姜丛^1,,
蒋长胜^1,,,
尹凤玲^1,,,
张延保¹,
毕金孟²,
龙锋³,
司政亚⁴,
尹欣欣^1,5
1. 中国地震局地球物理研究所, 北京 100081
2. 天津市地震局, 天津 300201
3. 四川省地震局, 成都 610041
4. 北京市地震局, 北京 100080
5. 甘肃省地震局, 兰州 730000

基金项目: 国家自然科学基金（41804094，U2039204），国家科技基础资源调查专项课题（2018FY100504），中国地震科学实验场专项（2019CSES0106，2019CSES0105）联合资助


详细信息

作者简介: 姜丛, 女, 1997年生, 硕士研究生, 主要从事地震监测技术研究.E-mail: 994569363@qq.com

通讯作者: 蒋长胜, 男, 1979年生, 博士生导师、研究员, 主要从事地震监测技术与地震预测理论研究.E-mail: jiangcs@cea-igp.ac.cn; 尹凤玲, 女, 1984年生, 副研究员, 主要从事地球动力学与地震活动性研究.E-mail: yinfengling@cea-igp.ac.cn

中图分类号: P315

收稿日期:2021-06-07
修回日期:2021-08-16
上线日期:2021-09-10

A new method for calculating b-value of time sequence based on data-driven (TbDD): A case study of the 2021 Yangbi M_S6.4 earthquake sequence in Yunnan

JIANG Cong^1,,
JIANG ChangSheng^1,,,
YIN FengLing^1,,,
ZHANG YanBao¹,
BI JinMeng²,
LONG Feng³,
SI ZhengYa⁴,
YIN XinXin^1,5
1. Institute of Geophysics, China Earthquake Administration, Beijing 100081, China
2. Tianjin Earthquake Agency, Tianjin 300201, China
3. Sichuan Earthquake Agency, Chengdu 610041, China
4. Beijing Earthquake Agency, Beijing 100080, China
5. Gansu Earthquake Agency, Lanzhou 730000, China


More Information

Corresponding authors: JIANG ChangSheng,E-mail:jiangcs@cea-igp.ac.cn ; YIN FengLing,E-mail:yinfengling@cea-igp.ac.cn

MSC: P315

--> Received Date: 07 June 2021
Revised Date: 16 August 2021
Available Online: 10 September 2021


摘要
摘要:时间序列的b值在天然地震和工业开采诱发地震的危险性分析中具有重要的应用潜力，但长期以来受到计算规则设置的人为主观性、计算结果的可靠性和时序对突变识别精度不高等问题影响，制约了不同结果的可比较性和共识性科学认识的提炼.本文借鉴基于数据驱动（data-driven）的地震活动参数计算思路，采用连续函数形式的OK1993模型、时间轴随机段落划分、贝叶斯信息准则模型选择等技术环节，构建了基于数据驱动的时间序列b值计算新方法TbDD.利用合成地震目录的理论测试，并分别与固定地震数目的窗长和步长、固定地震数目的步长和累积窗长等传统的固定窗口法进行了比较研究.结果表明，TbDD方法可较好地还原合成地震目录的b₀值输入参数，在计算规则设置的客观性和对b值突变过程的准确识别上具有明显优势.此外，我们还对新近发生的2021年5月21日云南漾濞M_S6.4地震序列进行了实际案例应用.结果显示，此次序列的b值在M_S6.4主震前为0.7左右、震前20 h出现了约0.1幅度的下降，表明在序列发生前震区的差应力水平较高.而b值在M_S6.4主震发生后起伏明显、逐渐增加至0.8左右，这一现象可能与震区在主震后早期较为剧烈的应力调整有关.进一步针对随机模型的数量以及时间轴的随机段落划分设置对TbDD方法b值计算结果的影响程度进行了测试，发现b值受随机模型数量影响较小、具备较强的稳定性，时间轴的随机段落划分设置可影响b值时序微观起伏变化的识别.本文发展的TbDD方法在对时间序列b值计算的准确性、余震趋势跟踪的高精度要求，以及工业开采诱发地震风险管控等领域有较好的应用潜力，所获得的2021年云南漾濞M_S6.4地震序列的b值计算结果也对理解此次地震序列的孕育过程有参考价值.
关键词: 数据驱动/
地震活动/
OK1993模型/
时间序列分析/
b值

Abstract:The b-value of time sequence has great potential in application for the seismic hazard analysis of natural earthquakes and industrial-mining induced earthquakes. However, it has long been subject to the subjectivity in the artificial selection of calculation rules, the reliability of calculation results and low accuracy of mutation recognition in time sequence, which restricts the comparability of different results and the refinement of scientific consensus. In this work, we apply the data-driven approach to seismic parameter calculation to construct a new Time-sequence b-value Data Driven method, TbDD for short. We use technologies such as the Ogata-Katsura 1993 (OK1993) model of continuous function, random partition of time axis and model selection based on the Bayesian information criterion (BIC) in TbDD at the same time. We use synthetic earthquake catalogs which are used in TbDD for theoretical verification, and compare the results with those produced by the traditional methods of fixed window length and step length with the same number of earthquakes, the accumulated window and step length with fixed number of earthquakes. The results show that TbDD can well recover the initial input parameter of b₀ when utilizing the synthetic catalogs. It has a significant advantage in the objectivity of calculating rules setting and the accurate identification of the b-value mutation process. We apply the new approach and the corresponding model selection rules to an actual earthquake catalog of the M_S6.4 Yangbi, Yunnan earthquake of May 21, 2021. The results show that the b-value remains around 0.7 before the M_S6.4 main shock and the decline in the amplitude of about 0.1 appears 20 hours before the M_S6.4 main shock, which indicates that the differential stress level of this region is high. The b-value changes obviously after the M_S6.4 main shock, gradually increases to about 0.8, which may be related to the early stress adjustment in the seismic zone after the main shock. Further, we test TbDD by using random models and find that the b-value acquired by TbDD has strong stability and is less affected by the number of random models. When we test TbDD by using random partition of time axis, we find that the settings of random partition of time axis can affect the identification of micro-fluctuations in b-values of time sequence. The TbDD method developed in this paper has a great application potential in the fields with high requirement for calculation accuracy, such as aftershock trend track, and the seismic risk management of industrial-mining induced earthquakes. In the end, the b-value of time sequence calculated from the M_S6.4 Yangbi, Yunnan earthquake in 2021 is of reference value for the understanding of the seismogenic process of the earthquake.
Key words:Data-driven/
Seismic activity/
OK1993 model/
Time sequence analysis/
b-value



PDF全文下载地址:

http://www.geophy.cn/data/article/export-pdf?id=46b448a3-a6bb-4dd3-b1d8-427cac3d1e6b