【作者】 万柯松；

【导师】 傅容珊； 倪四道；

【作者基本信息】 中国科学技术大学 ， 固体地球物理学， 2008， 博士

【摘要】 随着数字地震台网的急剧增加和基于互联网的数据共享,产生了巨大的地震波数据积累。这种数据积累给地震学带来的优势不仅仅体现于其射线覆盖率和地震定位的准确性,而且还方便研究者应用特殊的分析方法提取来自地球深部更为细微结构的信息。本文利用GRSN、FNET、ORFEUS、CDSN等台网所接收到的波形数据,对昆仑山地震的破裂过程、苏门答腊地震的P波波形混杂、欧洲南部410km地震波速度间断面和西太平洋下地幔D”层的剪切波速度异常等问题进行了详细的地震学研究。2001年11月14日的昆仑山地震(Ms8.1),是我国大陆近半个世纪以来最强烈的地震,其地震破裂过程具有极长的走滑破裂带(～400km)和超过剪切波速度的破裂速度。本文以地表野外观测到的破裂带和INSAR同震破裂成像的结果为断层约束,通过改进的后向褶积(back-projection)方法,从位于地震震中西北的GRSN台网和震中东南的ISN台网两个角度联合对昆仑山地震的破裂过程进行了成像研究。结果显示,从破裂速度的角度出发,昆仑山地震可以分为四段,其破裂速度分别为2.3km/s、5.5km/s、4.3km/s和4.1km/s。除第一段的破裂速度低于Rayleigh波速度外,其它几段的破裂速度都超过了地壳中剪切波的波速,这种现象被称为超剪切破裂。这些速度段的空间分布与破裂带断层的产状有着良好的对应关系。该结果证实了陈晓非教授提出的超剪切速度破裂与断层地表出露的关系很可能是解释昆仑山地震中由sub Rayleigh速度破裂转变为超剪切速度破裂的转换机制。苏门答腊地震中PP等后续震相对P波震相的污染给地震学研究,特别是对该地震震源过程的研究带来了一定困难。为解决这一问题,本文提出了一种基于不同震相慢度的循环叠加方法,以期将这些震相分离开来,同时还构造了一系列理论地震图用于检验该方法的正确性。结果表明,分离出的波形清晰度高,失真较小,将分离波形按走时叠加得到的波形重构也与原始输入波形相符。本文利用通过这一检验后的波形分离方法对2005年3月28目的Nias地震和2004年12月26日的Sumatra地震分别进行了P波波形的提取。进一步的研究还显示,对于Sumatra地震的超长破裂带(＞1200km),本方法还存在一定不足,不过尚在可接受范围内。通过对Nias、地震和2006年7月17日的Java地震持续时间进行的详细分析显示,这两个地震的持续时间分别为～120s和～190s。由此推断,后一地震引发海啸的可能较大,这与实际情况相符。本文还利用GRSN和ORFEUS台网数据对欧洲南部地幔中的410km地震波速度间断面的精细结构进行了研究。对P波在该间断面绕射波形的观测和理论分析显示,单纯的尖锐速度结构或是线性渐变的速度过渡层产生的绕射波形都与观测到的振幅变化规律不符。我们利用F—K方法对各种不同速度结构的410km间断面模型计算了理论地震图,并将计算结果的振幅变化规律与实际数据做了详细的比较。结果显示,这一地区的410km间断面并不十分尖锐,其过渡带厚度约为20km。而最符合实际数据提供的振幅变化规律的结构模型,为具有50%的线形速度渐变层和50%的速度阶跃。这一结果与Gaherty(1999)根据矿物学实验提出的地震波速度随深度变化的规律相似,可看做是后者的一个近似的地震观测学证据。这一点表明,这个速度变化层可能的地球化学解释是由上地幔中橄榄岩和榴辉岩的特定比例造成的。针对西太平洋剪切波低速区边缘所引起的复杂剪切波波形,本文还相应于何玉梅等(2006)提出的D”层速度结构模型提出了能适于更大震中距范围内数据的模型。本文提出的模型在核幔边界上方30km～230km处存在2%高速异常层,而去除了何玉梅等模型中该区域存在超低速区的结构。这种模型与钙钛矿在较高温情况下相变所产生的速度结构十分类似。我们推测,后钙钛矿相变是引起该区域剪切波波形复杂化的原因。在本文的最后,我们研究了2000年以来斐济—汤加地区发生的14个震深大于180km的地震。采用其在中国数字地震台网上接受到的数据,我们发现在某些台站,尤其是KMI台站上接收到的数据波形中ScS震相存在明显延迟,导致S+ScS的整体波形形状发生变化。为解释D”层对数据波形的影响,我们按层析模型的异常分布构建了D”层的剪切波速度异常模型。对这些模型的地震图合成表明,靠近震源一侧为约—3%低速异常的模型可能解释数据波形的变化。本文还对异常区的边界范围进行了确认,由于研究区域所处的特殊位置,这一结果对确定太平洋低速异常区的边界提供了重要证据。更多还原

