The three components of the crustal magnetic field provide essential constraints on the structure of the source layer as well as its age. In marine equatorial regions where the mid-ocean ridges are oriented north-south, the traditional total field anomalies are tiny, making them essentially undetectable. Still, the vector components of the equatorial magnetic field are detectable and can thus provide the only means of dating the oceanic crust there. However, collecting vector magnetic anomalies is difficult as it requires excellent knowledge of the three orientation angles. The existing vectorial systems are either installed on the carrying platform (airplanes or ships), thus suffer from significant magnetic contamination, or they are towed behind ships but suffer from poor constraints on the heading direction. Here we present the first aero-towed vector magnetic system (AeroVmag) that we have recently developed in order to reduce the magnetic contamination level to essentially null while maintaining excellent knowledge of the orientation angles. The system contains three independent sensors: a vector magnetometer, a scalar magnetometer, and a dual-GNSS/INS orientation sensor (accuracy of 0.02º). We tested the system by collecting data at a low altitude (100 m) above the northeast part of the Sea of Galilee, Israel. Data were collected at a high (200 Hz) sampling rate along a dense grid of profiles that allowed us to evaluate the error levels of our measurements. The scalar results compare favorably with an earlier sea surface total field survey. Together with the vector data, our observations also unravel the location of the main segment of the Dead Sea Transform fault that straddles the survey area. To conclude, this new system will allow the collection of cost-effective and accurate vector magnetic anomaly data. We expect this system to be most valuable over ice-covered, equatorial, and strongly magnetized regions.

How to cite: Granot, R. and Edut, R.: An aero-towed vector magnetic system, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-15157, https://doi.org/10.5194/egusphere-egu23-15157, 2023.

Identification of lineament structure plays a vital role in determining the metallogenic area and distribution of the geologic structure. Edge detection methods are mostly used to recognize the lineaments and define the geologic boundaries. Cooperatively using edge detection results of the gravity, magnetic and remote sensing data to recognize lineaments would obtain more geologic information. In this paper, new edge detectors of potential field derivatives are proposed to determine the sources’ boundary, named second tilt derivative, tilt of vertical derivative, and normalized second vertical derivative, respectively. Presented approaches are characterized by producing zero amplitude over sources’ edges and equalizing anomalies from different depths. Compared with original edge detection techniques including other second derivative methods, synthetic examples reveal significant superiorities of suggested approaches in providing more accurate and sharper edges and are especially effective in distinguishing superimposed anomalies. The experiments also demonstrate that the normalization to the edge detectors will make images cleaner and geologic edges more easily captured. Applied to airborne gravimetric and magnetic data in the Pobei area (NW China), the proposed methods display more geologic details and lineaments. Canny, Sobel, and Prewitt operators are applied to extract boundaries of remote sensing image. Lineaments picked by the three different types of data are combined collectively to get a comprehensive lineaments structure interpretation.

How to cite: Li, G. and Liu, S.: Identifying the Lineament Structure Cooperatively Using the Airborne Gravimetric, Magnetic and Remote Sensing Data: A Case Study from the Pobei Area, NW China, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-10565, https://doi.org/10.5194/egusphere-egu23-10565, 2023.

    Magnetic anomalies commonly contain anomalies generated by crustal rocks that have variable mineral compositions and natural remanent magnetizations. Understanding the magnetic susceptibility, remanent magnetization, magnetization direction, and distribution is important for studying the spatial location, formation, and evolution of underground rocks. However, superposition of magnetic anomalies leads to nonnegligible errors of inversion and interpretation. To overcome the interpretation problems caused by source interference, it is necessary for the target magnetic anomaly to be extracted from the observed magnetic anomaly data. Different methods have been developed to separate the magnetic anomalies of different sources using the spectral differences of regional and residual anomalies. Such methods, which include matched filtering, Wiener filtering, and wavelet analysis, have been successfully applied to solve many geological problems. However, these methods cannot extract the anomalies caused by the interference of sources at similar depths because the spectra of the target and residual magnetic anomalies are similar. Effective techniques to obtain additional magnetic information regarding the distribution of rocks at different layers and with different magnetization directions remain lacking.

