«STANDARDIZED CATCH RATE OF SHORTFIN MAKO (ISURUS OXYRINCHUS) AND BIGEYE THRESHER (ALOPIAS SUPERCILIOSUS) CAUGHT BY SÃO PAULO LONGLINERS OFF SOUTHERN ...»
SCRS/2007/084 Collect. Vol. Sci. Pap. ICCAT, 62(5): 1542-1552 (2008)
STANDARDIZED CATCH RATE OF SHORTFIN MAKO (ISURUS
OXYRINCHUS) AND BIGEYE THRESHER (ALOPIAS SUPERCILIOSUS)
CAUGHT BY SÃO PAULO LONGLINERS OFF SOUTHERN BRAZIL
Bruno L. Mourato1, Alberto F. Amorim2, Carlos A. Arfelli2
SUMMARYThe CPUE must be standardized in order to gather estimations that potentially could be used as indexes of relative abundance. The standardized CPUE of shortfin mako (Isurus oxyrinchus) and bigeye thresher (Alopias superciliosus) sharks was estimated from the data of the Santos and Guaruja tuna longline fishery that operated in the southwest Atlantic from 1971 to 2006. The modeling procedures followed a Generalized Linear Model (GLM) approach assuming a log-normal error distribution. The final model included “year”, “quarter” and “target” as main factors. For the both species the deviance analysis showed that the “year” factor was the most significant followed by “target” and “quarter” factors. As a result, the CPUE gradually decreased for the shortfin mako shark from 1971 to 2006 and showed a slight decline for the bigeye thresher shark from 1978 to 2006.
RÉSUMÉ La CPUE doit être standardisée afin de rassembler des estimations qui pourraient potentiellement être utilisées comme indices de l’abondance relative. La CPUE standardisée du requin taupe bleue (Isurus oxyrinchus) et du renard à gros yeux (Alopias superciliosus) a été estimée à partir des données de la pêcherie palangrière thonière de Santos et Guaruja qui opérait dans l’Atlantique Sud-Ouest de 1971 à
2006. Les procédures de modélisation ont suivi une approche de modèle linéaire généralisé (GLM) postulant une distribution d’erreur long-normale. Le modèle final incluait « année », « trimestre » et « cible » comme facteurs principaux. Pour les deux espèces, l’analyse des déviances a indiqué que le facteur « année » était le plus important, suivi des facteurs « cible » et « trimestre ». En conséquence, la CPUE a graduellement chuté pour le requin taupe bleue de 1971 à 2006, et a fait l’objet d’une légère baisse pour le renard à gros yeux de 1978 à 2006.
RESUMENLa CPUE debe ser estandarizada con el fin de reunir estimaciones que puedan utilizarse potencialmente como índices de abundancia relativa. La CPUE estandarizada del marrajo dientuso (Isurus oxyrinchus) y el zorro ojón (Alopias superciliosus) fue estimada a partir de los datos de las pesquerías de palangre de túnidos de Santos y Guaruja que operaban en el Atlántico sudoeste desde 1971 hasta 2006. Los procedimientos de modelación siguieron un enfoque de Modelo lineal generalizado (GLM) asumiendo una distribución de error log-normal. El modelo final incluía como factores principales “año”, “trimestre” y “objetivo”. Para ambas especies, el análisis de desvianza mostraba que el factor “año” era el más importante seguido de los factores “objetivo” y “trimestre”.
Como resultado, la CPUE descendió gradualmente para el marrajo dientuso desde 1971 hasta 2006 y mostró un descenso ligero para el zorro ojón desde 1978 hasta 2006.
1 M.Sc. student -Instituto de Pesca – APTA – SAA, Santos (SP), FAPESP Schorlarship, email@example.com 2 Instituto de Pesca – APTA – SAA, Santos (SP), firstname.lastname@example.org, email@example.com
Since the beginning national longliners settled in São Paulo State (Santos and Guaruja cities) caught and commercialized sharks (Sadowsky and Amorim, 1977; Amorim and Arfelli, 1992; Arfelli and Amorim, 1994;
Amorim et al., 1998). In the 1971-06 period there were some changes the species targeted by these longliners (Mourato et al. 2004). Sharks participation increased since the 70’s decade which represented about 20%. It was double in 80’s and 90’s (40%), and also increasing in 2000’s (50%) (Amorim, 1992; Amorim et al., 2002).
