Jiajia Sun
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Jiajia Sun has a total of 65 co-authors. Here are the top co-authors by number of shared publications:
- Xiaolong Wei
Department of Earth and Planetary Sciences, Stanford University, Stanford, USA
12 shared publications - Yaoguo Li
Colorado School of Mines
8 shared publications - Jiefu Chen
Department of Electrical and Computer Engineering University of Houston Houston Texas USA
5 shared publications - Sihong Wu
Department of Earth and Atmospheric Sciences, University of Houston, Houston, TX, USA
4 shared publications - Xuqing Wu
University of Houston
4 shared publications Sihong Wu
University of Houston Department of Earth and Atmospheric Sciences Houston, TX
3 shared publications
Latest Publications
- Open Access
- 19 October 2026
- Published by Oxford University Press (OUP) in Geophysical Journal International
CURIE: A continental-scale probabilistic magnetotelluric reference model of USArray data for the contiguous United States
Summary: Magnetotellurics (MT) provides critical constraints on Earth’s electrical conductivity structure and thereby illuminates lithosphere and mantle processes. The EarthScope USArray MT program provides a unique opportunity to investigate large-scale tectonic and compositional variations across the contiguous United States (U.S.). However, uniform MT interpretation is hindered by data quality and the computational and initialization demands of 3D inversion. To provide a physically consistent and uncertainty-aware continental-scale baseline, we develop CURIE (continental U.S. 1D resistivity inference ensemble), comprising two complementary probabilistic reference model ensembles obtained by independently inverting the Zxy and Zyx components from 1773 MT stations using unsupervised, physics-driven normalizing flows. For each station and mode, CURIE provides posterior resistivity distributions, uncertainty estimates, and predicted MT responses. We assess the 1D assumption using filtered data misfit, mode differences, and phase tensor skew and ellipticity. Most stations yield filtered misfits below 4, indicating that 1D models reproduce the dominant responses. Low phase-tensor skew and ellipticity broadly coincide with low-misfit regions across the Interior Plains and the Intermontane Plateaus, whereas elevated indicators and misfits along tectonic and continental margins identify areas where 1D interpretation requires greater caution. At shallow depths, CURIE recovers resistivity patterns associated with major geological provinces; at mid- to lower-lithospheric depths, it resolves the broad contrast between the tectonically active western U.S. and the stable eastern craton. Differences between the Zxy and Zyx results along the Pacific and Gulf coasts, and in Florida further reflect orthogonal sensitivities to conductivity gradients and indicate 3D structural orientations, with some near-coast misfits influenced by the coast effect. The resulting dataset provides a continent-wide probabilistic reference that offers a first-order view of the U.S. electrical structure while supporting data quality assessment, identification of potential 3D effects, and development of initial or reference models for future 3D inversion and geodynamic interpretation.
- 19 October 2026
- Published by Oxford University Press (OUP) in Geophysical Journal International
Open AccessCURIE: A continental-scale probabilistic magnetotelluric reference model of USArray data for the contiguous United States
Summary: Magnetotellurics (MT) provides critical constraints on Earth’s electrical conductivity structure and thereby illuminates lithosphere and mantle processes. The EarthScope USArray MT program provides a unique opportunity to investigate large-scale tectonic and compositional variations across the contiguous United States (U.S.). However, uniform MT interpretation is hindered by data quality and the computational and initialization demands of 3D inversion. To provide a physically consistent and uncertainty-aware continental-scale baseline, we develop CURIE (continental U.S. 1D resistivity inference ensemble), comprising two complementary probabilistic reference model ensembles obtained by independently inverting the Zxy and Zyx components from 1773 MT stations using unsupervised, physics-driven normalizing flows. For each station and mode, CURIE provides posterior resistivity distributions, uncertainty estimates, and predicted MT responses. We assess the 1D assumption using filtered data misfit, mode differences, and phase tensor skew and ellipticity. Most stations yield filtered misfits below 4, indicating that 1D models reproduce the dominant responses. Low phase-tensor skew and ellipticity broadly coincide with low-misfit regions across the Interior Plains and the Intermontane Plateaus, whereas elevated indicators and misfits along tectonic and continental margins identify areas where 1D interpretation requires greater caution. At shallow depths, CURIE recovers resistivity patterns associated with major geological provinces; at mid- to lower-lithospheric depths, it resolves the broad contrast between the tectonically active western U.S. and the stable eastern craton. Differences between the Zxy and Zyx results along the Pacific and Gulf coasts, and in Florida further reflect orthogonal sensitivities to conductivity gradients and indicate 3D structural orientations, with some near-coast misfits influenced by the coast effect. The resulting dataset provides a continent-wide probabilistic reference that offers a first-order view of the U.S. electrical structure while supporting data quality assessment, identification of potential 3D effects, and development of initial or reference models for future 3D inversion and geodynamic interpretation.
