SIZE-FRACTIONATED PRIMARY PRODUCTIVITY 371 



the stations were sigiiificaiitly different. Samples also differed 

 significantly on 8 dates, but not always the same dates as those for 

 station effects. The four dates with no AT showed no significant 

 differences between stations or samples at o: = 0.05. 



In an attempt to see which variables were contributing most to 

 the observed MANOVA results, we carried out univariate three-way 

 mixed-model ANOVA tests on each size fraction separately. The 

 results of these tests, shown in Table 3, indicate that the 10- to 

 64-jum fraction is consistent over all stations and samples. The <10- 

 and > 64-/um fractions, however, form the basis of distinctions 

 between stations. This is not wholly unexpected in view of the fact 

 that the medium fraction, on the average, contributed only about 

 12% to the sum or productivity rates. Note also that sample variances 

 are significant for the < 10-^m fraction but not for the other 

 fractions. 



Univariate ANOVA tests were also run on the sums of the 

 fractions and the unfractionated totals of the productivity measure- 

 ments. A three-way mixed-model ANOVA was again used for the 

 sums of the fractions, but a two-way mixed model was required for 

 the unfractionated totals because of unequal cell numbers. The 

 results are shown in Table 4. Only the sample factor was not 

 significant for the sums of the fractions. The ANOVA results for the 

 unfractionated totals showed that stations were not significantly 

 different, however. 



To estimate relative variations within a station, we determined the 

 coefficients of variation. These values ranged from 0.118 to 224.2%, 

 with one exceptional value at 1217%. This considerable variability is 

 not totally unexpected since patchiness of plankton is well docu- 

 mented (e.g., Piatt, Dickie, and Trites, 1970; McAlice, 1970). 

 Furthermore, methodological problems are indicated by the fact that 

 the sums of the fractions and the unfractionated total are not equal 

 for any given experiment. This disparity has been documented in 

 other studies (e.g., Rodhe, Vollenweider, and Nauwerck, 1958; 

 McCarthy, Taylor, and Loftus, 1974). 



Chlorophyll a 



The unfractionated chlorophyll a concentrations ranged from ~2 

 to 13 mg/m^ . Little difference could be seen among the stations for 

 either unfractionated or fractionated concentrations. As with the 

 productivity measurements, the < 10-/jm fraction was relatively low 

 and constant throughout the year (ranging between 0.5 and 

 3.5mg/m'^). The > 64-/im fraction was responsible for the large 

 seasonal fluctuations observed (ranging between 1 and 11 mg/m'^). 



