FISHERY BULLETIN: VOL. 70, NO. 2 



single sample estimates the structure of the 

 assemblage. The phytoplankton association 

 within the Subarctic Pacific had a low diversity 

 and significant concordance between species. As 

 a result, the species showed a high degree of 

 consistency of relative abundances within sam- 

 ples. In every sample Nitzschia turgiduloides 

 was the numerically dominant species, Denticula 

 seminae was the second dominant, and one of 

 two species, Chaetoceros atlanticus or DactyU 

 iosolen mediterraneus was third in abundance. 

 Thus, a single sample appeared to give a pre- 

 cise estimate of the structure of a less diverse 



assemblage, even though the large between-sam- 

 ple variability decreased the precision of the 

 estimate of absolute abundances of single 

 species. 



In contrast, the phytoplankton association in 

 the upper 50 m of the Central Pacific had a high 

 diversity and lacked concordance between spe- 

 cies. At 10, 35, and 50 m, respectively, three, 

 seven, and five species were dominant in at least 

 one sample. Thus, a single sample from a di- 

 verse assemblage gave an imprecise estimate of 

 the relative abundances of the component spe- 

 cies. 



Table 3. — Confidence interval about single samples. (95% confidence intervals about single samples, x, calculated 

 from the expression 0.3x — 3.2a; and compared with the population mean density as estimated by the mean of 10 

 samples, X.) 



> Substations e and h separated by 3.5 nautical miles. 



ANALYSIS OF PATTERNS 



In the analysis of patchiness and its causal 

 factors, the size and shape of a patch often re- 

 ceives primary consideration. This approach 

 is hampered by the difficulty of accurately de- 

 fining a patch, particularly where, as in the 

 ocean, one can rarely see the patch as a physical 

 entity. An alternate approach is to examine 

 the scale on which a population shows consistent 

 spatial distribution, regardless of the degree of 

 contagion. Since the detection of nonrandom- 

 ness depends upon the interaction of the size 

 and distribution of the samples with the pop- 

 ulation distribution, if the scale of sampling is 



systematically altered, the observed population 

 variance may change, and those sampling scales 

 which produce maximum variances may indi- 

 cate scales of heterogeneity in the population 

 distribution. 



The six sets of 10 and 11 samples were con- 

 sidered as sets of 45 and 55 pairs of samples 

 separated by intervals of 0.5, 1.0, 1.5, . . . 10.5 

 miles. For all nonrandomly distributed species, 

 the variance was calculated between each pos- 

 sible pair of samples and averaged for each in- 

 terval. Thus, for the set of 10 Subarctic sam- 

 ples, in which three pairs were separated by 0.5 

 mile, four pairs by 1.0 mile, two pairs by 1.5 

 miles, etc., s^o.s is an average of three variances, 

 s^i.o an average of four variances, s-1.5 an aver- 



368 