- 7 October 2026
- Published by Society of Exploration Geophysicists in The Leading Edge
Demystifying and unifying ELBO for Probabilistic Deep Learning Inversion
Variational inference has emerged as a computationally efficient framework for solving Bayesian inverse problems. Rather than relying on random sampling in high-dimensional model spaces, it reformulates probabilistic inference as an optimization problem, marking a fundamental shift from sampling-based to optimization-based approaches. Its objective, the evidence lower bound (ELBO), is central to modern deep generative approaches, but its form can look different from one model class to another. This article derives the ELBOs step by step for two representative models used in probabilistic geophysical inversion: conditional variational autoencoders (cVAEs) and normalizing flows (NFs). The cVAE objective appears as a reconstruction term minus a latent-space Kullback–Leibler (KL) divergence, whereas the NF objective appears as data likelihood plus prior plus entropy. We show that these are not separate principles. Both are direct substitutions into the same generic ELBO identity. The apparent difference comes from the modeling choice of what counts as the observed variable and what is the hidden variable: in cVAE training, the hidden variable is the latent code; in unsupervised NF inversion, the hidden variable is the Earth model itself.
- 7 October 2026
- Published by Society of Exploration Geophysicists in The Leading Edge
Demystifying and unifying ELBO for Probabilistic Deep Learning Inversion
Variational inference has emerged as a computationally efficient framework for solving Bayesian inverse problems. Rather than relying on random sampling in high-dimensional model spaces, it reformulates probabilistic inference as an optimization problem, marking a fundamental shift from sampling-based to optimization-based approaches. Its objective, the evidence lower bound (ELBO), is central to modern deep generative approaches, but its form can look different from one model class to another. This article derives the ELBOs step by step for two representative models used in probabilistic geophysical inversion: conditional variational autoencoders (cVAEs) and normalizing flows (NFs). The cVAE objective appears as a reconstruction term minus a latent-space Kullback–Leibler (KL) divergence, whereas the NF objective appears as data likelihood plus prior plus entropy. We show that these are not separate principles. Both are direct substitutions into the same generic ELBO identity. The apparent difference comes from the modeling choice of what counts as the observed variable and what is the hidden variable: in cVAE training, the hidden variable is the latent code; in unsupervised NF inversion, the hidden variable is the Earth model itself.
- Open Access
- 26 September 2026
- Published by Oxford University Press (OUP) in Geophysical Journal International
- Vol. 247 (3)
Probabilistic inversion of electromagnetic data using Stein variational gradient descent
SUMMARY: Geophysical inversion is inherently non-unique, and efficient uncertainty quantification remains a major challenge for large-scale applications. Conventional Bayesian approaches, such as Markov chain Monte Carlo (McMC), provide rigorous posterior estimates but are often computationally prohibitive, while deterministic inversion methods lack uncertainty characterization. In this study, we develop a Stein variational gradient descent (SVGD) framework for 1-D geophysical Bayesian inversion. SVGD is a variational inference method that approximates the posterior distribution using an ensemble of interacting particles, which are iteratively updated through gradient-based optimization. The proposed framework is evaluated using airborne electromagnetic data in both synthetic and field scenarios. Synthetic tests demonstrate that SVGD effectively recovers subsurface resistivity structures while capturing associated uncertainty. The influence of particle number on posterior resolution and computational cost is analysed, and comparison with McMC highlights differences in posterior characteristics and computational efficiency. Application to field data from the Mississippi Alluvial Plain shows that the SVGD posterior mean agrees well with deterministic inversion results, while providing the uncertainty estimates. In addition, noise sensitivity tests are conducted to investigate the impact of data errors on inversion behaviour. We find that increasing noise level degrades resolution and shifts the posterior toward prior-dominated regimes. Overall, the SVGD framework provides a flexible and scalable approach to probabilistic geophysical inversion, with strong potential for extension to higher dimensional problems.
