Search Results (26)

Arduino-based creepmeters utilizing a Linear Variable Differential Transformer (LVDT) and ultrasonic sensors were fabricated to monitor displacement changes along the creeping segment of the West Valley Fault (WVF) in southeastern Metro Manila, Philippines. Along with a custom-assembled, Arduino-based rain gauge, these instruments were initially intended to prevent data gaps during the COVID-19 pandemic when commercial data recorders experienced operational downtime. However, they have since proven to be cost-effective alternatives for determining short-term slip rates and monitoring displacement variations driven by episodic and seasonal precipitation changes. The LVDT creepmeter provides higher accuracy for displacement and slip rate determination. Conversely, the ultrasonic creepmeter is better suited for tracking abrupt displacement changes and, to some extent, longer-term displacement trends as it is more sensitive to environmental conditions. Deploying low-cost monitoring instruments in active fault regions bridges critical data gaps and improves the understanding of creep triggers and mechanisms. Although vertical creep occurs along pre-existing tectonic features of the WVF creeping segment, our creepmeter monitoring reveals sustained, accelerated creep within its southern portion. This localized movement is driven primarily by nontectonic forces—chiefly groundwater extraction, with episodic and seasonal precipitation influences. Consequently, this implies a continued ground rupture hazard and the potential for induced seismicity. Full article

The Anninghe Fault Zone is an active, deep–large fault in southwestern China, with a history of multiple strong earthquakes. To reveal the temporal patterns of seismicity and improve medium- to short-term earthquake frequency prediction, this study constructs a quarterly seismic frequency sequence (M ≥ 3.0) from May 1972 to September 2025 and applies the SARIMA (seasonal autoregressive integrated moving average) model for modeling and prediction. The hypothesis is that the frequency sequence exhibits modelable seasonality, trends, and nested periodic structures. The ADF test and Ljung–Box test confirm that the sequence is stationary and non-white noise, satisfying the prerequisites for SARIMA modeling. The centered moving average method is used to extract short-term (1 year), medium-term (5 years), and long-term (10 years) periodic components, and corresponding SARIMA models are constructed. Results show that the medium-period model ARIMA(2,0,1) × (1,0,0)₂₀ achieves the best prediction accuracy (RMSE = 0.6868, MAE = 0.6143), followed by the short-period model, while the long-period model yields slightly higher errors. All selected models pass residual white noise tests and parameter significance tests, and exhibit good robustness under different training–test splits. The main innovations are: (1) the first systematic application of SARIMA to earthquake frequency prediction in the Anninghe Fault Zone, and (2) a preliminary physical interpretation of multi-scale periodic components (e.g., seasonal loading, strain accumulation fluctuations). This method offers significant application value in regions with sparse seismic networks or limited precursory data, providing a new statistical tool for regional seismic hazard assessment and disaster mitigation planning. Full article

This study presents a comprehensive multi-parameter analysis of seismo-ionospheric responses to the Mw 7.7 Myanmar earthquake on 28 March 2025, using GNSS-based Total Electron Content (TEC) data, seismic b-value trends, and acoustic gravity wave (AGW) signatures. A significant negative TEC anomaly (~30 TECU below the statistical threshold) was detected on 25 March, three days before the mainshock under geomagnetically quiet conditions, indicating a lithospheric origin. Concurrent variations in the Ionospheric Disturbance Index (IDI) and Rate of TEC Index (ROTI) indicate pronounced background departures and enhanced short-term variability during the preparation phase. Temporal b-value analysis shows a consistent decline from 1.12 to 0.58 across the 30-year to 6-month windows, with the lowest values clustering near the epicenter, indicating progressive stress accumulation. Spatial b-value mapping further reveals a low b-value zone overlapping the region of TEC depletion, while the Relative Seismic Hazard Index (RSHI) highlights high-hazard zones aligned with the epicentral area. Kernel density estimation (KDE) supports this coupling by showing a dominant low-b, low-vTEC cluster, consistent with linked lithospheric stress and ionospheric depletion. Overall, the integrated GNSS and seismic analyses demonstrate the value of multi-domain observations for characterizing earthquake preparation processes, highlighting a coherent physical linkage between crustal stress accumulation and ionospheric depletion that can enhance short-term seismic hazard assessment. Full article

