270 
BULLETIN OF THE BUREAU OF FISHERIES 
Two sources were available for air (and water) temperatures: The Climatological 
Data, published by the United States Weather Bureau since 1908, and daily air and 
water temperatures (unpublished) taken at Seward by the United States Coast and 
Geodetic Survey since June, 1925. 
Since only air temperatures are available for these months previous to 1926, a 
correlation was made between the air and surface water temperatures which were 
taken daily at the same hour at Seward, to determine the degree of relationship exist- 
ing between the two. These data, covering a period of 5 years, gave a coefficient of 
correlation of 0.93; to test the significance of this relationship t was computed, and 
was found to be 11.017, when a t of only 2.878 was equivalent to a probability of 
0.01, proving the correlation to be highly significant. This shows that air temperatures 
give an accurate index to surface-water temperatures, and that the use of air tempera- 
tures as an index to conditions on the spawning beds is justified. 
Unfortunately the series of air temperatures for Cordova, Latouche, and Seward 
are not complete. (Table 4.) Figure 5 shows, however, that the temperatures for 
these points are comparable from year to year. The temperatures for the Prince 
William Sound area, as given in the following discussion, are averaged from those 
which are available for each year from these three localities. The dotted line in 
Figure 5 shows the air temperatures for Seward, taken from unpublished United 
States Coast and Geodetic Survey data (from which 5.40 degrees have been subtracted 
to make them comparable to the United States Weather Bureau statistics). 
The means of the vertebral counts of the various year classes were therefore 
correlated with the average air temperatures for March, April, May, and June from 
Seward, Cordova, and Latouche by the formula: 
S(xy ) 
VS(* 2 ) • S(y 2 ) 
(Fisher, 1930) 
giving a coefficient of correlation of —0.85. The significance of this coefficient was 
tested by the method of Fisher for small samples (1930, p. 159) if n' be the number 
of pairs of observations: 
<=7i=rV^=2 
giving a 1 of 5.807 which has a probability much less than 0.01 (Fisher’s tables) 
showing that the coefficient of correlation —0.85 is definitely significant. 
Table 5. — Correlation of air temperature and average number of vertebrae in Prince William Sound 
Year 
1919 
1920 
1921 
1922 
1923 
1924 
1925 
1926 
1927 
Air temperat ure * 
40.07 
38.98 
41.83 
38.58 
42.23 
42.25 
40.34 
46.20 
40.63 
Mean of vertebrae 
52. 870 
53. 150 
52. 963 
52. 719 
52. 821 
52. 456 
52.784 
Number of specimens: 
Elrington Passage 
12 
11 
107 
73 
58 
70 
9 
343 
6 
Naked Island - 
4 
3 
45 
16 
3 
3 
2 
McClure Bay 
7 
74 
27 
32 
29 
1 
2 
Eshamy Bay - - 
5 
56 
14 
8 
Macleod Harbor _ 
4 
5 
142 
50 
599 
34 
1 
6 
16 
350 
7 
Port Chalmers 
2 
94 
3 
Zaikoff Bay 
2 
35 
7 
Port Fidalgo 
4 
5 
54 
Glacier Bay 
5 
5 
168 
19 
Shelter Bay 
1 
86 
9 
Point Helen 
70 
2 
Sleepy Bay 
2 
173 
24 
Total _ 
23 
20 
286 
135 
130 
263 
95 
1,972 
111 
1 Average for March, April, May, and June from Seward, Cordova, and Latouche. 