- 26 September 2026
- Published by Oxford University Press (OUP) in Geophysical Journal International
- Vol. 247 (3)
Open AccessProbabilistic inversion of electromagnetic data using Stein variational gradient descent
SUMMARY: Geophysical inversion is inherently non-unique, and efficient uncertainty quantification remains a major challenge for large-scale applications. Conventional Bayesian approaches, such as Markov chain Monte Carlo (McMC), provide rigorous posterior estimates but are often computationally prohibitive, while deterministic inversion methods lack uncertainty characterization. In this study, we develop a Stein variational gradient descent (SVGD) framework for 1-D geophysical Bayesian inversion. SVGD is a variational inference method that approximates the posterior distribution using an ensemble of interacting particles, which are iteratively updated through gradient-based optimization. The proposed framework is evaluated using airborne electromagnetic data in both synthetic and field scenarios. Synthetic tests demonstrate that SVGD effectively recovers subsurface resistivity structures while capturing associated uncertainty. The influence of particle number on posterior resolution and computational cost is analysed, and comparison with McMC highlights differences in posterior characteristics and computational efficiency. Application to field data from the Mississippi Alluvial Plain shows that the SVGD posterior mean agrees well with deterministic inversion results, while providing the uncertainty estimates. In addition, noise sensitivity tests are conducted to investigate the impact of data errors on inversion behaviour. We find that increasing noise level degrades resolution and shifts the posterior toward prior-dominated regimes. Overall, the SVGD framework provides a flexible and scalable approach to probabilistic geophysical inversion, with strong potential for extension to higher dimensional problems.
- 1 January 2026
- Published by Society of Exploration Geophysicists in Geophysics
- Vol. 91 (1) , G47-G61
Creating geologically informed training models for deep learning inversion: Applications to potential-field data
Recent advancements in deep learning-based geophysical inversion have drawn considerable attention. Most of these inversions are supervised, which requires the creation of training models that capture as much prior geological information as is available in an area of interest. However, creating such geologically informed training models is challenging because some geological knowledge is difficult to be expressed in mathematical terms. Moreover, geological prior information is not always tied to specific spatial locations. To address these challenges, a novel method based on alpha shapes, a concept from computational geometry, was proposed to generate training models that can easily integrate five key types of geological prior information, namely, (1) top boundaries, (2) dip angles, (3) surface outcrop contacts, (4) mineralization zones intersected by drillholes, and (5) measured physical property values on rock samples. Three distinct scenarios were presented to demonstrate how the proposed method can be used to systematically generate geologically informed training models. It was also shown that deep generative models, such as the conditional variational autoencoder, trained on these geologically informed models, can not only output inversion results that align with prior geological knowledge but also quantify the associated uncertainties. To validate our approach, it was applied to a set of magnetic measurements collected in Qinghai Province, China, for the exploration of Cu-Mo critical mineral deposits. The resulting susceptibility models reveal a major dipping structure that is consistent with surface geology and the magnetic data. The proposed method offers a flexible and unified framework for generating geologically informed training models for deep learning-based geophysical inversions.
