STANDARDIZED CATCH RATES FOR BIGEYE TUNA (THUNNUS OBESUS) FROM THE PELAGIC LONGLINE FISHERY IN THE NORTHWEST ATLANTIC AND THE GULF OF MEXICO

Transcription

1 SCRS/2007/069 Collect. Vol. Sci. Pap. ICCAT, 62(2): (2008) STANDARDIZED CATCH RATES FOR BIGEYE TUNA (THUNNUS OBESUS) FROM THE PELAGIC LONGLINE FISHERY IN THE NORTHWEST ATLANTIC AND THE GULF OF MEXICO John Walter 1, Mauricio Ortiz and Craig Brown SUMMARY Two indices of abundance of bigeye tuna from the United States pelagic longline fishery in the Atlantic are presented; an index of number of fish per thousand hooks estimated from numbers of bigeye tuna caught and reported from Pelagic longline logbook data from and a biomass index (dressed weight) per thousand hooks estimated from the dealer weight-out data for the period The standardization analysis procedure included the following variables; year, area, quarter of the year, gear characteristics (light sticks, etc) and fishing characteristics (operations procedure, and target species). Standardized indices of abundance by year and quarter of the year were also constructed from the dealer weight-out data for Standardized indices were estimated using Generalized Linear Mixed Models under a delta lognormal model. Both indices indicate an overall decline since the mid-1980s with slight increases since RÉSUMÉ Ce document présente deux indices d abondance du thon obèse de la pêcherie palangrière pélagique des Etats-Unis dans l Atlantique : un indice du nombre de poissons par mille hameçons, estimé d après le nombre de thons obèses capturés et déclarés dans les livres de bord des palangriers pélagiques de 1986 à 2006 et un indice de biomasse (poids manipulé) par mille hameçons, estimé d après les données des pesées des négociants pour la période La procédure d analyse de standardisation incluait les variables suivantes : année, zone, trimestre, caractéristiques des engins (baguettes lumineuses, etc.) et caractéristiques de pêche (procédure des opérations et espèces cibles). Les indices standardisés de l abondance par année et trimestre ont également été élaborés d après les données des pesées des négociants pour Les indices standardisés ont été estimés à l aide de Modèles Linéaires Généralisés Mixtes (GLMM) dans le cadre d un modèle delta lognormal. Les deux indices indiquent un déclin global depuis le milieu des années 1980, avec de légères augmentations depuis RESUMEN Se presentan dos índices de abundancia de patudo de la pesquería de palangre pelágico de Estados Unidos en el Atlántico; un índice en número de ejemplares por mil anzuelos, estimado a partir del número de patudos capturados y comunicados en los datos de los cuadernos de pesca de la pesquería de palangre pelágico de 1986 a 2006, y un índice de biomasa (peso eviscerado) por mil anzuelos estimado a partir de los datos de venta al peso de los comerciantes para el periodo El procedimiento de análisis de estandarización incluía las siguientes variables: año, zona, trimestre del año, características del arte (bastones luminosos, etc.) y características de la pesca (procedimiento de las operaciones y especies objetivo). También se obtuvieron los índices estandarizados de abundancia por año y trimestre del año utilizando los datos de venta al peso de los comerciantes para el periodo Los índices estandarizados se estimaron utilizando modelos lineales generalizados mixtos con un enfoque de modelo delta-lognormal. Ambos índices indican un descenso global desde mediados de los ochenta con ligeros incrementos desde KEYWORDS Bigeye tuna, abundance indices, catch/effort, catch rate standardization, Generalized Linear Model, pelagic longline fisheries 1 NOAA Fisheries, Southeast Fisheries Center, Sustainable Fisheries Division, 75 Virginia Beach Drive, Miami, FL, , USA. E- mail: john.f.walter@noaa.gov 445

