非平稳过程的最优线性滤波在唐山地震地下水位观测资料中的应用

为了压低干扰和有效地提取信息,本文采用Kalman滤波方法,处理了京津张唐地区17口深水井1970～1981年间的观测资料。结果表明,1976年7月28日的唐山大地震前,天津地区6口水井水位有2-3年的异常下降,下降的开始时间以3-4公里/月的速度传播。天津的6口井位于大震破裂面的西南端。大震后至今,水位无明显的异常变化,与6级以上的强余震无明显关系。北京地区的11口水井水位与大震似无明显关系。本文所采用的Kalman滤波中,每一步都采用了自回归方法和递归滤波方法计算了一步转移矩阵的元素和控制项,因而转移矩阵和控制项是随时间变化的,更适于地震前兆资料的适时处理。

APPLICATION OF OPTIMAL LINEAR FILTERS OF NONSTA-TIONARY PROCESS TO WELL WATER LEVEL DATA BEFORE AND AFTER TANGSHAN EARTHQUAKE

By use of Kalman filtering for signal detection, data of seventeen well water levels before and after Tangshan earthquake of July 28, 1976 were analysed. These data were observed from 1970 through 1981.Results of filtering indicate that well water level anomalous changes in Tianjin area have continued about 2-3 years before the Tangshan earthquake. Anomalous changes seem to be spreading with a speed of about 3-4 km per month. The well water levels in Beijing area did not show conspicuous anomalous changes. These anomalous changes disappeared after the Tangshan earthquake.The Transition matrix and control term in our filter are functions of time t, so that this method is applicable to real processing of seismic precursoty data specially.