Towards Data-Driven Evaluation of Site Effects From Coda of Strong Motion Timeseries Using Recurrent Neural Networks

It is known that coda of strong motion records are products of numerous scatterings of body and surface waves within the subsurface soil structure. Several studies have successfully simulated coda wave envelopes by modelling the energy decay. However, due to limitations of quantifying soil heterogeneities, deterministic simulation of scattered wavefields is much more challenging. Therefore, the reverse problem of estimating non statistical properties of subsurface structure from coda waves remains a theoretical potential. On the other hand, machine learning techniques have proven useful in dealing with problems of similar nature, where a theoretical solution is imaginable yet hard to achieve due to a great number of unknown variables. This study utilizes artificial neural networks to propose a new approach of evaluating site effects from coda waves, with the future prospect of obtaining the similar results from microtremor records. A Long Short-Term Memory recurrent neural network is designed using Tensorflow 2 library in Python language. The study utilizes a strong motion dataset consisting of about 60000 3-component records as well as borehole data at 464 stations of Kiban-Kyoshin Network across Japan. The prediction input is coda wave timeseries of strong motion records, defined based on a parametric energy criterion, and all 3 seismograph components EW, NS and UD are used as parallel sequential features. In the first step, the prediction target is a vector of 3 site effect proxies namely, time-averaged shear-wave velocities for the upper 30-m depth Vs30 and the upper 10-m depth Vs10 and predominant frequency f0. In this step, different model parameter combinations are tested to ensure the basic model’s ability in extracting site-specific information from the input coda waves. One of the combinations is then used in the second step, in which the prediction target is surface to downhole ratio of Fourier Amplitude Spectra. For each of the 3 components EW, NS and UD, 100 identical networks are trained to each predict the desired ratio at a certain target frequency. Accuracy of test sample predictions confirms applicability of the proposed approach as well as its potential for future works on microtremor timeseries instead of coda waves.

An Evaluation of Strong-Motion Parameters at the S-net Ocean-Bottom Seismograph Sites Near the Kanto Basin for Earthquake Early Warning

We analyzed strong-motion records at the ground and borehole in and around the Kanto Basin and the seafloor in the Japan Trench area from three nearby offshore earthquakes of similar magnitudes (Mw 5.8–5.9). The seafloor strong-motion records were obtained from S-net, which was established to enhance tsunami and earthquake early warnings after the 2011 great Tohoku-oki earthquake disaster. The borehole records were obtained from MeSO-net, a dense network of seismometers installed at a depth of 20 m in the Tokyo metropolitan area. The ground records were obtained from the K-NET and KiK-net networks, established after the 1995 great Hanshin-Awaji earthquake disaster. The MeSO-net and S-net stations record the shakings continuously, while the K-NET and KiK-net records are based on triggering thresholds. It is crucial to evaluate the properties of strong motions recorded by the S-net for earthquake early warning (EEW). This paper compared the peak ground accelerations (PGAs) and peak ground velocities (PGVs) between the S-net and K-NET/KiK-net stations. Because the MeSO-net records were from the borehole, we compared the PGAs and significant durations of the low-frequency motions (0.1–0.5 Hz) between the S-net and MeSO-net stations from identical record lengths. We found that the horizontal PGAs and PGVs at the S-net sites were similar to or larger than the K-NET/KiK-net sites for the S wave. In contrast, the vertical PGAs and PGVs at the S-net sites were similar to or smaller than those at the K-NET/KiK-net sites for the S wave. Particularly, the PGAs and PGVs for the P-wave parts on the vertical-component records of S-net were, on average, much smaller than those of K-NET/KiK-net records. The difference was more evident in the PGAs. The average ratios of S-wave horizontal to vertical PGAs were about 2.5 and 5 for the land and S-net sites, respectively. The low-frequency PGAs at the S-net sites were similar to or larger than those of the MeSO-net borehole records. The significant durations between the two-networks low-frequency records were generally comparable. Quantification of the results from a larger dataset may contribute to ground-motion prediction for EEW and the design of the offshore facilities.