2 1. Introduction Relative abundance indices are critical indicators of stock status as well as essential inputs into stock assessments. Fishery catch per unit effort (CPUE) indices operate under the basic assumption that catch per unit effort is proportional to stock size. Numerous factors including changes in vessel characteristics, fishing locations and methods, stock size and distribution and targeting of alternative species erode this basic relationship and complicate the interpretation of raw CPUE. Provided that data on covariates that influence the relationship between CPUE and abundance exist, these can be incorporated into standardized measures that presumably more accurately reflect population relative abundance (Maunder and Punt 2004). The report presents updated standardized CPUE estimates in numbers of bigeye tuna obtained from pelagic longline logbook data (PLL) for years 1986 to 2006 and in weight from dealer weigh-out data (DLS) for the years We use a delta lognormal approach implemented as a generalized linear mixed model in SAS using methodology similar to Ortiz and Diaz (2003) and Ortiz (2004). We also present nominal catch rates by geographic area and as a function of various explanatory variables and length-frequency histograms and plots of mean length by year and area. 2. Materials and methods Data for this analysis comes from the United States Atlantic and Gulf of Mexico Pelagic longline fishery described in detail by Hoey and Bertolino (1988). Swordfish, yellowfin and bigeye tuna are the predominant target species. The pelagic longline fishing grounds of the US fleet extends from the Grand Banks in the North Atlantic to 5-10 south of the Equator, mainly in the Western Atlantic including the Caribbean Sea and the Gulf of Mexico (Fig. 1).The fishery has operated under several time-area restrictions since 2000 due to management regulations related to swordfish and other species (Federal Register 2000, Fig. 1). These restrictions included two permanent closures to pelagic longline fishing, one in the Gulf of Mexico known as the Desoto Canyon, effective since 1 st November 2000, and the second permanent closure on the Florida East Coast effective since 1 st March In addition, three time-area restrictions were also imposed for the pelagic longline gear in the US Atlantic coast: the Charleston Bump, an area off the South Carolina coast closed from 1 st February to 30 th April starting in year 2001, The Bluefin tuna protection area off the South New England coast closed from 1 st June to 30 th June starting in 1999, and the Grand Banks area that was closed from 17 July 2001 to 9 January 2002 as a result of an emergency rule implementation (Cramer 2002). Catch and effort data is available from three sources: pelagic logbooks reported daily by vessel captains (Scott et al 1993, Cramer and Bertolino 1998, Ortiz et al 2000), pelagic observer program data collected beginning in 1992 by onboard fishery observers on approximately 4-5% of the Pelagic Longline trips (Lee and Brown 1999) and dealer weigh-out data sheets submitted at the completion of each trip. Dealer weight-out data consists of summarized catch per total effort for an entire trip so that spatial and temporal information on fishing locations and catch rates represent averages for an entire trip. In contrast, logbook and observer daily reports generally have spatially and temporally explicit information on individual longline sets. U.S. Pelagic logbook data are available from However, from 1986 to 1991, submission of logbooks was voluntary, and became mandatory in Weigh-out data collection began in 1981 and observer data collection began in In this paper we use both vessel logbook reports and dealer weigh-out data. This data and further descriptions of the data collection and processing are available from the National Marine Fisheries Service Southeast Fisheries Science Center ( The Pelagic Longline Logbook data comprises a total of 292,507 recorded sets from 1986 through Each record contains information of catch by set, including: date-time, geographical location, catch in numbers of targeted and bycatch species, number of hooks, light stick and various other gear parameters for each set as well as environmental conditions such as temperature. Various data restrictions were necessary to eliminate incomplete or erroneous records or records that were non-standard such as very short sets of fewer than 100 hooks, weight of fish incorrectly recorded as number or set with zero fish of any species captured. Restricting the logbook data only to sets from 1986 onward, those with greater than 100 hooks per set, and with complete catch, effort, location and date information, resulted in a total of 266,296 sets, of which 74,933 (28.1%) reported positive catches of bigeye tuna. The pelagic dealer weigh-out data comprises a total of 40,025 records. Similarly we applied a set of data restrictions that eliminated incomplete or clearly erroneous records resulting in a total of 37,040 reports of which 14,764 (39.8%) reported positive catches of bigeye tuna. 446

