6 th International Conference & Exposition on Petroleum Geophysics “Kolkata 2006” (1257) Parallel 2D Pre-Stack Imaging On PARAM Padma R. Rastogi*, A. Ray, S. Uppadhyay Centre for Development of Advanced Computing, Pune University Campus, Pune, India Summary Obtaining high-resolution images of the underground geological structures using seismic reflection data in prestack or poststack domain is crucial for exploration of oil and gas deposits. In petroleum industry, seismic migration is one of the most compute intensive steps in seismic data processing sequence. It accurately images the reflected and diffracted energy and aids in delineating the structural features of the underground geological formations. Migration techniques are highly compute and input/output (I/O) intensive and therefore requires high performance computing. In the last decade the development of parallel distributed computing platforms, related system software and programming environments have made it possible to use parallel codes for high resolution imaging. Centre for Development of Advanced Computing (C-DAC), Pune has developed the OpenFrame architecture for scalable parallel computing applications. Various seismic migration algorithms have been developed and implemented for imaging purposes. In this presentation we will discuss the parallel implementation achieved on C-DAC’s PARAM Padma platform for two totally different 2D pre-stack migration algorithms. In both cases parallelisation has been done using MPI message passing environment and MPI I/O for parallel file reading and writing. The first algorithm is Finite difference based depth migration in frequency - space domain (ω -x depth migration) and the second one is based on integral or summation approach migration in time (Kirchhoff time migration). The benchmarking results for both the algorithm clearly shows that large-scale problems can be solved by implementation of highly efficient and scalable codes. These codes can be easily ported across cluster of workstations. Introduction Seismic imaging is a form of echo-reconstructive technique based on experiments, in which a certain earth volume is illuminated by an explosive or vibratory source, and the backscattered energy by the inhomogeneties of the medium is recorded on the surface in digital form. The inhomogeneties act as reflecting surfaces, which cause signal echoing; the echoes are then recorded at the surface and processed through a “computational lens” defined by a propagation model to yield an image of the inhomogeneties. Seismic Migration is a wave equation based technique that attempts to remove all the distortions from the reflection records by migrating events to their true spatial locations. Migration technique positions dipping and diffracted events observed on seismic section to their true positions in the subsurface resulting in an image, which has greater spatial resolution. It is quite logical to term migration as an inverse process in which recorded events are propagated back to the corresponding reflector locations. There are many ways to migrate seismic data. The numerical techniques employed can generally be separated in three broad categories, namely: summation or integral methods such as Kirchhoff migration (Schneider, 1978); finite difference methods (Claerbout, 1976, 1985); and transform methods such as f-k migration (Stolt, 1978; Gazdag, 1978; Gazdag and Sguazzero, 1984). All these migration methods make use of some approximation to the scalar wave equation. The choice of a migration method to a particular data set depends upon the complexity of the velocity model. Migration methods (f-k migration) that are computationally fast can only accommodate velocity variations with depth. Other methods, e.g. Kirchhoff method, finite difference method, PSPI (Phase Shift Plus Interpolation) method, can also handle lateral velocity variations, but require large computational resources in terms of speed, memory and I/O. Migration can be performed in time or in depth. In the presence of strong lateral velocity variations, time migration followed by time to depth conversion does not image the reflected energy to its true subsurface position. Depth migration is essential in these cases. Depth migration compensates for ray bending, lateral velocity pull-ups and structure. A natural advantage of depth migration is that the output image is displayed in depth and therefore can directly be utilized for geologic interpretation. Migration can be carried out in prestack or poststack domain. The poststack methods are applied to stacked data and are based upon the exploding reflector concept. Prestack migration methods are applied to data before stack and based on downward continuation of the wavefield (Claerbout, 1985). As the common-midpoint stack does not correctly stack dipping event, poststack migration cannot image the dipping event correctly. Though, it is cheaper in terms of computation but inferior to the prestack migration.