- 1 January 2026
- Published by Society of Exploration Geophysicists in Geophysics
- Vol. 91 (1) , G47-G61
Creating geologically informed training models for deep learning inversion: Applications to potential-field data
Recent advancements in deep learning-based geophysical inversion have drawn considerable attention. Most of these inversions are supervised, which requires the creation of training models that capture as much prior geological information as is available in an area of interest. However, creating such geologically informed training models is challenging because some geological knowledge is difficult to be expressed in mathematical terms. Moreover, geological prior information is not always tied to specific spatial locations. To address these challenges, a novel method based on alpha shapes, a concept from computational geometry, was proposed to generate training models that can easily integrate five key types of geological prior information, namely, (1) top boundaries, (2) dip angles, (3) surface outcrop contacts, (4) mineralization zones intersected by drillholes, and (5) measured physical property values on rock samples. Three distinct scenarios were presented to demonstrate how the proposed method can be used to systematically generate geologically informed training models. It was also shown that deep generative models, such as the conditional variational autoencoder, trained on these geologically informed models, can not only output inversion results that align with prior geological knowledge but also quantify the associated uncertainties. To validate our approach, it was applied to a set of magnetic measurements collected in Qinghai Province, China, for the exploration of Cu-Mo critical mineral deposits. The resulting susceptibility models reveal a major dipping structure that is consistent with surface geology and the magnetic data. The proposed method offers a flexible and unified framework for generating geologically informed training models for deep learning-based geophysical inversions.
- 17 October 2025
- Published by Elsevier in Computers & Geosciences
- Vol. 207
Invertible neural network for real-time inversion and uncertainty quantification of ultra-deep resistivity measurements
- 17 October 2025
- Published by Elsevier in Computers & Geosciences
- Vol. 207
Invertible neural network for real-time inversion and uncertainty quantification of ultra-deep resistivity measurements
- 10 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. PP (01962892) , 1
Generalized Transitional Markov Chain Monte Carlo Sampling Technique for Bayesian Inversion of Electromagnetic Data
In the context of Bayesian inversion for scientific and engineering modeling, Markov chain Monte Carlo sampling strategies have become the benchmark due to their flexibility and robustness in dealing with arbitrary posterior probability density functions (PDFs). However, these algorithms been shown to be inefficient when sampling from posterior distributions that are high-dimensional or exhibit multi-modality and/or strong parameter correlations. In such contexts, transitional Markov chain Monte Carlo (TMCMC) provides a more efficient alternative. Despite the recent applicability for Bayesian updating and model selection across a variety of disciplines, TMCMC may require a prohibitive number of tempering stages when the prior PDF is significantly different from the target posterior. Furthermore, the need to start with an initial set of samples from the prior distribution may present a challenge when dealing with implicit priors, e.g. based on feasible regions. Finally, TMCMC can not be used for inverse problems with improper prior PDFs that represent lack of prior knowledge on all or a subset of parameters. A generalization of TMCMC is proposed that alleviates such challenges and limitations towards providing a more robust and efficient tempering sampling strategy. We present convergence analysis, proving that the distance between the intermediate distributions and the target posterior distribution monotonically decreases as the algorithm proceeds. We also demonstrate the advantages of the proposed generalization through a series of test problems and an engineering application in the oil and gas industry.
- 10 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. PP (01962892) , 1
Generalized Transitional Markov Chain Monte Carlo Sampling Technique for Bayesian Inversion of Electromagnetic Data
In the context of Bayesian inversion for scientific and engineering modeling, Markov chain Monte Carlo sampling strategies have become the benchmark due to their flexibility and robustness in dealing with arbitrary posterior probability density functions (PDFs). However, these algorithms been shown to be inefficient when sampling from posterior distributions that are high-dimensional or exhibit multi-modality and/or strong parameter correlations. In such contexts, transitional Markov chain Monte Carlo (TMCMC) provides a more efficient alternative. Despite the recent applicability for Bayesian updating and model selection across a variety of disciplines, TMCMC may require a prohibitive number of tempering stages when the prior PDF is significantly different from the target posterior. Furthermore, the need to start with an initial set of samples from the prior distribution may present a challenge when dealing with implicit priors, e.g. based on feasible regions. Finally, TMCMC can not be used for inverse problems with improper prior PDFs that represent lack of prior knowledge on all or a subset of parameters. A generalization of TMCMC is proposed that alleviates such challenges and limitations towards providing a more robust and efficient tempering sampling strategy. We present convergence analysis, proving that the distance between the intermediate distributions and the target posterior distribution monotonically decreases as the algorithm proceeds. We also demonstrate the advantages of the proposed generalization through a series of test problems and an engineering application in the oil and gas industry.
