Non-linear optimization by generalized neighborhood algorithm (GNA) and its application for magnetotellurics (MT)...

Area of Science:

• Geophysics
• Computational Science

Background:

• Population-based optimization algorithms are used for non-linear inverse problems, relying on exploration and exploitation.
• Current methods often yield unrealistic solution uncertainties due to population convergence.

Purpose of the Study:

• To improve the estimation of solution uncertainties in inverse problems.
• To enhance model space exploration for more representative uncertainties.

Main Methods:

• A modified Generalized Neighborhood Algorithm (GNA) divides search agents into two groups: one exploring the whole model space, the other focusing near promising solutions.
• The algorithm's tuning parameter controls the extent of neighborhood search.

Main Results:

• GNA successfully inverted synthetic magnetotelluric (MT) data, accurately recovering models and fitting responses.
• Application to field data showed good agreement between inverted subsurface resistivity and known geology.

Conclusions:

• The proposed GNA method provides more representative uncertainties for inverse problems.
• It offers a robust and effective tool for geophysical data inversion, particularly for magnetotelluric data.