Wyniki wyszukiwania - BazTech

Ograniczanie wyników

1 Acta Geophysica

1 2022

Znaleziono wyników: 1

Liczba wyników na stronie

Wyniki wyszukiwania

Sortuj według:

Ogranicz wyniki do:

An effective estimate for selecting the regularization parameter in the 3D inversion of magnetotelluric data

Ghaedrahmati Reza, Moradzadeh Ali, Moradpouri Farzad

Acta Geophysica

2022

Vol. 70, no 2

609--621

The magnetotelluric (MT) inverse problem is a nonlinear and strongly ill-posed problem. Therefore, to avoid the problem of non-uniqueness of response, this problem is mainly solved by Tikhonov regularization method. The purpose of this study is to present a suitable method for selecting the regularization parameters in the 3D MT inverse problem, with regard to the accuracy and speed of the inversion. In this research, the regularization parameter is simply estimated in each iteration of inversion as the ratio of the data misfit to sum of the data misfit and model norm in the pre-iteration. This scheme is applied in the well-known 3D inversion algorithm, WSInv3DMT, instead of the discrepancy principle method. The accuracy of this scheme is assessed by performing the inversion on synthetic models and real data. Results from the inversion for the synthetic and real data indicate that the data misfit and the model norm are reduced with an acceptable rate during the inversion operation. The inverse model has been smoothly converged to an appropriate model and that unrealistic structures have not been included in the model. The results also show that estimation of the regularization parameter by the discrepancy principle method and continuing the inversion to achieve the target data misfit may lead to the production of a model with non-realistic structures, while in the proposed scheme the inversion has not encountered this problem and it converges to an appropriate model after fewer iterations of inversion. In addition, the results show that the time consumed for the inversion of a set of real data with 41 stations and 16 measurement frequencies would decrease up to 27 percent compared to the time devoted for inverting the same set of data by the discrepancy principle method. Also the inversion does not deviate toward unrealistic models and it closely converges to the model of real geological structures.