NOTE Lazzari et al. : Physical environment and recruitment variability of Clupea harengus 
379 
and medium and high values of the environmental 
factor were combined. The 2x2 tables were tested 
for independence by using Fisher’s exact test because 
one cell often had zero observations (Zar, 1984). 
A (-test for unequal variances (Zar, 1984) was used 
to determine whether the mean environmental fac- 
tors differed between years of good and poor recruit- 
ment. The hypothesis that the mean environmental 
factor was higher (or lower, depending on the rela- 
tionship with the age-2 estimates) during higher than 
expected recruitment years (1966, 1970, 1983, 1988, 
n= 4) than during lower than expected recruitment 
years (1971, 1972, 1974, 1978, n- 4) was tested. 
Results 
Our examination of SST, percent sky cover, relative 
humidity, air temperature, solar radiation, and wind 
speed and direction revealed widespread coherence 
at several locations in the Gulf of Maine. The null 
hypotheses of independence were rejected in all 
months for the wind directions recorded every 3 hours 
between Portland and Boston, August through De- 
cember 1980-89 (P<0.001, n>2,200). Wind direction 
at both locations showed a definite seasonal trend 
from May into September when more southerly winds 
predominated. Monthly mean wind speeds for the 
period 1961-90 between Portland and Boston were 
always significantly greater at Boston in all months 
((-test, P<0.001, n- 30), except in January and Octo- 
ber. In addition, significant Pearson correlations 
(P<0.001, n=30) were found for all monthly means of 
solar radiation (r 2 =0.80), total sky cover (r 2 =0.79), 
air temperature (r 2 =0.95), relative humidity 
(r 2 =0.68), and precipitation (r 2 =0.91) between Port- 
land and Boston for the same period. 
The null hypothesis of independence was rejected 
for the age-2 Atlantic herring abundance estimates 
and ten environmental factors with 3x3 contingency 
tables. These environmental factors were November 
storms, March sunshine, and October sea surface 
temperature, October and first quarter (January- 
March) barometric pressure, December and August- 
September Lasker events, the number of days of 
southwesterly winds in September, the total num- 
ber of days of southeasterly winds between August 
and December, and the number of storm days with 
southeasterly winds in November. 
Five environmental factors were associated with 
either high or low age-2 abundance in the 2x2 con- 
tingency table analyses (Table 2). Low abundance 
was associated with reduced sunshine in March and 
with fewer days of southeasterly winds from August 
to December. Associations with high age-2 abundance 
Table 2 
Probabilities of 2 x 2 contingency table (Fisher’s exact chi- 
square test) results for the environmental factors associ- 
ated with the age-2 Atlantic herring estimates, ns = not 
significant. 
2x2 contingency tables 
Low age-2 High age-2 
estimate estimate 
Environmental 
factor 
High 
factor 
Low 
factor 
High 
factor 
Low 
factor 
November storms 
ns 
ns 
0.020 
ns 
March sunshine 
ns 
0.012 
ns 
ns 
September SW wind 
ns 
ns 
0.002 
ns 
Aug-Dec SE wind 
ns 
0.005 
ns 
ns 
November SE storms 
ns 
ns 
0.026 
ns 
occurred with more November storms, with more 
days of southwesterly winds in September, and with 
more days of southeastern storms in November. 
Comparison of environmental factors (mean val- 
ues) between the four best and four worst recruit- 
ment years revealed only March sunshine and three 
southern wind direction factors were significantly 
different. The amount of monthly March sunshine 
and the number of days of southwesterly winds in 
September, the number of days of southeasterly 
winds from August to December, and the number of 
storm days with southeasterly winds in November 
were all significantly higher during good recruitment 
years than during lower recruitment years (Table 3). 
The four strongest year classes (1966, 1970, 1983, 
1988) were associated with greater than 50% of the 
possible March sunshine and were associated in 
eleven out of twelve cases with above average Sep- 
tember southwestern, August-December southeast- 
ern winds and November southeastern storms (Fig. 
3). In eight cases, conditions were greater than 50% 
above normal for the entire period. The only excep- 
tion during the four years of above average recruit- 
ment occurred in 1970 when August-December 
southeastern winds were extremely low. 
The four worst year classes (1971, 1972, 1974, 
1978) were associated with less than half of the pos- 
sible March sunshine and with average or below av- 
erage September southwesterly, August-December 
southeasterly winds and November southeasterly 
storms in 9 of 12 cases. In eight of these cases, condi- 
tions were at least 25% below normal for the entire 
period. However, two below average year classes 
(1979 and 1984) were produced despite the fact that 
all three of the significant wind factors were above 