3 2.1 Dependent variables Two dependent variables were used for the analysis, catch per unit effort (1000 hooks) in number (CPUEN) and in weight (CPUEW). CPUEN was obtained from the logbook data as the number of bigeye tuna caught per 1000 hooks for each longline set. This number included kept as well as live and dead discarded bigeye tuna, however discarded bigeye tuna represented less than 5% of the total number of caught tuna. CPUEW was obtained from the DLS weigh-out database as the total dressed carcass weight of bigeye tuna landed for an entire trip divided by the total number of hooks set for the trip. For comparison nominal catch rates of bigeye tuna caught per 1000 hooks was also obtained from the DLS database. Note that the DLS recorded tuna did not include discarded fish. 2.2 Factors Five fixed factors and a random effect of year and interactions between the factors were evaluated in the analysis. Eight geographical areas (area) of longline fishing have been traditionally used for classification; these include: the Caribbean, Gulf of Mexico, Florida East coast, South Atlantic Bight, Mid-Atlantic Bight, New England coastal, northeast distant waters, the Sargasso Sea, and the offshore area (Figure 1). Calendar quarters (qtr) were used to account for seasonal fishery distribution through the year (Jan-Mar, Apr-Jun, Jul-Sep, and Oct- Dec). Other factors included in the analyses of catch rates included; the use and number of light-sticks (lightc) expressed as the ratio of light-sticks per hook, and a variable named operations procedures (OP), which is a categorical classification of US longline vessels based on their fishing configuration, type and size of vessel, main target species, and area of operation(s). Fishing effort is reported as number of hooks per set, and nominal catch rates were calculated as number of bigeye tuna caught per 1000 hooks for each observation. The U.S. Atlantic longline fleet targets mainly swordfish and yellowfin tuna, but other tuna species are also targeted including bigeye tuna and albacore (to a lesser extent, some of the trips-sets target other pelagic species including sharks, dolphin and small tunas). We define targeting (targ) as a categorical variable with four levels based on the proportion of the number of swordfish caught to the total number of fish per set, with four discrete target categories corresponding to the ranges 0-25%, 25-50%, 50-75%, and %. As fishing practices targeting swordfish generally diverge from those targeting bigeye this provides an empirical means to determine whether vessels are targeting particular species. Note that there is a record for target species within the PLL database, however it is deemed to be uninformative (L. Beerkircher, SEFSC pers comm.) Factors related to bigeye-specific fishing strategy such as longer gangions, more hooks between floats and longer floatlines were examined. Due to changes in the units with which gangion length were recorded this variable proved impossible to use at this time and no clear relationship between floatline length or number of hooks between floats were observed. Indices of abundance of bigeye were estimated by a generalized linear modeling approach assuming a delta lognormal model (Lo et al. 2002). The standardization procedure splits the dependent data into two parts, one of which models the proportion of positive sets which is assumed to have a binomial error distribution. The second part models the mean catch rate (CPUEN or CPUEW) of successful sets and assumes a lognormal error distribution. The standardized index is the product of the two. Parameterization of the model used the GLM structure where the proportion of successful sets is a linear function of the fixed factors the random year effect and random year-interaction effects when the year term was within the interaction. The logit function was selected as link between the linear factor component and the binomial error. Analyses were done using Glimmix and Mixed procedures from the SAS statistical computer software (SAS Institute Inc., 1997, Littell et al. 1996). A step-wise regression procedure was used to determine the set of factors and interactions that significantly explained the observed variability. We used criterion which assumes that the difference in deviance between two consecutive models follows a Chi-square distribution. The deviance is essentially a measure of the residuals explained by the model. Using this statistic, with degrees of freedom equal to the number of additional parameters estimated minus one, a Chi-square test was constructed which indicates if the additional factor is or is not statistically significant (McCullagh and Nelder, 1989). Deviance analysis tables were constructed for both components of the delta model: Proportion of successful trips/sets, and mean catch rate in both numbers and weights of positive trips/sets. Each deviance table includes the deviance explained by the additional factor or interaction, the overall percent explained by each factor, and the Chi-square probability test. Final selection of explanatory factors was conditional to: a) the relative percent of deviance explained by the added factor, normally factors that explained more than 5-10% of deviance were included and b) the Chi-square significant test and c) if an interaction was deemed significant, the main effects of each component of the 447

