362 



FISHERY BULLETIN OF THE FISH AND WILDLIFE SERVICE 



Table 2. — Size and timing of Kvichak red salmon escape- 

 ments at Igiugig, 1955 through 1959 



26 



30 



20 24 



28 



June 



July 



Aug. 



Figure 7. — Daily escapements of red salmon up the 

 Kvichak River, 1955 through 1959. 



spring, a few fish migrate upi'iver; adult red sal- 

 mon have been reported taken in gill nets on upper 

 Iliamna Lake as early as June 6. 



There has been no apparent correlation between 

 the duration and tlie size of the escapements. The 

 main part of the small 1955 and the large 1956 

 escapements each passed in 19 days, and neither the 

 start nor the end of the migration in tlie 2 years 

 was more than 3 days apart. 



FACTORS INFLUENCING VISUAL COUNTS 



To a large extent, the accuracy of the basic 

 sample counts depends on ( 1 ) individual counters, 

 (2) migration intensity, (3) weather conditions, 

 and (4) disrupted migration patterns. Once the 

 relationsliips of these factors to the counts are 

 recognized and understood, they can be taken into 

 consideration in improving the counting method. 



Errors Introduced by Individual Counters 



Counting errors are known to result from indi- 

 vidual differences in seeing, counting, and record- 

 ing numbers of passing salmon. In 1957, 32 counts 



of 5 minutes' duration were taken by 2 observers 

 counting simultaneously from the same tower. 

 One counter (A) participated in all counts, while 

 the other counter was taken from a group of three 

 men designated XYZ. The data obtained are 

 listed in table 3, where the difference in numbers of 

 fish counted by XYZ is expressed as a plus or 

 minus percentage of variation from A's count. 



The range of variations between the 5-minute 

 counts of A and one of the other three counters ex- 

 tends from — 22.1 to + 17.9 percent. By combining 

 two consecutive counts to form standard 10-minute 

 counts, the range is shortened from —7.8 to -M0.7 

 percent. Further reductions occur when the 

 counts are totaled and the differences calculated. 

 For the two groups of 5-minute counts, the total 

 differences are but -f3.5 and —5.3 percent, respec- 

 tively. The totals of all 32 counts differ by only 

 — 1.0 percent. Consequently, errors occurring in 

 counts between paired observers, operating under 

 a variety of observation conditions, tended to can- 

 cel out. Such counting errors, therefore, appar- 

 ently occurred randomly and probably did not 

 bias estimates of the escapement. 



Errors Associated With Migration Intensity 



As intensity of the migration increases, fish must 

 be tallied more rapidly. An increase in migration 

 intensity might indicate an increase in counting 

 errors. The comi^arative data (fig. 8), however, 

 show only a slight correlation between migration 

 intensity and percentage of counting variation, 

 even thougli greater variations in total numbers of 

 fish counted did occur. This indicates that counts 

 taken during increasingly heavy runs do not neces- 

 sarily inject an increasing number of errors into 

 the calculations. 



The five comparative counts with extreme varia- 

 tions exceeding a plus or minus 10 percent ap- 



