利用高密度強地動觀測網開發震源機制初動解自動解算系統

Abstract

臺灣低價位高密度的強地動觀測網由大約700個地震站所組成，主要由P-Alert所組成，此地震網同時接收來自中央研究院地球科學研究所 (Institute of Earth Sciences, Academia Sinica, IESAS) 和國家地震工程研究中心 (National Center for Research on Earthquake Engineering, NCREE) 70個寬頻站的寬頻訊號，提供了良好的測站包覆性，有助於監測地震和決定震源機制解。在過去，地震網監測到的P波初動被用來決定震源機制解。本研究建立一套完整流程系統，當地震發生時，首先挑波器會自動的決定P波到時及判斷初動為上動或下動，並以中央處理器(Central Processing Unit, CPU) 平行計算，再利用等時差 (Equal Differential Time, EDT) 曲面定位法配合格點搜尋法來進行地震定位，同時利用OpenACC平行套件使用圖形運算處理器 (Graphic Processing Unit, GPU) 平行化加速定位的運作，最後利用基因演算法 (Genetic Algorithm , GA) 決定震源機制解，在CPU及GPU平行化計算下，這個系統可以在3到5分鐘內計算出震源與其機制解。

A dense low cost seismic network has been installed in Taiwan with about 700 stations. This network also receives 70 broadband seismic signals from Institute of Earth Sciences, Academia Sinica (IESAS) and National Center for Research on Earthquake Engineering (NCREE). It provides a good stations coverage for the earthquake monitoring as well as focal mechanism determination in Taiwan region. First-motion polarities from this network are used to determine focal mechanism. When an earthquake occurs, automatic picker will determine P arrivals and polarities of the first motion using Central Processing Unit (CPU) parallelization. Then, Equal Differential Time (EDT) surface and grid search methods are used to determine hypocenter location. The OpenACC parallel computing toolkit and Graphic Processing Unit (GPU) are used to reduce processing time. Finally, Genetic Algorithm (GA) method is used to determine focal mechanism. By CPU and GPU parallelization, this system can determine the hypocenter and focal mechanism within 3 to 5 minutes after an earthquake occurrence.