Scholarlink

<p>This study presents a data-driven Sharp Boundary Inversion (SBI) algorithm for one-dimensional magnetotelluric data to objectively delineate abrupt subsurface resistivity transitions. Conventional smoothness-regularized inversions suffer from the unnatural blurring of discontinuous features, such as faults and lithological contacts. To overcome this limitation, the proposed SBI framework models these discontinuous features by locally relaxing smoothing constraints. The specific locations and relaxation strengths are determined within a novel two-stage optimization architecture guided by Akaike Bayesian Information Criterion (ABIC). In this scheme, an outer loop leveraging the Optuna hyperparameter optimization framework explores the optimal locations and relaxation parameters of sharp boundaries, and an inner Gauss–Newton loop iteratively updates the resistivity model, concurrently determining the global smoothing hyperparameter via ABIC minimization. This integrated scheme autonomously determines the optimal boundary depths and localized smoothing penalties without relying on subjective manual tuning or detailed a priori information. Numerical experiments demonstrate the stability of the algorithm, localizing boundaries with a maximum deviation of a single 50-m layer under severe 10% noise conditions. Crucially, the SBI faithfully reconstructs complex subsurface structures featuring both smooth gradients and abrupt jumps, without generating the staircase artifacts that are common to global sharpening techniques. Released as an open-source tool, this algorithm provides a highly robust and objective foundation for subsurface structure interpretation in electromagnetic geophysics.</p> Graphical abstract <p></p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Objective and automated determination of sharp resistivity boundaries in one-dimensional magnetotelluric inversion

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月20日

查看全文
  • Keiichi Ishizu,
  • Yukitoshi Fukahata

摘要

This study presents a data-driven Sharp Boundary Inversion (SBI) algorithm for one-dimensional magnetotelluric data to objectively delineate abrupt subsurface resistivity transitions. Conventional smoothness-regularized inversions suffer from the unnatural blurring of discontinuous features, such as faults and lithological contacts. To overcome this limitation, the proposed SBI framework models these discontinuous features by locally relaxing smoothing constraints. The specific locations and relaxation strengths are determined within a novel two-stage optimization architecture guided by Akaike Bayesian Information Criterion (ABIC). In this scheme, an outer loop leveraging the Optuna hyperparameter optimization framework explores the optimal locations and relaxation parameters of sharp boundaries, and an inner Gauss–Newton loop iteratively updates the resistivity model, concurrently determining the global smoothing hyperparameter via ABIC minimization. This integrated scheme autonomously determines the optimal boundary depths and localized smoothing penalties without relying on subjective manual tuning or detailed a priori information. Numerical experiments demonstrate the stability of the algorithm, localizing boundaries with a maximum deviation of a single 50-m layer under severe 10% noise conditions. Crucially, the SBI faithfully reconstructs complex subsurface structures featuring both smooth gradients and abrupt jumps, without generating the staircase artifacts that are common to global sharpening techniques. Released as an open-source tool, this algorithm provides a highly robust and objective foundation for subsurface structure interpretation in electromagnetic geophysics.

Graphical abstract