378 
Fishery Bulletin 95(2), 1 997 
<D 
U) 
< 
1 
1965 1967 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 
Year class 
Figure 2 
Age-2 Atlantic herring stock size estimates (x 10 6 ) for the 1965 through 1989 year 
classes of the coastal Atlantic herring stock. 
368°) for each month from 1965 to 1989. The num- 
ber of storm days, determined by daily average wind 
speeds greater than 5 m/s from each direction were 
also tabulated monthly. The number of major storms, 
defined as the number of periods of three or more 
consecutive days with mean wind speeds greater than 
5 m/s were tabulated monthly. The number of major 
storms, defined as the number of periods of three or 
more consecutive days with mean winds in excess of 5 
m/s were tabulated monthly and for the August-De- 
cember period. The number of Lasker events, defined 
as the number of consecutive days when mean winds 
were less than 5 m/s for four or more consecutive days 
(Pauly, 1989) were tabulated monthly for August-Sep- 
tember and for the August-December period. 
We examined SST, percent sky cover, relative hu- 
midity, air temperature, solar radiation, and wind 
speed and direction at several locations in the Gulf 
of Maine to determine the similarity of environmen- 
tal conditions within the Gulf. Wind data recorded 
at the Portland, ME, and Boston, MA, airports every 
3 hours from August through December 1980-89 
were compared for direction with contingency tables. 
Winds were converted from vector variables into one 
of four directions from compass headings of north- 
easterly, southeasterly, southwesterly, and north- 
westerly for each 3-h period. Monthly means of envi- 
ronmental variables including percent sky cover, rela- 
tive humidity, air temperature, solar radiation, and 
wind velocity recorded by the National Weather Ser- 
vice, Logan Airport, Boston, MA, 
were compared by using Pearson 
corelation analyses and /-tests 
with those recorded at the Port- 
land airport for the period 1961- 
90 (n=30). 
Similarity of historic seawater 
temperature data between several 
sites along the Maine coast, in 
Massachusetts Bay, and in waters 
off New Brunswick (Canada) were 
tested by using Pearson correlation 
analyses. Monthly and yearly 
mean sea surface temperatures 
recorded at Boothbay Harbor 
(BBH) were compared with simi- 
lar temperatures recorded in St. 
Andrews, New Bruswick, between 
1921 and 1969, in Eastport be- 
tween 1930 and 1971, Bar Harbor 
between 1947 and 1971, and in 
Portland between 1922 and 1971. 
The 25 years of environmental 
data and age-2 population esti- 
mates for the Gulf of Maine were 
partitioned into three clusters with the KMEANS 
cluster analysis procedure from SYSTAT (Wilkinson, 
1991). This analysis is used to divide a series of data 
into a selected number of clusters in order to reduce 
the within-group sums of squares to a minimum 
value. The independence of periods of low, medium, 
and high values of the environmental variables were 
tested against the observed versus the expected age- 
2 population estimates with 3x3 contingency tables. 
As a selection criterion to identify important vari- 
ables, we used a significant (P<0.05) result for the 
3x3 contingency tables. The existing low, medium, 
and high cells were then combined to form 2x2 con- 
tingency tables from these results. All environmen- 
tal factors of interest and the age-2 abundance esti- 
mates were tested again in four ways. To test the 
hypothesis that high estimates were associated with 
low values of an environmental factor, low and me- 
dium age-2 estimates and medium and high values 
of the factor were combined. To test the hypothesis 
that high estimates were associated with high val- 
ues of an environmental factor, low and medium age- 
2 estimates and low and medium values of the envi- 
ronmental factor were combined. To test the hypoth- 
esis that low estimates were associated with high 
values of an environmental factor, medium and high 
age-2 estimates and low and medium values of the 
factor were combined. To test the hypothesis that low 
estimates were associated with low values of an en- 
vironmental factor, medium and high age-2 estimates 