4 interaction were included. Once a set of fixed factors was specified, all possible 1 st level interactions were evaluated, in particular random interactions between the year effect and other factors. Standard indices of abundance were calculated as the product of the year effect least square means (LSmeans) from the binomial and the lognormal components of the delta lognormal model. LSmeans estimates included a weight proportional to the observed margins of the input data, to account for the unbalanced distribution of the data. Lognormal estimates also included a bias back-transformation correction factor as describe by Lo et al. (1992). Variances were obtained by the delta method for the approximate variance of the product of two random variables (Zhou 2002). Quarterly indices of abundance were obtained as the least square mean estimates obtained by defining a new variable as year*qtr however the quarterly index was only estimated for years as years had several quarters with zero catch. Note that indices in both weight and number are not age or size-specific. 2.3 Time-area restrictions Because of the time-area restrictions mentioned above, it was important to evaluate their possible effect on catch rates of bigeye tuna. We did so in a manner similar to Ortiz (2004) by constructing a separate standardized catch indices for only the areas that are outside of any of the closed areas and comparing these indices to the all-area indices. For the Pelagic Logbook data, it was possible to assign most of the longline sets to specific latitudelongitude positions and evaluate historic catch trends in and out of the time-area restrictions. In contrast, the DLS data represents total catch for a trip with only a general geographic zone that is larger than the time-area closures. Therefore it was not possible to precisely allocate the bigeye tuna catch to closure or non-closure locations. An approximate solution was to assign an average latitude-longitude by linking the weight-out records to the logbook set by set information using the an unique trip identifier number. However, prior to 1996 the trip identifier did not exist so it was not possible to assign these trips to closed or open areas. 3. Results and discussion 3.1 Differences from previous standardization This analysis differs from Ortiz (2004) in several substantive ways. First sample sizes are larger in this study because Ortiz (2004) removed all observations if they lacked an operations code and if they were of OP code 1. Since 2002 and increasing percentage of the total logbook records do not have an OP code, potentially resulting in the removal of 20-30% of the total logbook records. For this reason I kept all unknown OP code records except OP code 1. In addition several logbook records appeared to have anomalous numbers of bigeye tuna and inspection of the datasheets indicated that weight of fish was recorded as number. In this paper the presented standardized catch rates are for all areas, both inside and outside of the area closures. In general, however, the catch rates trends differ little between indices including or excluding the closed areas (Figure 13). Particularly the DLS index is essentially the same because DLS records prior to 1996 cannot be assigned a latitude and longitude and likely due to the loss of spatial resolution when the location of all trips within a set was averaged. Lastly the models selected differ from Ortiz (2004) in that no interaction terms were deemed significant. For the time period , an average of 11,369 longline sets were recorded in the pelagic logbook database. These comprise the most complete records of US pelagic longline CPUE available for the Atlantic and Gulf of Mexico. During this time period, there has been a reduction in the total numbers of vessels in the fishery, with a concomitant increase in the number of hooks set per vessel. This has kept overall fishing effort relatively constant. However, effort has been displaced from a series of time area closure areas (Figure 1 and 2). Figure 2 presents the geographic distribution of fishing effort and nominal catch rates for bigeye tuna from the pelagic logbooks for eight selected years from 1987 through The plotted values are the sum of annual log(hooks) deployed and the mean catches in number per 1000 hooks calculated on grids of approximately 40x40 nautical miles. The plots show the substantial reduction of fishing effort, particularly after the implementation of time/area closures. 3.2 Nominal catch rates Nominal catch rates by area and year indicate higher catch rates in the Mid-Atlantic Bight, Northeast Coast and Northeast distant waters (Figures 3-6). Catch rates are very low in the Gulf of Mexico throughout the entire time period. In recent years, however, catch rates on the Florida East coast have increased primarily to vessels displaced from closed areas in the Florida Straits to the waters north of the Bahamas where bigeye tuna appear to be more abundant (Figure 2). Comparison of the DLS and the logbook standardized catch rates in number of 448

5 bigeye tuna per 1000 hooks shows fairly close agreement as would be expected since the logbook data comes from individual set reports and the DLS data are trip summaries (Figure 7). Deviations may be due to released fish which were included in the logbook CPUEN. 3.3 Model selection The fixed factors in the analysis were chosen through deviance table analysis (Tables 1 and 2) and no fixed factor interactions were deemed significant. Random effects interactions were chosen on the basis of likelihood ratio tests and reduction in AIC values (Tables 3 and 4). The final models selected for the binomial and lognormal components for the CPUEN were as follows (note that factor OP was included as a fixed factor in the proportion positive model because it was included in the interactions): Proportion Positive= Year Area qtr targ OP Year*area Year* OP Year*qtr log(cpuen) = Year Area qtr targ OP Year*area Year* OP Year*qtr The final model selected for the binomial and lognormal components for the CPUEW in biomass were: Proportion Positive = Year Area targ OP Year*area Year*op Year*targ log(cpuew) = Year Area targ qtr OP Year*area Year*OP Year*targ Year*qtr The final model selected for the binomial and lognormal components of the quarterly CPUEW index in biomass were: Proportion Positive = YearQtr Area targ OP YearQtr*area YearQtr*targ log(cpuew) = YearQtr Area targ OP YearQtr*area YearQtr*targ 3.4 Standardized catch rates Histograms of nominal CPUEN from the logbooks and CPUEW indicate no severe departures from the assumed lognormality (Figure 8). Diagnostic plots for the delta lognormal model fit to the Pelagic Logbook data of residuals for positive CPUE, chi-square residuals of proportion positive by year, Q-Q plot of cumulative normalized for positive CPUE and a histogram of residuals for positive CPUE indicate no severe violations (Figure 9). The same plots for CPUE from the DLS weigh-out data also indicate fairly well-behaved data with no clear patterns or trends in the residuals (Figure 10). Standardized catch rates of bigeye tuna numbers and biomass per 1000 hooks (Figures 11, Tables 3 and 4) indicate a decline since 1987, a stable trend from and slight increases since Ortiz (2004) noted a similar decreasing trend for the period with both logbook and weigh-out indices. Nominal catch rates show less of a trend and diverge from standardized indices in the early part of the time series being higher in the logbook number index (Figure 11A) and lower than the DLS weight index (Figure 11B). Standardized catch rates in weight from the DLS data by quarter or season of the year were also created which indicated seasonal variation in catch rates (Figure 12, Table 5) with catch rates highest in the fall. Both standardized catch in number and weight show very similar patterns (Figure 13) reflective of the absence of a trend in mean size over the time period for the entire fishery (Figures 14-16). Mean sizes, do, however differ between regions with the larger bigeye tuna in the Mid-Atlantic and South Atlantic Bight (Figure 14). 3.5 Closed Area analysis Standardized indices were constructed for all areas and for just areas the areas that have been open to fishing throughout the entire time period (Figure 17A, B). Removing all observations that occurred in a area that was closed as some time reduced the number of observations in the Logbook database by almost ½ for the time period but the indices show similar trends. With the reduction in overall effort and in particular in the closed areas, the total number of observations has decreased since 1998 but, again the trends between the open areas and all the areas appear similar. The DLS indices for all areas and for just the open areas do not differ. This 449

