Rose et al. : Effective herding of flatfish by cables with minimal seafloor contact 
139 
180°W 1 70°W 160°W 
Figure 3 
Fishing locations (•) in the eastern Bering Sea for the 2006 tests of 
the effects of raised sweeps on flatfish herding. Regions shaded with 
diagonal lines are areas of trawl closures around the Pribilof Islands 
and in Bristol Bay. Contour lines indicate depths. 
floor contact. This was mounted in 
a protective sled, which was towed 
both behind the sweeps, to show 
interactions between the sweeps 
and the seafloor, and, separate 
from the trawl, across the track of 
a previous haul, to show marks left 
on the seafloor. These observations 
were made only on sweeps with the 
20-cm disks. The sled was also 
towed across tracks from previous 
trawl tows with conventional and 
modified sweeps and was equipped 
with a video camera for detailed 
imagery. 
To estimate the proportional 
change in catch due to the sweep 
modifications, the difference be- 
tween the natural logarithms of 
the catch weights from modified 
and unmodified trawls (Log Dif) 
was calculated for each species 
from each twin-trawl haul: 
Log Dif = In (Catch modlfied ) - 
In (Catch unmodified ). (1) 
This statistic, equivalent to the log- 
arithm of the ratio between catches with modified and 
unmodified nets, was appropriate because absolute catch 
sizes were uncontrolled and varied widely. A statistic 
based on subtracting the untransformed trawl catches, 
like that for an ordinary paired t-test, would have varied 
proportionally to absolute catch rates, whereas catch 
ratios, as measured by Log Dif were independent of the 
fish densities encountered during each tow. Averages 
and confidence intervals of Lo gDif were computed for 
each species and sweep modification. To report these 
results as ratios, the averages and confidence intervals 
were then back-transformed with the exponential func- 
tion. Catch results were only used for species with more 
than a minimal catch (>10 fish) in both nets. The null 
hypothesis that the sweep modifications did not affect 
catch was tested with a t-test of whether average Lo gDif 
was different from 0, equivalent to a paired t-test for 
differences between the log-transformed catches. 
To test whether the sweep modifications affected the 
size-selectivity for different fish species and to minimize 
variability, we pooled fish into three size classes for 
each species, except for arrowtooth flounder, where a 
wide size range made four size classes more appropri- 
ate. The size-class boundaries were set so that approxi- 
mately one-third (one-fourth for arrowtooth flounder) of 
the fish in the combined control catches were in each 
category. To maintain consistency with the weight- 
based analysis of overall catch, and because the Alaska 
trawl fleet classifies fish sizes by weight, the boundaries 
of the size classes were defined by individual weights 
instead of lengths, and the catches of each size class 
were computed as weights, instead of numbers. Length- 
weight functions from the annual Bering Sea shelf trawl 
survey (NMFS, unpubl. data 1 ) were used to convert the 
sampled lengths to their corresponding weights. The 
catch of each size class was estimated by expanding the 
proportion of that size class, by weight, from the sample 
of catch for that species. As with the total catch data, 
averages and confidence intervals were calculated. We 
used analysis of variance to test for differences between 
size classes for each combination of species and for each 
sweep modification. 
Results 
From 6 to 23 September 2006, 61 successful twin trawl 
hauls were conducted, including 19, 26, and 16 hauls 
with experimental sweep clearances of 5, 7.5, and 10 cm, 
respectively. Depths at these tow sites (Fig. 3) ranged 
from 70 to 117 m, and bottom temperatures ranged from 
2.5° to 5.5°C. 
Sonar imagery during towing showed that unmodified 
sweeps produced a continuous cloud of disturbed sedi- 
ment due to contact with the seafloor. Variation in the 
density of that cloud appeared to result from contact 
with high and low spots on the seafloor, and rapid oscil- 
lation of strong and weak cloud intensity appeared to 
be due to vibration of the sweeps. In contrast, the sedi- 
ment cloud from the modified sweep appeared only di- 
rectly behind the disk cluster. The only clouds from the 
1 NMFS, Alaska Fisheries Science Center, RACE Division, 
7600 Sand Point Way NE, Seattle, WA 
