Table S. --Subplot portion of a split-plot analysis of variance of catch of juvenile sockeye 

 salmon (age and I combined) in tow nets in three lake basins of the Naknek system, 1962-64. 

 Depth of tow (shallow and deep) and depth- time interaction are tested. 



!No 10- to 20-ft tows made in midperiod in 1962. 

 2 No samples from area N-6 in late time period. 

 *lndicates 101 level of significance. 



periods each season (pre-26July, 27 July to 10 August, 

 after 10 August — the main treatment effect), and the 

 various number of areas within each lake (the block 

 effect). Because of unequal numbers of observations 

 per cell, the analysis was done with untransformed 

 data consisting of one observation per cell — the mean 

 for the area-time period-depth. In only two instances 

 did a significant difference appear in the subplot 

 treatments involving depth of tow (Table 5), i.e., there 

 were no consistent significant differences in catches of 

 juvenile sockeye salmon in surface versus deep tows. 

 Because of the indicated lack of difference between 

 surface and deep tows in the lakes with the largest 

 catches and the most sampling. I have assumed that 

 the surface catches in 1961 reasonably represent the 

 abundance in the surface to 20-ft depth. Pella (1968) 

 did not find a significant difference in abundance of 

 juvenile salmon with depth in Lake Aleknagik. 



A two-way analysis of variance among areas and 

 times within lakes of average catches of juvenile sock- 

 eye salmon in tow nets was made for those lakes with 

 the most useful data — Coville Lake, Iliuk Arm, and 

 South Bay (Table 6). These lakes had the most sam- 

 ples and usually had the largest catches and the 

 greatest changes in abundance. Only averaged paired 



tows (one shallow and one deep for the same night and 

 area) were used in the analysis. The analysis was done 

 with the same untransformed data as in the split-plot 

 analysis. However, the error terms used for the F tests 

 were obtained by using the individual catches in each 

 area-time cell (resulting in more degrees of freedom 

 than in the split-plot analysis) as suggested by Scheffe 

 (1959). 



Statistically significant effects of areas, time, and 

 area-time interaction on the abundance of juvenile 

 sockeye salmon occurred in less than half the tests (17 

 of 39). Although the effects of areas and times were 

 frequently not statistically significant, the differences 

 observed were usually consistent from year to year 

 and agreed with the observed changes (such as inter- 

 lake migrations) and with the observations that num- 

 bers of age fish increase during the first part of each 

 season and decrease later each season. I have, there- 

 fore, presented the quantitative results of the tow net 

 sampling in general summaries consisting of bar and 

 line graphs. 



Trends in Abundance for the Entire System 



Some stocks of juvenile sockeye salmon in the 

 Naknek system begin to migrate oceanward as soon as 



11 



