The determination of a representative mean temperature ( T ) 

 is, of course, the critical factor in this part of the problem. The 

 temperature structure of the ocean is certainly not constant, partic- 

 ularly closer to the surface; therefore, semi-synoptic temperature 

 fields should be used if possible. The only place where sufficient 

 data is available for reliable analysis on a daily basis is at the 

 surface so it was decided to use the FNVvF Sea Surface Temperature 

 (SST) analyses based on 84 hours of reported ship engine injection 

 temperatures at the top of the layer. In order to include a part Oi 

 the deeper temperature structure, the SST field is presently combined 

 with a climatological field at 200 meters depth to obtain 



Finally, this field is modified empirically in areas vvhere salinity 

 considerations are known to be important (Gyashio, Greenland, 

 Labrador currents) . This in effect corrects the ocean temperatures 

 for salinity much as the meteorologist corrects atmospheric temperatures 

 for moisture content when he uses the concept of ' virtual temperature." 



