Electromagnetic Signal Associated with Seismic Waves: Case Study in the North Central Algeria Area

Abstract

This study aimed to investigate the temporal variations in electromagnetic signals related to crustal tectonic activity in northern Algeria, using an intermittent magnetotellurics (MT) measurement system that was launched in December 2014. On 23 December 2014, a moderate earthquake (Mw = 4.9) occurred at 08:00:21 (UT) in the village of Hammam Melouane, exactly 36.3 km northeast of the magnetotellurics station. This seismic event provided a unique opportunity to observe, clearly and distinctly, a co-seismic electromagnetic signal from the natural electromagnetic field in the magnetotellurics time series. Analysis of the recorded signals revealed that the co-seismic electromagnetic signal was synchronized with the P-wave arrival time. Based on the time–frequency misfit criteria analysis method, we conducted a quantitative comparison between the co-seismic telluric and magnetic horizontal component time series, in addition to the horizontal components of a seismic trace record, which were collected from the seismological station nearest the magnetotelluric site. Good waveform similarity was found between the electromagnetic and seismic signal components, with a high temporal frequency misfit envelope and a high fit value. After the arrival of the seismic waves, the north–south co-seismic electric and magnetic component shifted to a higher frequency compared to the east–west component, which was similar to the seismic velocity components. Thus, to investigate the generating mechanism, a one-dimensional geoelectric model was derived using the magnetotelluric data, and the polarization trajectories of the co-seismic electromagnetic signal were analyzed. The results indicated that the electrokinetic effect was likely the mechanism generating the observed co-seismic electromagnetic signals.

Appendix A

In this section, we outline the data processing and inversion of the collected magnetotelluric records to derive the 1D geoelectric model in the vicinity of the MT station. To this end, we selected the time-series records of 7–8, 11–12, and 14–15th December 2014 due to their high quality. Then, we carried out the time series processing to extract the MT transfer functions, i.e., the apparent resistivity and the impedances phases in a 1000–0.001 Hz frequency band through the robust processing code of Jones and Jödicke (1984). The obtained apparent resistivity and phase of the MT impedances are of acceptable quality despite the presence of anthropogenic noise and use of a single measurement station. The apparent resistivity and phase data of the four impedance tensor elements are illustrated in Fig. 7, while the measurement axes were rotated to the N65° E direction. The apparent resistivity curves of the diagonal elements are quite low compared to those of the anti-diagonal elements, indicating a non-3D underlying structure and the presence of low distortion, except at very low frequencies (< 0.01 Hz). 

Fig. 7

Apparent resistivity and phase data of the impedance tensor off-diagonal elements (a, b) and diagonal elements (c, d). Here, the measuring axes were rotated to the N65° E direction, i.e., the x-axis in the N65° E and the y-axis in the N155° E directions

To more formally identify dimensionality and strike, the phase tensor method was used (Caldwell et al., 2004). This phase tensor (PT) is defined by the ratio between the imaginary and real parts of the MT impedance tensor. The PT is therefore unaffected by magnetotelluric distortion, and thus it represents only the regional geoelectric structure. For each frequency, the PT can be represented by an ellipse, whose major axis gives the strike direction (with 90° ambiguity). So, in case of a layered terrain, the ellipse becomes a circle. Additionally, to distinguish between a 2D and 3D earth, Caldwell et al. (2004) suggested a parameter called β-skew, in which a small β-skew (|β|< 5°) indicates a 2D geological structure; otherwise the structure is 3D. Moreover, a parameter named ellipticity (λ) was proposed to distinguish a 1D structure (λ < 0.1) from a 2D structure (λ > 0.1). Thus, the selected structure applied to our data showed that the PT method reveals a stratified ground up to the frequency of 1 Hz (high frequencies), which corresponds to a ~ 1 km penetration depth, then 2D beyond (low frequencies); however, we note a strike change around the 0.03 Hz frequency corresponding to a ~ 2 km depth of penetration. Certainly, the strike changes from the N20° E direction between the frequencies of 1 Hz and 0.03 Hz, to the N65° E direction for low frequencies. The latter direction is consistent with observed regional geological discontinuities on the geological map (major NE–SW faults; Fig. 1).

With data submitted by the single MT station, we attempted to approximate the regional geoelectric structure by a 1D model obtained by inverting the apparent resistivity and the phase data of the two principal components of the MT impedance tensor. Hence, the orientation was set to N65° E; i.e., the measurement axes were rotated in this direction, in which the x-axis was oriented to N65° E and the y-axis to N155° E. After that, the apparent resistivity and phase data are then recalculated in the new axis system, while the impedance tensor xy components are defined by Ex parallel and Hy perpendicular to the strike. Also, the values corresponding to the electrical polarization (or TE mode) and the yx component are defined by Ey perpendicular and Hx parallel to the strike, and the magnetic polarization (or TM mode). Then, the apparent resistivity and phase data of the two principal components of the impedance tensor were inverted using the Marquardt algorithm (Marquardt, 1963). This inversion algorithm is used to iteratively improve an initial model by minimizing the root mean square (RMS) between the observed and predicted (model response) apparent resistivity and phase data. Lastly, the iterative process is stopped and the solution is then given by the obtained resistivity model only when the RMS reaches an acceptable value. After several trials, we selected a launching model consisting of a minimum of layers, which will be improved by the inversion algorithm, obtaining a 1D resistivity model comprising five layers and an investigation depth estimated at 10 km (Fig. 5).