论文部分内容阅读
评估了在世界22个地区利用地震预测CN算法进行的中期地震预测。区域的选择取决于可得到的地震目录的质量。由于CN算法的参量已事先确定并发表,因而对地震的预测完全可以重复。总的检验结果如下:在发布警报的26％的时间内,预测的24次地震中有11次确实发生了,占46％。预测的总误差可以用警报总时间和预报失误率之和R表征。我们得到R=0.26+(1-0.46)=0.8。该R值比对加州地区的回顾性预测的总误差高2倍。预测的统计显著性到目前为止为91％;如果以后几年预测质量仍保持不变的话,统计显著性将超过96％。
The mid-term earthquake prediction using the CN prediction algorithm in 22 regions of the world was evaluated. The choice of area depends on the quality of the available seismic inventory. Since the parameters of the CN algorithm have been determined and published in advance, the prediction of the earthquake can be completely repeated. The overall test results are as follows: Within 26% of the time the warning was issued, 11 of the 24 predicted earthquakes did occur, accounting for 46%. The predicted total error can be characterized by the sum R of the total alarm time and the forecast error rate. We get R = 0.26 + (1-0.46) = 0.8. This R value is two times higher than the total error in retrospective projections in California. The statistical significance of the forecast has so far been 91%; the statistical significance will exceed 96% if the forecast quality remains unchanged in the next few years.