be shown by the number and distribution of color 

 combinations of tagged fish on the spawning 

 grounds. Theoretically, with no segregation the 

 frequency distributions of color combinations 

 occurring on particular spawning grounds should 

 be similar to those of the overall tagged popula- 

 tion. 



Because of the size and depth of many spawning- 

 areas we could not recapture or see all tagged fish 

 present. If the likelihood of recapture or observa- 

 tion were equal for all tag color combinations, 

 spawning ground recoveries should have reflected 

 the relative proportion of tags from each day of 

 tagging. Minor disagreement between actual and 

 expected tag recovery proportions could have 

 occurred as a result of sampling variation and 

 would not necessarily indicate segregation. On 

 the other hand, any substantial disagreement 

 could be taken as a sign of segregation. 



Actual tag recoveries were compared with the 

 expected recoveries for major spawning grounds 

 of the Naknek system in two ways: (1) By 

 graphical comparison of the pattern of weighted 

 spawning ground recoveries by day of tagging with 

 the pattern of the daily escapement passing the 

 tagging site, and (2) by statistical comparison 

 using contingency x 2 analysis (Snedecor, 1956) 

 to test the hypothesis of like tag recovery distribu- 

 tions between spawning areas. 



It would have been desirable to tag a given 

 proportion of the run each day so that individual 

 spawning populations would have been tagged in 

 proportion to their daily abundance. This would 

 have made it possible to compare directly the 

 spawning ground recoveries for each day of tagging 

 with the appropriate daily escapement size. This 

 could not be done, however, because the size of 

 the daily escapement, which was counted upstream 

 from the tagging site, was unknown before each 

 day's tagging. 6 Salmon in the escapement were 

 counted daily from observation towers located 

 on each bank of the river 1% miles upstream from 

 the tagging site (fig. 2). Counting migrating 

 sockeye salmon from towers has been proved a 

 reliable method of assessing the size of daily 

 escapements in Bristol Bay 7 (Becker, 1962). 



The number <>f recoveries was dependent on the 

 number of fish tagged (/-0.949,d.f. = 21, P (0.001)), 

 but because a different proportion of the run was 

 tagged each day, recoveries were weighted to make 

 them directly comparable to the daily escapement 

 size. Weighting was accomplished by adjusting 

 the number of fish tagged each day to a standard 

 proportion of the daily escapement. The standard 

 selected was the proportion tagged (0.58 percent) 

 on July 5, the day when the greatest number of 

 fish were tagged (table 1). Weighted spawning 

 ground recoveries for each day's tagging were ob- 

 tained as follows: 



W= 



Q.5SA 



6 Estimates of the daily escapement size were provided by the Alaska 

 Department of Fish and Game (table 1). 



' Star! of the Administration of Alaska Commercial Fisheries. 195f>. 

 Progress report and recommendations for 1957, 34 pp. [Processed.] 



Where : 



W= Weighted recoveries 

 A = Actual recoveries 

 0.58 = Proportion of the escapement tagged on 

 day of greatest tagging 

 X= Proportion of the run tagged on day in 

 question 



For any spawning area, the proportion of adjusted 

 recoveries for each day of tagging is thus directly 

 comparable to the daily escapement. I obtained 

 the graphical pattern of tag recoveries for selected 

 spawning areas by plotting the percentage of the 

 total adjusted recoveries by day of tagging and 

 compared this pattern with that of the daily 

 escapement, which was obtained by plotting the 

 percentage of the total escapement on each day of 

 tagging. The amount of disagreement or simi- 

 larity between the two graphs indicated the extent 

 of segregation. 



We recovered adequate tags to make this 

 comparison for most of the known major spawning 

 grounds of the Naknek system; however, for 

 some areas the difficulty encountered in sampling 

 the spawning populations resulted in insufficient 

 recoveries to make comparisons. These included 

 the beach spawning areas of Grosvenor Lake, 

 the tributary streams of the Savonoski River, 

 and a "suspected" beach spawning area in Iliuk 

 Arm of Naknek Lake (fig. 2). For most of the 

 small lateral tributary streams with small spawn- 

 ing escapements, tag recoveries were so few that 

 streams had to be grouped to provide sufficient 

 recoveries for comparison. 



Some problems were encountered in identifying 

 tag color combinations. In shallow-water streams 



SALMON SPAWNING POPULATIONS IN NAKNEK RIVER 



467 



