<![CDATA[This study investigates the application of the Adam Gradient Descent Optimizer (AGDO) for one-dimensional (1D) magnetotelluric (MT) inversion. The algorithm was evaluated using a benchmark function, synthetic MT datasets generated from layered Earth models with 5% Gaussian noise, and field MT data from the Cloncurry region, Australia. For the synthetic datasets, AGDO successfully reconstructed the target resistivity structures, yielding RMSE values ranging from 0.0599 to 0.0603 across four models with different resistivity configurations. For the field datasets, the inversion produced RMSE values between 0.0225 and 0.0582, indicating good agreement between the observed and calculated responses. The ensemble of models obtained from 100 independent runs clustered around the best-fit solutions, demonstrating stable convergence and consistent inversion results despite different random initial populations. These results indicate that AGDO provides an effective and reliable optimization framework for one-dimensional MT inversion and represents a promising alternative for addressing the nonlinear and non-unique nature of the problem.]]>
Copyrights © 2026
