Earthquake Early Warning Research Articles

Earthquake multi-classification detection based velocity and displacement data filtering using machine learning algorithms.

Earthquake is one of the natural disasters that have a big impact on society. Currently, there are many studies on earthquake detection. However, the vibrations that were detected by sensors were not only vibrations caused by the earthquake, but also other vibrations. Therefore, this study proposed an earthquake multi-classification detection with machine learning algorithms that can distinguish earthquake and non-earthquake, and vandalism vibration using acceleration seismic waves. In addition, velocity and displacement as integration products of acceleration have been considered additional features to improve the performances of machine learning algorithms. Several machine learning algorithms such as Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), and Artificial Neural Network (ANN) have been used to develop the best algorithm for earthquake multi-classification detection. The results of this study indicate that the ANN algorithm is the best algorithm to distinguish between earthquake and non-earthquake, and vandalism vibrations. Moreover, it's also more resistant to various input features. Furthermore, using velocity and displacement as additional features has been proven to increase the performance of every model.

Real-time prediction of earthquake potential damage: A case study for the January 8, 2022 MS 6.9 Menyuan earthquake in Qinghai, China

It is critical to determine whether a site has potential damage in real-time after an earthquake occurs, which is a challenge in earthquake disaster reduction. Here, we propose a real-time Earthquake Potential Damage predictor (EPDor) based on predicting peak ground velocities (PGVs) of sites. The EPDor is composed of three parts: (1) predicting the magnitude of an earthquake and PGVs of triggered stations based on the machine learning prediction models; (2) predicting the PGVs at distant sites based on the empirical ground motion prediction equation; (3) generating the PGV map through predicting the PGV of each grid point based on an interpolation process of weighted average based on the predicted values in (1) and (2). We apply the EPDor to the 2022 MS 6.9 Menyuan earthquake in Qinghai Province, China to predict its potential damage. Within the initial few seconds after the first station is triggered, the EPDor can determine directly whether there is potential damage for some sites to a certain degree. Hence, we infer that the EPDor has potential application for future earthquakes. Meanwhile, it also has potential in Chinese earthquake early warning system.

Comparing and integrating artificial intelligence and similarity search detection techniques: application to seismic sequences in Southern Italy

SUMMARYUnderstanding mechanical processes occurring on faults requires detailed information on the microseismicity that can be enhanced today by advanced techniques for earthquake detection. This problem is challenging when the seismicity rate is low and most of the earthquakes occur at depth. In this study, we compare three detection techniques, the autocorrelation FAST, the machine learning EQTransformer, and the template matching EQCorrScan, to assess their ability to improve catalogues associated with seismic sequences in the normal fault system of Southern Apennines (Italy) using data from the Irpinia Near Fault Observatory (INFO). We found that the integration of the machine learning and template matching detectors, the former providing templates for the cross-correlation, largely outperforms techniques based on autocorrelation and machine learning alone, featuring an enrichment of the automatic and manual catalogues of factors 21 and 7, respectively. Since output catalogues can be polluted by many false positives, we applied refined event selection based on the cumulative distribution of their similarity level. We can thus clean up the detection lists and analyse final subsets dominated by real events. The magnitude of completeness decreases by more than one unit compared to the reference value for the network. We report b-values associated with sequences smaller than the average, likely corresponding to larger differential stresses than for the background seismicity of the area. For all the analysed sequences, we found that main events are anticipated by foreshocks, indicating a possible preparation process for main shocks at subkilometric scales.

Enhancing Earthquake Detection from Orbit

A new application of machine learning boosts scientists’ ability to use data from satellite navigation systems to detect and warn of earthquakes.

