大地电磁阻抗张量的改进截断最小二乘估计

Improved LTS estimation of MT impedance tensor

近年来,稳健性估计方法已被广泛地应用于大地电磁(MT)阻抗张量估计中.和传统的最小二乘估计相比,它能较好地改善MT阻抗估计的抗噪性.然而,这两类方法的抗干扰能力及其估计结果的置信度仍不能满足当下资源探测"攻深找盲"的需求.本文介绍了一种改进的截断最小二乘估计(LTS)方法.我们首先通过稳健的马氏距离来识别并剔除品质极差的观测数据,然后通过广义的复数截断最小二乘估计出阻抗张量初始值,最后用重加权最小二乘方法(RLS)来改进阻抗张量的统计有效性.一系列的模型数据和野外数据试算表明我们提出的这个方法是有效的.改进截断最小二乘估计方法不仅具备传统LTS崩溃点高的优点,而且因为采用了重加权最小二乘方法,能够取得较高的统计有效性.此外在LTS算法的计算过程中,我们使用了一些加速策略,表现出很高的计算效率.

The robust estimation of magnetotelluric impedance tensor has been commonly used in MT prospecting of the earth electrical structure worldwide. Compared with ordinary least square method,these methods can highly improve the robustness of the MT impedance tensor estimation. However, they can ＇ t satisfy geophysicists＇ increasing needs of high confidence in the accuracy of the estimator and strong resistance to cultural electromagnetic noise. An improved least trimmed squares（ LTS） estimate is described which not only preserves the high breakdown point of traditional LTS estimate, but also achieves higher precision.Extremely bad observations are firstly detected and removed through the robust Mahalanobis distance. The impedance tensor is then obtained employing a generalized complex LTS estimate and improved by a reweighted least squares（ RLS） procedure. The performance of this method is tested by a variety of model data and field magnetotelluric data. What ＇s more, it ＇s computational efficient,making it applicable to the big data analysis of long time MT survey.