8th. World Congress on Computational Mechanics (WCCM8) 5th European Congress on Computational Methods in Applied Sciences and Engineering (ECCOMAS 2008) June 30 –July 5, 2008 Venice, Italy WAVE ACOUSTIC PROPAGATION FOR GEOPHYSICS IMAGING, FINITE DIFFERENCE vs FINITE ELEMENT METHODS COMPARISON AND BOUNDARY CONDITION TREATMENT Anne-Cecile Lesage 1 , Mauricio Araya-Polo 1 and Guillaume Houzeaux 1 1 Barcelona Supercomputing Center CASE Department - Nexus II Campus Nord UPC - Barcelona - Spain {anne-cecile.lesage, mauricio.araya}@bsc.es - http://www.bsc.es Key Words: Geophysics Prospection, PDE, 3D Wave Acoustic Propagation, Finite Difference, Finite Element, Boundary Conditions, High Performance Computing, Parallelism. ABSTRACT Imaging techniques for geophysic prospection of sea bottom are extremely demanding in terms of mathematical methods and computational resources [8]. This is because the measurements are going deeper than before, thus making the structures identification a hard task, and the datasets to be com- puted huge. Besides, the current trend is to analyze the images in three dimensions (3D) [4], adding an extra difficulty to the process. Currently, the prospection process is highly automatized by computer programs, where these programs not only implement and solve the mathematical model, but also carry the burden of the datasets manipulation, particularly in pre and post processing. All of these demands (complex mathematical models to be solved and huge datasets to be manipulated) lead us to high per- formance computing (HPC) environments, which are mainly available by supercomputers composed by thousands of computational nodes, thus efficient parallelization of those computer programs is required. Figure 1: Marmousi test case. Impulse response test at t =0.36s on a 2D cut Geophysic prospection of the sea bottom widely and recently use isotropic acoustic wave propagation [4]. From the mathematical modeling point of view two crucial points have to be considered. The first point is the numerical method used to solve the particular PDE of acoustic. In this paper, we compare two methods: Finite Difference (FD) [3] and Finite Element (FE) [6]. Their drawbacks and advantages are exposed, specifically under the light of HPC implementation. FD is a classic method,