38 



Fishery Bulletin 104(1) 



likely causes variable bridle and footrope contact with 

 the bottom. Vessel heave at the starboard trawl block 

 was used as a proxy for sea state. Heave, pitch, and 

 roll data were collected at 1-sec intervals with a heave 

 sensor (VT TSS, DMS-25, Watford, UK) mounted in the 

 bridge along the centerline of the vessel. Heave data 

 at the starboard block were then predicted, given the 

 X, y, z coordinates of the block from the heave sensor, 

 as distance (cm) from its equilibrium position. In the 

 analyses, the standard deviation of the heave was used 

 as the index of sea state. 



Net crabbing was subjectively assessed on a four-point 

 scale by a single observer, then numerically coded as 

 follows; 1) none — the net trailed straight behind the 

 vessel; 2) slight — the warp could be seen entering the 

 water between the side rail and the aft gantry; 3) mod- 

 erate — the point of entry of the warp into the water was 

 blocked from view by the aft gantry; and 4) severe — the 

 warp was observed entering the water behind the stern 

 ramp. Conditions usually remained constant and one 

 observation was made per towing-mode treatment. How- 

 ever, in some instances conditions changed rapidly and 

 warranted more than one code. In such cases the aver- 

 age of the observation codes was used in the analyses. 



Bottom current direction and velocity (cm/sec) were 

 measured at 10-sec intervals with an oceanographic 

 current meter (Nobska, MAVS-3, Woods Hole, MA) 

 moored in the vicinity of our trawling activity three 

 meters from the bottom. The current data were parti- 

 tioned into two directional components, one parallel to 

 the course of the vessel and the other perpendicular, 

 and averaged for each sample treatment. 



Data analyses 



Comparison of the means and standard deviations among 

 treatments Our null hypothesis, that the three treat- 

 ments had the same effect, was tested with a Krus- 

 kal-Wallis one-way ANOVA for all measures of trawl 

 performance. We selected this test because of the skewed 

 nature of the data. To describe the effect of the three 

 treatments on trawl geometry features the mean and 

 standard deviation (SD) were calculated for the wing 

 spread, door spread, and headrope height off-bottom 

 from each treatment in each haul. 



To describe the bottom-tending performance of the 

 footrope we calculated the following statistics for each 

 treatment: 



1 mean footrope distance off-bottom — the sum of mean 

 distances off-bottom along the footrope, from five 

 footrope BCS units; 



2 standard deviation of the footrope distance off- 

 bottom — the sum of standard deviations along the 

 footrope, from five footrope BCS units; and 



3 symmetry of the footrope distance off-bottom — the 

 sum of the absolute difference between the means 

 of the two wing positions and the absolute differ- 

 ence between the means of the two corner positions 

 on the footrope. 



To describe the bottom-tending performance of the 

 lower bridle, we considered only the BCS unit positioned 

 40 m from the wing tip. The variability in bottom-tend- 

 ing performance of the bridles 40 m from the wing tip 

 may have the highest impact on the catchability of the 

 trawl because the BCS at the 25 m position was always 

 on the bottom and the BCS at the 50 m position was 

 always off the bottom. Performance of the bridles at the 

 40 m position was characterized by 



1 mean bridle distance off-bottom — the sum of the 

 mean distances off-bottom from the BCS units lo- 

 cated 40 m forward of the wing tip on both lower 

 bridles; 



2 standard deviation of the bridle distance off-bot- 

 tom — the sum of the standard deviations from the 

 BCS units located 40 m forward of the wing tip on 

 both lower bridles; and 



3 symmetry of the bridle distance off-bottom — the 

 absolute difference between means from the BCS 

 units located 40 m forward of the wing tip on both 

 lower bridles. 



Assessment of the effect of environmental factors If 



differences among towing mode treatments were found 

 by the ANOVA (P<0.05), then the effects of heave, 

 crabbing, and bottom current on gear performance 

 within each treatment were explored. Multiple regres- 

 sion analyses were performed for each of the treatment 

 statistics with environmental variables as dependents. 

 At the start of the analyses, the models included all 

 four dependent variables (heave, crabbing, current par- 

 allel, and current perpendicular to the tow direction). 

 Variance inflation factors (VIFs) were calculated for 

 all variables to test for multicollinearity (Neter et al., 

 1996). Dependent variables in all models had VIFs 

 ranging from 1 to 1.4, indicating that no serious multi- 

 collinearity existed among dependent variables. Models 

 were simplified (backward deletion) until all P-values 

 of the individual slopes were lower than 0.05. This 

 procedure enabled us to establish which variables had 

 a statistically significant impact on performance of the 

 trawl within each treatment. 



Results 



A total of 21 successful hauls were performed during 

 the experiment, yielding 42 possible treatment sets 

 for analyses. The number of successful treatment sets 

 included in the analyses varied for each of our perfor- 

 mance statistics as a result of either malfunctions or 

 poor survey trawl performance caused by conflicts with 

 abandoned fishing gear. 



Comparison of the means and standard deviations 

 among treatments 



Trawl geometry A total of 41 treatment sets were used 

 to analyze the six trawl geometry statistics (means and 



