中长期地震预测中的PI算法改进研究及应用

Application of Improved PI Algorithm in Medium to Long-term Earthquake Prediction

图像信息学PI(Pattern Informatics)算法是一种基于统计物理学的地震预测新方法,近年来得到了较大发展。本文探索把此方法与地震活动性网格点密集值方法相结合,并尝试用于华北地区中长期地震预测。在预测中,使用1970—2011年ML≥3.0区域地震目录,针对MS≥5.0预测"目标震级",采用15年尺度的地震目录滑动时间窗,均为3年尺度的地震活动"异常学习"时段和"预测时间窗",结合一定时空及震级范围内地震的数量和震中分布因素,进行地震危险性概率计算。对该方法的预测效果使用Molchan图表法进行统计检验。结果表明,此方法在某些方面优于PI算法,且在地震趋势分析和中长期预测方面有较好的应用潜力。

Pattern informatics（PI） method is a new way to forecast earthquake based on statistical physics,which has seen great development in recent years.In this paper,this method,combined with the method of density at grid nodes of seismic activity,which is used to study the prediction of long-term prediction in north China.The probibilitic results of future seismic evaluation of the occurrence of ML≥5.0 earthquakes has been shown based on the data of ML≥3.0 events in earthquake catalog from 1970 to 2011,time window of 16 years,3 years of ＂anomalous investigation＂ and ＂prediction time term＂ and combinated with some factors,such as earthquakes numbers in the space-time and magnitude range and distribution of the events.The statistic investigation for the prediction ability is shown by Molchan chart method.It is shown that this combined method is better than PI method in some aspects and could be applied in the analysis of the future seismicity and middle to long term prediction.