基于Hyposat和HypoDD的山东地区地震波一维速度模型

本文收集了2009—2016年山东地区M≥0.1的1 554次地震事件及其走时数据,对该地区地震波一维速度模型进行了研究。首先,使用走时曲线拟合、折合走时曲线分析等方法给出了山东地区地震波一维速度结构模型的初始模型以及参考变化范围;然后,精选55次震相较为丰富的地震事件,用Hyposat方法进行批量重定位,根据残差以及地球物理背景给出P波参考模型,之后对理论波速比加入合理的扰动,对所有P波和S波速度模型组合进行批量地震重定位,根据残差结果极小值获得了最优的P波和S波速度模型;最后,利用多重检验方法对最优模型进行检验,包括天然及非天然地震事件检验、3种定位水平差异检验、4种定位深度差异检验以及乳山和长岛震群的双差定位检验,最终获得了较为可靠的山东地区地震波一维速度模型:vP1=6.13km/s,vP2=6.88km/s,vP3=7.93km/s,H1=22km,H2=33km,vS1=3.54km/s,vS2=3.95km/s,vS3=4.46km/s。另外,本文发现,单纯型等非线性算法的定位结果对于速度模型依赖性较低,反之Geiger类型的线性迭代定位方法则较为依赖速度模型。

One-Dimension Seismic Velocity Model of Shandong Province Based on Hyposat and HypoDD

We collected earthquakes with local magnitude above 1.0 and travel time data in Shandong area from the year 2009 to 2016,and calculated one-dimension seismic velocity model.The process mainly included three steps:1)Initial model.:to fit the travel time data and analyze the reduced travel time curve,eventually to get the initial model and its span;2) Best model:to relocate the selected events of all possible models based on Hyposat software,according to the RMS and geophysical background,to get the best model of P phase and the possible model of S phase through adding a reasonable variable to theoretical seismic velocity ratio,then to relocate events of all models according to the RMS,and finally to reveal the best model;3) Model test:to test the best model by using the test of natural earthquake event,blast event,horizontal epicenter comparison of three location methods,depth comparison of four location methods,HypoDD epicenter comparison of Rushan and Changdao earthquake clusters,and ultimately to get the best model.The one-dimension seismic velocity model in Shandongis:VP1=6.13 km/s,VP2=6.88 km/s,VP3=7.93 km/s,H1=22 km,H2=33 km,VS1=3.54 km/s,VS2 =3.95 km/s,VS3 =4.46 km/s.Besides,the location accuracy of nonlinear algorithm such as simplex generally is less dependent on the seismic velocity model;whereas the linear iterative positioning method based on the Geiger algorithm is more dependent on the seismic velocity model.