6 is due to the inability to assign DLS data to locations prior to 1996 so that the data is the same and to the lack of spatial precision in assigning locations after Literature cited BROWN, C Standardized catch rates of yellowfin tuna, Thunnus albacares, based upon the United States pelagic longline observer program data, Collect. Vol. Sci. Pap. ICCAT, 62, in press (SCRS/ 2007/056). CRAMER, J Large Pelagic Logbook Newsletter NOAA Tech. Mem. NMFS SEFSC 471, 26 pages. CRAMER, J. and A. Bertolino Standardized catch rates for swordfish (Xiphias gladius) from the U.S. longline fleet through Collect Vol. Sci. Pap. ICCAT, 49(1): CRAMER, J. and M. Ortiz Standardized catch rates for bigeye (Thunnus obesus) and yellowfin (T. albacares) from the U.S. longline fleet through Collect. Vol. Sci. Pap. ICCAT, 49(2): FEDERAL REGISTER Atlantic Highly Migratory Species; Pelagic Longline Management; Final rule. 50 CFR part 635. Vol. 65, No. 148 August 1, HOEY, J.J. and A. Bertolino Review of the U.S. fishery for swordfish, 1978 to Collect. Vol. Sci. Pap. ICCAT, 27: LEE, D.W. and C.J. Brown Overview of the SEFSC Pelagic Observer Program in the northwest Atlantic from Collect. Vol. Sci. Pap. ICCAT, 49(4): LITTELL, R.C., G.A. Milliken, W.W. Stroup and R.D. Wolfinger SAS System for Mixed Models, Cary NC: SAS Institute Inc., pp. LO, N.C., L.D. Jacobson, and J.L. Squire Indices of relative abundance from fish spotter data based on delta-lognormal models. Can. J. Fish. Aquat. Sci. 49: MAUNDER, M. and A. Punt Standardization of catch and effort data: a review of recent approaches. Fisheries Research 70: MCCULLAGH, P. and J.A. Nelder Generalized Linear Models 2 nd edition, Chapman & Hall. ORTIZ, M Standardized catch rates for bigeye tuna (Thunnus obesus) from the Pelagic longline fishery in the Northwest Atlantic and the Gulf of Mexico. ICCAT SCRS/04/133. ORTIZ, M. and G. A. Diaz Standardized catch rates for yellowfin tuna (Thunnus albacares) from the U.S. pelagic longline fleet. Collect. Vol. Sci. Pap. ICCAT, 56(2): ORTIZ, M., J. Cramer, A. Bertolino and G.P. Scott Standardized catch rates by sex and age for swordfish (Xiphias gladius) from the U.S. longline fleet Collect. Vol. Sci. Pap. ICCAT, 51(5): PINHEIRO, J.C. and D.M. Bates Mixed-effect models in S and S-Plus. Statistics and Computing. Springer-Verlag New York, Inc. SAS INSTITUTE INC SAS/STAT Software: Changes and Enhancements through release Cary NC: SAS Institute Inc., pp. SCOTT, G.P., V.R. Restrepo and A. R. Bertolino Standardized catch rates for swordfish (Xiphias gladius) from the US longline fleet through Collect. Vol. Sci. Pap. ICCAT, 40(1): ZHOU, S Estimating Parameters of Derived Random Variables: Comparison of the Delta and Parametric Bootstrap Methods. Transctions of the American Fisheries Society. 131:

