SCRS/2003/072 Col. Vol. Sci. Pap. ICCAT, 56(6): 1391-1401 (2004) APPLICATION OF MULTIFAN-CL IN THE STOCK ASSESSMENT OF ALBACORE D. García 1, V. Restrepo 2, H. Arrizabalaga 3, C. Palma 2, I. Mosqueira 1, V. Ortiz de Zárate 4 SUMMARY Following SCRS recommendations, an attempt to fit MULTIFAN-CL to albacore stocks is made with the purpose of illustrating the potential use of this software in future stock assessments. In general, the use of MULTIFAN-CL looks very promising both for the southern and the northern stock, and it is suggested that an informal group carry out the work intersessionally. RÉSUMÉ Suite aux recommandations du SCRS, on a tenté d ajuster MULTIFAN-CL aux stocks de germon dans le but d illustrer l utilisation potentielle de ce logiciel dans de futures évaluations de stocks. En règle générale, l utilisation de MULTIFAN-CL paraît très prometteuse à la fois pour le stock sud et pour le stock nord, et il est suggéré qu un groupe informel réalise les travaux durant la période intersessions. RESUMEN Siguiendo las recomendaciones del SCRS, se intento ajustar MULTIFAN-CL a los stocks de atún blanco con el fin de ilustrar la utilización potencial de este programa para futuras evaluaciones de stocks. En general, la utilización de MULTIFAN-CL parece muy prometedora tanto para el stock del Norte como para el del Sur, y se sugirió que un grupo informal siguiera con esta tarea en el periodo intersesiones. KEYWORDS Integrated statistical methods, catch at age, Thunnus alalunga. 1 AZTI. Txatxarramendi irla. 48395 Sukarrieta. Basque Country (Spain). 2 ICCAT. Corazon de María 8. 28002 Madrid. (Spain). 3 AZTI. Herrera Kaia z/g. 20110 Pasaia. Basque Country (Spain). 4 IEO. Apdo. 240. 39080 Santander. (Spain). 1391
1 Introduction For several years, the SCRS has recommended that species groups consider using "integrated statistical methods" for stock assessment. In order to facilitate the beginning of ICCAT work in this direction, the Secretariat contacted Dr. Dave Fournier to set up test datasets for use of the MULTIFAN-CL software. Dr. Fournier visited the Secretariat in March 2003, as well as seven scientists who helped create the datasets and ran the software. One of the datasets created in March was for south Atlantic albacore. The analyses presented here are for that stock, as well as for the north Atlantic stock. In both cases, the analyses are very preliminary. MULTIFAN-CL is a flexible tool for modelling and, for that reason, requires careful work in terms of defining fleets, regions, etc. Thus, what is presented here is considered as a first step in the introduction of this software to the Albacore Species Group and it is hoped that the Group will set up a work program to make improvements to the datasets in the future. The model in MULTIFAN-CL is a length based, age structured statistical model (Fournier et al. 1998) that includes spatial structure, fish movement and tagging data analysis (Hampton and Fournier 2001). The model is fit to the time series of catch and size composition data, and may also be fit to tagging data simoultaneously (Kleiber et al. 2003). MULTIFAN-CL is now used routinely for tuna stock assessments by the Oceanic Fisheries Programme (OFP) of the Secretariat of the Pacific Community (SPC) in the western and central Pacific ocean (WCPO). 2 Materials and Methods 2.1. ALB-S Data and Model Definition The example set up for the southern stock was simplistic in that it lacked spatial structure and contained only four fleets. These were (1) Japanese longline, (2) Chinese Taipei longline, (3) South Africa + Namibian baitboats, and (4) "Others". Nominal catches for these four fleets were taken from Task I, on a monthly basis. Monthly effort data were obtained from Task II catch/effort, raised so as to match the Task I catches. For fleet 3, only South African C/E data were used, as these were more common throughout the series. Fleet 4 was comprised of a variety of gears and fisheries (primarily longline) and no effort series was associated to its catches. Size samples were obtained from the Task II database. When catch-at-size data were available, they were divided by an