within the range of variability in the bay. Phosphate concentrations in the 
microcosms tended to be higher than in the bay during the latter part of 
September and October. Comparison between the chlorophyll a graphs in 
Figure 24-5 and in Figure 24-3 indicates that the timing of the phytoplankton 
blooms may have been different in the two years. Nevertheless, the overall 
behaviour of the tanks and the bay is difficult to distinguish. Figure 24-5 also 
gives some indication of the patchiness and other variability apparent in 
Narragansett Bay. 
The variability and scatter of the magnitudes of individual measurements 
both in the bay and in the tanks show that it is difficult to distinguish tank 
behaviour from bay behaviour by using comparisons of single variables. In 
addition, various of the single variables are correlated, making statistical 
analysis of single variables less rigorous. Accordingly, multivariate statistical 
comparisons were attempted. 
A stepwise discriminant analysis (11) was performed on a weekly data set 
from the microcosms and bay input water collected from August 17 to 
December 6, 1976, to observe the replicability of the microcosms. The first 
two axes of this analysis explained 84 percent of the variation in the data set 
(Figure 24-6). The first five variables in the order of their importance 
(nitrate-nitrite, phosphate, ammonia, silicate and zooplankton) accounted for 
99 percent of the variation explained in the first two axes (Table 24-5). 
Chlorophyll concentration explained so little of the variation (less than 1 
percent) that the analysis did not include it. Generally microcosms 1,5,6,7, and 
8 were more similar to each other while microcosms 2,3,4, and 9 and bay input 
water showed a greater individuality. If the microcosms were exactly similar, 
10 percent of each group would be classified into itself and each of the other 
nine groups. In fact, microcosm 3 classified 47 percent to itself, microcosm 9 
classified 53 percent to itself and bay classified 44 percent to itself, indicating 
that these microcosms and the bay had the most individuality (Table 24-5). 
The individuality of microcosms 3,9 and bay, as indicated mainly by 
differences in nutrient concentrations in the discriminant analysis, was not 
reflected in phytoplankton species as analyzed by correspondence analysis (2) 
(Figure 24-7). The nine microcosms were not distinguishably different in their 
species composition from August to December. Initially all microcosms and the 
bay were tightly clustered on the lower right hand side of Figure 24-7. A 
bloom in November was reflected in a greater variability in microcosm location 
and species location (left hand side of Figure 24-7), but at no time was there a 
characteristic species or species group which caused clearly different 
microcosm location. 
371 