- Open Access
- 25 June 2025
- Published by Oxford University Press (OUP) in Geophysical Journal International
- Vol. 242 (3)
Variational inference for geophysical Bayesian inverse problems using normalizing flows: an unsupervised approach to electromagnetic data inversion
SUMMARY: Inversion of geophysical data usually exhibits strong non-uniqueness, arising from sparse data coverage, limited number of measurements, inherent nonlinearity of governing physical laws, noise and other factors. Methods based on Monte Carlo sampling are commonly used to explore the posterior model distributions, but these approaches are computationally demanding. Variational inference (VI) provides an alternative by transforming a high-dimensional sampling problem into an optimization problem, thereby significantly reducing the computational time. However, conventional VI methods, which typically use simple distribution families, like Gaussians, to approximate the posterior, may lack flexibility necessary to capture the complexity of the posterior distributions. Normalizing flows (NFs), a type of deep generative models, address this limitation by transforming a simple initial distribution into a highly complex target distribution through a sequence of invertible and differentiable transformations. In this study, we develop an NF-based VI method and apply it to electromagnetic (EM) data. This approach allows for explicit integration of prior knowledge and reference models into the inversion process. Both synthetic tests and field applications on EM data demonstrate that NF-based inversion effectively recovers the posterior model distribution in a more efficient manner, while providing excellent data fitting performance. Unlike many other machine learning algorithms, NFs do not require a training set, making it highly transferable across various inversion problems with minimal adjustments. The proposed NF-based method offers a more robust and computationally efficient solution to uncertainty quantification and shows great potential to be extended to solve 3-D geophysical Bayesian inversions, a major challenge that the geophysical community has faced for decades.
- 25 June 2025
- Published by Oxford University Press (OUP) in Geophysical Journal International
- Vol. 242 (3)
Open AccessVariational inference for geophysical Bayesian inverse problems using normalizing flows: an unsupervised approach to electromagnetic data inversion
SUMMARY: Inversion of geophysical data usually exhibits strong non-uniqueness, arising from sparse data coverage, limited number of measurements, inherent nonlinearity of governing physical laws, noise and other factors. Methods based on Monte Carlo sampling are commonly used to explore the posterior model distributions, but these approaches are computationally demanding. Variational inference (VI) provides an alternative by transforming a high-dimensional sampling problem into an optimization problem, thereby significantly reducing the computational time. However, conventional VI methods, which typically use simple distribution families, like Gaussians, to approximate the posterior, may lack flexibility necessary to capture the complexity of the posterior distributions. Normalizing flows (NFs), a type of deep generative models, address this limitation by transforming a simple initial distribution into a highly complex target distribution through a sequence of invertible and differentiable transformations. In this study, we develop an NF-based VI method and apply it to electromagnetic (EM) data. This approach allows for explicit integration of prior knowledge and reference models into the inversion process. Both synthetic tests and field applications on EM data demonstrate that NF-based inversion effectively recovers the posterior model distribution in a more efficient manner, while providing excellent data fitting performance. Unlike many other machine learning algorithms, NFs do not require a training set, making it highly transferable across various inversion problems with minimal adjustments. The proposed NF-based method offers a more robust and computationally efficient solution to uncertainty quantification and shows great potential to be extended to solve 3-D geophysical Bayesian inversions, a major challenge that the geophysical community has faced for decades.
