Inversion of Magnetotelluric Data using Bayesian Neural Networks
Dhruv Poddar, Rohan Sharma, Divakar Vashisth
备注：
Preprint of the extended abstract submitted to the 3rd Indian Near Surface Geoscience'24 Conference & Exhibition by AF Academy and the European Association of Geoscientists & Engineers (EAGE)
Magnetotelluric (MT) inversion is a key technique in geophysics for imaging deep subsurface resistivity structures. However, the inherent ill-posedness and non-uniqueness of inverse problems make them challenging to solve. While supervised deep learning approaches have shown promise in this domain, their predictions are typically deterministic and fail to capture the associated uncertainty, an essential factor for decision-making. To address this limitation, we explore the application of Bayesian Neural Networks (BNNs) for MT inversion with uncertainty quantification. Specifically, we train a Bayesian Convolutional Neural Network (BCNN) on a synthetically generated MT dataset. The BCNN effectively recovers resistivity profiles from apparent resistivity data, with the predicted means closely matching the ground truth across both the training and test sets, while also providing uncertainty estimates quantified within 3 standard deviations from the mean. These results underscore the potential of BNNs to enhance deep learning-based geophysical inversion frameworks by incorporating principled uncertainty quantification.
Multiphysics inversion with variable complexity of receiver-function, surface-wave dispersion and magnetotelluric data reduces uncertainty for lithosphere structure
3-D Magnetotelluric Deep Learning Inversion Guided by Pseudo-Physical Information
Key Laboratory of Earth Exploration and Information Techniques of Ministry of Education, College of Geophysics, Chengdu University of Technology（教育部地球探测与信息技术重点实验室，成都理工大学地球物理学院）；College of Computer Science and Cyber Security, Chengdu University of Technology（成都理工大学计算机科学与网络security学院）
Physically Guided Deep Unsupervised Inversion for 1D Magnetotelluric Models
Universidad Industrial de Santander（桑坦德工业大学）；King Fahd University of Petroleum and Minerals（法赫德国王石油与矿产大学）；College of Petroleum Engineering and Geosciences（石油工程与地球科学学院）
3-D Magnetotelluric Investigations for geothermal exploration in Martinique (Lesser Antilles). Characteristic Deep Resistivity Structures, and Shallow Resistivity Distribution Matching Heliborne TEM Results
在一项面向可再生能源开发的全球法国计划框架内，通过综合多种方法对马提尼克岛（法国小安的列斯群岛）进行了广泛调查（2012年至2013年），目的是精确确定先前强调的潜在地热资源（Sanjuan等人，2003年）。在所采用的常用调查方法中，我们在马提尼克岛三个最有前途的地热田上方进行了三次大地电磁（MT）调查，即Anses d 'Arlet、Montagne Pel{é}e和Pitons du Carbet远景区。共采集了约100个MT台站，表现出单维或多维行为和静态偏移效应。在处理数据与远程参考，三维MT反演的四个复杂的元素MT阻抗张量没有预静位移校正，已被执行为每个扇区，提供三个三维电阻率模型下降到约12至30公里的深度。通过生成三维电阻率模型，包括马提尼克岛周围从海岸到200公里距离的水深测量，在三维反演中考虑了海岸效应。模型的正演响应用于计算海岸效应系数，该系数在3-D反演过程中应用于计算的MT响应，以与观测数据进行比较。每个区段的三维电阻率模型都继承了不同地质历史，反映了其三维电阻率分布和与火山历史有关的特殊性。特别是，与Montagne Pel{é}e层状火山有关的地热田，其特征是准普遍存在的导电层和相当单调的典型电阻率分布，这使得地热目标的解释变得困难。相反，Anse d 'Arlet地区的电阻率分布完全不同，地热目标被认为与沿主结构轴延伸的不太深的电阻性侵入体相连。除了…
EFKAN: A KAN-Integrated Neural Operator For Efficient Magnetotelluric Forward Modeling
College of Big Data and Software Engineering, Zhejiang Wanli University（大数据与软件工程学院，浙江万里大学）；Key Laboratory of Geoscience Big Data and Deep Resource of Zhejiang Province, School of Earth Sciences, Zhejiang University（浙江省地球科学大数据与深部资源重点实验室，浙江大学地球科学学院）
Bayesian full waveform inversion with learned prior using deep convolutional autoencoder
University of Texas at Austin（德克萨斯大学奥斯汀分校）；Halliburton Landmark（哈里伯顿陆盾）；Idaho National Laboratory（爱达荷国家实验室）；School of Earth Sciences, Yunnan University（云南大学地球科学学院）；Peking University（北京大学）
Bayesian geoacoustic inversion using mixture density network
Research Institute of Petroleum Exploration & Development PetroChina（中国石油勘探开发研究院）；College of Meteorology and Oceanography National University of Defense Technology（国防科技大学气象海洋学院）
Enhancing Deep Learning based RMT Data Inversion using Gaussian Random Field
Indian Institute of Technology (Indian School of Mines)（印度理工学院（印度矿业学院））；UiT The Arctic University of Norway（挪威北极大学）；Indian Institute of Petroleum & Energy（印度石油与能源研究院）；Transmute AI Lab (Texmin Hub)（Transmute AI实验室（Texmin中心））
本文针对深度学习 RMT 反演中的分布外泛化问题，提出用 Gaussian Random Fields 生成地下电阻率模型并端到端训练网络，在合成与现场数据中验证了其抗噪和泛化优势。
Surrogate-assisted Bayesian inversion for landscape and basin evolution models
University of New South Wales（新南威尔士大学）；University of Sydney（悉尼大学）；SRM Institute of Science and Technology（SRM科学技术研究所）
景观演化模型的复杂性和计算成本高的性质在未知参数的推断和优化方面提出了重大挑战。贝叶斯推断提供了一种用于未知模型参数的估计和不确定性量化的方法。在我们以前的工作中，我们开发了并行回火贝叶斯作为参数估计和不确定性量化的荒地景观演化模型的框架。并行回火Bayeslands具有高性能计算功能，具有数十个并行运行的处理核心，以提高计算效率。虽然我们使用并行计算，但该过程仍然具有计算挑战性，因为需要绘制和评估数千个样本。\textcolor{black}{In large-scale landscape and basin evolution problems, a single model evaluation can take from several minutes to hours, and in some instances, even days. Surrogate-assisted optimisation has been used for several computationally expensive engineering problems which motivate its use in optimisation and inference of complex geoscientific models.}代理模型的使用可以通过开发计算成本低廉的模型来模拟昂贵的模…