    Unlike existing regional-residual separation methods used for separating superimposed magnetic anomalies caused by sources with a large depth separation, this study focuses on magnetic anomalies generated by variability of the magnetic parameters and source interference with and without depth differences. We propose a new and useful method for extracting a target magnetic anomaly from an observed magnetic anomaly field. An optimization scheme is proposed for approximating the low-rank component of an observed magnetic anomaly field on the basis of low-rank theory. The magnetic dipole layout is added as a constraint based on the assumed source location. The optimal magnetizations of the magnetic dipoles are then obtained to minimize the objective function. The sum of the magnetic anomalies generated by the magnetic dipoles is calculated as the target magnetic anomaly. The synthetic and field data experiments indicate that the proposed method can accurately and robustly recover target magnetic anomalies. In the field data experiments, the magnetization information of the various isolated sources is derived via 3D fuzzy C-means inversion using the extracted magnetic anomalies. The results show that the proposed method can extract the geometric and physical information of each target magnetic source, even when the observed magnetic anomaly field is generated by various superimposed sources with target source interference at similar depths. The proposed method has the potential for dealing with the separation problems of potential field data with different scales, including the separation of the geomagnetic core field and the lithospheric magnetic field as well as the extraction of target magnetic anomalies from satellite magnetic measurements. Therefore, this approach could be of great importance for geological investigations and mineral exploration.

How to cite: Zhu, D., Hu, X., and Liu, S.: Extraction of targeted source information from superimposed magnetic anomalies, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-11664, https://doi.org/10.5194/egusphere-egu23-11664, 2023.

Filtering is a fundamental procedure that precedes further quantitative interpretations. In the potential fields case, filtering is used to separate and discriminate the different contributions of a given dataset. In this note we describe a potential field filtering technique based on a “Extremely Compact Sources” (ECS) approach. ECS filtering method allows us to solve the problem of the interfering anomalies, that could hide the real amplitude and shape of the single contributions. The interference phenomena may involve the superimposition of a regional field generated by deep sources in the study area on local anomalies, or the superimposition of anomalies having similar wavelengths. While many methods have been developed during the years to try to separate regional from local fields, fewer methods have been developed to address the separation of interfering  anomalies caused by similar sources. The ECS technique exploits the inherent ambiguity of potential fields to retrieve an extremely compact source model in which sources are well separated each one from other. In this way, the filtering process can be done through a simple "muting" process (setting the physical property of the cells relative to the unwanted contributions to 0) directly in the source domain. We show applications of the ECS technique to both synthetic and real anomalies to prove the validity of the methodology for both separation of interfering anomalies and filtering of regional fields.

How to cite: Maiolino, M. and Florio, G.: ECS (“Extremely Compact Sources”): a new method for potential field data filtering., EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-7911, https://doi.org/10.5194/egusphere-egu23-7911, 2023.

Spectral analysis, based on Fourier Transform, provides a high-resolution analysis in frequency domain but it has not resolution in the space domain. Due to this lack of space resolution, also celebrated methods such as the Spector and Grant’s one, cannot yield information about the position of the source identified in the frequency domain. We propose to fix these issues by resorting to a scalogram analysis, obtained through the continuous wavelet transform of the potential fields, using the Morlet analyzing wavelet. In the scalogram it is indeed possible distinguishing and locating the source contributes for both their space and scale contents. The depths to top and bottom of the potential fields causative sources are investigated locally along bounded subvolumes, subareas, and scale-profiles on the 3D scalogram. The application of such local spectral analysis to synthetic examples and real data leads to results in good agreement  with the known information about the causative sources, providing simultaneously good space and scale resolutions.

How to cite: Abbas, M. A., Milano, M., and Barbolla, D. F.: Depth estimation of gravity and magnetic sources from Wavelet transform spectral analysis, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-15151, https://doi.org/10.5194/egusphere-egu23-15151, 2023.

