PI算法用在天山地震带的回溯性研究

Retrospective Study on the Application of Pattern Informatics Method in Tianshan Seismic Zone

摘要:

以天山地震带为研究区域，以该地区1994-2014年发生的M_S6.0以上强震为研究对象进行回溯性研究。以截止震级M_C为M_S4.0，空间网格间距0.4°×0.4°，异常学习时间窗为15年，预测时间窗为4年，系统计算天山地震带1994年以来的"地震热点"图像，分析研究时段内强震前后的PI图像演化过程，并采用ROC方法和R值评分检验，对图像信息方法在天山地区的预测效能进行评分。结果表明：（1）在回溯性研究时段内发生的强震都分布在地震热点上；（2）研究时段内发生的6次6.5级以上地震前均出现不同程度的连续热点异常演化过程；（3）ROC检验和R值评分结果显示，PI算法优于随机预测，PI算法在文章中采用的模型参数对天山地震带M_S ≥ 6.0地震具有较好的中长期预测效果。

Abstract:

Pattern informatics (PI) method is an earthquake prediction method based on statistical physics, which has been an active research area in recent years. This method can help to determine whether the seismicity in a grid during a time period deviates from its history and from the average state of the selected study region. In this paper, earthquakes above M_S6.0 during 1994-2014 in the Tianshan seismic zone are retrospectively analyzed with the PI method. The results show that (1) most of the studied earthquakes occurred in the seismic regions (hotspots); (2) six earthquakes above M_S6.5 have obvious continuous evolution process; (3) the PI method results in reasonable medium and long term prediction of M_S ≥ 6.0 earthquakes in the Tianshan seismic zone.