Photonic Integrated Sensing and Communication System Harnessing Submarine Fiber Optic Cables for Coastal Event Monitoring

Information and knowledge explosions boost the development of telecommunication infrastructures, especially for submarine fiber optic cables. These underwater fibers not only undertake most of the international data transmission on this planet, but also offer new opportunities to study earth science and human marine activities. The emerging technology of integrated sensing and communication is putting this concept into practice. In this article, we present work on coastal event monitoring with photonic integrated sensing and communication technology using submarine telecommunication optical cables. Ocean wave analysis, earthquake detection, and marine ferry monitoring are highlighted for detailed analysis. With the proposed integrated system, we successfully observed ocean waves with different frequencies and velocities, microseisms within 200 km, Mw 4.3 - Mw 6.3 earthquakes in the Far East, and vessel sailing activities. Enabled by the novel integrated technology and mass submarine fiber networks, more research works on photonic integrated sensing and communication systems aiming at ocean science and marine engineering are expected for future insights.

Rupture process and aftershock focal mechanisms of the 2022 M6.8 Luding earthquake in Sichuan

According to the China Earthquake Networks Center, a strong earthquake of M6.8 occurred in Luding County, Ganzi Tibetan Autonomous Prefecture, Sichuan Province, China (102.08°E, 29.59°N), on September 5, 2022, with a focal depth of 16 km. Rapid determination of the source parameters of the earthquake sequence is vital for post-earthquake rescue, disaster assessment, and scientific research. Near-field seismic observations play a key role in the fast and reliable determination of earthquake source parameters. The numerous broadband seismic stations and strong-motion stations recently deployed by the National Earthquake Intensity Rapid Report and Early Warning project have provided valuable real-time near-field observation data. Using these near-field observations and conventional mid- and far-field seismic waveform records, we obtained the focal mechanism solutions of the mainshock and M ≥ 3.0 aftershocks through the waveform fitting method. We were further able to rapidly invert the rupture process of the mainshock. Based on the evaluation of the focal mechanism solution of the mainshock and the regional tectonic setting, we speculate that the Xianshuihe fault formed the seismogenic structure of the M6.8 strong earthquake. The aftershocks formed three spatially separated clusters with distinctly different focal mechanisms, reflecting the segmented nature of the Xianshuihe fault. As more high-frequency information has been applied in this study, the absolute location of the fault rupture is better constrained by the near-field strong-motion data. The rupture process of the mainshock correlates well with the spatial distribution of aftershocks, i.e., aftershock activities were relatively weak in the maximum slip area, and strong aftershock activities were distributed in the peripheral regions.

Research on Sparse Denoising of Strong Earthquakes Early Warning Based on MEMS Accelerometers.

In view of the fact that the noise in the same frequency band as the useful signal in the MEMS acceleration sensor observation data cannot be effectively removed by traditional filtering methods, a denoising method for strong earthquake signals based on the theory of sparse representation and compressive sensing is proposed in this paper. This skillfully realized the separation of strong earthquake signals from noise by adopting a fixed dictionary and utilizing sparse characteristics. Furthermore, considering the weakness of the sparse denoising method based on the fixed dictionary in the high signal-to-noise ratio, a spare denoising method based on learning an over-complete dictionary is proposed. Through the initial given seismic data, the ideal over-complete dictionary is trained to achieve seismic data denoising. In addition, for the interference waves of non-seismic events, this paper proposes an idea based on sparse representation classification to remove such non-seismic interference directly. Combining the ideas of noise reduction and non-seismic event elimination, we can obtain a standard sparse anti-interference denoising model for earthquake early warning. It's innovative that this model implements the sparse theory into the field of earthquake early warning. According to the experimental results, in the case of heavy noise, the denoising model based on sparse representation can reach average SNR of 8.73 and an average MSE of 29.53, and the denoising model based on compression perception can reach average SNR of 7.29 and an average MSE 41.34, and the denoising model based on learning dictionary can reach average SNR 11.07 and average MSE 17.32. The performance of these models is better than the traditional FIR filtering method (average SNR -0.73 and average MSE 260.37) or IIR filtering method (average SNR 4.73 and average MSE 73.95). On the other hand, the anti-interference method of the sparse classification proposed in this paper can accurately distinguish non-seismic interference events from natural earthquakes. The classification accuracy of the method based on the noise category of the selected test data set reaches 100% and achieves good results.