The Longmen Shan Fault Zone is marked by intricate geological structures and frequent seismic activity, which gives rise to persistent seismic hazards. To tackle the challenge of capturing the multi-temporal characteristics of earthquake frequency, this study combines machine learning with time series analysis to conduct earthquake frequency prediction research. Based on the 1970–2023 seismic dataset from the China Earthquake Networks Center, the seismic records were structured into four temporal scales: daily, weekly, monthly and quarterly. The minimum completeness magnitude (Mc) was determined as M3.0 by applying the G–R relationship. After conducting white noise tests and data normalization, ACF and PACF were utilized to select the optimal time-step parameters for the LSTM model. Considering the inherent characteristics of the seismic data, the 99th percentile of the frequency series was set as the threshold, and an auxiliary parameter was introduced to label high-frequency earthquake days for the construction of the LSTM model. Upon the completion of LSTM model fitting, heteroscedasticity tests were performed on the residuals between the predicted and observed values. Confirming the presence of significant heteroscedasticity, the GARCH model was incorporated to process these residuals, thus establishing a complete LSTM-GARCH coupled model. The results reveal that seismic activity in this region is normally low-frequency with occasional high-frequency occurrences. The proposed model achieves R² above 0.80 across all four temporal scales, accompanied by superior performance in all error metrics. This study validates that the LSTM-GARCH model can effectively extract the multi-scale patterns of earthquake frequency, with the best performance observed at the daily scale. Ablation experiments further demonstrate that this coupled model outperforms both the ARIMA and single LSTM models, providing reliable technical support for short-to-long-term earthquake prediction and regional disaster risk assessment. Full article

Forecasting post-seismic landslide displacement is challenged by the difficulty in distinguishing short-term acceleration from creep and the risk of spatiotemporal leakage. To address this, an interpretable deep-learning framework is developed, integrating SBAS-InSAR time series with an Attention-enhanced Gated Recurrent Unit (Attention-GRU). Prior to modeling, a multi-stage preprocessing strategy, including empirical mode decomposition, is applied to mitigate noise and delineate active deformation zones. Unlike standard architectures, the model’s temporal attention mechanism adaptively amplifies critical precursory acceleration phases. Furthermore, a strict landslide-object-based partitioning strategy is employed to rigorously mitigate spatiotemporal leakage. The framework was evaluated in the Le’an Town landslide cluster using multi-source data. Targeting identified hazardous regions, the method achieved an R² of 0.93 and reduced MAPE by 42.7% relative to the SVR baseline. This reflects a location-specific predictive capability, within active zones rather than regional generalization. SHapley Additive exPlanations (SHAP) further confirmed the model captures physical relationships, such as sensitivity to 25–35° slopes and vegetation degradation. Ultimately, the proposed framework offers a transparent, physically interpretable tool for operational hazard mitigation. Full article

The b-value is a critical parameter for gauging seismic activity and is essential for seismic hazard assessment, monitoring stress evolution in focal zones, and forecasting major earthquakes. The minimum magnitude of completeness ($M_c$), a key indicator of the completeness of an earthquake catalog, reflects the monitoring capability of a seismic network and serves as a crucial foundation for the accurate calculation of the b-value. We began by integrating multi-source earthquake catalogs for mainland China using the nearest-neighbor method. Building on this, we employed a combination of partitioned time-series analysis and a grid-based spatial scanning technique to systematically investigate the spatiotemporal evolution of the $M_c$ and the b-value across mainland China and its adjacent regions. Our findings indicate the following: (1) Since the 1980s, the overall trend of $M_c$ has shifted from high and unstable values to low and stable ones. However, significant earthquake events can cause a notable short-term increase in the $M_c$. (2) The b-value exhibits strong fluctuations, primarily influenced by the dual effects of the tectonic stress field and catalog completeness. These fluctuations are particularly pronounced in highly active seismic regions such as the Sichuan–Yunnan area and Taiwan, whereas the western Tibetan Plateau has consistently maintained a low b-value. (3) The spatial distributions of both the $M_c$ and the b-value are markedly heterogeneous. By developing a unified and complete earthquake catalog for mainland China, our research highlights the qualitative leap in monitoring capabilities brought about by the continuous densification and technological upgrading of seismic networks. This dataset provides a solid foundation for future seismological research, disaster prevention practices, and especially for the development of AI-based earthquake prediction models. Full article

