The future of exploration is challenging because the shallow water exploration is slowing down since everything seems to be discovered, explained Dr. San Luis. The boundary conditions in Gulf of Mexico are challenging. It is a competitive environment with significant remaining reserves but exploration involves a cost of over 150 MMUS$ and the data is multi-client.

The Repsol company requires an operator to develop a strong competitive advantage based in proprietary assets to directly find some natural resources but hydrocarbons are produced from reservoirs under the earth.

The opportunities for direct observation of subsurface exploration require HPC support. There are some geologic structures that form traps which hinder marine acquisition. There are lots of unprocessed data of what is discovered under the earth, explained Dr. San Luis.

Seismic imaging is a grand challenge for HPC. It solves an inverse problem: it creates a model from the observed data. There is no unique solution and the company is dealing with an iterative process that depends on human interaction for intermediate evaluations and iterations. The algorithms are based on partial differential equations.

Subsalt plays an important role in the Gulf of Mexico. Therefore, Repsol has to search for alternatives for imaging in the Gulf of Mexico. Alternatives for subsalt imaging are the one way wave equation with dip limitation; the Kirchhoff PSDM which has problems with complex structures; and the comparative generic algorithm.

The future of subsalt imaging is RTM, however, the speaker explained. The Kaleidoscope project partners have started to work with the idea of no computer limitation in order to have a complete RTM.

IBM and the Barcelona Supercomputer Center are working in collaboration to promote R&D and CSIC is also helping. Repsol is the sponsor of the project. The methodology involves a model, an algorithm and petascale computing capacity including multi-element processors, given that CPU frequency has reached a limit. Any multi-element processor will be difficult to program.

The Cell/B.E. Processor will be used which was co-developed by Sony, IBM and Toshiba, and was presented in February 2005. Every processor is initially composed by 9 subunits that work independently, explained Dr. San Luis. The Mare Nostrum Supercomputer which figures in the top 20 of supercomputers will also support the project.

The algorithm is the next generation of imaging tools and provides:

• anisotropic one-pass and two-pass to profile migration
• plane-wave migration and migration in tilted co-ordinates

The model involves depth and time migration and takes into account the importance of the velocity model.

Repsol has the practical experience to bring the project to a success. The project is very intensive and encompasses a simultaneous research in hardware and software to achieve a petascale solution to seismic imaging.

The software research focuses on the quality of algorithms avoiding shortcuts or trade-offs to ensure the maximum possible image quality regardless of the computing power required, Dr. San Luis concluded.