基于站间天然电磁场单位脉冲响应的大地电磁时间序列去噪方法

针对天然大地电磁场信号在人文活动密集地区易受噪声干扰的问题,本文提出利用两个同步测点天然电磁场时间序列之间的单位脉冲响应,合成本地点受干扰时段的数据,从而去除大地电磁噪声.首先,选择高信噪比时段的数据,采用最小二乘法,估算本地点与参考点之间的单位脉冲响应,再根据卷积定律,结合参考磁场合成本地点的磁场和电场.最后用合成数据替换含噪声时段数据,实现时间域去噪.实测高信噪比数据和含噪数据的处理结果表明,该方法可以高精度合成本地点磁场与电场信号,有效去除本地点电场和磁场噪声,包括相关噪声,提高大地电磁数据质量.

A new method of noise deletion in magnetotelluric time-series based on impulse response of inter-station transfer function

A new method to construct synthetic natural electric and magnetic time series is proposed, in which the time series of local sites are derived using an impulse response and a reference (STIR). This method based on the assumption that the external source of the magnetic field is uniform, and the electric and magnetic fields acquired on the surface satisfy a time-independent linear relation in the frequency domain. According to the convolution theorem, we can synthesize natural electric and magnetic time series using the impulse responses of inter-station transfer function with a reference. Applying this method, two impulse responses need to be estimated:the quasi-MT impulse response tensor and the horizontal magnetic impulse response tensor. These impulse response tensors are related to the local horizontal electric and magnetic components with the horizontal magnetic components at a reference site, respectively. Some clean segments of time series are selected to estimate impulse responses by using the least-square (LS) method. A test with good quality of MT data shows that the resulted synthetic time-series agree with the natural electric and magnetic time series. For contaminated MT examples, when this method is used to remove noise present at the local site, the scatters of MT sounding curves are remarkably reduced and the data quality are improved.