Scalable machine learning solution for commercial scale three dimensional geophysical inversions

Souvik Mukherjee¹, Santi Adavani^2*, Alan Morgan³, William N. Barkhouse⁴, Ronald S. Bell⁴, Peter G. Lelievre⁵, Colin G. Farquharson⁶

¹EmPact-AI, USA, ^2*RocketML, now at S2 Labs, USA, ³Bell Geospace, USA, ⁴Drone Geoscience, USA, ⁵Mt. Allison University, Canada, ⁶Memorial University, Canada

Abstract

Application of artificial intelligence (AI) and machine learning (ML) based workflows and methodologies for geophysical data processing, imaging, and interpretation are active focus areas in industry and academia. While much progress has been made to demonstrate applicability in many use cases, key bottle neck for widespread commercial use has been the prohibitively high computational cost involved in applying the method for large scale three dimensional inverse problems.

Key changes to the form of the simulated input data used for training and the corresponding design of the architecture of the hidden layers enable approximately O(n) (where n is the number of layers in the network) reduction in the computational complexity of the training architecture. Combined with multi-GPU Distributed Deep Learning (DDL) algorithms optimized specifically for training large scale ML data, this results in significant improvements in resolution of inversion results relative to conventional least squares imaging, while computational efficiency improves by order of magnitude compared to several commonly used open-source ML architectures and platforms.

When deployed for inversion of dense, closely spaced high resolution handheld magnetometer data collected over a buried pipe in a field in Texas, the resolved three-dimensional geometry and location using the new algorithm showed over 6-fold improvement compared to conventional three-dimensional least squares inversion. When applied to an 18-fold larger data set collected by a drone-based magnetometer over a field in California, the buried complex metallic pipe like structure was resolved using little over 2 days of compute time. Similar exercise undertaken in google collab GPU platform using state-of-the-art google tensorflow would have taken 3 – 6 months to complete, suggesting a 50 – 100-fold improvement in computational efficiency.

The method was also benchmarked against Los Alamos National Laboratory’s (LANL) open-source seismic full waveform inversion (FWI) dataset. LANL trained 24000 seismic data sets simulated from various 2D velocity models using 32 P100 Tesla GPU machines in 2 hours. When inferenced on 6000 previously unseen test models, the root mean square error (RMSE) in the inverted normalized velocity models was 0.018. The current workflow on the same data set achieved a comparable RMSE of 0.012 on 6000 unseen test models after training 24000 models in 50 minutes using just 4 GPU (V100) machines, achieving nearly 20-fold improvement in computational efficiency.

In addition to magnetic and seismic data, the method is being developed for applications to electromagnetic and full tensor gravity gradiometer (FTG) data. Given the significant improvements in resolution and computational efficiency, it is expected that successful ground truth based field trials of AI based geophysical data inversion has the potential to unlock several new application areas while dramatically improving the business impact of such applications in existing ones.

How to cite: Mukherjee, S. and Lelievre, P.: Scalable machine learning solution for commercial scale three dimensional geophysical inversions, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-8297, https://doi.org/10.5194/egusphere-egu23-8297, 2023.

Applications of Machine Learning to the geosciences are increasing in numbers during the last two decades because of its computation power. In this work we propose a method to estimate the basement depth from gravity data using a supervised machine learning approach. We used the Bishop synthetic model to represent a variety of examples of input – target (i.e., gravity data – depth of the basement) training dataset. We so generated a large set of examples, using an overlapping moving window along the profiles in the N-S and E-W direction, associating the corresponding depth values of the basement to each of the windows. Due to its data-driven nature, the neural networks perform better as the number of examples provided in the training phase increases. However, increasing the number of examples leads to a higher computational cost in terms of speed and hardware needed. In the Big Data era this is not a huge issue, thanks to the increasingly present services of cloud computing. We found a good compromise on an average machine between speed and performance by using about 300k examples. A trial-and-error approach was used to find the hyperparameters that have the best compromise between performance and computation time.

