Pierson, Tick, and Baer 



SO, the running time we have been able to achieve with a rather obsolete com- 

 puter was 50 seconds for a 3 -hour time step. It was estimated that the running 

 time could be reduced to about 35 seconds if the program is recoded in machine 

 language. 



The North Pacific, on the other hand, does present some very difficult pro- 

 gramming problems. The amount of active data is approximately one million 

 numbers. Even with the latest computers having large core memories, large 

 word sizes, and packing, it would not be possible to hold all the data in the main 

 store. Use of external lower speed mass storage is required. In addition, for 

 reasons indicated in previous sections, the program itself will be very large. 



Since the basic nature of the operations required is that of having a matrix 

 first in row order for growth and dissipation and then in column order for prop- 

 agation, a careful layout of the data on this external storage medium is essential 

 if the program is not to spend an inordinate amount of time exchanging the data 

 between the main memory and the mass memory. One of the latest of the new 

 generation of computers will be required for operational use. 



Making the computations for a global grid is intrinsically no more difficult 

 than that for the Pacific. The same techniques for moving data off and to the 

 external memory will work for this larger grid. There will be considerably 

 more programming to deal with the various marginal seas and with island arcs, 

 and for special propagation procedures in the various parts of the world, but 

 none of these present very great obstacles. The running time for the calculation 

 at this stage becomes an important factor. Extrapolating from the past assures 

 one that the needed computer speeds will be available by the time the program 

 is ready. 



A global forecast system using spacecraft measurements as input involves 

 the use of a large data network and a great deal of message switching and dis- 

 play capability. The time constraints and the volume of computation to be han- 

 dled will require a data system of a very sophisticated and highly automatic 

 variety. These requirements, however, are not unique to the oceanographic 

 forecasting. The space program, and even the planned management information 

 systems, all require the same type of performance within the same type of con- 

 straints. One may also expect that the technology and experiences will be avail- 

 able at the time the system is ready for full-scale implementation. 



The preparation of the wind fields present the usual problems of dealing 

 with data from a remote source. The errors in the format have to be discovered 

 and corrected, and data must be removed from the transmission medium into 

 the computer with a minimum of human intervention. These, of course, are now 

 standard problems, and most any organization can deal with them. 



DATA FROM A SPACECRAFT 



It seems possible to measure a wave property which will be related to 

 either the wind speed or to the wave height at a point on the ocean by means of a 

 radar scatterometer on a polar orbiting spacecraft. Such a radar scatterometer 



524 



