270 Uiiircrsitij of CaUfaniia I'lihlical i<nis in Zoologij [Vol. lo 



Moreover, this is not an isolatetl case; similar differences have been 

 repeatedly observed (Michael and McEwen, 19] 5, table I, water- 

 samples 3329-30, 3369-70, and 3377-78). 



These facts emphasize the necessity of fairly continnons and mi- 

 merous observations at each position. They demonstrate that even a 

 very crude approximation toward a biological or planktonic "census 

 of the sea" can be obtained, if at all, only from a vast number of 

 simultaneous observations well distributed throughout the i-egion to 

 which the census refers. Even granting that plankton is "uniformly 

 distributed" if "physical conditions" are uniform, the main errors 

 can not, as intimated by Johnstone (1908, pp. 119-121), be eliminated 

 by refinement of apparatus. The employment of nets whose filtration 

 "constants" are known even to within an infinitesimal quantity can 

 not eliminate the effects of unavoidable variation in temperature and 

 salinity, not to mention those of other envii'oiimental elements. 



2. Mean Annual Valites 



The average of a large number of observations of temperature or 

 salinity well distributed throughout each year of a given period is 

 required for an exact determination of the mean annual temperature 

 or salinity corresponding to that period. If the observations are not 

 well distributed or if they are missing for certain months, the simple 

 average of the observations will usually give an erroneous estimate. 

 However, by means of graphical interpolation, as shown by plate 26, 

 figures 45 and 46, approximate estimates can be made of the true 

 values for each month of what corresponds to an average year of the 

 pei'iod, and the average of these results is the best approximation to 

 the mean annual temperature or salinity that can be obtained from 

 the data. The relation of the mean annual temperature to depth 

 shown by plate 34 and the corresponding salinity relation shown by 

 plate 35 were determined in this way from plate 26. The extreme 

 monthly averages indicated by the dotted lines were also determined 

 from the same curves. Since little if any seasonal variation in tem- 

 perature or salinity occurs below four hundred meters, the simple 

 average of all values observed at each of these depths was plotted, and 

 the curves (pi. 36, figs. 96-97) drawn through the points representing 

 the observed values give the relation between the mean annual tem- 

 perature and salinity to depth, in the interval from four liundred to 

 fourteen hiuidred meters. 



