASPIC to process-error estimators.
Version 7 of ASPIC implements several improvements; most notably, the inclusion of MAP estimation. It also includes true maximum-likelihood estimation, as well as the (equivalent) nonlinear least squares. The software can be run by drag-and-drop, command line, or by scripts (batch files); special features are provided to summarize runs made from scripts.
Over the years, the R output file from ASPIC has become comprehensive. It can be used as input for the analyst's R scripts for automating graphics or other post-run analyses.
The allowable lengths of data series (data extents) were increased considerably in ASPIC 7. If still larger data extents are needed, please contact the author.
Cadrin, S. X., S. H. Clark, et al. 1999. Application of catch-survey models to the northern shrimp fishery in the Gulf of Maine. North American Journal of Fisheries Management 19: 551-568.
Fletcher, R.I., 1978. On the restructuring of the Pella–Tomlinson system. Fish. Bull. 76: 515–521.
Lotka, A. J. 1924. Elements of physical biology. Reprinted 1956 as “Elements of mathematical biology” by Dover Press, N.Y., 465 p.
Nelder, J. A., and R. Mead. 1965. A simplex method for function minimization. Comp. J. 7:308–313.
Pella, J. J. 1967. A study of methods to estimate the Schaefer model parameters with special reference to the yellowfin tuna fishery in the eastern tropical Pacific ocean. Doctoral Dissertation, University of Washington, Seattle. 156 p.
Pella, J. J. and P. K. Tomlinson. 1969. A generalized stock production model. Bull. Inter-Am. Trop. Tuna Comm. 13:419–496.
Polacheck, T. , R. Hilborn, and A. E. Punt. 1993. Fitting surplus production models: comparing methods and measuring uncertainty. Canadian Journal of Fisheries and Aquatic Science 50:2597–2607.
Prager, M. H. 1994. A suite of extensions to a nonequilibrium surplus-production model. Fish. Bull. 92374-389.
Prager, M. H. 2002. Comparison of logistic and generalized surplus-production models applied to swordfish, Xiphias gladius, in the north Atlantic Ocean. Fish. Res. 58: 41–57.
Prager, M. H., C. P. Goodyear, and G. P. Scott. 1996. Application of a surplus production model to a swordfish-like simulated stock with time-changing gear selectivity. Transactions of the American Fisheries Society 125~729--740.
Prager, M. H., and C. P. Goodyear. 2001. Effects of mixed-metric data on production model estimation: simulation study of a blue-marlin-like stock. Trans. Am. Fish. Soc. 130: 927–939.
Prager, M. H., C. E. Porch, K. W. Shertzer, and J. F. Caddy. 2003. Targets and limits for management of fisheries: A simple probability-based approach. N. Am. J. Fish. Manage. 22: 349–361.
Shertzer, K. W., and M. H. Prager. 2002. Least median of squares: A suitable objective function for stock assessment models? Can. J. Fish. Aquat. Sci. 59: 1474–1481.
Quinn, T. J., and R. B. Deriso. 1999. Quantitative fish dynamics. Oxford University Press, NY.
Schaefer, M. B. 1954. Some aspects of the dynamics of populations important to the management of the commercial marine fisheries. Bull. Inter-Am. Trop. Tuna Comm. 1(2):27–56.
Schaefer, M. B. 1957. A study of the dynamics of the fishery for yellowfin tuna in the eastern tropical Pacific Ocean. Bull. Inter-Am. Trop. Tuna Comm. 2:247–268.
Schnute, J. 1977. Improved estimates from the Schaefer production model: theoretical considerations. J. Fish. Res. Board Can. 34:583–603.
Williams, E. H., and M. H. Prager. 2002. Comparison of equilibrium and nonequilibrium estimators for the generalized production model. Can. J. Fish. Aquat. Sci. 59: 1533–1552.