670 



Fishery Bulletin 94(4). 1996 



survey (Hickford and Schiel, 1995) and from gill nets 

 that were used for behavioral observations. By record- 

 ing the morphological features of the catch, along with 

 the form of entanglement, the primary factors that de- 

 termine the vulnerability of individual species to par- 

 ticular mesh sizes could be identified. Analysis of the 

 quantity and quality of the catch landed from gill nets 

 set for various periods should yield an optimum set 

 time that will maximize landings and reduce wastage. 



Materials and methods 



The gillnet catch analyzed in this study was pooled 

 from several experiments. Consequently, the result- 

 ing sampling design is not orthogonal. The netting 

 was done on rocky reefs around the Kaikoura Penin- 

 sula on the east coast of New Zealand's South Island 

 (42°25'S, 173°42'E) from 8 January 1993 to 26 Feb- 

 ruary 1993. The nets (Table 1) were set from a 6- 

 meter runabout and hauled in by hand. Each net was 

 set in a random direction and the ends were anchored 

 with weights and marked with surface buoys. At least 

 10 meters separated any two nets. The nets were set 

 on the bottom at depths ranging from 3 to 15 meters 

 and for periods of 11-17 hours. At all sites the benthic 

 habitat type had been described (Hickford and Schiel, 

 1995) and the fish populations had been surveyed 

 with visual transects by divers immediately before 

 the nets were set. At the end of all sets, the nets were 

 placed in bins and brought back to the laboratory with 

 fish still entangled in the mesh for analysis of the catch. 

 The combined catch of 114 net sets of three mesh sizes 

 over a single known habitat type (rocky pinnacles, 

 mixed algae [Hickford and Schiel, 1995]) was analyzed. 

 As each fish was removed from the net, its species 

 and fork length (mm) were recorded as was the 

 method by which each fish had become trapped in 

 the mesh. If a fish was held by the mesh encircling 

 its body between the posterior edge of its operculum 

 and the base of its pectoral fin, it was determined to 



have been "gilled." If the mesh encircling the body 

 was posterior to the base of the pectoral fin, the fish 

 was determined to have been "wedged." If a fish was 

 held because mesh had snagged an appendage, such 

 as the fins, spines, teeth, or maxilla, or if the fish's 

 struggling had simply enveloped it in the mesh, it 

 was described as "tangled." Careful handling of the 

 nets resulted in very few "drop-outs" from the net. 

 However, any fish that were loose in the net were 

 excluded from subsequent entanglement analysis. 



Entanglement data were collated for each species 

 in each mesh size. Because the species composition 

 of individual net sets was so variable and because 

 many species were caught only in a small proportion 

 of sets, the analyses of entanglement data were re- 

 stricted to the five most commonly caught species. 

 This produced a 5x3x3 contingency table, in which 

 the number of fish caught were categorized accord- 

 ing to species, mesh size, and entanglement mode. 

 This table was analyzed by using a log-linear model 

 that required thirteen iterations for the G-value to be 

 minimized ( I AG I < 0.001; Sokal and Rohlf, 1981). The 

 odacid Odax pullus was the only species caught in large 

 enough numbers across most net sets for individual 

 statistical analysis of entanglement to be done. 



The duration of each net set was recorded. Two set 

 times were chosen for analysis: a 6-h daytime set 

 from late morning to late afternoon and a 15-h night 

 set from late afternoon to early morning. The num- 

 ber offish and number of species caught during each 

 set were compared. A comparison of the capture rates 

 of common species was also made between day and 

 night sets and between mesh sizes. Each fish caught 

 was given a condition index according to the degree of 

 damage it had sustained while in the net (Table 2). 



Results 



The 114 net sets caught 1,165 fish from 14 families 

 (Table 3). The odacid Odax pullus (46% of the total 



