216 University of California Publications in Zoology [Vol. 15 



higher level. Moreover, even when the cable is vertical and careful 

 testing indicates that the apparatus is working properly, the ther- 

 mometers occasionally reverse and the water bottle occasionally closes 

 at a higher level. 



On those few occasions when such an error in depth is immediately 

 indicated by the temperature reading, the observation is, of course, 

 repeated. In the majority of cases, however, a depth error made 

 during any particular cruise can be discovered only after all the 

 observations have been critically examined. Since the apparatus 

 employed is designed to insure the simultaneous reversal of ther- 

 mometers and closure of the water bottle, determination of the source 

 and magnitude of temperature errors requires con.sideration of the 

 corresponding salinities and densities in situ; similarly, determination 

 of the source and magnitude of salinity errors requires consideration 

 of the corresponding temperatures. 



As indicated on page 212, correlation tables are drawn up between 

 temperature and depth, salinity and depth, and temperature and 

 salinity, and, if necessary, the results are compared with previous data 

 obtained under similar conditions. These tables, together with the 

 fact that the density in situ must increase with an increase in depth 

 (Michael and McEwen. 1915, p. 25), and that temperature and salinity 

 curves must have certain well defined limits for any particular month 

 and position (McEwen, 1916). form our main criteria for isolating 

 erroneous observations. 



Among the observations thus isolated, the temperatures are usually 

 so correlated with the corresponding salinities as to sugge.st errors in 

 depth rather than in either temperature or salinity. Accordingly, the 

 depth is regarded as unknown and is determined indirectly from the 

 temperature, and checked by a similar determination from the cor- 

 responding salinity. If both agree, it is concluded that the ther- 

 mometers reversed and the water bottle closed at the depth thus indi- 

 cated, and the data are so tabulated. U.sually. only a few erroneous 

 observations remain which cannot be thus corrected, and these are 

 rejected if it is clear that the values will be misleading; otherwise, 

 they are retained but questioned. It may be well to add that defective 

 plankton hauls, while corrected for depth errors when possible, are 

 never rejected. 



