Lazzari: Dynamics of larval fish abundance in Penobscot Bay, Maine 
83 
and temporal patterns of the larval fish assemblage and 
environmental data (Gauch, 1982). Varimax rotation was 
performed on factors from the PCAs in both cases because 
rotated solutions tend to extract components that cor- 
relate highly with a smaller number of variables than 
unrotated PCAs (Stevens, 1986). This step aided in the 
interpretation of the factors. Rotated factor scores from 
the PCA were used as the dependent variables in the 
MAN OVA because they were uncorrelated, thus satisfying 
the assumption of independence for parametric statistical 
tests. 
Environmental data were log-transformed to satisfy 
assumptions of univariate normality and then summa- 
rized by PCA. The resulting factor scores were grouped 
by month, and means were compared by MANOVA to as- 
sess the null hypothesis that environmental variables did 
not differ among months. This analysis was followed by a 
Tukey-Rramer multiple range test to detect which month- 
ly means differed. This test controls for increases in the 
type-I error rate associated with unequal sample sizes 
(Day and Quinn, 1989). Temporal changes in environmen- 
tal variables were interpreted by evaluating product mo- 
ment correlations of the environmental variables and fac- 
tor scores from PCA. I considered only variables that were 
significantly correlated with an individual factor in inter- 
preting that factor (significance levels were adjusted for 
multiple comparisons, where p '= 1 - ( 1 - p) 1,k and k equals 
the number of comparisons [Sokal and Rohlf, 1981]). The 
results of these analyses were used to discriminate among 
months. 
The analysis of larval fish assemblage composition was 
similar to that performed for environmental data. However, 
fish taxa not present in 5% of the samples were eliminat- 
ed to reduce the influence of rare taxa (taxa that were ex- 
cluded were present in five of 102 samples in 1997 and two 
of 40 samples in 1998) as potential outliers (Gauch, 1982). 
A matrix of angularly transformed [arcsine of the square 
root of a proportion (Sokal and Rohlf, 1981)] relative abun- 
dances for the m = 12 taxa in n = 102 samples in 1997, 
and the m = 13 taxa in n = 40 samples in 1998 were con- 
structed and summarized by PCA. Varimax rotated factor 
scores generated from PCA of the relative abundance ma- 
trix were compared among months by using MANOVA 
(unbalanced design). The null hypothesis was that mean 
factor scores (and taxonomic composition) did not change 
among months. This was followed by a Tukey-Rramer test 
of significantly affected factors to identify the extent to 
which monthly means differed. 
Factor scores were then grouped by location and served 
as the dependent variables in MANOVA to test for dif- 
ferences in sample location (upper, mid, or lower bay). 
The null hypothesis was that mean factor scores (and 
taxonomic composition) did not change among locations. 
A Tukey-Rramer test was performed to identify location 
mean factor scores that differed significantly. Temporal 
and spatial changes in the taxonomic composition of lar- 
val fish assemblages were then interpreted by identify- 
ing which factors were affected by month and location 
treatments and by evaluating product-moment correla- 
tions of angularly transformed relative abundance data 
Table t 
Mean (± standard error) environmental variables for the 
entire study period, reported by month for 1997 and 1998. 
Temperature (°C) and salinity recorded as mean for top 
20 m of the water column and volume displacement (Pvol) 
in mL of unidentified plankton per 100 m 3 . 
Month 
Temperature 
Salinity 
Depth (m) 
Pvol 
1997 
April 
3.37 (0.13) 
30.56(0.13) 
74.0 (9.4) 
17.3 (1.7) 
May 
5.49 (0.12) 
30.24(0.08) 
58.0 (11.8) 
4.2 (0.5) 
June 
8.52 (0.22) 
30.44 (0.05) 
46.8(8.2) 
13.7 (2.1) 
1998 
March 
2.41 (0.01) 
29.90(0.05) 
38.8(18.8) 
34.4 (0.5) 
April 
4.23 (0.14) 
30.14(0.05) 
56.5 (10.7) 
25.5 (0.9) 
and significant PCA factor scores (i.e. “loadings” on fac- 
tors significantly affected by time and location; Pielou, 
1984). The relationships of the factor scores from PCA 
of the relative larval fish abundance matrix and the log- 
transformed environmental variables were examined by 
CCA. 
Standard lengths, or notochord lengths, from 1997 and 
1998 were compared among cruises (sample dates) for 
each species by using Kolmogorov-Smirnov two-sample 
tests. The null hypothesis was that size distribution did 
not differ significantly between sample dates. 
Results 
A total of 102 plankton net samples and CTD measure- 
ments were collected during biweekly sampling between 4 
April and 25 June 1997 in Penobscot Bay. Sixteen stations 
were sampled on five of the seven sampling trips; how- 
ever, inclement weather (wind speeds >25 knots) during 
the first week of April (4-7, 971) and from late April to 
early May (29 April-1 May 97III) limited the number of 
stations that could be sampled during these trips to 10 
and 12, respectively. The number of cubic meters of water 
filtered per 20-minute tow varied from 137.5 to 755.7 m 3 
depending on station location and sample date. The total 
volume of filtered in 1997 was 44,622 m 3 . 
In 1998, 40 plankton net samples and CTD measure- 
ments were collected during approximately biweekly sam- 
pling between 18 March and 30 April 1998. Ten stations 
were sampled on each of the four sampling trips. The num- 
ber of cubic meters of water filtered during each 20-minute 
tow varied from 54.2 to 582.5 m 3 depending on station lo- 
cation and sample date. The total volume of water filtered 
in 1998 was 12,459 m 3 . 
Environmental data 
Station depth ranged from 20 m at mean low water at 
the upper bay station R4 to 91 m at station 02 (Table 1). 