What Is Fling Step? Its Theory, Simulation Method, and Applications to Strong Ground Motion near Surface Fault Ruptures

ABSTRACT We present the theory of the fling step and a theoretical method for simulating accurately the near-fault strong motions, and apply it to reproduce various strong-motion records near surface faults. Theoretically, the fling step is the contribution of the static Green’s function in the representation theorem (Hisada and Bielak, 2003), and we show that this theory holds for any seismic velocity structure. We first demonstrate the validity of this theory using theoretical solutions of a circular fault model in a homogeneous full-space. Next, we apply the theory to layered half-spaces, present a theoretical method based on the wavenumber integration method, and introduce various techniques to simulate the near-fault ground motions including fling steps with high accuracy. Finally, we demonstrate the effectiveness of the method by reproducing various strong-motion records near surface fault ruptures and discuss the characteristics of near-fault strong motions including the fling step and the forward directivity pulse. We made all of the software and data used in this article available on the internet.

Rupture process of the 7 May 2020 Mw 5.0 Tehran earthquake and its relation with the Damavand stratovolcano, and Mosha Fault

<p>In May 2020 an earthquake with Mw 5.0 struck at ~40 km east of Tehran metropolis and ~15 km south of the Damavand stratovolcano. It was responsible for 2 casualties and 23 injured. The mainshock was preceded by a foreshock with Ml 2.9 and followed by a significant aftershock sequence, including ten events with Ml 3+. The occurrence of this event raised the question of its relation with volcanic activities and/or concern about the occurrence of larger future earthquakes in the capital of Iran. Tehran megacity is surrounded by several inner-city and adjacent active faults that correspond to high-risk seismic sources in the area. The Mosha fault with ~150 km long is one of the major active faults in central Alborz and east of Tehran. It has hosted several historical earthquakes (i.e. 1665 Mw 6.5 and 1830 Mw 7.1 earthquakes) in the vicinity of the 2020 Mw 5.0 Tehran earthquake’s hypocenter. In this study, we evaluate the seismic sequence of the Tehran earthquake and obtain the full moment tensor inversion of this event and its larger aftershocks, which is a key tool to discriminate between tectonic and volcanic earthquakes. Furthermore, we obtain a robust characterization of the finite fault model of this event applying probabilistic earthquake source inversion framework using near-field strong-motion records and broadband seismograms, with an estimation of the uncertainties of source parameters. Due to the relatively weak magnitude and deeper centroid depth (~12 km), no static surface displacement was observed in the coseismic interferograms, and modeling performed by seismic records. Focal mechanism solution from waveform inversion, with a significant double-couple component, is compatible with the orientation of the sinistral north-dipping Mosha fault at the centroid location. The finite fault model suggests that the mainshock rupture propagated towards the northwest. This directivity enhanced the peak acceleration in the direction of rupture propagation, observed in strong-motion records. The 2020 moderate magnitude earthquake with 2 casualties, highlights the necessity of high-resolution seismic monitoring in the capital of Iran, which is exposed to a risk of destructive earthquakes with more than 10 million population. Our results are important for the hazard and risk assessment, and the forthcoming earthquake early warning system development in Tehran metropolis.</p>

Magnitude Calculation without Saturation from Strong-Motion Waveforms

ABSTRACT After destructive earthquakes, it is a challenge to estimate magnitude rapidly and accurately for dissemination to emergency responders and the public. Here, we propose criteria to calculate peak ground displacement (PGD) from strong-motion records, which can be used to calculate unsaturated event magnitude. Using collocated strong-motion and Global Navigation Satellite Systems observations of five major earthquakes in Japan, we demonstrate the effectiveness and accuracy of our strategy. Our results show that, with the right filtering criteria, PGD estimated from strong-motion acceleration waveforms is consistent with geodetic estimates. The methodology, however, does not allow for calculation of reliable estimates of coseismic deformation or other ground displacement metrics. We demonstrate a simulated real-time magnitude estimation that suggests it is feasible to generate an unsaturated magnitude estimate in real time from near-field strong-motion records. These findings have important implications for early warning and emergency response in seismically active areas, especially where real-time strong-motion data are more widely available than geodetic measurements.