According to Anon (2002), shorfin mako, Isurus oxyrinchus (Rafinesque, 1809) and bigeye thresher, Alopias superciliosus (Lowe, 1839) sharks were discussed in the ICCAT Data Preparatory Meeting for Atlantic Shark Stock Assessment, in Halifax of September 2001. Catch and effort data of commercial fleet can be used, depending on standardization of CPUE. The CPUE must be standardized in order to gather estimations that potentially could be used as indexes of relative abundance the CPUE. Those indexes can help on the task of assessing the stock biomass. Generalized linear models (GLM) have been often used to standardize catch rate data. In this paper GLM is used in an attempt to estimate indexes of relative abundance of shorfin mako and bigeye thresher for the Southern Brazil. The reliability of the estimates and their potential usefulness for assessment analyses is discussed.
2. Material and Methods
2.1 Data and variables of models Data from São Paulo longline fleet operating off Southern Brazil from 1971 to 2006 were analyzed. Landing records of shortfin mako and bigeye thresher sharks in weight (kg) were collected from commercial sheets of fishing companies. Fishing effort in number of hooks collected from logbooks was obtained from Instituto de Pesca. The data were aggregated for each year and quarter and the catch rate for both species in the dth time
strata (Ud) is:
where C is the catch (kg) and f is the nominal effort (number of hooks).
Besides “year” and “quarter” we used information on the proportion of sharks in the total catch per month to build the “target” factor. Quartiles of the proportions were used to code four levels. According to Arfelli (1996) Santos longline fleet has been fishing off the southeastern coast of South America in the subtropical area (17-35°S / 27-52°W) (Figure 1). However, detailed information about the area were the longlines were deployed by set was not available for the whole period, hence area was not taken into account in the analysis.
A Generalized linear model (McCullagh and Nelder, 1989) was developed using the software package S-Plus.
The approach used to standardize the catch rate is in line with that described by Gavaris (1980), hence we
assumed that catch rate follows a lognormal distribution. The equation for standardizing CPUE was as follows:
where ln is a natural logarithm; CPUE as defined as kilograms per 1000 hooks; u is a overall mean; Year is a effect of year; Target is a effect of target; Quarter is a effect of quarter and; e is a normal error term.
The model for the shortfin mako shark including the temporal series (1971-2006) and for the bigeye thresher shark corresponding the period between 1978 and 2006. All variables were defined as categorical (i.e.
1543 factors). A stepwise approach was used to identify variables that affect the catch rate. The initial model is saturated; it contains all variables and second level interactions among them. Then the model was simplified by excluding terms that were not statistically significant. The criteria used to select the terms were the Akaike Information Criterion – AIC (Akaike, 1974). Finally the distribution of residuals was used to check if the assumption about the lognormal distribution. Diagnostics methods described in McCullagh and Nelder, (1989) and Ortiz e Arocha (2004) were also used to verify the goodness of the model fitting. A pseudo-R2 coefficient was calculated as the fraction of the total deviance explained by the model. This is a measure of the explanatory power of the model.
Assuming that the coefficients estimated for the levels of the “year” factor reflects the annual changes of the biomass they can be used to estimate indexes of relative abundance. Than the standardized catch rate was calculate using the back-transformed calculation of year coefficients by the inverse of the link function.
3. Results and discussion
Although the catch statistics for sharks is dubious for some of the longliners fisheries in the Atlantic Ocean, the data about shorfin mako and bigeye thresher sharks landings in the Santos and Guaruja harbors are reliable. The shorfin mako shark has one of the best meats and it was never rejected by this fleet (Amorim and Arfelli, 1992).
On the other hand the meat of bigeye thresher has low value and and fins have regular price, so they were rejected in the beginning of 70’s. In the last three decades their commercialization improved and the fishermen unload all the shark species (Amorim et al., 1998). In general that fleet did not practice the finning because they have market for the meat. The mentioned fleet sometimes caught thresher shark (Alopias vulpinus) but is very rare. They are included in by the thresher common name.
