376 ZEEMAIM AND GRUNEWALD 



univciriate analysis of variance. The models were designed to factor 

 out effects of repetitive sampling to estimate true differences. The 

 statistical results seem contradictory, however. When the three size 

 fractions were analyzed simultaneously, power-plant operations 

 appeared to be having an effect, however small. Differences among 

 samples from the same station were also found to be significant when 

 the plant was in operation. This indicates patchy distribution of 

 phytoplankton. When the sums of the size fractions were subjected 

 to ANOVA techniques, station differences were once again seen, but 

 sample effects were not observed. The unfractionated productivity 

 measurements, on the other hand, showed no differences among 

 stations. Sample effects in the last analysis could not be tested. 



One explanation for the observed results is that perhaps the 

 size-fractionation method produces artifacts. The sums of the 

 fractions and the unfractionated totals are not the same in any given 

 experiment. It is, therefore, not unreasonable to assume that this 

 could influence the outcome of statistical tests. Holmes and 

 Anderson (1963) and Lasker and Holmes (1957) showed how 

 variable retention can be on different filters. Evidence that size 

 fractionation is responsible for much of the variance in productivity 

 among stations comes from the ANOVA results. Analysis on the 

 unfractionated totals showed no significant differences among 

 stations, yet the ANOVA on the sums of the fractions showed that 

 differences did exist. If the methodology were as precise as desired, 

 the results would agree. 



According to Strickland and Parsons (1972), the precision of the 

 radiocairbon technique is ~7% for two replicates at the 

 1.5 mg C m~^ hr~' level and ~8% at the 25 mg C m~^ hr~^ level. 

 For six replicates the precision is increased to about 5 and 4% for the 

 two levels, respectively. From the coefficients of variation in this 

 study, it was evident that any attempt to characterize differences in 

 productivities among the sites was going to be complicated by large 

 variations, possibly caused by patchiness and analytical error. 



Differences in productivity rates between the precondenser 

 (FRB) and the postcondenser stations (KD) were small when 

 averaged over the entire year. Percent differences (in comparison to 

 FRB) were -9.08, +5.32, -12.12, -2.62, and -9.27% for the 

 small-, medium-, and large-size fractions, the unfractionated total, 

 and the sum of the fractions, respectively. Even though these values 

 showed a general decrease at KD, we do not feel they are biologically 

 significant in view of the methodologies involved in their determina- 

 tion. The slight increase observed for the medium-size fraction, if 

 real, was possibly caused by breakage of filamentous or colonial algae 