- Open Access
- 15 May 2025
- Published by Copernicus GmbH
An improved characterization of the subsurface in the Limerick Basin (Ireland) using deep generative model-based 2D gravity inversion constrained with drill hole and petrophysics data
Mineralization in the Limerick Basin, located in southwest Ireland, uniquely associates with volcanic rocks, unlike other mineralized zones in the Irish midlands, where mineral systems align with large-scale normal faults. To better visualize the subsurface structures influencing Limerick’s mineralization, we conducted 2D gravity inversion incorporating geological and petrophysical constraints.Conventional methods of deterministic inversion involve adding a model norm term to the data misfit term in the objective function to regularize an ill-posed problem and obtain stable solutions. While previous studies on constrained deterministic inversion have modified the model norm to include prior information or constraints in geophysical inversion, the complex nature of geological priors makes encoding this information mathematically and computationally challenging. To tackle this problem, we implemented a deep generative model, specifically a conditional variational autoencoder (cVAE)-based inversion framework, to incorporate structural constraints derived from drill hole and petrophysical data.Initially, we tested this framework on a synthetic case by training the cVAE on thousands of 2D density models comprising geological features analogous to the field geology and populated with density values consistent with the drill core measurements acquired from the study area. Artificial drill holes were created to fix the depths of geological units at the drill hole contact points across all training models, ensuring that the predicted models adhered to prior constraints. Following training, we tested the network on some test data, which showed that the predicted models successfully captured the structural and petrophysical property constraints. The geometries of the geological features were also well recovered.We applied this method to gravity data from a NW-SE trending profile crossing the western part of Limerick Syncline. Thousands of density models were generated using drill hole data, incorporating measured rock densities for training. Since the profile’s central and deeper sections lacked sufficient geological data for direct validation of the results, we implemented a hypothesis-testing approach. In each hypothesis, geological features were added to the training density models based on prior geological knowledge of the study area. If simulated data from an inverted model failed to match field data, more geological features were added to the training models in the next hypothesis, and the workflow was repeated to achieve a low data misfit.The inversion provided three key insights into the study area’s geology. First, it identified potential volcanic intrusions in the southern Limerick Syncline, possibly extending from depths below the basement. Second, it estimated the syncline’s geometry in areas with limited geological constraints. Third, it revealed a sharp vertical displacement in stratigraphy, indicating a potential south-dipping fault in the northwest portion of the syncline. This fault may have influenced mineralizing fluid migration, playing a critical role in mineral deposit localization.
- 15 May 2025
- Published by Copernicus GmbH
Open AccessAn improved characterization of the subsurface in the Limerick Basin (Ireland) using deep generative model-based 2D gravity inversion constrained with drill hole and petrophysics data
Mineralization in the Limerick Basin, located in southwest Ireland, uniquely associates with volcanic rocks, unlike other mineralized zones in the Irish midlands, where mineral systems align with large-scale normal faults. To better visualize the subsurface structures influencing Limerick’s mineralization, we conducted 2D gravity inversion incorporating geological and petrophysical constraints.Conventional methods of deterministic inversion involve adding a model norm term to the data misfit term in the objective function to regularize an ill-posed problem and obtain stable solutions. While previous studies on constrained deterministic inversion have modified the model norm to include prior information or constraints in geophysical inversion, the complex nature of geological priors makes encoding this information mathematically and computationally challenging. To tackle this problem, we implemented a deep generative model, specifically a conditional variational autoencoder (cVAE)-based inversion framework, to incorporate structural constraints derived from drill hole and petrophysical data.Initially, we tested this framework on a synthetic case by training the cVAE on thousands of 2D density models comprising geological features analogous to the field geology and populated with density values consistent with the drill core measurements acquired from the study area. Artificial drill holes were created to fix the depths of geological units at the drill hole contact points across all training models, ensuring that the predicted models adhered to prior constraints. Following training, we tested the network on some test data, which showed that the predicted models successfully captured the structural and petrophysical property constraints. The geometries of the geological features were also well recovered.We applied this method to gravity data from a NW-SE trending profile crossing the western part of Limerick Syncline. Thousands of density models were generated using drill hole data, incorporating measured rock densities for training. Since the profile’s central and deeper sections lacked sufficient geological data for direct validation of the results, we implemented a hypothesis-testing approach. In each hypothesis, geological features were added to the training density models based on prior geological knowledge of the study area. If simulated data from an inverted model failed to match field data, more geological features were added to the training models in the next hypothesis, and the workflow was repeated to achieve a low data misfit.The inversion provided three key insights into the study area’s geology. First, it identified potential volcanic intrusions in the southern Limerick Syncline, possibly extending from depths below the basement. Second, it estimated the syncline’s geometry in areas with limited geological constraints. Third, it revealed a sharp vertical displacement in stratigraphy, indicating a potential south-dipping fault in the northwest portion of the syncline. This fault may have influenced mineralizing fluid migration, playing a critical role in mineral deposit localization.
