ORIGINAL ARTICLE Parallel visualization of seismic wave propagation Huai Zhang Æ Shaolin Chen Æ Shi Chen Æ Si Chen Æ Huimin Jing Æ David A. Yuen Æ Yaolin Shi Received: 19 March 2008 / Accepted: 20 May 2008 / Published online: 14 June 2008 Ó Springer-Verlag 2008 Abstract Today parallel visualization of massive datasets from observation and numerical simulation of seismic waves is one of the major goals of geoscience community. A majority of these datasets are time-varying volume data (TVVD), also known as 4D field data. The difficulty of visualizing them on distributed parallel system mainly lies in the algorithm designing for distributed preprocessing of raw datasets, hierarchical point-to-point or collective communication implementation based on distributed data allocation, synchronous volume rendering techniques. In this work we present viable solutions for preprocessing of raw data sets, novel algorithms of parallel rendering and display matrix. Our main objective is focused on the par- allel visualization of results coming from full 4D seismic wave propagation simulations. Keywords Massive data sets  3D seismic wave propagation  Parallel visualization  Parallel rendering  High-resolution display matrix Introduction Rapid development of high-performance computing and information technology, which are powerful tools in the earth science community currently, provided geoscientists with huge amount of data from observation and numerical simulation everyday. In recent years, great achievements were achieved in cyber-infrastructure of geological research, such as digital seismic stations network, high precision GPS network, GIS system, In SAR technology, various meteorological satellite systems, experimental instruments and facilities, etc. (Zhang et al. 2002a, b; Shen et al. 2003; Chen et al. 2003). Meanwhile, modern high- performance computing technologies improved the geo- scientist’s capabilities to detect a more accurate picture of the physics, and encouraged them to address the nature of strongly nonlinearly coupled or complex physical phe- nomena by investigating stronger nonlinearities by high- resolution numerical models (http://www.earthscope.org, http://www.geodynamics.org). Numerical simulation of global and regional seismic waves involves several scientific tasks, such as selection of the proper numerical method for a given physical problem (Komatitsch and Tromp 2002a, b), the effective methods that can deal with absorption boundary conditions with hierarchical/heterogeneous material structure in massively parallel simulation models (Komatitsch and Tromp 2003). Seismic wave propagation in heterogeneous medium and basin structures (Wu et al. 2002; Mou and Pei 2003; Dong 2005), wave reflection and dispersion problems in lateral hierarchical space (Dong and Li 2004) , etc. Visualizing the numerical simulation results of global or regional seismic wave propagation is one of the typical implementations of parallel visualization of massive datasets in solid earth science (Tsuboi et al. 2004). These simulation results are always consisted of a large series of large scale time- varying volume data which record the simulated displace- ment and velocity fields of each time step. They are usually scalar or tensor point data on millions or even hundreds of H. Zhang (&)  S. Chen  S. Chen  H. Jing  Y. Shi Laboratory of Computational Geodynamics, Graduate University of Chinese Academy of Sciences, 19A Yue Quanlu Ave., 100049 Beijing, China e-mail: huaizhang@gmail.com S. Chen  D. A. Yuen Department of Geology and Geophysics and Minnesota Supercomputing Institute, University of Minnesota, Twin Cities, Minneapolis, MN 55455, USA 123 Vis Geosci (2008) 13:117–124 DOI 10.1007/s10069-008-0014-y