A computer model for generating world ocean sound velocity profile (SVP) information is presented. It employs a ’’least‐squares’’ predictor to combine National Oceanographic Data Center (NODC) archival SVP data with any amount of available new sound velocity measurements that might be available. A technique is presented in the paper for the analysis of NODC World Ocean SVP data which is highly efficient. The technique is called empirical orthonormal function (EOF) analysis and it is capable of a very large compaction of the data set. This method provides a very compact presentation of the total statistical nature of the SVP data bank. The end result is a computer model which permits the optimum utilization of all archival data and any new data at a given place and time in world oceans to produce a new complete SVP. Of even greater significance is the fact that the form of the predicted SVP profile is such that it is easily employed in any propagation loss prediction model that is currently in use.