- 16 February 2025
- Published by American Geophysical Union (AGU) in Journal of Geophysical Research: Solid Earth
- Vol. 130 (2)
Characterizing Marine Magnetic Anomalies: A Machine Learning Approach to Advancing the Understanding of Oceanic Crust Formation
Linear magnetic anomalies (LMA), resulting from Earth's magnetic field reversals recorded by seafloor spreading serve as crucial evidence for oceanic crust formation and plate tectonics. Traditionally, LMA analysis relies on visual inspection and manual interpretation, which can be subject to biases due to the complexities of the tectonic history, uneven data coverage, and strong local anomalies associated with seamounts and fracture zones. In this study, we present a Machine learning (ML)‐based framework to identify LMA, determine their orientations and distinguish spatial patterns across oceans. The framework consists of three stages and is semi‐automated, scalable and unbiased. First, a generation network produces artificial yet realistic magnetic anomalies based on user‐specified conditions of linearity and orientation, addressing the scarcity of the labeled training dataset for supervised ML approaches. Second, a characterization network is trained on these generated magnetic anomalies to identify LMA and their orientations. Third, the detected LMA features are clustered into groups based on predicted orientations, revealing underlying spatial patterns, which are directly related to propagating ridges and tectonic activity. The application of this framework to magnetic data from seven areas in the Atlantic and Pacific oceans aligns well with established magnetic lineations and geological features, such as the Mid‐Atlantic Ridge, Reykjanes Ridge, Galapagos Spreading Center, Shatsky Rise, Juan de Fuca Ridge and even Easter Microplate and Galapagos hotspot. The proposed framework establishes a solid foundation for future data‐driven marine magnetic analyses and facilitates objective and quantitative geological interpretation, thus offering the potential to enhance our understanding of oceanic crust formation.
- 16 February 2025
- Published by American Geophysical Union (AGU) in Journal of Geophysical Research: Solid Earth
- Vol. 130 (2)
Characterizing Marine Magnetic Anomalies: A Machine Learning Approach to Advancing the Understanding of Oceanic Crust Formation
Linear magnetic anomalies (LMA), resulting from Earth's magnetic field reversals recorded by seafloor spreading serve as crucial evidence for oceanic crust formation and plate tectonics. Traditionally, LMA analysis relies on visual inspection and manual interpretation, which can be subject to biases due to the complexities of the tectonic history, uneven data coverage, and strong local anomalies associated with seamounts and fracture zones. In this study, we present a Machine learning (ML)‐based framework to identify LMA, determine their orientations and distinguish spatial patterns across oceans. The framework consists of three stages and is semi‐automated, scalable and unbiased. First, a generation network produces artificial yet realistic magnetic anomalies based on user‐specified conditions of linearity and orientation, addressing the scarcity of the labeled training dataset for supervised ML approaches. Second, a characterization network is trained on these generated magnetic anomalies to identify LMA and their orientations. Third, the detected LMA features are clustered into groups based on predicted orientations, revealing underlying spatial patterns, which are directly related to propagating ridges and tectonic activity. The application of this framework to magnetic data from seven areas in the Atlantic and Pacific oceans aligns well with established magnetic lineations and geological features, such as the Mid‐Atlantic Ridge, Reykjanes Ridge, Galapagos Spreading Center, Shatsky Rise, Juan de Fuca Ridge and even Easter Microplate and Galapagos hotspot. The proposed framework establishes a solid foundation for future data‐driven marine magnetic analyses and facilitates objective and quantitative geological interpretation, thus offering the potential to enhance our understanding of oceanic crust formation.