7 Table 1. Deviance analysis table of BET CPUE in number of tuna/1000 hooks from the pelagic logbooks. Percent of total deviance refers to the deviance explained by the full model; p value refers to the Chi-square probability test between two consecutive models. Bigeye Tuna Logbook Catch (Numbers of fish) Model factors positive catch rates values d.f. Residual deviance Change in deviance % of total deviance year % year area % year area qtr % year area qtr targ % year area qtr targ op % year area qtr targ op lghtc % year area qtr targ op lghtc targ*lghtc % year area qtr targ op lghtc qtr*lghtc % year area qtr targ op lghtc qtr*op % year area qtr targ op lghtc year*targ % year area qtr targ op lghtc year*lghtc % year area qtr targ op lghtc qtr*targ % year area qtr targ op lghtc op*lghtc % year area qtr targ op lghtc area*op % year area qtr targ op lghtc targ*op % year area qtr targ op lghtc year*qtr % year area qtr targ op lghtc area*qtr % year area qtr targ op lghtc year*op % year area qtr targ op lghtc area*lghtc % year area qtr targ op lghtc area*targ % year area qtr targ op lghtc year*area % Model factors proportion positives d.f. Residual deviance Change in deviance % of total deviance year % year area % year area qtr % year area qtr targ % year area qtr targ op % year area qtr targ op lghtc % year area qtr targ op lghtc targ*lghtc % year area qtr targ op lghtc qtr*targ % year area qtr targ op lghtc op*lghtc % year area qtr targ op lghtc qtr*lghtc % year area qtr targ op lghtc year*targ % year area qtr targ op lghtc area*op % year area qtr targ op lghtc year*lghtc % year area qtr targ op lghtc qtr*op % year area qtr targ op lghtc year*qtr % year area qtr targ op lghtc targ*op % year area qtr targ op lghtc area*lghtc % year area qtr targ op lghtc year*op % year area qtr targ op lghtc area*qtr % year area qtr targ op lghtc area*targ % year area qtr targ op lghtc year*area % 451

8 Table 2. Deviance analysis table of BET CPUE in kg of tuna/1000 hooks from the dealer weight out data. Percent of total deviance refers to the deviance explained by the full model; p value refers to the Chi-square probability test between two consecutive models. Model factors positive catch rates values d.f. Residual deviance Change in deviance % of total deviance p Year % < Year Area % < Year Area Targ % < Year Area Targ Qtr % < Year Area Targ Qtr Op % < Year Area Targ Qtr Op Area*Op % < Year Area Targ Qtr Op Targ*Qtr % Year Area Targ Qtr Op Qtr*Op % Year Area Targ Qtr Op Year*Targ % Year Area Targ Qtr Op Targ*Op % Year Area Targ Qtr Op Area*Qtr % Year Area Targ Qtr Op Year*Qtr % Year Area Targ Qtr Op Year*Op % < Year Area Targ Qtr Op Area*Targ % < Year Area Targ Qtr Op Year*Area % < Model factors proportion positives d.f. Residual deviance Change in deviance % of total deviance p Year % < Year Area % < Year Area Targ % < Year Area Targ Qtr % < Year Area Targ Qtr Op % < Year Area Targ Qtr Op Targ*Op % < Year Area Targ Qtr Op Area*Op % Year Area Targ Qtr Op Targ*Qtr % < Year Area Targ Qtr Op Qtr*Op % < Year Area Targ Qtr Op Year*Targ % Year Area Targ Qtr Op Year*Qtr % Year Area Targ Qtr Op Area*Targ % < Year Area Targ Qtr Op Area*Qtr % Year Area Targ Qtr Op Year*Op % < Year Area Targ Qtr Op Year*Area % <

9 Table 3. Random factor interaction table for BET CPUE in # of tuna/1000 hooks from the logbook data. Percent of total deviance refers to the deviance explained by the full model; p value refers to the likelihood ratio test between two consecutive models. Bigeye tuna GLM Mixed Models Deviance -2 REM Log likelihood Akaike's Information Criterion Schwartz's Bayesian Criterion Likelihood Ratio Test Proportion positives Year Area Qtr Target OP Year Area Qtr Target OP Year*area Year Area Qtr Target OP Year*area Year*Target #NUM! Year Area Qtr Target OP Year*area Year*Target Year*op * Year Area Qtr Target OP Year*area Year*op Year*targ Year*qtr Positive catch rates Year Area Qtr Target OP Year Area Qtr Target OP Year*area Year Area Qtr Target OP Year*area Year*Target Year Area Qtr Target OP Year*area Year*Target Year*op * Year Area Qtr Target OP Year*area Year*op Year*Target Year*qtr

