场地效应研究的传递台站线性反演法——以芦山地震为例

采用线性反演法研究了地震动的场地效应，并针对线性反演法的局限性，提出了“传递台站”的方法来扩展其应用.通过传递台站，地震事件不再局限于必须被参考台记录到，台站也不再局限于须与参考台站同时获取到地震记录，改进后可大大增加数据样本，使得结果更趋稳定.同时，对于原来难以参与分析的台站，也可得到其场地反应.研究中，还采用了基于遗传算法的非参考台方法，进一步分析了场地反应和验证所提方法的合理性.将所提方法应用于2013年芦山7.0级地震周边区域，分析了这一地区的强震台站的场地反应和地震波衰减因子Q_s值.研究结果显示，当采用一个和二个传递台站进行分析时，所能利用的地震事件和所能求解场地反应的台站数目逐步增加，随着数据样本的增加，分析结果趋于稳定.研究区域的Q_s值拟合结果为Q_s（f）=334.16f^0.73.遗传算法结果表明，线性反演法中所选取的参考台站有其自身的场地反应.当利用遗传算法获得的参考场的场地反应去校正线性反演法的结果时，对于多数场地，两者的结果符合得很好，说明了遗传算法的结果可以合理地认为是场地较为真实的反应.

Site effects estimation by transfer-station generalized inversion method-A case study of the Lushan earthquake

Site effects play a very important role in characterizing seismic ground motion. In this study, the generalized inversion technique (GIT) is applied to estimate site effects. Moreover, the GIT is modified to improve its analytical ability. GIT needs a reference station located at a rock site as a standard. Its site effect is considered to be a constant. For the same earthquake, the record spectrum of an interested station is divided by that of the reference station, and the source term is eliminated. Thus site effects and the attenuation can be acquired. In the GIT process, the amount of earthquake data available in analysis is limited to that recorded by the reference station, and the stations of which site effects can be estimated are also restricted to those stations which recorded common events with the reference station. In order to improve the limitation of the GIT, a modified GIT is put forward in this study, namely, the transfer-station generalized inversion method (TSGI). Comparing with the GIT, this modified GIT can be used to enlarge data set and increase the number of stations whose site effects can be analyzed. And this makes solution much more stable. To verify the results of GIT, a non-reference method, the genetic algorithms (GA), is applied to estimate absolute site effects. On April 20, 2013, an earthquake with magnitude of M_S7.0 occurred in the Lushan region, China. After this event, more than several hundred aftershocks with M_L≥3.0 occurred in this region. The purpose of this paper is to investigate the site effects and Q_s factor for this area based on the aftershock data. Our results show that when the TSGI is applied instead of the GIT, the total number of events used in the inversion increases from 31 to 54 and the total number of stations whose site effect can be estimated increases from 29 to 37. By adding another transfer station in the TSGI, the number of events and stations analyzed rise to 64 and 40 respectively. Most stations exhibit site amplifications. The fitting of Q_s factor is Q_s(f)=334.16f^0.73. As can be seen from the inversion procedure, site effects estimated by the GIT or TSGI are relative to the reference site. The results of GA demonstrate that the site response of the reference station is not a constant, which is the main reason for the difference between the results of GIT and GA. When the results of GIT are corrected with the site effect of the reference station obtained from GA, these two results agree very well for most of the stations. This indicates that the site effects obtained by GA can represent the absolute site effects, and the results of GIT or TSGI are reliable if the reference site is void of amplification, for instance, the reference station is located at a real rock site.