DISRAY: A distributed ray tracing by map-reduce

详细信息    查看全文

• 作者：Afsaneh Mohammadzaheri ; Hossein Sadeghi ; ^{sadeghi@um.ac.ir} ; ^{sadeghimail@yahoo.com} ; Sayyed Keivan Hosseini ; Mahdi Navazandeh
• 关键词：Seismic ray tracing ; Tomography ; Distributed computing ; High performance computing
• 刊名：Computers and Geosciences
• 出版年：2013
• 出版时间：March, 2013
• 年：2013
• 卷：52
• 期：Complete
• 页码：453-458
• 全文大小：1388 K

文摘

Seismic ray tracing is one of the most important tools in seismology which is required in many applications such as body wave tomography and exploration seismology. Due to the large amount of data introduced in global tomography, most studies are still using one-dimensional (1D) ray tracing methods. Furthermore, since three-dimensional (3D) global studies generally use initial models obtained by 1D ray tracing algorithms, such algorithms are essentially desired. Regarding new advances in High Performance Computing (HPC) as well as the importance of 1D models in global tomography, this study tries to improve the 1D ray tracing algorithm through an innovative programming paradigm, Map-Reduce. Based on this, we present a novel algorithm, DISRAY, for the distributed implementation of 1D ray tracing shooting methods. It effectively decreases computational effort, and therefore accuracy could be improved. The algorithm implements the shooting method based on the geometrical ray theory inside a radially homogeneous earth model. Travel times, epicentral distances, and ray paths of the first arrival phases and other later phases from take-off angles 0¡ã to 180¡ã are calculated. This algorithm can efficiently be extended to more complex models. The proposed algorithm is well remarked for its efficiency and simplicity.