Enhanced Regional Earthquake Catalog with Alaska Amphibious Community Seismic Experiment Data

Abstract The Alaska Amphibious Community Seismic Experiment (AACSE) comprised 75 ocean-bottom seismometers and 30 land stations and covered about 650 km along the segment of the subduction zone that includes Kodiak Island, the Alaska Peninsula and the Shumagin Islands between May 2018 and September 2019. This unprecedented onshore-offshore dataset provided an opportunity to compile a greatly enhanced earthquake catalog for the region by both increasing the number of detected earthquakes and improving the accuracy of their source parameters. We use all available regional and AACSE campaign seismic data to compile an earthquake catalog for the region between Kodiak and the Shumagin Islands including the Alaska Peninsula (51° N–59° N, 148° W–163° W). We apply the same processing and reporting standards to additional picks and events as the Alaska Earthquake Center currently uses for compilation of the authoritative regional earthquake catalog. Over 7200 events (both newly detected and previously reported) have been processed with AACSE data. We added about 30% more events, 60% more phase picks, lowered the magnitude of completeness by about 0.2 on average across the region, and improved location errors. All data have been published in public data archives. In addition, we test the machine-learning earthquake detection and picking algorithm EarthquakeTransformer (EQT) on the AACSE seismic dataset, comparing EQT-determined P and S picks with the new catalog. EQT is entirely trained on land data, whereas AACSE is amphibious. Overall, EQT finds 59% of P and 63% of S arrivals in the catalog within 300 km epicentral distance. The percent of catalog picks detected by EQT varies inversely with earthquake epicentral distance, and EQT performs particularly poorly on data from earthquakes recorded by instruments in the outer rise.

EPick: Attention-based multi-scale UNet for earthquake detection and seismic phase picking

Earthquake detection and seismic phase picking play a crucial role in the travel-time estimation of P and S waves, which is an important step in locating the hypocenter of an event. The phase-arrival time is usually picked manually. However, its capacity is restricted by available resources and time. Moreover, noisy seismic data present an additional challenge for fast and accurate phase picking. We propose a deep learning-based model, EPick, as a rapid and robust alternative for seismic event detection and phase picking. By incorporating the attention mechanism into UNet, EPick can address different levels of deep features, and the decoder can take full advantage of the multi-scale features learned from the encoder part to achieve precise phase picking. Experimental results demonstrate that EPick achieves 98.80% accuracy in earthquake detection over the STA/LTA with 80% accuracy, and for phase arrival time picking, EPick reduces the absolute mean errors of P- and S- phase picking from 0.072 s (AR picker) to 0.030 s and from 0.189 s (AR picker) to 0.083 s, respectively. The result of the model generalization test shows EPick’s robustness when tested on a different seismic dataset.

Earthquake Event Recognition on Smartphones Based on Neural Network Models.

Using sensors embedded in smartphones to study earthquake early warning (EEW) technology can effectively reduce the high construction and maintenance costs of traditional EEW systems. However, due to the impact of human activities, it is very difficult to accurately detect seismic events recorded on mobile phones. In this paper, to improve the detection accuracy of earthquakes on mobile phones, we investigated the suitability of different types of neural network models in seismic event detection. Firstly, we collected three-component acceleration records corresponding to human activities in various scenarios such as walking, running, and cycling through our self-developed mobile application. Combined with traditional strong-motion seismic event records fusing typical mobile phone accelerometer self-noise, all records were used for establishing the training and testing dataset. Finally, two types of neural network models, fully connected and convolutional neural networks, were trained, validated, and tested. The results showed that the accuracy rates of the neural network models were all over 98%, and the precision rate for seismic events and the recall rate for non-earthquake events could both reach 99%, indicating that the introduction of neural networks into the earthquake recognition on smartphones can significantly enhance the accuracy of seismic event recognition. Therefore, we can exceedingly reduce the amount of data transmitted to the processing server, further lowering the load on the server processor and effectively increasing the lead time at each target site for an EEW system.

