about 0.3 to 1.2 mps. The width of the stream was 

 divided into four equal segments and the middle 1.2 m 

 of each segment was sampled with a 1.2-m-wide fyke 

 net. Two sampling schemes were used: (1) regularly, 

 the site that passed the most water (and caught the 

 most fish) was fished as an index; and (2) at intervals, 

 based on observed changes in the character of the mi- 

 gration, nets were fished at the four sites following a 

 modified Latin-square design (Cochran and Cox, 

 1957) so that the number offish migrating in the entire 

 stream during the period could be estimated. In the 

 Latin-square scheme catches were classified accord- 

 ing to site, time of day, and days — factors considered 

 to have the greatest influence on variability of the in- 

 dividual catches. Estimates of the numbers offish that 

 migrated through the sites were obtained by fitting a 

 multiplicative model 8 to the observed catches, esti- 

 mating numbers migrating through unsampled site- 

 time of day-day strata from parameter estimates of the 

 model, and then summing over all strata (sampled and 

 unsampled), and finally expanding this total to account 

 for the proportion of the river sampled by the nets. 



The estimated outmigration of juvenile sockeye 

 salmon during the Latin-square and the number caught 

 in the index site during the same period (the index site 

 was fished continuously) were used to estimate the 

 portion of the total migrants captured at the index site. 

 This figure, the index catch expansion factor, is used 

 to estimate the number migrating when only the index 

 net was fished. 



The estimate of the numbers of juvenile salmon mi- 

 grating out of Coville Lake during the period sampled 

 each year is based on a combination of the Latin- 

 square estimates and the index catches. For periods 

 when migration estimates from Latin-square sampling 

 were made, the daily migration was estimated by di- 

 viding the expanded Latin-square estimate by the 

 number of days involved; when only index sampling 

 was done, the daily migration was estimated by ex- 

 panding the catch in the index net by the index catch 

 expansion factor. The index catch expansion factor 

 was used up to the halfway date toward the next 

 Latin-square period and then the factor for the next 

 period was used. When no sampling was done for a 

 day or days, the average of the preceding and the fol- 

 lowing estimates was used. The estimated total migra- 

 tion for the season is the sum of the estimates for each 

 day. 



In Grosvenor River juvenile salmon were sampled 

 with fyke nets at two general locations. Recently 

 emerged fry were sampled with the 1-m fyke nets in 

 the shallow water along shore near the outlet of the 

 lake. Older fish were sampled with the 1.2-m nets 

 which were attached to a cable at a point about 2 miles 

 below Grosvenor Lake, just above the island in Gros- 



"The model was developed by Jerome J. Pella of the National 

 Marine Fisheries Service, Auke Bay Fisheries Laboratory, and a 

 full description and analysis of the model and its application will be 

 published soon. 



venor River. The river was about 78 m wide and 0.6 to 

 2 m deep where the cable crossed and the water veloc- 

 ity was from 0.6 to 1.2 mps where the fyke net was 

 fished. 



For purposes of analysis, I have summarized the 

 data by 10-day intervals; 1 August was arbitrarily 

 selected as the starting date. 



Migration from Coville Lake to 

 Grosvenor Lake 



1961. — In 1961 the sampling of juvenile sockeye 

 salmon migrating down Coville River was exploratory 

 and intended mainly to determine the timing and the 

 age classes involved. The sampling was done in two 

 periods — early (18 May to 1 1 June) and late (27 July to 

 10 September). During the early period the small 

 (1-m-square) fyke nets and seines were used and both 

 age and age I fish were caught. During the late 

 period the fishing was mainly with the 1-m-square fyke 

 net and mostly age fish were caught. 



Because so many age fish appeared to be involved 

 in the summer migration, I have made an order of 

 magnitude approximation of the number that migrated 

 from Coville Lake to Grosvenor Lake from 27 July to 

 10 September. Knowledge gained in subsequent years 

 makes the following assumptions reasonable: (1) the 

 fyke nets caught 4% of the juvenile sockeye salmon 

 migrating down Coville River during the periods 

 fished (based on portion of river sampled); (2) the rate 

 of catch during the time fished each day was typical of 

 the whole day; and (3) the catch per day can be aver- 

 aged for 10-day periods. Using these assumptions, I 

 estimated that in 1961 about 5.6 million age sockeye 

 salmon migrated from Coville Lake to Grosvenor 

 Lake between 22 July and 10 September (Table 12). 



1962. — In 1962 the migration of sockeye salmon 

 down Coville River was sampled from 29 May to 15 

 September and more systematically than in 1961. A 

 cable was installed across the river near Grosvenor 

 Lake where the river was about 24 m wide. The four 

 6-m sites were established on the cable and the large 

 fyke nets were fished in the middle of each site. Nets 



Table 12. --Estimated number of age sockeye 

 salmon migrating from Coville Lake to Grosvenor 

 Lake by 10-day periods between July 22 and 

 September 10, 1961. 



Fish migrating 

 Period each period 



July 22-51 

 August 1-10 

 August 11-20 

 August 21-30 

 August 31-September 10 

 Total 



871,000 



1,058,000 



3,459,000 



96,000 



71,000 



5,555,000 



26 



