BRISTOL BAY AND ALASKA PENINSULA SALMON STATISTICS 61 



1903, the Kvichak catch continued to increase, and the trend reached its peak in 

 1915 and 1916, since when there has been a material drop, although this is by no 

 means as marked as in the case of tiio Nushagak catch. The trend at Egegilc has 

 been much the same as that in the Kvichak, with the exception of a drop in the 

 years 1902 to 1906. The peak of the Egegik trend came in 1916, since which time 

 the trend has slowly but unmistakably declined. The Ugashik trend is quite different, 

 rising to its highest peak m 1901 and then falling gradually until 1909, since which 

 time it showed a gradual l)ut steady recovery, which was broken sharply in 1925. 

 Without much doubt the early drops in the trends at Ugashik and Egegik were due 

 to reduced mtensity of fishing, as shown by corresponding decrease in the number of 

 gill nets. For some reason the early development of the fishery in these two districts 

 was arrested for a time but was resumed later. 



It is apparent that there has been a certain amount of independence in the trend 

 of the red salmon fisheries in the four districts. With the possible exception of the 

 Ugashilx, however, they all show a present tendency to drop. In the case of the 

 Ugashik it would appear, from the raw data presented in Figure 6, that this stream, 

 too, is entering a period of decreased productivity. While the depletion of the 

 Nushagak is much more pronoxmced than in the other districts, it is quite evident 

 that the red-salmon catch in the entire Bristol Bay region is distinctly on the decline. 



We will now examine the short-time fluctuations, as distmguished from the long- 

 time fluctuations, or "secular" changes indicated by the trends. We have discussed 

 above the general miportance of a knowledge of the character of these short-time 

 fluctuations. To be more explicit, it is important that we know (1) whether or not 

 there is any regularity in these short-time changes — whether they occur in cycles or 

 not; (2) what the interval of the cycles is, if the changes are cyclic in character; 

 (3) whether there are sudden or progressive changes in the nature of the fluctuations, 

 and (4) whether there is any correlation in the fluctuations in different streams. 



Our interpretation of the facts disclosed by an analysis of the fluctuations will 

 depend, as m the case of any statistical analysis of such data, upon an understanding 

 of the biological and economic factors that may affect them. We are concerned here 

 chiefly with the discovery of the facts about the fluctuations in the catch of salmon 

 and must leave the consideration of the true causative factors for future treatment. 



The study of the short-time fluctuations has been based on the percentage 

 deviation of the yearly catch from the trend or moving average by fives. In this 

 method, adequately described by Chaddock (loc. cit.), the percentage deviation for 

 any year is the algebraic value of the catch minus the trend, divided by the value 

 of the trend. When the catch is greater than the trend, the deviations have a 

 positive value and a negative value when less than the trend. Such treatment does 

 two important things to our data — it removes the effect of the long-time, secular 

 fluctuations, which might accentuate or destroy any correlation that might exist 

 between two series of data, and it makes it possible to compare more fairly and 

 more directly the fluctuations at very different levels of abundance, whether in 

 different streams or in the same stream at different periods. For example, if the 

 trend in one series of data was at 1,000,000 and in another series of data was at 



