13

observations  tend  to  cluster  at  some  ’’standard”  levels.  Since  older  oceanographic
techniques  were  not  conducted  to  meeting  target  depths,  the  spikes  on  the  histogram  may
indicate  target  depth  rather  than  those  actually  sampled.

Latitude

Latitude

Fig.8:  Percentage  of  rejected  observations  (red)  and  number
of  observations  in  1-degree  latitude  zones  (blue).

An  example  of  our  quality  check  for  the  area  100-110°E  30-20°S  in  the  Pacific  Ocean  is
shown  in  Fig-6.  The  procedure  described  above  effectively  removes  outliers.  Thus,  a  couple
of  profiles  with  obviously  erroneous  sample  depths  could  not  be  identified  as  outliers  by  the
statistical  check  for  T  and  S  only,  but  were  rejected  by  the  statistical  check  on  sample  depth.
The  statistical  check  was  repeated  three  times.  Each  time  the  percentage  of  outliers  per
cruise  was  computed.  A  certain  fraction  of  cruises  was  found  to  have  a  much  higher
percentage  of  rejected  observation.  Such  cruises  were  flagged  and  excluded  from  the  further
analysis.  The  percentage  of  rejected  observations  (Fig.  7)  varies  with  depth,  depending  on