Seismic data, used mainly in the petroleum industry, commonly present sizes of tens of gigabyte, and, in some cases, hundreds. This work presents propositions for manipulating these data in order to help overcoming the problems that application for seismic processing and interpretation face while dealing with file of such magnitude. The propositions are based on reorganization and compression. The knowledge of the format in which the data will be used allows us to restructure storage reducing disc- memory transference time up to 90%. Compression is used to save storage space. For data of such nature, best results in terms of compression rates come from techniques associated to information loss, being clustering one of them. In this work we present an algorithm for minimizing the cost of clustering a set of data for a pre-determined number of clusters. Seismic data have space coherence that can be used to improve their compression. Combining clustering with the use of space coherence we were able to compress sets of data with rates from 7% to 25% depending on the error associated. A new file format is proposed using reorganization and compression together.