We used, as a testing dataset, the gravity data due to a surface modelled from the Himalaya region DEM, with noisy and noise-free data. We found that this avoided overfitting and helped to verify the ability of the trained network to generalize to other cases, even with noisy data. The method was successfully applied also to a real dataset case: the isostatic anomaly of the Yucca Flat sedimentary basin (Nevada, USA) showing good agreement with previous inverse-modelling of the data, even if the author consider a set of layers with increasing density vs. depth, while in our case we used a mean density contrast of -0.7 g/cm³.

How to cite: Vitale, A., Gianluca, G., and Fedi, M.: Supervised machine learning to estimate the basement depth by gravity data, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-10036, https://doi.org/10.5194/egusphere-egu23-10036, 2023.

We propose a joint inversion method for potential fields aiming at recovering reliable models of complex source distributions, as those involving either shallow or deep-seated bodies. In this case, anomalies are characterized by different wavelength-contents, so that we may try to invert jointly the field and its higher-order vertical derivatives. To accomplish this task, we adopt a sequential strategy with a cross-gradient constraint. In this way, we can decouple the combined objective function into three terms: the field, its vertical-derivative and the cross-gradient constraint. For either separate or joint inversion we used a modified focusing algorithm, able to produce compact—source models and to incorporate different types of a-priori information (softer or harder constraints) to better address the ambiguity. The softer constraints include the model weighting function. Specifically, we used the inhomogeneous form of the model weighting function. On the other hand, we introduced harder constraints in the form of a reference model, which allows introducing other information from geology, previous geophysical interpretations or from well logs. The strength of the method is the applicability on both gravity and magnetic field to investigate different scenarios from small-scale (cavity detection) to basin-scale (resources exploration). In all the proposed cases, we obtained a significant model of the different sources at any depths. This is further demonstrated by a strong decrease in the cross-gradient values and a meaningful clusterization in the cross-plot of physical parameters.

How to cite: Bianco, L., Tavakoli, M., Vitale, A., and Fedi, M.: A multi-order joint inversion for potential field modelling., EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-6843, https://doi.org/10.5194/egusphere-egu23-6843, 2023.

Gravity inversion is a process that obtains the spatial structure and physical properties of underground anomalies using surface collected gravity anomaly data. The principle of gravity inversion based on deep learning (DL) is to learn the mapping between gravity anomaly data and geological models by training a neural network with geological models as labels. However, using DL inversion requires generating a large amount of training data for each geological target, resulting in a significant consumption of time and storage space. We propose to use a neural network to approximate the expensive forward computation with a fast evaluation alternative. After training, the network can reproduce gravity anomalies at any observation point. To evaluate the accuracy of the forward model, we use the gravity anomalies predicted by the forward network for inversion network training. In addition, to mitigate the problem of poor generalization of existing DL inversions, we propose to use multi-task learning (MTL). Learning multiple related tasks simultaneously improves the generalization ability of the model, thus improving the performance of the main task. In this paper, a multi-task UNet3+ network is proposed to realize anomaly bodies localization and density reconstruction simultaneously. The test results on the synthetic dataset show that the gravity anomalies predicted by the forward network can be successfully inverted, and the multi-task approach can predict the subsurface geology more accurately than the single-task UNet3+. To further illustrate the effectiveness of the algorithm, we apply the method to the inversion of the San Nicolas deposit in central Mexico, and the inversion results are consistent with known geological information.

How to cite: Lv, M. and Zhang, Y.: Fast forward approximation and multi-task inversion of gravity anomaly based on UNet3+, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-8211, https://doi.org/10.5194/egusphere-egu23-8211, 2023.

Since the European Union is foreseeing a great increase in the demand of critical raw materials in the coming decades, new ore deposits will have to be explored (European Commision 2020). This has given an incentive to renew the mineral exploration also in Latvia, with a few of the magnetic anomalies known to be caused by various metal ore deposits, all of which lie in the Proterozoic crystalline basement. One of these anomalies lie in the northern part of Latvia – the Strenči magnetic anomaly. A geophysical exploration, involving magnetic, gravity and seismic exploration, as well as core drilling and geochemical analysis, was carried out in the late 20^th century (Vetrennikov et al. 1986). Since then, no exploration has been done in the area. But the advent of powerful open-source modelling software which runs on consumer-grade computers has presented an opportunity to build new geophysical models based on old and new data.

