中文摘要：

利用从属函数、变差率和趋势速率方法提取甘肃模拟水位观测资料的地震前兆异常。结果表明：（1）1987年1月至2011年9月，4项水位资料共出现31次从属函数异常、29次变差率异常和23次趋势速率异常。其中，22次从属函数异常、21次变差率异常和14次趋势速率异常很好地对应了甘肃及邻区M。5．0以上地震。通过分析异常特征及其与地震的关系，提出中强地震时间预测的指标与方法，得出甘肃模拟水位从属函数计算的经验常数以及变差率和趋势速率的异常阈值；（2）用从属函数、变差率和趋势速率方法定量提取异常结果较为理想，通过对异常识别方法的效能评价，得到的R值均大于凡值，表明该方法信度至少为97．5％，说明3种方法具有实效性。

英文摘要：

Using the methods of subordinate function, variation rate and trend rate, we extracted the earthquakes pre- cursory anomalies from observation＇ data of analogue water level in Gansu Province. The result indicated that： （ 1 ） 4 items of water level data included 31 subordinate function anomalies, 29 variation rate anomalies and 23 trend rate anomalies from Jan. , 1987 to Sep. , 2011. Among them, 22 subordinate function anomalies, 21 variation rate anomalies and 14 trend rate anomalies corresponded with the Ms 〉 5 earthquakes that ever occurred in Gansu Province and its adjacent region. By the analysis of the features of these anomalies and their relation to the earthquakes, the method of time prediction for the moderate-strong and strong earthquakes were proposed, and the empirical constant, anomaly threshold value of variation rate and trend rate were obtained from subordi- nate function of analogue water level in Gansu Province. （2） Using the methods of subordinate function, varia- tion rate and trend rate, abnormal result which was quantitatively extracted was better. R value was greater than R0 value obtained by the effectiveness evaluation of abnormal identification method. The results showed that con- fidence level of this method is at least 97.5 %, which indicated these three methods are effectiveness.