基于数据驱动的时间序列b值计算新方法(TbDD):以2021年云南漾濞M_(S)6.4地震序列为例

A case study of the 2021 Yangbi M_(S)6.4 earthquake sequence in Yunnan

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b 0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_(S)6.4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_(S)6.4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_(S)6.4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_(S)6.4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.

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 b 0 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_(S)6.4 Yangbi,Yunnan earthquake of May 21,2021.The results show that the b-value remains around 0.7 before the M_(S)6.4 main shock and the decline in the amplitude of about 0.1 appears 20 hours before the M_(S)6.4 main shock,which indicates that the differential stress level of this region is high.The b-value changes obviously after the M_(S)6.4 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_(S)6.4 Yangbi,Yunnan earthquake in 2021 is of reference value for the understanding of the seismogenic process of the earthquake.