Earthquake early warning in countries where damaging earthquakes only occur every 50 to 150 years – The societal perspective

Earthquake Early Warning (EEW) systems can trigger actions just before the shaking hits and hence mitigate the impact of damaging earthquakes. Recent studies have mainly focused on technical challenges, performance or on the attitudes of the public towards EEW systems in countries where EEW systems are already in operation or where damaging earthquakes are frequent. Building on these efforts, from the societal perspective we assess how different message elements influence the intention to take action and ability to grasp relevant information within a few seconds. Further, we explore the public attitudes towards EEW systems in a country where damaging earthquakes are expected less frequently. In a survey (N = 596, between-subjects experiment) targeting the Swiss public, we assess EEW system preferences and test different versions of EEW messages to identify elements which trigger people to take immediate actions and can be correctly recalled. Our main findings are that i) the public attitudes in countries with moderate seismic hazard are similar to attitudes in countries with high seismic hazard; ii) pictograms trigger people to protect themselves on the spot whereas maps prompt the public to look for further information or to warn others; iii) the designs preferred by the public are not always those that actually trigger them to take action; and iv) people tend to react proportionally to the hazard level. We show that insights from studies focusing on societal issues related to EEW systems can help operators define technical settings and to address specific concerns or likely misconceptions in education campaigns.

A study on the effect of input data length on a deep-learning-based magnitude classifier

Abstract. The rapid characterisation of earthquake parameters such as its magnitude is at the heart of earthquake early warning (EEW). In traditional EEW methods, the robustness in the estimation of earthquake parameters has been observed to increase with the length of input data. Since time is a crucial factor in EEW applications, in this paper we propose a deep-learning-based magnitude classifier based on data from a single seismic station and further investigate the effect of using five different durations of seismic waveform data after first P-wave arrival: 1, 3, 10, 20 and 30 s. This is accomplished by testing the performance of the proposed model that combines convolution and bidirectional long short-term memory units to classify waveforms based on their magnitude into three classes: “noise”, “low-magnitude events” and “high-magnitude events”. Herein, any earthquake signal with magnitude equal to or above 5.0 is labelled as “high-magnitude”. We show that the variation in the results produced by changing the length of the data is no more than the inherent randomness in the trained models due to their initialisation. We further demonstrate that the model is able to successfully classify waveforms over wide ranges of both hypocentral distance and signal-to-noise ratio.

A study on small magnitude seismic phase identification using 1D deep residual neural network

Reliable seismic phase identification is often challenging especially in the circumstances of low-magnitude events or poor signal-to-noise ratio. With improved seismometers and better global coverage, a sharp increase in the volume of recorded seismic data has been achieved. This makes handling seismic data rather daunting by using traditional approaches and therefore fuels the need for more robust and reliable methods. In this study, we develop 1D deep Residual Neural Network (ResNet), for tackling the problem of seismic signal detection and phase identification. This method is trained and tested on the dataset recorded by the Southern California Seismic Network. Results demonstrate that the proposed method can achieve robust performance for the detection of seismic signals and the identification of seismic phases. Compared to previously proposed deep learning methods, the introduced framework achieves around 4% improvement in earthquake detection and a slightly better performance in seismic phase identification on the dataset recorded by Southern California Earthquake Data Center. The model generalizability is also tested further on the STanford EArthquake Dataset. In addition, the experimental result on the same subset of the STanford EArthquake Dataset, when masked by different noise levels, demonstrates the model’s robustness in identifying the seismic phases of small magnitude.

