FISHERY BULLETIN: VOL. 69, NO. 4 



of 5 days during the peak of the fishing season. 

 This procedure eliminated some of the variability 

 present at the beginning or the end of the fishing 

 season due to irregular entries or departures of 

 the salmon. The choice of 5 days was made in 

 order to avoid too complicated a scheme, and 

 often more than half of the total Nushagak catch 

 was taken during the time period considered. 



In this two-way classification with two rows 

 corresponding to drift net and set net and five 

 columns corresponding to the time periods, the 

 catch in 1 day and by a given type of gear is: 



Cii = fij  n  Nj  e,j , 



where fa = the number of fishing units of 

 type i operated on day ;, 



n = the coefficient of catchability of 

 gear type (' for all size groups, 



Nj = the average stock of salmon en- 

 countered by the gear on day j, 

 and 



en = error term. 



If r . . is the coefficient of catchability of a unit 

 of a theoretical average of all types of gear, 

 one can write ai = n/r . . Similarly, if the av- 

 erage stock size encountered by the gear during 

 the entire period is defined as N . ., one has 

 Pj = Nj/N . . Finally, the error term was con- 

 sidered log-normal (Beverton and Holt, 1957). 



The random variable Yn — log {dj/fn) can 

 be written then as 



Yij = m + ai + bj + e,j. 



Since we have only two types of gear, log r-y — 

 log Vi ~ a2 — ai. An estimate of ai can be ob- 

 tained directly from a linear hypothesis program, 

 such as BMD 05V (Di.xon, 1965) , under the con- 

 straint tti + a2 = 0. The results expressed 

 as arithmetic ratios are listed in Table 5. 



Fishing Power of 5J^-Inch Gill Nets 



Two size groups of fish iiredominate in all 

 Bristol Bay sockeye salmon fisheries (Mathisen, 

 Burgner, and Koo, 1963). The 3-ocean fish 

 measure on the average from 5 to 6 cm longer 

 than the 2-ocean fish, and the males of both size 

 groups are between 2 and 3 cm larger than the 

 females. 



Between years there are i)ronounced differ- 



ences in the proportion of 2- and 3-ocean fish and, 

 to a much smaller extent, in the sex ratio of the 

 total runs. Since during the middle part of the 

 Nushagak fishery considered here, the mesh size 

 of the gill nets remained stable at 51/2 inches, 

 the total fishing mortality generated by one unit 

 of gear changed from year to year primarily 

 with changes in the relative proportion of 2- 

 and 3-ocean fish and males and females. Con- 

 sequently, the coefficient of catchability must be 

 determined by length or age groups, and sep- 

 arately for males and females. There are only 

 5 years, 1946-1950, with records of catch and 

 escapement when sailboats were used together 

 with linen gill nets. Conversion to powered fish- 

 ing boats was largely accomplished by 1954, al- 

 though a shift in boat types continued. At the 

 same time nylon gill nets came into universal use. 

 Added to these changes were modifications of 

 boundary lines of the fishing districts. There- 

 fore, the rate of the present-day fishing of the 

 gear in Bristol Bay is not comparable with that 

 which prevailed during the middle period of the 

 Nushagak fishery. 



Data on catch and escapement and the cor- 

 responding length-frequency distributions for 

 Nushagak from 1946-1950 are available (Math- 

 isen et al.. 1963). The escapements were esti- 

 mated visually and may not be too accurate. 

 But in 1 year, 1946, when an independent esti- 

 mate could be made from a tagging experiment, 

 the correspondence was remarkably great 

 (Mathisen, 1969) . Effort during the same years 

 is listed in Taljle 5. On the assumption that set 

 net effort can be converted into drift net effort 

 and that all units of gear were fishing simul- 

 taneously on the same stock, it is a straightfor- 

 ward matter of computing the coefficient of 

 catchability for each centimeter group and sep- 

 arately for males and females from the expres- 

 sion on page 754 used in reverse. 



There were rather large year-to-year varia- 

 tions; therefore the following smoothing process 

 has been applied to the data. An arithmetic 

 mean value for each centimeter grou]) was found 

 for the 5 years considered. A moving average 

 of 5's of these arithmetic means provided the 

 final values in the selection curves in Figure 4. 

 The (lip in the selection curve for males is con- 



756 