【Abstract】 Benefit form the building of digital seismic networks and the Internet sharing,a great deal of seismology data is available.It would advantage us not only in numbers, but also in quality.With special analysis methods for these seismic arrays,we can get more details of seismic signals,which infer more structures in the Earth.In this research,we analysis data from networks include GRSN,FNET,ORFEUS,CDSN and etc to solve several problems:the rupture process of the Kokoxili earthquake,the waveform mixing of the Sumatra earthquake,the fine structure of the 410km discontinuity beneath the south Europe,and the S velocity anomalies of the lowermost mantle beneath the western Pacific.Long strike-slip co-seismic rupture zone（＞400km）has been observed after the 14^thNovember,2001 Kokoxili earthquake.The rupture speed of this earthquake has been determined to exceed the shear wave velocity in the crust.Based on the results of field investigations and INSAR reevaluations,we used an improved back projection method to image the rupture process of this earthquake.As a result,the rupture process of the Kokoxili earthquake can be divided with rupture velocity 2.3km/s,5.5km/s,4.3km/s and 4.1km/s in 4 segments.Rupture started at sub-Rayleigh wave speed and became supershear when it approached the main Kunlun fault at the south Bukadabanfeng.The 4 segments divided with rupture speed can correspond well with the fault segments,which is Taiyanghu segment,the two parts of the Kusaihu segment and the Kunlun Pass segment.It seems that the mechanism of sub-Rayleigh rupture transferring to supershear can be explained with the surface touching of the fault.In the Sumatra earthquake,the long series of P wave train mixing with PP phase bring us a lot of trouble.We use a stacking method divide the mixing phases.To check the method,we calculated a group of synthetic seismograms and applied the dividing method on them.In result,the divided phases were clear,and the reconstructed seismograms fit well with the input seismograms.Using this method, we divided P waveforms from the wave train of the 28^thMar,2005 Nias earthquake and the 26^thDecember,2004 Sumatra earthquake.Further researches shows that the huge rupture zone of the Sumatra earthquake still affect the result,but the effect is acceptable.As an application,we divided the P waveform form 17^thJuly,2007 Java earthquake to get the actual extended time of the event.Using this we can get this parameter rapidly and actually even for large events.The May 18,1998,earthquake in south Italy（Mw5.4）produced a strong topside reflection off the 410-km discontinuity which was recorded on a multitude of seismic arrays throughout the south Europe.Data from GRSN and ORFEUS networks provide a detailed look at the 410-km structure.The salient features of the data set are（1） amplitude of P410 is 1/2 that of P phase in distance 16°～14°;（2）P410 match P phase together at distance 14°;（3）amplitude of P410 is 2/3 that of P phase in distance 14°～11°;（4）P410 die away rapidly in distance 11°～10.5°;（5）P410 disappears when distance little than 10.5°.These features are best modeled by a 20km thick 410-km model which contains a 50%part of linear gradient and a 50%part of velocity jump. The model is similar with a mineralogical model for Olivineα-βtransition.In the last part of the article,we observed the lowermost mantle beneath the western pacific with seismic wave trains.We calculated the ScS-S residuals from CDSN data,and found strong S velocity abnormal（-3s）in the southwestern part of the research area,where the northeastern part is a weak abnormal area.We use 2D finite-difference programs to calculate some synthetics with 3 typical mantle models: mantle plumes with -3%S velocities,the edge of mantle plumes with -3%S velocities in the right half,and a -30%ULVZ in the bottom of the mantle.Comparing the synthetics with the data,we considered that the source of the S abnormal can be explained that the area is in the edge of the Pacific mantle Plume.更多还原