Four types of gillnets were constructed for this study. Each gillnet was 91 m (300 ft) long. Two complete sets of nets (48 nets) were constructed. All reported gillnet characteristics are nominal. The standard flatfish, lead-added, and dual leadline nets were identically designed except for the construction of the floatline. Each type was constructed of light green (mesh size: 178 mm (7 in)), monofilament mesh webbing with a diameter of 0.47 mm, twenty-five meshes deep. The leadline was 91 m (50 fm) of 23 kg/183 m (50 lb/100 fm) leadline. The floatline of the standard flatfish net consisted of 91 m (50 fm) of 13 mm (0.5 in) diameter foamcore float line with built-in floatation (1.7 oz./yd (52.5 g/m)). The leadadded design was made with a floatline the same as the standard flatfish net, but with flat pieces of lead weight wrapped around the floatline every 9 m (5 fm). The dual leadline net was made with the floatline and leadline consisting of 91 m (50 fm) of 23 kg/ 183 m (50 lb/100 fm) leadline. This net had no floatline as it is normally defined. The standard cod net was used to determine if cod were present in the study area. It was designed following industry practice, and differed from the other three nets in color, twine diameter, leadline weight, and hanging ratio. It was constructed of light green monofilament mesh webbing (mesh size: 178 mm (7 in)) with a diameter of 0.57 mm, twenty-five meshes deep. The floatline was 91 m (50 fm) of 9.5 mm (0.375 in) twisted polyethylene (PE) floatline with one deepwater gillnet float every fathom, or fifty floats per net. Each float provided approx. 3 oz. (85 g) of flotation. The leadline was 91 m (50 fm) of 29 kg/ 183 m (65 lb/100 fm) leadline. Eight nets of the same design were tied into a string; one string of each design was set in the same general location. The geographical arrangement of the strings was changed each time the nets were hauled, based on a modified Latin square design to reduce bias. In general, strings were set and hauled following normal commercial fishing practice. However, soak times were limited to overnight (~24 h), shorter than standard when targeting flatfish. This shorter soak time was selected to allow more rapid testing and to increase survival of discarded fish. A 'set' was defined as each instance of a net being hauled and its catch quantified. Strings were fished on consecutive days whenever possible.
Testing was halted for safety reasons and scientific validity when weather conditions were poor. Bottom temperatures were collected by probes attached to nets during the May 2001 and February 2002 testing periods. Soak durations were defined as the difference between the time when the setting of the nets began until the end of the hauling of the nets. On trips where the gear was set and not hauled, set times were recorded by the vessel captain. When only the time that setting ended was recorded, an estimate of the begin time was made by using other set durations for that vessel. When no set time was recorded, soak times were estimated using water temperatures collected by sensors attached to three of the four nets, if available. Durations were used to normalize catches to lb/hr.
Modified box-and-whisker plots were constructed for catch rates (lb/hr) of cod and yellowtail flounder separately above and below minimum landing size (MLS) and for winter flounder above MLS. Box-and-whisker plots give a visual representation of the distribution of the catch rates for each net by set. The box ends are defined as the first and third quartile of all observed catch rates for that net. The median is a solid line through the box. The mean is represented by a plus sign. The whiskers at either end extend to the most extreme data point, except where those points exceed 1.5 times the length of the quartile box. More extreme points are shown as solid dots. Box-and-whisker plots typically cannot be used for hypothesis testing.
To determine the appropriate statistical test, Bartlett's test for homogeneity of variance (Sokal and Rohlf 1995) was first employed; catch rates (lb/hr) were found to be heteroscedastic, making use of ANOVA or t-tests for catch comparisons inadvisable unless transformed. As an alternative to transforming data, the non-parametric randomization test (Sokal and Rohlf 1995, Rago 2004) was chosen to compare the catch of several species and size groups in each experimental net (lead-added and dual leadline) against the control, the standard flatfish net. Using this method, mean differences in catch rates were compared set-by-set for cod, yellowtail and winter flounders above and below MLS. Sets that had zero catches in all four designs were excluded from analysis.
The observed mean difference between the catch rate in each experimental string and the standard flatfish string for each set was compared to a distribution of 1000 or more differences determined from random assortments of the pool of catch data. The p-value was defined as the percentage of the mean differences more extreme than the observed difference. Length-frequencies of target species were pooled and compared between the standard net and each of the control nets using the Kolmogorov-Smirnov test (Sprent 1989). Sample sizes were adjusted for cluster effects following the methods of Pennington et al. (2001). One day of filming was conducted with an underwater remotely-operated vehicle (ROV) to examine the underwater profile of individual nets.