186 



Fishery Bulletin 97(1), 1999 



Materials and methods 



Field sampling 



Each month between July 1990 and June 1994, 

 Manomet Center for Conservation Sciences (under 

 contract to the NEFSC) placed observers on board 

 demersal trawlers working throughout the northeast- 

 ern United States. The selection of vessels and trips 

 undertaken did not adhere to any consistent survey 

 design but was based on changing national, man- 

 agement and stock assessment priorities (Murawski 

 et al., 1995) in addition to varying concentrations of 

 fishing fleets in particular areas and times. As a con- 

 sequence, the observer coverage of these trawlers was 

 highly uneven throughout the period; some areas and 

 times received high sampling and others received very 

 little or no sampling. Of relevance to this paper are the 

 observer data gathered in the southern New England 

 and mid-Atlantic areas shown in Figure 1. 



During observed trips, the contents of the codend 

 from each tow were emptied onto the deck and sorted 

 by the crew. If the entire catch could not be sampled, 

 data were collected from a representative subsample. 

 The data collected from each tow that were of rel- 

 evance to this paper were the following; area, depth, 

 date, time and tow duration, the weights of retained 

 and discarded individuals of various species in the 



catch, the number of species caught, and the total 

 weights of all species retained or discarded. Because 

 of various operational and logistic reasons, not all 

 tows from all trips were sampled for the weights of 

 retained and discarded species; however, the data 

 examined in this paper include only those tows that 

 were completely sampled. 



Analysis of data 



Because of the lack of a consistent survey design in 

 this observer program, the data generated could not 

 be regarded as randomly collected, independent 

 samples of the trawling effort of the region, and thus 

 violated the basic assumptions required for conven- 

 tional statistical analyses (see also Kennelly et al., 

 1997). Nevertheless, because of the size of the data 

 set and its extensive spatial and temporal coverage, 

 large subsets of data could be extracted for many 

 areas and months and therefore permitted the iden- 

 tification of certain key areas, depths, and times of 

 consistently high rates of scup discard (see "Results" 

 section). The first and broadest examination of the 

 data was to plot the discard rates for scup for all the 

 areas sampled in the region to identify those areas 

 that experienced consistently high rates. Next, dis- 

 card rates for scup from each sampled tow in each of 

 these identified areas was plotted against depth to 



80" w 



76" W 



, , I.,., . . 



74" W 

 , ,.J., . , 



72" W 



70" W 68" W 



66 W 



4-— 



40 N - 



Long Island »JLy;_r,61, 

 New York City 



V. ^(^ K'M 



38° N 



Chesapeake Bay ; v .  ■^ 



36° N - 



Cape Hatteras T^ 





50f 500f 



Figure 1 



Map showing the various NMFS statistical areas examined in the present paper Areas labelled in 

 bold are those with sufTicient observer coverage to be examined in Figure 2. 



