Advances in transient electromagnetic methods

As a result of recent developments in both computer and measurement hardware and mathematical algorithms, there have been significant advances in traditional data acquisition instrumentation, particularly in sensor capabilities, computer modelling, inversion algorithms and software, and machine-learning-based noise reduction and data interpretation methods for transient electromagnetic (TEM) methods. Drones are also becoming more available and powerful, so drone-based survey methods and instruments are also being actively developed. However, there are still some problems that need to be solved through active research. Three-dimensional (3D) inversions are still very computationally intensive. Processing data in real time is not trivial, especially for large-volume data collected during aerial surveys. Therefore, there is still a heavy reliance on 1D interpretation methods. Data collected from surveys in urban areas are prone to significant cultural noise, which needs to be properly denoised before interpretation. The theme of this special issue, ‘Advances in transient electromagnetic methods’, focuses on recent research and presents new theories and applications with the aim of presenting recent advances in addressing key issues in time-domain geophysical electromagnetic methods. The collection of 13 papers mainly covers the following four subtopics. Li et al. (2023a) present an approach to detect unfavourable geological boulders in tunnels using a novel TEM configuration. In their innovative method, an electric dipole is inserted into a hole drilled from the tunnel face in the direction of the tunnel construction. Electromagnetic data are then collected at the tunnel face. Through extensive forward modelling, Li et al. successfully demonstrate the feasibility of this configuration for accurate exploration of small, highly conductive boulders. Lei et al. (2023) conduct a comprehensive study of the shadow effect and source overprint effect in short-offset TEM (SOTEM) methods using forward modelling. Their study provides two important results: (i) the relative anomaly caused by the SOTEM shadow effect is about twice as large as the anomaly caused by the SOTEM source overprint effect; and (ii) the source overprint effect manifests itself predominantly in the all-time ∂bz/∂t data and the early ex data. Yu et al. (2023) present a neural network architecture specifically designed for processing 2D TEM data. Their proposed structure integrates a classical convolutional neural network denoising autoencoder with a gated recurrent neural network autoencoder. This combined architecture enables the direct input of 2D TEM response data as images into the network, significantly enhancing data processing efficiency compared to traditional single-time-channel processing methods. Peng et al. (2023) present a novel algorithm designed for data denoising in the context of the airborne TEM method. Their algorithm, known as noise-whitening-based weighted nuclear norm minimization, focuses on recovering valuable anomalous information that may be obscured by noise in late time gates. To tackle this issue, the algorithm uses weighted nuclear norm minimization, which seeks to find a low-rank approximation of the data matrix while taking into account the estimated noise characteristics. Zhao et al. (2023) propose a novel deep-learning method for wavefield transformation using U-Net. Zhao et al. establish a nonlinear mapping from the electromagnetic diffusion field to the pseudo seismic wave field by training the network with the same velocity model. This innovative approach eliminates the need to solve a set of ill-posed linear equations that are typically encountered in traditional methods. Li et al. (2023b) develop a 1D inversion method capable of handling the effects of transmitting-current waveforms, including waveform repetitions. Their research demonstrates that the magnetic induction, bz, is less susceptible to background noise but more influenced by full-waveform effects compared to its time derivative, ∂bz/∂t. Li et al. then apply the 1D full-waveform inversion method to delineate the geological structures present at the Nantong mudflat in China. Xian et al. (2023) present an approach aiming to improve the resolution of thin layers by introducing an adaptive roughness matrix calculation method. This approach is specifically developed for the semi-airborne TEM method and undergoes comprehensive evaluation using synthetic models and field data. Lv et al. (2023) introduce an innovative approach that leverages transfer learning for the fusion of multisource geophysical data. Their method involves using the ResNet50 network to extract initial features from multisource geophysical images. To further enhance the features and reduce their dimensionality, Spearman correlation analysis and zero phase component analysis are employed. Finally, fusion rules are applied to obtain a fused image that combines the information from different geophysical sources. Wu et al. (2022) conduct a field case study using bz data measured by a superconducting quantum interference device magnetometer. The collected bz data are used to generate apparent-resistivity section views, which are then employed to identify ore-bearing strata and ore-controlling structures in the Baiyun gold deposits located in China. Wang et al. (2023) provide a summary of the relationships between the horizontal and vertical components of the impulse response ∂bz/∂t. Wang et al. apply these relationships to the Heicigou gold deposit located in east Kunlun, China. The findings of their study reveal that the ore-bearing alteration zone in the deposit is wide above and narrow below. This zone is situated between the upper and lower walls of the fault and exhibits characteristics of medium resistivity. Yang et al. (2023) conducted a field case study focusing on the detection of coal-mining goaf areas using a semi-airborne TEM method. They employ a high-precision wavefield imaging method for data processing and imaging. Through their study, the researchers successfully identify and characterize the location and interface information of the coal mine goaf areas. Mohamed et al. (2023) carry out a study at the archaeological site of Tell el-Rub'a in Egypt, where they uncover buried historical harbours using frequency-domain terrain conductivity measurements and TEM sounding. To process and interpret the collected data, Mohamed et al. employ a 1D inversion method. Song et al. (2023) conduct a case study focused on the application of SOTEM surveys for investigating urban active faults in Hebei, China. The study successfully identifies two buried faults through the analysis of 1D inversion results. The papers in this special issue primarily emphasize data processing and inversion techniques for SOTEM and semi-airborne TEM, along with the application of machine-learning techniques for TEM data de-noising and imaging. In the case studies presented, the mainstream approach for interpreting field data still revolves around 1D data imaging and inversion. It is worth noting that while 3D inversion is a widely discussed topic in TEM research, this special issue does not include any articles specifically addressing this aspect, suggesting that 3D inversion is not yet practicable enough to be useful for real-life situations. Conflict of interest statement. The authors declare that there are no conflicts of interest related to the research, authorship, and publication of this article.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.