贝叶斯概率统计预测模型在三峡工程水库地震重点监视区地震趋势性的应用与检验

Application and Test of Bayesian Probabilistic Statistical Prediction Model in the Seismic Trend of the Key Monitoring Area of the Three Gorges Reservoir Seismic

摘要:

水库地震是由人类工程活动引发的地震活动，具有群发性、高频次和大烈度特点，对其进行预测研究有助于评估水库地震灾害风险。本研究将三峡工程水库地震重点监视区划分为1 km×1 km的若干预测单元，运用贝叶斯概率统计预测模型，依据库水深度、岩体类型、区域构造应力状态、断层活动性、岩溶发育程度、库水透水深度、与库水沟通关系、地震活动背景8个诱震因素来预测诱发震级的概率。结果表明：在重点监视区，以碳酸盐岩为主的库岸附近区域，诱发地震概率较高。仙女山断裂北端、高桥-牛口断裂近库岸段存在诱发中强震（6.0>M3≥4.5）的可能。而三峡坝址区、远离库岸区域不发震或诱发微震的概率较高。此外，为了验证预测模型的准确性，本研究还使用三峡水库175 m蓄水以来实际地震情况对预测模型进行检验，检测结果表明实际地震的发生地点及震级与预测结果存在明显的一致性。表明采用精细合理划分方式的贝叶斯概率统计预测方法在水库诱发地震危险性评定方面具有很好的效果。本研究的概率预测结果对于确定地震风险区域、制定地震预警和防范策略，以及保障水库工程的安全运行具有重要参考意义。

Abstract:

The reservoir earthquake is an earthquake activity caused by human engineering activities. It has the characteristics of group occurrence, high frequency and high intensity. The prediction and research on it is helpful to evaluate the risk of reservoir earthquake disaster. In this study, the earthquake monitoring area of the Three Gorges Project reservoir is divided into several prediction units of 1 km × 1 km. The Bayesian probability statistical prediction model is used to predict the probability of induced magnitude based on eight induced factors : reservoir water depth, rock mass type, regional tectonic stress state, fault activity, karst development degree, reservoir water permeable depth, communication relationship with reservoir water, and seismic activity background. The results show that in the key monitoring area, the probability of induced earthquakes is higher in the area near the reservoir bank dominated by carbonate rocks. There is a possibility of inducing medium-strong earthquakes ( 6.0 > M3 ≥ 4.5 ) in the northern part of Fairy mount fault and the near reservoir bank section of Gaoqiao-Niukou fault. The probability of non-earthquake or induced micro-earthquake in the Three Gorges dam site area and far away from the reservoir bank area is higher. In addition, in order to verify the accuracy of the prediction model, this study also uses the actual earthquake situation since the 175 m impoundment of the Three Gorges Reservoir to test the prediction model. The test results show that the location and magnitude of the actual earthquake are obviously consistent with the prediction results. It shows that the Bayesian probability statistical prediction method with fine and reasonable division method has a good effect in the risk assessment of reservoir induced earthquake. The probability prediction results of this study have important reference significance for determining the earthquake risk area, formulating earthquake early warning and prevention strategies, and ensuring the safe operation of reservoir projects.