区域地震预测策略和预测效能评价技术研究

发布时间：2018-06-21 15:34

本文选题：效能评价技术 + 参考模型　；参考：《中国地震局地球物理研究所》2017年硕士论文

【摘要】：在区域地震危险性分析中,地震预测效能评价、参考模型的构建和区域地震预测策略研究是重要的三个基本问题。而在当前我国地震预测研究和预测预报业务工作体系的“转型升级”过程中,迫切需要构建不同时间尺度的、效能可评价的地震预测模型,开展地震预测效能评价和区域地震预测策略研究。与此相对应,目前国际上正在开展的以严格的统计检验、“竞赛式”预测模型构建等为主的“地震可预测性合作研究”(CSEP)计划,以及“可操作的地震预测”(OEF)等工作对我国相关工作具有重要的借鉴意义。本论文针对上述三个基本问题,借鉴国际经验并结合中国区域地震预测预报实际,开展适合我国区域地震活动特点的地震预测效能评价、参考模型构建和比较研究、区域地震预测策略研究,试图为未来构建新型的区域地震预测预报体系提供科学技术参考。针对上述三方面问题,分别选用了南北地震带(21.0°～43.0°N,97.0°～107.0°E)和“晋冀蒙交界”地区(37.0°～41.0°N,110.0°～116.0°E)为研究区,其中南北地震带为全球“地震可预测性合作研究”(CSEP)计划中国检验区(Chinese Testing Region),而晋冀蒙交界地区先后20余次被划入中国地震局年度重点地震危险区,是华北地区最重要的天然地震预测试验场。数据使用了中国地震台网提供的1970年以来的“全国统一正式目录”。在数据处理中,为获得“背景地震”(background sesimicity)使用了时空ETAS模型和随机除丛法进行了筛选。为选定晋冀蒙交界地区的地震序列目录,采用了区分地震序列与周围背景地震的时-空天然分布界限的“自然边界法”,即通过纬度-时间图、经度-时间图以及震中分布图相结合的方式。为确保地震序列的目录完整性,研究中还使用了“震级-序号”法确定每个地震序列的最小完整性震级(Mc)。针对预测效能评价及其技术问题,论文考虑到当前全球“地震可预测性合作研究”(CSEP)计划和“可操作的地震预测”(OEF)前沿动态,开展了 CSEP计划中通用的Molchan图表法、区域技能评分、N-test、T-test等地震预测检验技术的引入研究和应用。上述方法将分别解决概率预测与危险区划分等“二元”预测结果检验、连续多时段预测的整体评价、地震发生率的预测检验、不同预测模型的比较研究等问题。上述检验方法一方面在优化和构建预测模型中起到重要作用,另一方面可量化评价预测模型和预测结果,这些预测效能评分方法的引入将对解决国内相对单一方法的预测结果评价发挥重要作用。针对预测模型构建中所必须的参考模型问题,论文将全球CSEP计划中国检验区—南北地震带地区为研究区,构建适合该区域地震活动特点的效能可评价的参考模型。参照原理简单、物理意义明确的构建原则,选用简单平滑模型(Triple-S model)和相对强度(RI)模型作为参考模型,使得“参考模型”成为其它地震预测模型继续发展的“竞技标靶”,并在考虑随机除丛等影响因素情况下,与“地震矩加速释放”(AMR)预测模型进行了比较研究。研究中还使用了时空ETAS模型进行除丛处理后获得背景地震目录,用于AMR模型的预测效能检验。相关研究结果表明,构建的两种参考模型(Triple-S模型和RI模型)可适用南北地震带地区的中期尺度预测。“加速矩释放”(AMR)模型的实际预测效果低于参考模型,表明该模型当前的模型设计尚无法满足区域地震预测的需要。针对区域地震预测策略研究问题,论文利用时间序列ETAS模型和“瘦化算法”,对晋冀蒙交界地区开展了连续滑动预测试验及系统的预测效能评估。研究结果表明,ETAS模型和“瘦化算法”对该地区的余震发生率具有一定预测能力,预测“过少”的比例较低,但预测“过多”的比例较高;在主震发生后的早期阶段预测效果较好,但预测效能随着序列的持续时间显著下降;适度增加预测时间窗长可改善预测“过少”的情况,但对预测“过多”的情况改善不明显。此外,高阶余震激发能力偏弱,可能是晋冀蒙交界地区ETAS模型和“瘦化算法”预测效能偏低的原因之一。作为余震短期发生率预测的可能策略,建议在该地区的ETAS模型和“瘦化算法”预测应用中,采用3天的预测时间窗、仅对序列早期阶段进行应用,且重点关注“不低于”相应预测地震数目的底线思维,可能更具有现实意义。
[Abstract]:In the analysis of regional seismic risk, the evaluation of earthquake prediction efficiency, the construction of reference model and the study of regional earthquake prediction strategy are three important problems. In the process of "transformation and upgrading" of the current earthquake prediction research and prediction business system in our country, it is necessary to construct different time scales and evaluate the effectiveness. The earthquake prediction model, the evaluation of the effectiveness of the earthquake prediction and the study of the regional earthquake prediction strategy. In contrast, the "seismic predictability cooperation research" (CSEP), and the "OEF", are being carried out by the strict statistical test, the "competition" prediction model construction and so on. In view of the above three basic problems, this paper, aiming at the above three basic problems, drawing on the international experience and combining with the reality of China's regional earthquake prediction, carries out the evaluation of earthquake prediction efficiency suitable for the characteristics of regional seismic activity in China, the reference model construction and comparative study, the study of regional earthquake prediction strategy, and the attempt to study the regional earthquake prediction strategy. In order to provide scientific and technical reference for the construction of a new regional earthquake prediction system in the future, the north and South seismic belts (21 to 43 degrees N, 97 to 107 degrees E) and "Jin Ji Meng junction" region (37 to 41 degrees N, 110 to 116 degree E) are selected as the research areas, and the north and South seismic belt is the global "earthquake precondition" CSEP (CSEP) plans the China inspection area (Chinese Testing Region), and the border area of the Jin Hebei Mongolia border area has been divided into more than 20 times in the annual key earthquake hazard area of the China Earthquake Administration. It is the most important natural earthquake prediction test field in North China. In the data processing, the space-time ETAS model and random cluster method are used to select the "background earthquake" (background sesimicity). The "natural boundary method" which distinguishes the natural distribution boundary between the earthquake sequence and the surrounding