ABSTRACT 



A first investigation, based on actual infrared data, has been made of the 

 feasibility of observing sea surface temperatures or temperature gradients from a 

 satellite. It has been found that the satellite-measured large-scale patterns are 

 generally persistent, with only limited day-to-day changes, and that the smaller 

 scale features are similar to those observed in intensive conventional measurement 

 programs over relatively restricted areas. These and other evidence suggest that: 



1) The satellite data provide good measurements of the gradients of sea 

 surface temperatures. 



2) At least one conventional sea surface temperature measurement is 

 required to serve as a benchmark if good absolute values of sea surface temperatures 

 are required. This is due to uncertainties in atmospheric attenuation of infrared 

 radiation and in sensor calibration and degradation for the TIROS radiometer. 



3) In many cases, the satellite data appear to be detecting synoptic scale 

 changes which take place over a period of one or more days. 



A critical problem is the detection and elimination of data points where 

 clouds prevent an uncontaminated vievi? of the sea surface. It is found that, in day- 

 time, the TIROS Channel 5 data, converted to an albedo, are suitable for detecting 

 cloud contaminated points. No method for reliable cloud detection at night is 

 presently apparent. 



It is found that other sources of cloud cover information, such as conventional 

 meteorological observations and satellite TV data, are insufficient for determining 

 all the significant clear areas for satellite observations of sea surface temperature. 

 This suggests that comprehensive studies or operational analyses of sea surface 

 temperatures will require the processing of relatively extensive quantities of 

 satellite radiation data, using these data themselves to determine the daytime cloud- 

 free areas. In this manner maximum use can be made of whatever cloudfree areas do 

 exist both on an individual day basis, and for the averaging of data for short periods of 

 more or less successive days. A single case (for a period of six days) is analyzed 

 using this approach, and strongly substantiates these conclusions. It is recommended 

 that a more extensive set of pilot studies, using the approach of the processing of 

 relatively extensive samples of data, be conducted. v 



