大地电磁数据的Occam反演改进

IMPROVEMENT OF OCCAM'S INVERSION FOR MT DATA

Occam 反演以其稳定的收敛性,在大地电磁(MT)数据解释中有广泛的应用。但是其每次迭代均用一维搜索方法求拉格朗日因子μ值,需要许多次正演计算,速度非常慢。在讨论了 Occam 反演中数据拟合差随μ变化的基础上,本文采用了μ值在一定步长下逐次递减的求取方法,每次迭代只需一次正演,极大地提高了计算速度。另外,反演求得光滑模型而非最光滑模型,分辨率更高。理论及实际数据的反演试算均表明,和 Occam 反演相比,反演依然稳定,但速度更快,结果也更真实。

Occam's inversion is widely used in the explaination of Magnetotelluric (MT) data for its stable convergence.However,many times of for wards are required in each iteration to calculate Lagrangian multiplier μ while 1-D search method is used, so it is very slow.In this paper,based on the discussion of the variation of data misfit with Lagrangian multiplier μ in Occam's inversion,a new method,in which is decreased step by step with a constant step,is developed,It requirs only one forward in each iteration and speeds up the calculation greatly.Additionally,we obtain smooth model which has higher resolution than smoothest model.Trial calculations of the theoretical and field data show that the inversion converges stably but more quickly and the results are more accessible to the true model comparing to Occam's inversion.