Earthquake early warning systems based on low-cost ground motion sensors: A systematic literature review

Earthquake early warning system (EEWS) plays an important role in detecting ground shaking during an earthquake and alerting the public and authorities to take appropriate safety measures, reducing possible damages to lives and property. However, the cost of high-end ground motion sensors makes most earthquake-prone countries unable to afford an EEWS. Low-cost Microelectromechanical systems (MEMS)-based ground motion sensors are becoming a promising solution for constructing an affordable yet reliable and robust EEWS. This paper contributes to advancing Earthquake early warning (EEW) research by conducting a literature review investigating different methods and approaches to building a low-cost EEWS using MEMS-based sensors in different territories. The review of 59 articles found that low-cost MEMS-based EEWSs can become a feasible solution for generating reliable and accurate EEW, especially for developing countries and can serve as a support system for high-end EEWS in terms of increasing the density of the sensors. Also, this paper proposes a classification for EEWSs based on the warning type and the EEW algorithm adopted. Further, with the support of the proposed EEWS classification, it summarises the different approaches researchers attempted in developing an EEWS. Following that, this paper discusses the challenges and complexities in implementing and maintaining a low-cost MEMS-based EEWS and proposes future research areas to improve the performance of EEWSs mainly in 1) exploring node-level processing, 2) introducing multi-sensor support capability, and 3) adopting ground motion-based EEW algorithms for generating EEW.

Facilities management and earthquake risk reduction: The TURNkey project

Reducing vulnerability and improving resilience of the existing built environment to disaster events is a complex multidisciplinary challenge in which facilities managers need to contribute an understanding of the impact that a disaster event could have on both their hard and soft facilities. Facilities managers also need to identify potential operational, physical and corporate adaptations/mitigations that can support continued business function during and after a disaster event. This paper describes work in progress to develop an earthquake forecasting, early warning and rapid response system that business and critical infrastructure organisations can use as part disaster risk reduction and business continuity planning. Focusing on theory and methods, the paper considers the different aspects of resilience from a facilities management perspective and presents findings from a participatory action research programme that developed a series of use-cases to explore the potential impact of earthquake forecasting, early warning and rapid response on an organisation’s vulnerability and resilience to an earthquake; and identifying physical, operational and organisational mitigation interventions that can reduce an organisation’s disaster risk. The paper concludes the need for facilities managers to understand the different aspects of resilience and to apply the most appropriate to their hard and soft facilities management systems.

Spatiotemporal Graph Convolutional Networks for Earthquake Source Characterization.

Accurate earthquake location and magnitude estimation play critical roles in seismology. Recent deep learning frameworks have produced encouraging results on various seismological tasks (e.g., earthquake detection, phase picking, seismic classification, and earthquake early warning). Many existing machine learning earthquake location methods utilize waveform information from a single station. However, multiple stations contain more complete information for earthquake source characterization. Inspired by recent successes in applying graph neural networks (GNNs) in graph-structured data, we develop a Spatiotemporal Graph Neural Network (STGNN) for estimating earthquake locations and magnitudes. Our graph neural network leverages geographical and waveform information from multiple stations to construct graphs automatically and dynamically by adaptive message passing based on graphs' edges. Using a recent graph neural network and a fully convolutional neural network as baselines, we apply STGNN to earthquakes recorded by the Southern California Seismic Network from 2000 to 2019 and earthquakes collected in Oklahoma from 2014 to 2015. STGNN yields more accurate earthquake locations than those obtained by the baseline models and performs comparably in terms of depth and magnitude prediction, though the ability to predict depth and magnitude remains weak for all tested models. Our work demonstrates the potential of using GNNs and multiple stations for better automatic estimation of earthquake epicenters.

Developing the Earthquake Early Warning and Evacuation Systems (EEWES) for Schools in Ranau District

This article presents the main development components Early Earthquake Warning and Evacuation System (EEWES) for the Ranau District of Sabah. Many early earthquake warning systems have been developed since the last major earthquake in Ranau in 2015, but none has appropriately investigated the school children evacuation processes/exercises. The system was developed not just to expedite the evacuation process at these schools but also to help evacuate the school children and to get in touch with the children’s parents who are outstation. Those parents will be updated on the evacuation initiative that is taking place on a real-time basis. The system may also transfer the evacuation details to the local and federal emergency services and other agencies concerned with the evacuation exercises. The system mainly utilised the APIs released by regional disaster centres, which may deliver speedy transmission alerts. The future research direction is to incorporate the local tremor sensor details into the system.

