“地震可预测性合作研究”计划（CSEP）南北地震带试验区地震活动的一些统计地震学特征及其在地震预测中的应用

发布时间：2019-01-09 12:14

【摘要】：作为“地震可预测性合作研究”(Collaboratory for the Study of Earthquake Predictabiliy, CSEP)计划中国南北地震带试验区的准备工作,本论文选取研究范围为21°～41.5°N,97.50～107.5°E的区域,包括甘肃、四川、云南以及周边地区,分析了研究区以及各子区域的统计地震学特征。分析了研究区内完整性震级随时间变化的规律,结果显示1970年以来研究区整体的完整性震级可取作3.0。分析了1990年以来研究区地震活动的各态遍历性特征,发现不同子区域具有不同的各态遍历性特,这种各态遍历性明显受到2008年汶川地震序列的影响。作为基于Triple-S (Simple Smoothing Seismicity model)方法进行预测分析工作的发展,同时考虑在一个统一的物理背景下考虑地震活动的“涨落”和“平静”,使用PI算法来进行预测分析,其中预测时间为5年尺度,目标震级取6.0级。作为以往利用PI算法进行回溯性检验工作的发展,论文尝试利用PI算法开展向前预测检验。参数设置如下：截止震级3.0,目标震级6.0,空间网格0.2°×0.2°,“背景时间窗”10年,“异常学习时间窗”和“预测时间窗”5年。利用ROC(Recerver Operating Chracteristic)方法对结果进行统计检验,结果显示PI算法的预测效果均优于随机预测和仅考虑丛集性的“相对强度"RI算法。作为PI算法应用的发展,在理论方面,利用PI算法讨论了在一个具有自组织临界性(SOC)的系统中模型地震活动的可预测性问题。发展了基于砂堆原理的元胞自动机(CA)模型,并用在二维时空中分析地震危险性的退化PI算法进行分析。结果显示PI算法的预测效果优于随机预测。在实践方面,针对年度会商中危险区虚报率过高的问题,采用前兆反向追踪(RTP)技术的思路,结合2004年-2012年南北地震带的年度会商工作,利用PI算法和RI算法作为长期地震危险性的评估。结果显示在不降低正确预报的情况下,采用RTP算法的理论,通过PI/RI算法可以有效去除虚报的危险区,从而降低虚报率,表明最初被用来进行短期预报的RTP技术也可以为年度会商工作作出一定的贡献。
[Abstract]:As the preparatory work of the "earthquake predictability Cooperative study" (Collaboratory for the Study of Earthquake Predictabiliy, CSEP) program, this paper selects the area of 21 掳~ 41.5 掳N 97.50 ~ 107.5 掳E, including Gansu Province, for the experimental area of the north-south seismic belt in China. The statistical seismological characteristics of the study area and its sub-regions are analyzed in Sichuan, Yunnan and its surrounding areas. The rule of variation of integrality magnitude with time in the study area is analyzed. The results show that the integrality magnitude of the whole study area since 1970 is 3. 0. By analyzing the ergodic characteristics of earthquake activity in the study area since 1990, it is found that different sub-regions have different ergodic characteristics, which are obviously influenced by the Wenchuan earthquake sequence of 2008. As a development of prediction analysis based on Triple-S (Simple Smoothing Seismicity model) method and considering the fluctuation and calm of seismic activity in a unified physical background, PI algorithm is used to predict and analyze. The prediction time is 5 years and the target magnitude is 6.0. As a development of backtracking test using PI algorithm, this paper attempts to use PI algorithm to carry out forward prediction test. The parameters are as follows: as of magnitude 3.0, target magnitude 6.0, spatial grid 0.2 掳脳 0.2 掳, background time window 10 years, abnormal learning time window and prediction time window 5 years. The ROC (Recerver Operating Chracteristic) method is used to test the results. The results show that the prediction effect of the PI algorithm is better than that of the random prediction algorithm and the "relative strength" RI algorithm which only considers clustering. As a development of PI algorithm, the prediction of model seismic activity in a system with self-organized criticality (SOC) is discussed by using PI algorithm in theory. The (CA) model of cellular automata based on sand pile principle is developed and analyzed with the degenerate PI algorithm for seismic hazard analysis in two-dimensional space-time. The results show that the prediction effect of PI algorithm is better than that of random prediction. In practice, aiming at the problem of high false reporting rate of dangerous areas in annual meetings, the idea of (RTP) technology of precursor reverse tracking is adopted, and combined with the annual meeting work of the north-south seismic belt from 2004 to 2012, PI algorithm and RI algorithm are used to evaluate the long term seismic risk. The results show that the theory of RTP algorithm and PI/RI algorithm can effectively remove the dangerous area of false report and reduce the false report rate without reducing the correct forecast. It is shown that the RTP technique, which was initially used for short term forecasting, can also contribute to the annual conference.
【学位授予单位】：中国地震局地球物理研究所
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：P315.7

【参考文献】