一种利用特征值性质的MT阻尼最小二乘反演

阻尼最小二乘反演的方法总会得到一个最终迭代方程,该方程系数矩阵的特征值对反演结果有重要影响:小特征值相对大特征值对模型修正量的贡献更大,能够得到更高的分辨率,但同时更加不稳定。为此,以一维层状介质的大地电磁阻尼最小二乘反演为基础,采用特定单位的电阻率参数和层厚参数进行运算,同时改进阻尼因子的自适应调节方法,可以巧妙地利用小特征值能提高模型空间分辨率的特点,得到一种改进的阻尼最小二乘反演方法,并与传统阻尼最小二乘法进行了对比。合成数据测试结果表明,该方法能够得到电性陡变的反演模型,弥补了平滑模型反演不能有效识别电性突变界面的不足,对电性变化较为剧烈的地电结构的识别具有较好效果,分辨率有一定提高。

 Damping least squares inversion always involves a final iterative equation,and the eigenvalues of its coefficient matrix have an important effect on the inversion results.Namely,compared to large eigenvalues,small eigenvalues contribute more to model corrections,resulting in higher resolution,but leading to lower stability.An improved damped least square inversion method is proposed that is based on a one-dimensional magnetotelluric inversion scheme.The method cleverly uses the characteristic that small eigenvalues can improve the spatial resolution,which uses the resistivity and layer thickness parameters of a specific unit for inversion,while improving the adaptive adjustment method of the damping factor.Synthetic data test results showed that,the proposed method could invert a model with an electrical abrupt change,which makes up for the shortcoming of smooth model inversion,which cannot effectively identify an electrical abrupt interface.Compared with traditional methods,this method is more effective for the identification of electrical structures with geoelectrical abrupt change,enabling higher resolution.

国家自然科学基金项目(41674078)资助。