arbitrary number so as to obtain a sample size of similar magnitude to the actual sample sizes in neighboring years. For fleet 3, C. Smith kindly provided Southafrican data on a monthly basis for the period 1985-2000, as the Task II database only contained annual size frequency data. No size data was used for fishery 4. No tagging data or other auxiliary information was included in the model. The following are some of the main modeling choices that were made: - Initial age structure in equilibrium for 2 years. - Non-decreasing selectivity for fleets 1 and 2 (longline) - Fleet 4 selectivity is equal to fleet 2 selectivity (Others = Chinese Taipei) - Estimate length-dependent component of standard deviation of length at age. - Estimate von Bertalanffy k - Estimate a fixed M with a prior equal to 0.3 and a strong penalty. - Allow for time-series changes in catchability for fleets 1, 2 and 3. - Estimate seasonal changes in catchability for fleets 1, 2 and 3. - Assume a Beverton-Holt stock recruitment relationship 2.2 ALB-N Data and Model Definition The example for the northern stock also lacked spatial structure and contained five fleets: (1) all baitboats, (2) all trolling, (3) gillnet and midwater trowl, (4) Japanese longline and (5) Chinese Taipei lonline. 1392
All data were stratified quarterly. Nominal catches were taken from Task 1. Effort data were taken from Task II monthly catch/effort data. As several fleets were grouped into single fisheries, and in order to get an average measure of effort in an as long as possible series, it was computed an average CPUE for each fishery, and then calculate effort as Effort TOTAL = Catch TOTAL /CPUE MEAN The average CPUE was obtained with GLMs, unweighted, of the form: CPUE = Year + Quarter + Fleet + +Kind of effort + Year:Quarter Using S-Plus, assuming a quasi-likelihood family model with log-link and a variance function proportional to the mean (i.e., similar to a Poisson error structure). The following treatment of the data was used: - If monthly data were available for a given fleet, three monthly values were used corresponding to a given quarter. - In some cases, catch-effort data were available for a given fishery-quarter-year stratum but CATDIS contained no catches for that stratum. These data were excluded. - The catch for a given fishery-quarter-year stratum was zero, these data were excluded. - Some fleets were excluded from the analyses because they contained too few observations. Size samples were obtained from Task II database, in 2 cm intervals. When catch at size data were available, they were divided by the mean number of fish sampled by that fishery along the years, to obtain a similar magnitude to the actual samples of that fleet. When summing up the samples from different fleets to get the size samples for a given fishery, fleet specific size samples were summed weighted by a proportion reflecting the mean proportion of the catch in all years. In this preliminary analysis, no tagging information was used, allthough this tagging information was already prepared for future analyses. The following are some of the main modeling choices that were made: - Initial age structure in equilibrium for 5 years. Non-decreasing selectivity for fleets 4 and 5 (longline) Downweight Chinese Taipei length information, as it is not provided the sample size. Assume same selectivity for last 3 age classes in fishery 1 - Estimate length-dependent component of standard deviation of length at age. - Allow for time-series changes in catchability for fleets 1, 3, 4 and 5. - Estimate seasonal changes in catchability for all fleets. - Assume a Beverton-Holt stock recruitment relationship. 3 Results and Discussion Selected results for the southern stock are shown in Figures S-1 to S-5. Figure S-1 shows the overall fits to length frequency data for the three fisheries that had such data. Figure S-2 shows the input catches by fishery and the estimated ones (which match almost exactly). Figure S-3 shows the estimated catchabilities by fishery. Figure S-4 shows the estimates of selectivity, recruitment and growth. Figure S-5 shows the trends in biomass and fishing mortality relative to MSY levels. The estimated value of natural mortality was 0.25. Selected results for the northern stock are shown in Figures N-1 to N-6. Figure N-1 shows the overall fits to length frequency data for the five fisheries. Figure N-2 shows the input catches by fishery and the estimated ones. Figure N-3 shows the estimated catchabilities by fishery. Figure N-4 shows the estimates of selectivity, Figure N-5 shows the estimated recruitment and figure N-6 shows the trends in biomass and fishing mortality relative to MSY levels. 1393
