measuring ash-free dry mass on a V4 split of the 

 sample obtained with a Folsom splitter. A split 

 was filtered onto a pre-ashed Whatman GF/C 

 filter, rinsed with a solution of 6% aqueous 

 ammonium formate, and dried at 60°C. After 

 weighing, the sample was ashed in a muffle 

 furnace at 500°C for 4 hours then reweighed to 

 determine ash-free dry mass by difference. 



Standing stocks were expressed per unit 

 volume, using the volume filtered by each net as 

 determined from calibrated flow meters. 



Water bottle samples for chlorophyll a (Chla) 

 analyses were typically taken at 12-14 depths in 

 the upper 200 m of the water column. Following 

 filtration onto GF/C glass fiber filters, pigments 

 were extracted in refrigerated 90% acetone in 

 the dark and analyzed fluorometrically (Venrick 

 and Hayward 1984). Chla between the surface 

 and 150 m was integrated vertically using the 

 trapezoidal rule. A complete report of chlor- 

 ophyll data and detailed sampling information 

 may be found in SIO References 87-7 and 87-19 

 (Scripps Institution of Oceanography 1987a, b). 



RESULTS 



Several characteristics of the VERB system 

 were analyzed to determine its samphng bias, 

 precision, and estimated filtration efficiency. 



the 505 |jLm net, reflecting negligible clogging of 

 the finer mesh net. 



Wire Angles: The median wire angle (off the 

 vertical) was 7°. Eighty-six percent of the angles 

 were <12° (A'^ = 144). Apart from unusually 

 strong wind conditions when the net could not be 

 deployed, wh-e angles could generally be main- 

 tained at acceptable levels (< 10-12°) by maneu- 

 vering the vessel. 



202 fjim vs. 505 ^.m Nets: Retention charac- 

 teristics of the 2 nets were compared from the 

 >505 |xm fraction displacement volume. The 

 median ratio of the >505 [xm displacement 

 volume from the fine:coarse mesh nets was 1.17 

 (nonparametric 95% C.L.: 1.09-1.24, N = 123). 

 The departure from a ratio of 1.00 probably 

 reflects a difference between retention charac- 

 teristics of the 505 jjim plankton mesh in the field 

 and in the laboratory. Larger organisms appar- 

 ently show less escapement through 505 |xm 

 mesh under low pressure differential in the 

 laboratory, compared with the high pressure 

 differential across the mesh of a net under tow. 

 To avoid introducing a systematic bias, further 

 analyses are therefore restricted to the 2 frac- 

 tions (<505 iJim, >505 pim) collected by the 202 

 |xm net. 



Right/Left Bias: A series of replicated ver- 

 tical hauls was carried out with 202 ^lm nets 

 mounted on both sides of the VERB frame. This 

 was done over a 2V2-3 h interval on 2 occasions. 

 No bias was detected between the ash-free dry 

 mass (AFDM) retained by the left and right nets 

 on either 11 September {N = 9 hauls, P > 0.10, 

 Wilcoxon Signed Rank test) or on 26 September 

 (iV = 8 hauls, P> 0.10). 



Precision: Precision of replicated 202 ixm 

 vertical hauls, expressed as the coefficient 

 of variation (100 x SD/.f), was determined on 3 

 occasions, again over 2V2-3 hours. Coefficients of 

 variation of AFDM on 3 trials were 18.2% (A^ = 

 18 samples, 9 hauls), 14.1% (A^ = 16 samples, 

 8 hauls), and 17.0% (A^ = 13 samples, 13 

 hauls). 



Filtration Efficiency: Filtration efficiency of 

 the 202 |jLm VERB net was compared against 

 that of the 505 |i.m VERB net using the respec- 

 tive volumes of water filtered. Filtration effi- 

 ciency of the 202 p.m net averaged 99.4 ± 1.4% 

 (x ± 95% C.L., A^ = 126 comparisons) of that of 



Ash Content: The ash content of the 2 size 

 fractions did not differ significantly (P > 0.05). 

 Ash content as a percentage of total dry mass 

 averaged 16.0% ± 0.9% (x ± 95% C.L., A^ = 

 160) for the fraction <505 ^.m and 14.6% ± 0.7% 

 for the fraction >505 |xm. 



Standing Stocks: The relation between dis- 

 placement volume and AFDM appeared rela- 

 tively Unear when expressed on a log-log scale 

 (Fig. 3A). However, closer inspection revealed 

 that the slope of this relation differed for the 2 

 size fractions (P<0.001, Analysis of Covariance). 

 Hence a joint regression line was inappropriate. 

 When replotted separately on hnear axes, the 

 difference in these relations was more apparent 

 (Fig. 3B, C). A rectilinear function provided an 

 adequate least squares fit for the fi-action <505 

 |jLm (Table 1). However, the relation was non- 

 linear for the fraction >505 p.m. Use of a 

 curvilinear function (Table 1) decreased the 

 residual sum of squares and removed the serial 

 correlation of residuals that was apparent in a 

 straight line fit iP < 0.05). For the fraction >505 

 p.m, a linear relation was a particularly poor 



969 



