A development of grid search earthquake location program using OpenACC parallel computing toolkit andits application to earthquake early warning

• Main Authors: Hsi-An Chen, 陳璽安
• Other Authors: 吳逸民
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/w65kmk

碩士 === 國立臺灣大學 === 地質科學研究所 === 105 === The rapid and reliable hypocenter and magnitude estimation are two of the most issue for earthquake early warning (EEW) system after an earthquake occurs. In earthquake location aspect, for early warning purpose, the simplified one-dimensional (1-D) velocity model cannot reflect the complex structures in certain area and result in large location uncertainties. Therefore, an accurate location program based on three-dimensional (3-D) velocity model with a simple-calculating process is developed for earthquake location. First, we divided Taiwan region into 3-D grids, 1 km interval in longitude, latitude and depth, which are assumed as possible hypocenters. Thus, the study area consists of 6,880,000 grids. P wave travel times between each grid and seismic station are stored in a pre-calculated database using the 3-D velocity model and 3-D ray tracing method. When an earthquake occurs, an equal differential time (EDT) surface is then conducted in each different triggered station pairs. By comparing the residuals between EDT surfaces and each grid, the grid with the minimum residual will be considered as the hypocenter. In order to deal with massive computation load during the comparing process, we apply an OpenACC parallel computing toolkit to GPU (Graphic Processing Unit) to reduce the processing time. Records of 115 earthquakes and 10 TAIGER explosion events are used to test our location program performance. Results show our location program can provide a reliable hypocenter location in short times and thus it could improve the efficiency of the EEW application. The average location difference between result of this study and Central Weather Bureau (CWB) earthquake catalog for epicenter and depth are 4.77 km and 5.11 km respectively.