兴都库什—帕米尔—新疆中南部地区PI算法的回溯性研究

RETROSPECTIVE FORECAST TEST OF PI ALGORITHM IN XINDU KUSH VERSUS PAMIR VERSUS MIDDLE AND SOUTHERN REGION OF XINJIANG

为考察基于地震统计物理的图像信息学PI算法是否适用于兴都库什、帕米尔和新疆中南部地区的强震预测,并考察地震活动特征,将兴都库什—帕米尔—新疆中南部地区作为一个研究区整体,对该地区1995-2009年的MS6.5以上强震进行了回溯性预测研究。设定截止震级Mc为完整性震级MW4.9,空间网格间距为0.4°×0.4,°时间滑动步长为120天,选取10年尺度的＂异常学习时间窗＂和3年尺度的＂预测时间窗＂,分别对5个时间段进行计算。利用ROC图表法作为统计检验工具,使用R I算法作为预测结果参照的＂零假设＂。计算结果表明,PI算法远优于随机预测,可适用于该地区MS6.5以上强震的预测,但PI与RI算法的预测效果差距不大。这一方面说明该地区强震活动可能具有丛集性特征,另一方面,背景地震活动的非平稳性可能是造成PI算法与RI算法效果接近的原因。

If pattern informatics（PI） algorithm can be used on strong earthquakes forecast in Xindu Kush,Pamir and middle and southern region of Xinjiang,according to the seismic activity characteristics,this paper applied pattern informatics algorithm to the earthquakes with MS≥6.5 from 1995 to 2009 in Xindu Kush-Pamir-middle and southern region of Xinjiang.During computing,the cut-off magnitude is MW4.9,spatial grid is 0.4°×0.4°,temporal slip step is 120 days,abnormal study time window is 10 years,forecast temporal windows is 3 years,this paper computed five temporal segments.Used the ROC algorithm to test the result,the RI algorithm regard as reference of forecast result.The result show,pattern information algorithm is good than random forecast,it can be used to forecast strong earthquakes with MS≥6.5,the forecast effect is similar between PI algorithm and RI algorithm.One sides,the characteristics of strong earthquakes occurred in this region is crowd together,another sides,the instability of background seismic activity may be the reason.