10 Table 4. Random factor interaction table for BET CPUE in kg of tuna/1000 hooks from the dealer weight out data. Percent of total deviance refers to the deviance explained by the full model; p value refers to the likelihood ratio test between two consecutive models. Bigeye tuna GLM Mixed Models Deviance -2 REM Log likelihood Akaike's Information Criterion Schwartz's Bayesian Criterion Likelihood Ratio Test Proportion positives Year Area Qtr Target OP Year Area Qtr Target OP Year*area Year Area Qtr Target OP Year*area Year*op * Year Area Qtr Target OP Year*area Year*op Year*targ Year Area Qtr Target OP Year*area Year*op Year*targ Year*qtr #NUM! Positive catch rates Year Area Qtr Target OP Year Area Qtr Target OP Year*area Year Area Qtr Target OP Year*area Year*op Year Area Qtr Target OP Year*area Year*op Year*targ * Year Area Qtr Target OP Year*area Year*op Year*targ Year*qtr

11 Table 5. Nominal and standardized catch rates of bigeye tuna in #/1000 hooks (CPUEN) from the pelagic logbook data. Note that the nominal CPUE includes zero catches. Year Nominal CPUE Stand. CPUE Coeff Var Std Error Numb obs Index Upp CI 95% Low CI 95% Obs prop pos Table 6. Nominal and standardized catch rates of bigeye tuna in kg/1000 hooks (CPUEW) from the dealer weighout system (DLS). Year Nominal CPUE Standar d CPUE Coeff Var Std Error Numb obs Index Upp CI 95% Low CI 95% Obs prop pos

12 Table 7. Nominal and standardized catch rates of bigeye tuna in kg/1000 hooks (CPUEW) from the dealer weighout system (DLS) by year and quarter. Year quarter Nom- inal CPUE Stand. CPUE Coeff Var Std Error Numb obs Index Upp CI 95% Low CI 95% Obs prop pos 1984 Jan-Mar Apr-June Jul-Sept Oct-Dec Jan-Mar Apr-June Jul-Sept Oct-Dec Jan-Mar Apr-June Jul-Sept Oct-Dec Jan-Mar Apr-June Jul-Sept Oct-Dec Jan-Mar Apr-June Jul-Sept Oct-Dec Jan-Mar Apr-June Jul-Sept Oct-Dec Jan-Mar Apr-June Jul-Sept Oct-Dec Jan-Mar Apr-June Jul-Sept Oct-Dec Jan-Mar Apr-June Jul-Sept Oct-Dec Jan-Mar Apr-June Jul-Sept Oct-Dec Jan-Mar Apr-June Jul-Sept Oct-Dec Jan-Mar Apr-June Jul-Sept Oct-Dec Jan-Mar Apr-June Jul-Sept Oct-Dec

13 Table 7. Continued. Year quarter Nom- inal CPUE Stand. CPUE Coeff Var Std Error Numb obs Index Upp CI 95% Low CI 95% Obs prop pos 1997 Jan-Mar Apr-June Jul-Sept Oct-Dec Jan-Mar Apr-June Jul-Sept Oct-Dec Jan-Mar Apr-June Jul-Sept Oct-Dec Jan-Mar Apr-June Jul-Sept Oct-Dec Jan-Mar Apr-June Jul-Sept Oct-Dec Jan-Mar Apr-June Jul-Sept Oct-Dec Jan-Mar Apr-June Jul-Sept Oct-Dec Jan-Mar Apr-June Jul-Sept Oct-Dec Jan-Mar Apr-June Jul-Sept Oct-Dec Jan-Mar Apr-June Jul-Sept Oct-Dec

14 NED NEC 5 MAB 4 GOM 1 SAB 3 2 FEC SAR SAR CAR OFS OFS Figure 1. Geographical areas for the US Pelagic Longline fishery: CAR Caribbean, GOM Gulf of Mexico, FEC Florida east coast, SAB South Atlantic bight, MAB mid Atlantic bight, NEC North east coastal Atlantic, NED North east distant waters, SNA Sargasso Sea, and OFS Offshore waters. Shaded areas represent the current timearea closures affecting the pelagic longline fisheries. Permanent closures: (1) the DeSoto Canyon in the Gulf of Mexico and (2) The Florida east coast areas. Non-permanent closures: (3) the Charleston Bump area closed Feb- Apr, (4) the Bluefin tuna protection area closed in June, and (5) the Grand Banks closed since Oct