Back-projection results for the M w 7.5, 28 September 2018 Palu earthquake-tsunami

A back-projection technique allows seismologists to analyse rupture properties once seismic signals from a dense array of seismic networks are available. The observed waveforms are then traced back in space and time to the source region of an earthquake under investigation. In this study, the method utilised the back-projection image of the recorded high-frequency P-seismic waveforms filtered at 0.25-1.0 Hz by Multiple Signal Classification (MUSIC) processing to estimate the extent and the spread of earthquake rupture propagation of the M w 7.5 Palu event on 28 September 2018, generating a severe tsunami. This study aims to estimate rupture duration, its extent (the distance over which rupture propagates) and the corresponding speed, and rupture directivity (where most of the seismic energy propagates). The results revealed that the front of rupture propagates slightly offset southward at about 2.9 km/s over a distance of ~140 km away from the epicentre for about 49 s before slowing down at much smaller amplitudes after reaching the south end of Palu Bay. This finding is consistent with an NNW-SSE orientation of the active Palu-Koro Fault lying along the bay, suggesting that the strong ground motion is associated with the fault activities. This study has therefore substantial implications for enhanced earthquake and tsunami early warnings, helping the government and local authority build community resilience by warning people at risk from future possible earthquake and tsunami hazards.

Fractal Clustering as Spatial Variability of Magnetic Anomalies Measurements for Impending Earthquakes and the Thermodynamic Fractal Dimension

Several studies focusing on the anomalies of one specific parameter (such as magnetic, ionospheric, radon release, temperature, geodetic, etc.) before impending earthquakes are constantly challenged because their results can be regarded as noise, false positives or are not related to earthquakes at all. This rise concerns the viability of studying isolated physical phenomena before earthquakes. Nevertheless, it has recently been shown that all of the complexity of these pre-earthquake anomalies rises because they could share the same origin. Particularly, the evolution and concentration of uniaxial stresses within rock samples have shown the generation of fractal crack clustering before the macroscopic failure. As there are studies which considered that the magnetic anomalies are created by lithospheric cracks in the seismo-electromagnetic theory, it is expected that the crack clustering is a spatial feature of magnetic and non-magnetic anomalies measurements in ground, atmospheric and ionospheric environments. This could imply that the rise of multiparametric anomalies at specific locations and times, increases the reliability of impending earthquake detections. That is why this work develops a general theory of fractal-localization of different anomalies within the lithosphere in the framework of the seismo-electromagnetic theory. In addition, a general description of the fractal dimension in terms of scaling entropy change is obtained. This model could be regarded as the basis of future early warning systems for catastrophic earthquakes.

Implementation of Wireless Sensor Network (WSN) for Earthquake Detection

The current earthquake monitoring system uses a seismometer that can capture seismic vibrations very well but is expensive, heavy, and difficult to launch. Therefore, earthquake monitoring stations can only be launched in a few places in small numbers. This study aims to implement a Wireless Sensor Network (WSN) system for earthquake monitoring. The WSN system has advantages in cost, size, and ease of launch, so it is very appropriate to be used for this purpose. An earthquake detection sensor system has been designed in this study using a vibration sensor and a piezoelectric sensor. When an earthquake occurs, the resulting shock will trigger the vibration sensor and activate the sensor node. The shock data is then captured by the piezo sensor and processed by the microcontroller using Fast Fourier Transform (FFT) to determine the frequency value of the shock. The data is then sent to a gateway via a sensor network and uploaded to the Cayenne monitoring website. Operators can then view the data on the website. Three sensor nodes are implemented in this study. The test is done by placing those sensor nodes together in random positions. A shock is then given to the three sensor nodes, and the resulting data is then observed. The results show that the three sensors can detect, retrieve, process, and send shock data to the Cayenne monitoring website.