To examine the possibility of developing a modern geophysical model of the Stenči magnetic anomaly, an open-source software SimPEG (Simulation and Parameter Estimation in Geophysics) was used (Cockett et al. 2015). The input data was Total Magnetic Intensity (TMI) measurements - a combination of the data from the exploration in the last century and newly acquired data. Quality control measurements of the previous data revealed uncertainty of ±135 nT, giving a rather large uncertainty of approximately 6% for the combined input TMI data. Based on the previous geophysical core logging data it was determined that the magnetic ore maintains a large remanent magnetization which may severely impact the geophysical model (REF). Thus, two models were developed: the first one based on the inversion of the magnetic susceptibility, but the second one using Magnetic Vector Inversion which takes into account remanent magnetization.

The developed models and data from previous research allowed to conclude that the magnetic anomaly is caused by metamorphosed granulite facies crystalline basement rocks of the Proterozoic to a depth of five kilometers. It was also discovered that the magnetization vectors coincide with the general direction of the dip of the rock strata. It was calculated that the predicted ore body contains a significant amount of critical minerals used in renewable energy technologies. A conclusion was made that the previous research has gathered a wealth of data, that can be used in regional crystalline basement research but not for detailed geophysical exploration of anomalies due to the uncertainty of the old data. The use of open-source software has enabled a very cost-effective development of sophisticated geophysical models which are impaired only by the quality of input data. Development of such models may be the first step into geophysical exploration which may attract interest for further investment.

This research was funded by “MikroTik” and University of Latvia Foundation, project no. 2258, and by the University of Latvia grant No. AAp2016/B041//Zd2016/AZ03 project “Climate change and sustainable use of natural resources”.

How to cite: Brants, M. and Karušs, J.: Geophysical modelling of Strenči magnetic anomaly in Latvia using SimPEG open-source software, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-6865, https://doi.org/10.5194/egusphere-egu23-6865, 2023.

East China Sea and its adjacent areas are an important part of the circum-Pacific tectonic belt, and its fluctuation characteristics of Moho can provide an important basis for the study of the deep structure of the western Pacific. At present, the inversion of Moho depth based on potential field data is an important task. Meanwhile, its inversion accuracy is closely related to the gravity anomaly data quality, and to the density contrasts and inversion algorithm. We use the fast solution algorithm of forward problem for gravity field in a dual interface model to eliminate the gravity influence of terrain and sediments, and adopt the minimum curvature potential field separation method to remove the effect of residual geological bodies. Then we try to identify the Moho gravity anomaly as the regional field which has the strongest correlation with depths estimations from seismic data. Regression analysis and the "3σ" principle are used to delete the constraint points of Moho depth with large deviations, and the Bouguer plate formula is used to estimate the laterally variable density contrasts of Moho. Finally, the Moho depth in East China Sea and its adjacent areas is obtained by the dual-interface fast inversion algorithm, and the inversion deviations are mostly concentrated within 2 km. The inversion result shows that there is an obvious local uplift zone of Moho in East China Sea Basin, while the Okinawa Trough basin is located on a whole Moho uplift. The strike of the two uplift belts has the same change from NE to NNE.

How to cite: Wang, D., Wang, W., and Zhang, Y.: Moho Inversion of East China Sea and Its Adjacent Areas Based on Potential Field Data, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-14706, https://doi.org/10.5194/egusphere-egu23-14706, 2023.

Geothermal heat flux (GHF) is a critical basal boundary condition that exerts important influences on the initiation of flow of the Antarctic ice sheet and is related to crustal and lithospheric structure and composition. Despite its importance, our knowledge of Antarctic GHF heterogeneity remains limited and this hinders interdisciplinary efforts to better constrain Solid Earth influences on subglacial hydrology and ice sheet behaviour.

