Details of the Abstract
| Title of paper | Full Newton Inversion of 3D Magnetotelluric Data |
| List of authors | Varilsuha, D |
| Affiliation(s) | Istanbul Technical University |
| Summary | The three-dimensional (3D) electromagnetic (EM) inversion performed using the Gauss-Newton and the full Newton algorithms is compared to evaluate the performance and robustness. The inversion requires the calculation of the Hessian matrix, containing the second derivatives of the objective functional, and the gradient vector, composed of the first derivatives of the objective functional. The main purpose of this work is to investigate the benefits and computational aspects of the full Newton method which unlike the Gauss-Newton method includes the higher-order derivatives in the Hessian matrix. The findings of this study suggest that when only the conductivity parameters are considered in an inversion, the Newton method is computationally twice as expensive. At the same time, when the distortion parameters are considered to be affecting the data types and they are to be estimated in inversion, the computational cost increases to 2.5 times. On the other hand, the Newton method is found to be slightly smoother than the Gauss-Newton approach. The calculated higher derivatives in the Hessian matrix will be more significant when more non-linear data types, such as the phase and amplitude tensor, are to be used in inversion. |
| Session Keyword | 2.0 EM theory, modelling and Inversion |
| File upload |
2.0_full_newton_inversion_of_varilsuha_01.pdf
2P66.pdf |