Hydrocarbon exploration and production have long been a user and driver of sophisticated numerical algorithms and hardware. Advances in algorithms and processing play an essential role in finding new resource. As production in existing fields peak, continued progress in both exploration and development is necessary to provide additional resources. As the cost of exploration and production is high, even incremental advances in algorithms and computational speed can pay for themselves many times over.

In hydrocarbon exploration advanced computation plays an essential role in seismic imaging, interpretation, and modeling. Seismic imaging is difficult in areas with extreme lateral and vertical variations in seismic wave velocities. An example is sub-salt imaging, which, though challenging has resulted in significant discoveries in recent years. Currently, most imaging algorithms rely on compressional waves (P) and use only part of the amplitude information to resolve sub-surface variations. The next generation of algorithms will likely use shear (S) waves and converted phases recorded from a wide variety of acquisition geometries. The ultimate goal is to use all available information from the seismic trace. A second area of computational challenges lies in modeling and interpretation of the processed images. The goal of the interpretation is to create a geologic model and infer the presence of hydrocarbons. Currently, interpretation often requires considerable human oversight. It may be possible to automate much of this process with automated interpretation and prospect generation.

Development and production of hydrocarbons requires detailed knowledge of reservoirs and fluid/gas content. Resolution of geological properties is high near existing wells but sparse elsewhere. Statistical estimates guided by high-resolution seismic are often used but require multiples models to assess possible errors. Reservoir modeling requires tracking of a multiple-component system (e.g. water, oil, and gas) through a complex media. Typically, reservoir models use a relatively large cell size relative to the existing geologic knowledge due to computational bottlenecks. Finer-grained modeling may lead to more accurate results but will require considerable more computational power.