15 Figure 2. Spatial distribution of catch and effort in the US Pelagic longline fishery for selected years Scale is log(hooks set) per grid cell. Cell size is approximately 40 x 40 nautical miles. 459

16 Figure 3. Nominal bigeye tuna catch per 1000 hooks for each area from logbook data. Upper and lower limits are 95% confidence intervals. Numbers are the total number of daily logbook sets reporting positive bigeye catches. Figure 4. Nominal percentage of positive sets for bigeye tuna upper and lower limits are approximate 95% confidence intervals, based on normality. Numbers are the total number of daily logbooks reporting positive bigeye catches. 460

17 Figure 5. Nominal bigeye tuna catch per 1000 hooks for each area from dealer weight-out data. Upper and lower limits are 95% confidence intervals. Numbers are the total number of trips reporting positive bigeye catches. Figure 6. Nominal percentage of positive sets for bigeye tuna from dealer weight-out data. Upper and lower limits are approximate 95% confidence intervals, based on normality. Numbers are the total number of trips reporting positive bigeye catches. 461

18 comparison of nominal logbook and DLS number per 1000 hooks for positive records scaled mean Logbook DLS number year Figure 7. Comparison of nominal logbook and DLS number per 1000 hooks for positive records. A. B. C. D. Figure 8. Histograms of nominal catch rates for bigeye tuna from the U.S. Pelagic longline fishery. A. Log(catch per 1000 hooks) for positive logbook sets. B) Frequency distribution of proportion of positive trips. C) Log(weight per 1000 hooks) for positive dealer weigh out data, D) Frequency distribution of proportion of positive trips from dealer weigh out data. 462

19 Figure 9. Diagnostic plots for the delta lognormal model fit to the Pelagic Logbook data for CPUEN (A) residuals for positive CPUE and (B) chi-square residuals of proportion positive by year. (C) Q-Q plot of cumulative normalized for positive CPUE. (D) Histogram of residuals for positive CPUE. Figure 10. Diagnostic plots for the delta lognormal model fit to the dealer weight-out data for CPUEN (A) residuals for positive CPUE and (B) chi-square residuals of proportion positive by year. (C) Q-Q plot of cumulative normalized for positive CPUE. (D) Histogram of residuals for positive CPUE. 463

20 A. Bigeye Tuna Standardized Logbook CPUE US (95% C)I Scaled CPUE (fish/1000 hooks) CPUE obs B. Scaled CPUE (dressed wgt lbs/1000 hooks) Bigeye Tuna Standardized CPUE in weight US Longline Dealer weigh-out standardized observed Figure 11. Nominal and standardized catch rates for bigeye from the U.S. Pelagic longline fishery. A. Top, logbook data reported as number of fish per thousand hooks. B. Dealer weight out data reported as dress weight (lbs) per thousand hooks. 464

21 standardized CPUEW from DLS observed winter spring summer fall annual year*season Figure 12. Quarterly index obtained from dealer weight out data reported as dressed weight (lbs) per thousand hooks. Black line is the annual standardized index. 3.5 Scaled standardized CPUE to overlapping years 3.0 Index biomass Index Number Figure 13. Comparison of standardized catch rates for bigeye tuna from the U.S. Pelagic longline fishery logbook #/1000 hooks and dealer weighout kg/1000 hooks. Each index is scaled to the mean value for the continuous period

22 Figure 14. Histograms of skipjack tuna straight line fork length (cm) by area. Red lines and numbers indicate mean fork length. Figure 15. Histograms of skipjack tuna straight line fork length (cm) by year. Red lines and numbers indicate mean fork length. 466

23 A. FL, mm B. Caribbean Gulf of Mexico Florida East South Atl. Bight Mid-Atl. Bight NE Coast NE Distant Sargasso Sea South offshore FL, mm Figure 16. Box and whisker plots (median, 1 st, 3 rd quartile, minimum and maximum) of bigeye tuna straight line fork length by (A) area and (B) year. Box widths are proportional to sample size and dots represent lengths that are outside the 1.5*interquartile range. 467

24 A. B. scaled CPUEN from logbooks scaled CPUEN from logbooks all areas closed areas removed number all areas number closures removed all areas closed areas removed number closures removed number all areas Figure 17. Comparison of standardized indices constructed using observations from all areas and observations that occur in closed areas removed. A. logbook CPUEN in #/1000 hooks. B. Dealer weighout CPUEW in kg/1000 hooks. Note that prior to 1996 it was not possible to assign the DLS data to spatial locations total observations total observations 468