Within the framework of the 4D Antarctica ESA project we produced a new continent-wide aeromagnetic anomaly compilation for Antarctica, conformed at longer wavelengths with SWARM satellite magnetic data. It includes recent data collected after the ADMAP 2.0 compilation, over the Ross Ice Shelf, the Amundsen Sea Embayment and the Recovery and South Pole regions, as well as enhanced maps for the Gamburtsev Subglacial Mountains and Wilkes and Dome C regions, based on relevelling.

We applied Curie Depth Point (CDP) estimation using the centroid, modified centroid and fractal and defractal approaches. We tested different window sizes at continental scale and for detailed analysis (200x200 km; 300x300 km; 400x400 km) and centroid distances, and both automated ranges and hand-picked intervals over selected features. The estimates reveal regions of enhanced GHF along the coast of the Amundsen Sea Embayment, in general agreement with independent seismological estimates, and are interpreted as reflecting dynamic interactions between the West Antarctic Rift System and anomalously warm Pacific upper mantle at depth. A higher degree of continuity of potential thermal anomalies related to the Byrd Subglacial Basin is evident between the Thwaites and Pine Island catchments compared to a recent magnetic estimate (Dziadek et al., 2021). A large area of enhanced GHF under the Siple Coast ice streams and active subglacial lake districts is confirmed, but has lower values and greater complexity than previously imaged (Martos et al., 2017). This can be correlated with regions of thinner and thicker crust and different magnetic patterns as revealed from inversion of satellite and airborne gravity and aeromagnetic data respectively.

In East Antarctica, the new CDP estimates suggest that any Mesozoic to Cenozoic extension is restricted to upper crustal levels and is focussed in narrow regions. Intriguing, relatively shallow CDP anomalies (given their location within the composite East Antarctic craton) are revealed in the Dome C lake district and also Gamburtsev Subglacial Mountains lake district regions. These may speculatively stem either from intraplate Mesozoic to Cenozoic fault reactivation and/or enhanced intracrustal heat production. 

We conclude that our new Curie depth estimates yield geologically reasonable thermal boundary conditions, which can be used together with independent estimates derived e.g. from seismology, machine leaning and multi-variate analysis to initialise new thermal models that incorporate crust and lithosphere thickness variations and intracrustal composition (as a proxy for ranges of radiogenic heat production and thermal conductivity).

How to cite: Ferraccioli, F., Ford, J., Dziadek, R., Mather, B., Armadillo, E., Ebbing, J., Eagles, G., Gohl, K., Forsberg, R., Green, C., Fullea, J., and Verdoya, M.: New Curie depth estimates from satellite conformed aeromagnetic anomaly compilations and their implications for assessing Antarctic subglacial geothermal heat flux heterogeneity , EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-14724, https://doi.org/10.5194/egusphere-egu23-14724, 2023.

We estimate the depth of the Curie isotherm and the associated heat flow for the Circum-Arctic Region using a Monte Carlo Markov Chain approach. In the first step, Curie depths are determined where heat flow measurements are available. For the depth estimates, different parameters and concepts are tested (e.g. pure conduction compared to half-space cooling) in order to assess the uncertainty underlying the depth estimates, but also of the observing point. Hereby, we rely on existing models of the Arctic lithosphere including ArcCRUST and LithoRef18. Half of the calculated Curie depth points show a low sensitivity to the choice of the parameters and models and hence can be regarded as stable, representing the thermal field of the lithosphere and not local effects. Hence, we can use these points as constraints for the second step, where we invert an aeromagnetic anomaly map for both the Curie depth and susceptibility for the Circum-Arctic Region. The new model shows that in areas where reliable constraints exist, the magnetic inversion is preferring to explain the magnetic anomalies with lateral susceptibility distribution, reflecting hereby the main geological features of the region.

How to cite: Freienstein, J., Szwillus, W., and Ebbing, J.: Estimating Curie depth and Heat Flow in the Circum-Arctic Region, EGU General Assembly 2023, Vienna, Austria, 23–28 Apr 2023, EGU23-3308, https://doi.org/10.5194/egusphere-egu23-3308, 2023.