The quarterly means of nominal CPUE of shortfin mako and bigeye thresher sharks are presented in Figure 2a and 2b, respectively. The catch rates in the third quarter were usually higher than in the others for the both sharks, which corroborate the results of Costa et al. (1996), Amorim et al. (1998) and Mancini (2005). The third quarter is the most productive probably because the boats displace towards the subtropical convergence which appear in the south of the Brazilian waters in the austral winter (Olson et al., 1988). The subtropical convergence promotes an increase of phytoplankton biomass and consequently an increase of pelagic shark preys, as squids (Santos and Haimovici, 2002).
Figure 3 shows the distribution of the levels of the “target” factor by “year” of shortfin mako and bigeye thresher sharks. The level 4 is most representative in 80’s and the beginning of 90’s. In this period the tuna catch decreased and sharks were still very abundant, hence the sharks were the target species, mainly the blue shark (Amorim, 1992). The same happened to the tuna catches in the northeast (3-7ºS e 32-38ºW) from July 1983 to December 1988, with the increase of shark catches (Hazin et al., 1990). In the period 1995-1999 the level 4 was practically absent due the gear changes made in order to catch swordfish (Arfelli, 1996). After in the beginning of 2000’s the São Paulo (SP) fleet returned to direct at sharks, mainly the blue shark (Mourato et al., 2004).
Table 1 shows the deviance analyses of the selected model for shortfin mako shark. All the main factors are significant. Interactions were not significant and not included in the model. The “year” factor explained a large amount of the variation of the catch rate, followed by the “target” and the “quarter”. The proportion of the deviance explained by the model is about 0.70, as indicated by the pseudo-R2 calculation. Table 2 presents the deviance analyses of the selected model for bigeye thresher shark. Interactions were not significant and not included in the model. Analogous by the shortfin mako model, the “year” factor explained a large amount of the variation of the catch rate, followed by the “target” and the “quarter”. The pseudo-R2 calculation indicated that a great part of the proportion of the deviance was explained (0.76). Nevertheless these great values of pseudo-R2 was expected because the data were aggregated by year and quarter, hence a large part of catch rate variation was lost, what could explain the great value of pseudo-R2 in our analysis. This situation is not common. Most GLM analyses frequently result in low coefficients of determination (eg. Punt et al., 2000).
For the both models, the residuals distributions are homoscedastic and the fitting of the model seems to be not biased (Figure 4). Coefficient estimations for the both models are in Tables 3 and 4. Most the coefficients estimated for the “year” factor were not significant though most of standard errors were smaller than the coefficients estimated. Estimations for “quarter” factor were positive. Notice that the coefficients for third and fourth quarters were usually higher than other quarters, indicating that this is the most productive time for fishing sharks off the Southern Brazil.
1544 Figure 5 shows the scaled nominal and standardized CPUE for the shortfin mako. The standardized catch rate oscillated across the years, but it showed a slight decline in the period studied. According to Nakano (2002) the stock status of shortfin mako in the Atlantic Oean suffered a slight decrease over the past three decades, including the North and South units. Senba and Takeuchi (2005) standardized the shortfin mako CPUE and found a gradually decrease for the south Atlantic, based on the Japanese observer data.
Figure 6 presents the scaled nominal and standardized CPUE for the bigeye thresher shark. The standardized catch rate oscillated across the years; nevertheless it showed a slight decline in the 1978-2006 period. The high value in 2003 could be explained by the changes in the fishing areas.. Mancini (2005) related that the Santos and Guaruja longline fleet was the second fleet that most captured the bigeye thresher shark in Brazil in 2003. Baum et al. (2002) related a moderate decline of abundance of thresher sharks northwest Atlantic based on the U.S.
logbook and Canadian observer data.
If the catch rates estimated in this work are assumed to be indexes of relative abundance, the result suggests that the shortfin mako and bigeye thresher sharks from the South Atlantic unit are slightly affected by the fishery.
Despite that the decline of bigeye thresher shark biomass is less in comparison of shortfin mako biomass. The shortfin mako shark is one of the sub-target species for Santos and Guaruja tuna longline fisheries and its market value is among the highest of the pelagic sharks caught in the longline fisheries in Brazil. Although the results are speculative because the low fishing effort and the small operating area of the studied fleet, they might be taken into account when discussing the assessment of the South Atlantic stock of the shortfin mako and bigeye thresher sharks.