PI算法用于川滇-安达曼-苏门答腊地区7.0级以上强震危险性预测的回溯性检验

Sichuan-Yunnan versus Andaman-Sumatra:PI approach and retrospective forecast test

作为一个中长期地震预测方法,基于复杂系统统计物理的图象信息学PI算法近年来广受关注.针对7级以上强震成组和突发交替的川滇地区,考虑将与其构造和地震活动关系密切,且强震频发的安达曼-苏门答腊地区作为统一的强震预测研究区,使用PI算法进行MW7.0及以上预测"目标震级"的地震危险性分析.计算中使用了1973年以来的NEIC目录,采用10年尺度的地震活动"异常学习"时段和3年尺度"预测时间窗",对预测效果进行了ROC检验.回溯性研究显示,PI算法预测效果较好,表明将川滇-安达曼-苏门答腊地区作为统一的7级以上强震PI算法预测研究区在统计上具有合理性.从统计物理角度,研究区组合前后的各态遍历性曲线显示,组合后的研究区对PI算法的适用程度虽不优于单独考虑川滇地区,但优于安达曼-苏门答腊地区.PI预测图象显示,2008年汶川地震前可能存在中长期尺度的"前兆性"地震活动异常.

Pattern Informatics （PI） algorithm, a recently developed predictive algorithm based on the statistical physics of complex systems, has caused much attention in the study of earthquake predictability and time-dependent seismic hazard assesment. This paper applied the PI algorithm to the Sichuan-Yunnan region of southwest China, where strong earthquakes show complicated pat- tern, considering the teetonically correlated and active Andaman-Sumatra region and aiming at the forecast of ＇target earthquakes＇ no less than Mw7.0. NEIC catalogue since 1973 was used, with ＇anomaly training time window＇ being 10 years and ＇forecast time window＇ being 3 years, respectively. ROC test was used for evaluating the performance of the forecasts. Retrospective forecast test shows positive evidence for the application of PI algorithm to this region, and also indicates that considering the connection between the Sichuan-Yunnan region and the Andaman-Sumatra region is reasonable in the calculation. In the perspective of statistical physics, ergodicity shown by seismicity gets significantly improved for the Andaman-Sumatra region when the combination with the Siehuan-Yunnan region is considered, with an acceptable sacrifice of the ergodicity for the later. Hot spot picture given by PI algorithm shows some ＇ precursory＇ intermediate-term anomaly before the 2008 Wenchuan earthquake.