Bochenek et al.: Assessment of Stenotomus chrysops and Centropnstas striata discards in the Mid-Atlantic Bight 



by "the recorded tow time and speed." CPUE was then 

 calculated by using estimated swept area as the effort 

 term. Scope was calculated as "tow wire out" divided by 

 "average water depth." For geographic location, each tow 

 was assigned to a 10-minute square area (10-minute 

 latitude and longitude) (Powell et al. 3 ). 



For some analyses, data from the present study's 

 winter 2001 fishery and the NMFS observer database 

 were assigned to categories by codend mesh size: the 

 legal codend with 11.43-cm mesh; a composite codend 

 with 10.16-cm mesh followed by 11.43-cm mesh; codends 

 with meshes less than 6.35 cm; codends with meshes 

 between 6.35 cm and 10.16 cm; and codends having 

 some meshes greater than or equal to 12.7 cm. Gear 

 type was assigned to either a millionaire or large-mesh 

 box net based on the net styles used in our study and 

 interpretations of NMFS observer-recorded net descrip- 

 tions by knowledgeable fishermen. 



Statistical analysis 



Catch was evaluated by using the ratio of scup discards 

 to landings, total catch of all species, total discards of 

 all species, total scup discards, total scup landings, and 

 a comparison of whether the catch of scup per tow was 

 above or below the median for all tows in the study. In 

 addition, we examined the influence of fishing decisions 

 on discards 1) by distinguishing tows where scup dis- 

 cards exceeded scup landings from tows where landings 

 exceeded discards and 2) by distinguishing between the 

 scup catch of tows taken in the first and last half of 

 the trip. For the latter, we also analyzed tows by their 

 fractional position in the trip (whether a tow occurred at 

 the start of a trip, V4, V2, 3 Ai, or at the end of the trip). 

 This approach yielded results equivalent to the simpler 

 assignment of tows to the first and last half of the trip. 

 Only the results of the simpler analysis are presented. 

 Finally, we evaluated the impact of fishing decisions on 

 the length frequencies of scup caught. ANOVAs were run 

 by using ranked raw variables with class variables that 

 defined fishing practice (mesh size, gear, scope, effort), 

 time, and catch. The variable time was used to allocate 

 tows to three categories: 



1 Those trips from the present study taken from 1 to 

 24 January 2001 with a legal trip limit of 4536 kg 

 of scup; 



2 Those trips from the present study taken after 24 

 January 2001, with a legal trip limit of 454 kg of 

 scup; and 



3 Those scup trips taken in 1997-2000 from the NMFS 

 observer reports. 



Length frequencies were analyzed by ANOVA by using 

 the 25 th , 50 th , and 75 th percentiles and the mean as 

 descriptive variables. In initial analyses, the interaction 

 terms between mesh size or time and the other indepen- 

 dent variables were included. Interaction terms were not 

 significant more frequently than expected by chance and, 

 accordingly, were not included in our results. Significant 



differences identified by the ANOVA were further inves- 

 tigated by using Tukey's studentized range test and, for 

 covariates, by Spearman's rank correlation. 



Results 



Catch statistics — scup 



Ten trips were taken during our study and 62 tows were 

 successfully completed; 39 tows targeted scup and 12 

 tows targeted black sea bass (Table 1). For the remaining 

 tows, the captain targeted Loligo squid as part of the 

 normal fishing process and used a much smaller codend 

 mesh size. These LoZ/go-targeted tows were excluded 

 from further analyses. However, frequent changes in 

 target species emphasize the need for tow rather than 

 trip-aggregated data in discard analyses (Powell et al., 

 2004) because multiple targets within trips commonly 

 occur in Mid-Atlantic Bight fisheries. 



The majority of tows were taken in NMFS statistical 

 area 622. Scup-targeted tows occurred primarily dur- 

 ing daylight and at depths ranging from about 73.2 to 

 137.2 m in our study and from 54.9 to 109.7 m in the 

 NMFS observer data set. A few tows from both the NMFS 

 observer database and our study were deleted from the 

 analysis because the catch was released overboard rather 

 than brought onboard. Bycatch estimates from these 

 tows were assumed to be inaccurate in comparison to 

 other tows. This phenomenon occurs sporadically in many 

 fisheries (e.g., Roel et al., 2000). In our study, six scup- 

 targeted tows were disregarded for this reason. All four 

 participating boats had at least one trip where one tow 

 was released overboard. Observers reported that the 

 net was so full of fish, primarily scup, in these tows 

 that it could not be brought on deck. The catch for one 

 black-sea-bass-targeted tow was released overboard. In 

 addition, tows in which no discards were recorded were 

 not analyzed. Generally, such tows occurred when the 

 observer was asleep or sea conditions were too danger- 

 ous to collect data from the tow. Such tows did not occur 

 in our study but did occur sporadically in the NMFS 

 observer database. Regardless of the reason, we assumed 

 that any tow without recorded discards represented in- 

 complete sampling and, consequently, we discarded that 

 tow from further analysis (Powell et al. 3 ). Differences in 

 the tabulated number of observed tows and the number 

 of observed tows analyzed reflect the number of tows 

 excluded from the analyses for these two reasons. 



Length frequency — scup 



The length frequencies of landings and discards were 

 consistently significantly different (often PsO.0001) 

 (Fig. 1). The mean size of discarded scup was 17.7 cm 

 and ranged from 13.2 to 21.4 cm in our study. Fifty 

 percent of the scup discarded fell between 16.8 cm (25 th 

 percentile) and 18.5 cm (75 th percentile). In contrast, 

 the average size of scup landed was 24.2 cm and ranged 

 from 22.2 to 29.2 cm. Fifty percent of the scup landed 



