Golet et al.: Decline in condition of Thunnus thynnus in tine Gulf of Maine 



391 



annually between 1994 and 1996 in aerial surveys ( Lut- 

 cavage and Kraus, 1995), but catches in 2004 and 2005 

 decreased dramatically, and only 30% of the commercial 

 quota was landed in the New England region. Second, 

 coincident with the reduction in catch, over the past 

 decade fishermen and dealers have reported a decline 

 in fish quality irrespective of season. Fish landed in 

 September and October had the same somatic condi- 

 tion as those landed in June, indicating that northern 

 bluefin tuna are not establishing the fat reserves they 

 once did. Given that energy allocation is a key factor 

 in growth, maturation, reproduction, and migration in 

 long-lived fishes (Marshall et al., 1999; Rideout et al., 

 2005; Jorgenson et al., 2006), a decline in the somatic 

 condition of northern bluefin tuna would be expected to 

 affect the population. 



Catch rates of highly migratory species, especially 

 northern bluefin tuna, have fluctuated over the years 

 in many different regions of the world (Anderson and 

 Piatt, 1999; Ravier and Fromentin, 2001). These top 

 pelagic predators may have altered their distribution 

 because of environmental shifts (Anderson and Piatt, 

 1999), or may have suffered localized depletion because 

 of fishing pressures (Tiews, 1978; Fromentin and Pow- 

 ers, 2005). Although these causes may explain why 

 northern bluefin tuna distribution or abundance in the 

 Gulf of Maine has changed, they do not account for 

 the apparent decline in quality of those fish remaining 

 in the area. In this study, we examined records of the 

 fat and oil content and shape of northern bluefin tuna 

 captured in the Gulf of Maine from 1991 to 2004 in 

 order to investigate whether the observations of a de- 

 cline in quality (as advanced by commercial fishermen 

 and dealers) represents a significant change in somatic 

 condition of these fish. 



Materials and methods 



Fish condition is most often assessed through the use 

 of Fulton's K or linear regression, both of which give 

 a quantitative value to the physical condition of fish. 

 Such analyses were not possible in this study because 

 individual lengths and weights were not recorded for 

 many of the fish. As a substitute, we used grade data 

 from brokers in the commercial northern bluefin tuna 

 fishery who grade every fish before purchase. This pro- 

 cedure is quite involved and often requires schooling or 

 an apprenticeship to learn the trade. Grading involves a 

 qualitative assessment of the condition offish, defined by 

 the characteristics of freshness, color, fat and oil content, 

 and fish shape (Bestor, 2004). Fat grade is assessed by 

 evaluating the amount of marbling in a tail cut sectioned 

 between the third and fourth finlet, the thickness of the 

 midsection, and the amount of fat present in a small 

 core of muscle (near the mid-line) extracted for biopsy. 

 Shape grade is determined by the overall appearance of 

 the fish, the more rotund the better. A good quality fish 

 will receive high marks in all categories. Even though 

 different graders may use different terms, ranking of 



fish quality is consistent between experienced graders 

 (Footed). 



We used two of these characteristics, fat and oil con- 

 tent and fish shape, as proxies for fish condition. Fish 

 with large fat reserves and rotund appearance are pre- 

 sumably feeding in excess of their daily metabolic re- 

 quirements and hence, are in good condition. Fat and oil 

 content and fish shape are reasonable proxies to assess 

 condition because, unlike freshness and color, they can- 

 not be altered by either the time fishermen are at sea 

 or type of gear used. 



Detailed logbooks were obtained from a local fisher- 

 men's co-operative that consisted of 3834 observations of 

 fat and oil content and 3082 observations of shape from 

 fish landed between 1991 and 2004. Fish ranged in size 

 from the commercial minimum of 185 cm curved fork 

 length (CFL) to 300 cm CFL and weighed from 54 kg to 

 351 kg dressed (i.e., after head, gill plate, and internal 

 organs were removed). All of the grading was carried 

 out by the same individual using the same grade scale 

 for the entire 14-year period. 



To examine temporal trends in fish quality, as defined 

 separately by fat and oil content and fish shape, we 

 used multinomial logit regression with fat and oil grade 

 or shape grade as the dependent variable, and month 

 and year as independent variables. The multinomial 

 logit model estimates the probability of a fish being in 

 grade j in month m and year y as 



expujj(w,y)) 



'tj(m,y) = ^ 7— 



where »; (m.y) is a linear equation consisting of the vari- 

 ables for month (m) and year (y) and any month-year 

 interactions. 



The coefficients for these variables can take on dif- 

 ferent values for each grade (McCullagh and Nelder, 

 1989). For example, if we treat month as a categorical 

 variable, nAm,y) can be written as 



rij{m,y) = p„j+{p,j^jXm^) + {li2jXy) + {P3,^jXm^xy) 



where Bg, = the intercept for grade J; 



Bj j, : = the coefficient for the effect of month k on 

 grade j; 

 mf^ = an indicator variable denoting the month 



as a categorical variable; 

 B., = the effect of year on grade 7; and 



: the interaction between month k and year 



y- 



B 



S.k.J 



The model was fitted by using the multinomial command 

 in the NNET library of S-PLUS vers. 6.2 (Insightful 

 Corporation, Seattle, WA). The significance of each vari- 

 able was tested by using likelihood ratios for nested 



2 Foote, J. 2005. Personal commun. Jensen Tuna Inc., 8 

 Seafood Way, Boston, MA 02210. 