As mentioned before, the MULTIFAN-CL run presented here is mostly illustrative. Some of the results seem quite reasonable. For example, most of the estimated selectivities by fishery seem consistent with what was expected. On the other hand, the increasing trend in southern albacore recruitment over time seems suspect and is probably an artifact of data/model mis-specification. One of the main problems with the current definition of fisheries in southern albacore is that fishery 4 is a mix of many different gear types. It would be more appropriate to split this series into several fisheries. In addition, work is required to set up a dataset that more fully utilizes the available data (for instance, many available size and C/E samples were not used). For the northern stock, a significant reduction of the effort (and increase in cpue) since 1984 was observed in the input data. This may be affecting the increasing trends in recruitment and catchability of fishery 1 predicted by the model. In addition to this, the fit to fishery 1 size frequency data was not very good and the estimated selectivity for this fishery seem to be too high in the older ages. This suggests the need for a revision of the input data. The availability of tag recapture data (González-Garcés and Arrizabalaga 2002) and size distributions for some of the tagged fish makes possible to explore further interesting analyses including these data and assuming some spatial structure. Overall, the use of MULTIFAN-CL to assess albacore stocks looks very promising, and it is recommended that an informal group carry out the work inter-sessionally. References FOURNIER, D.A.;J. Hampton and J.R. Sibert (1998). Multifan-cl: A length-based, age-structured model for fisheries stock assessment, with application to south pacific albacore, thunnus alalunga. Can. J. Fish. Aquat. Sci. 55: 2105-2116. GONZÁLEZ-GARCÉS, A. and H. Arrizabalaga (2002). Update of albacore tag release-recapture information in the north atlantic and mediterranean for the period 1968-1999. Collect. Vol. Sci. Pap. ICCAT 54(5): 1454-1478. HAMPTON, J. and D.A. Fournier (2001). A spatially disaggregated, length-based, age-structured population model of yellowfin tuna (thunnus albacares) in the western and central pacific ocean. Mar. Freshw. Res. 52: 937-963. KLEIBER, P.;J. Hampton and D.A. Fournier (2003). Multifan-cl user's guide. : 110. 1394
Figure S-1. Summary length-frequency fits for S. Atlantic albacore. Top left: Japanese longline; top right: Chinese Taipei longline; left: Southafrican baitboat. Figure S-2. Observed and predicted catches for the 4 fisheries defined for S. Atlantic albacore. 1395
Figure S-3. Predicted catchabilities and effort deviations for S-Atl. albacore. 1 Selectivity 0.75 0.5 0.25 JapanLL ChTaiLL BB 0 1 2 3 4 5 6 7 8 Age Figure S-4. Estimates of selectivity, growth and recruitment for N-Atl. albacore. 1396
Figure S-5. Estimates of biomass and fishing mortality relative to MSY levels for S. Atl. albacore. Figure N-1. Summary length-frequency fits for N. Atlantic albacore. From top to down and from left to right: baitboat, troll, gillnet + midwater trowl, Japanes longline and Taiwanese longline. 1397
Figure N-2. Observed and predicted catches for the 5 fisheries defined for N. Atlantic albacore. 1398
Figure N-3. Predicted catchabilities and effort deviations for N. Atlantic albacore. 1399
Figure N-4 Estimates of selectivity by fishery. 1400
Figure N-5. Estimated recruitment for north Atlantic albacore. Figure N-6. Estimated biomass and fishing mortality relative to MSY for N. Atlantic albacore. 1401