earthquake is adopted to select the earthquake sequence catalogue in the boundary area of the Jin Ji Meng border area. In order to ensure the integrity of the seismic sequence, the "magnitude sequence number" method is used to determine the minimum integrity magnitude of each earthquake sequence (Mc). For the prediction effectiveness evaluation and its technical problems, the paper takes into account the current global "earthquake precondition". The CSEP plan and the frontier dynamic of "operational earthquake prediction" (OEF) have carried out the research and application of the universal Molchan chart, regional skill score, N-test, T-test and other seismic prediction techniques in the CSEP program. The above method will solve the "two yuan" prediction node, such as the probability prediction and the danger zone division, respectively. The overall evaluation, the prediction test of the earthquake occurrence rate, the comparison of the different prediction models, and so on. The above inspection method plays an important role in optimizing and constructing the prediction model on the one hand, and on the other hand, the prediction model and the pretest results can be quantified, and the introduction of these prediction effectiveness scoring methods will be introduced. It plays an important role in evaluating the prediction results of relatively single method in China. Aiming at the problem of the reference model in the construction of the prediction model, the paper takes the China inspection area of the global CSEP project as the research area of the north and South seismic belt, and constructs the effectiveness and evaluation reference model suitable for the characteristics of the seismic activity in this region. By using the simple and smooth model (Triple-S model) and the relative intensity (RI) model as the reference model, the "reference model" is used as the "competitive target" for the continued development of other earthquake prediction models, and the prediction of "acceleration release of earthquake moment" (AMR) is considered in the case of random cluster and other shadow factors. A comparative study of the model is carried out. The spatio-temporal ETAS model is also used to obtain the background seismic catalogue after the removal of the cluster, which is used to test the predictive effectiveness of the AMR model. The related research results show that the two reference models (Triple-S model and RI model) can be applied to the mid scale prediction of the north and South earthquake zone. The actual prediction effect of the (AMR) model is lower than the reference model, which indicates that the model design of the model is still unable to meet the needs of regional earthquake prediction. In view of the problem of regional earthquake prediction strategy, the paper uses the time series ETAS model and the "thinning algorithm" to carry out continuous sliding prediction experiment and system on the border area of Shanxi and Hebei Province. The results show that the ETAS model and the "thinning algorithm" have a certain predictive ability for the aftershock rate of the region, and the proportion of "too little" is low, but the proportion of "too much" is higher. The prediction effect is better in the early stage after the main shock, but the prediction efficiency is significant with the duration of the sequence. A moderate increase in the length of the prediction time window can improve the prediction of "too little", but it is not obvious to improve the prediction of "too much". In addition, the weak excitation capacity of the high order aftershocks may be one of the reasons for the low prediction efficiency of the ETAS model and the "thinning algorithm" in the border area of the Shanxi Hebei Mongolia border area. The strategy suggests that in the application of the ETAS model and the "thinning algorithm" prediction application in this area, the 3 day prediction time window is used only for the early stage of the sequence, and the bottom line thinking of "not less than" the corresponding prediction of the number of earthquakes may be of more practical significance.
【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】：2017
【分类号】：P315

【参考文献】