FISHERY BULLETIN: VOL. 75, NO. 2 



detail by, e.g., Ahlstrom (1954) and Fleminger 

 ( 1964). Very briefly, the net is 1 m in diameter at 

 the mouth and 5 m long, the filtering section 

 having a mesh size of about 0.5 mm. The net is 

 towed obliquely, from a ship traveling at a speed of 

 about 2 knots, from the surface down to a depth of 

 140 m and then returned to the surface. The 

 volume of water fitered varies from about 400 to 

 600 m 3 . 



Charts of the distribution of biomass for each 

 taxon have been given by Isaacs et al. (1969) for 

 the April and October cruises, by Isaacs et al. 

 ( 1971 ) for the January cruises, and by Fleminger 

 et al. ( 1974) for the July cruises. The station data 

 are held on a magnetic tape file at the Southwest 

 Fisheries Center, National Marine Fisheries 

 Service. 



DATA PROCESSING METHODS 



For the purposes of presenting summaries of 

 CalCOFI data in a compact form and to permit 

 some smoothing of the data by taking average 

 values, P. E. Smith's proposal for subdividing the 

 survey area into 23 zones was used in this study 

 (Figure 1 ). The extent of the survey and hence the 

 number of stations occupied varied from cruise to 

 cruise. The station patterns for the cruises 

 included in this study are given in Smith (1971), 

 and a summary showing the numbers of samples 

 in each zone is given in Table 2. 



The biomass data are available as grams/1,000 

 m 3 and estimated to two decimal places. The range 

 of estimates is from zero to over 5,000 g, and 

 within each taxon they are heavily positively 

 skewed. 



The results presented here were expressed in 

 terms of relative changes in biomass in time and 

 space within each taxonomic category, and exten- 

 sive averaging was employed. It was decided, 

 therefore, to apply a logarithmic transformation 

 to the original estimates. Averages based on log 

 transformed values are weighted in favor of the 

 more numerous low values as opposed to arith- 

 metic means, the values of which may be 

 determined largely by small numbers of high 

 estimates. 



In order to give zero a value on the transformed 

 scale it is normal to add 1 to the observation prior 

 to transformation. In this case, where the biomass 

 has been estimated to two decimal places, a 

 number of options is available, either 1.0, 0.1, or 

 0.01 can be added prior to transformation. Trials 



FIGURE 1. — A chart of the area of the CalCOFI survey showing 

 the grid of station positions on which were based the cruises 

 during the period 1955-59. Also shown is the subdivision of the 

 area into the standard zones used in this study. The well- 

 sampled zones for which annual means of biomass were calcu- 

 lated are marked with an asterisk (see Table 2). 



involving the calculation of means for each zone 

 for each cruise for a subset of the taxonomic 

 categories indicated that adding 1.0 produced a 

 considerable loss of resolution for means corre- 

 sponding to less than 1 g/1,000 m 3 , and adding 

 0.01 produced a resolution of low means that ap- 

 peared to be greater than was warranted by the 

 accuracy of the data. Therefore throughout this 

 study a transformation of the form 



Y = log 10 (10X + 1) 



358 