Microseismic signals induced by mining activities often have low signal-to-noise ratios, and traditional picking methods are easily affected by noise, making accurate identification of P-wave arrivals difficult. To address this problem, this study proposes an adaptive denoising algorithm based on wavelet-threshold-enhanced Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) and develops an automatic P-wave arrival picking method incorporating fractal box dimension features, along with a corresponding accuracy evaluation framework. The raw microseismic signals are decomposed using the improved CEEMDAN method, with high-frequency intrinsic mode functions (IMFs) processed by wavelet-threshold denoising and low- and mid-frequency IMFs retained for reconstruction, effectively suppressing background noise and enhancing signal clarity. Fractal box dimension is applied to characterize waveform complexity over short and long-time windows, and by introducing fractal derivatives and short-long window differences, abrupt changes in local-to-global complexity at P-wave arrivals are revealed. Energy mutation features are extracted using the short-term/long-term average (STA/LTA) energy ratio, and noise segments are standardized via Z-score processing. A multi-feature weighted fusion scoring function is constructed to achieve robust identification of P-wave arrivals. Evaluation metrics, including picking error, mean absolute error, and success rate, are used to comprehensively assess the method’s performance in terms of temporal deviation, statistical consistency, and robustness. Case studies using microseismic data from a mining site show that the proposed method can accurately identify P-wave arrivals under different signal-to-noise conditions, with automatic picking results highly consistent with manual labels, mean errors within the sampling interval (2–4 ms), and a picking success rate exceeding 95%. The method provides a reliable tool for seismic source localization and dynamic hazard prediction in mining microseismic monitoring. Full article

This study investigates the impact of nonlinear constitutive models on one-dimensional seismic site response analysis (SRA) for soft, reclaimed soil deposits in Saemangeum, South Korea. Two widely used models, MKZ and GQ/H, were applied to three representative soil profiles using the DEEPSOIL program. Ground motions were scaled to bedrock peak ground accelerations (PGAs) corresponding to annual return periods (ARPs) of 1000, 2400, and 4800 years. Seismic response metrics include the ratio of GQ/H to MKZ shear strain, effective PGA (EPGA), and short- and long-term amplification factors (F_a and F_v). The results highlight the critical role of the site-to-motion period ratio (T_g/T_m) in controlling seismic behavior. Compared to the MKZ, the GQ/H model, which features strength correction and improved stiffness retention, predicts lower shear strains and higher surface spectral accelerations, particularly under strong shaking and shallow conditions. Model differences are most pronounced at low T_g/T_m values, where MKZ tends to underestimate amplification and overestimate strain due to its limited ability to reflect site-specific shear strength. Relative to code-based amplification factors, the GQ/H model yields lower short-term estimates, reflecting the disparity between stiff inland reference sites and the soft reclaimed conditions at Saemangeum. These findings emphasize the need for strength-calibrated constitutive models to improve the accuracy of site-specific seismic hazard assessments. Full article

The Afghanistan earthquake of 21 June 2022 ruptured a ~10 km-long fault segment in the North Waziristan–Bannu fault system (NWBFS) located towards the north of the Katawaz Basin. The earthquake was shallow and reportedly caused widespread devastation. In this article, we investigated the long-term, i.e., geological and geomorphological, evidence of deformation along the earthquake segment. For comparison, we also studied the short-term space geodetic and remote sensing results documenting a visible offset between the fault traces. Focusing on the fault modelling and on the published results, it is thus clear that the earthquake rupture did not reach the surface; instead, it stopped in the shallow sub-surface at ~1 km depth. Moreover, the InSAR analyses show some technical issues, such as coherence loss, etc., likely due to severe ground-shaking leaving some gaps in the results; geological and geomorphological evidence complemented this information. As an outcome of this research, we confirmed that InSAR results could generally capture the overall fault geometry at depth, even in cases of blind faulting, whereas the detailed geometry of the tectonic structure, in this case with a right stepping en-echelon pattern, could be successfully captured by combining it with geological and geomorphological approaches and optical remote sensing observations. Accordingly, the right stepping fault generates a restraining bend in the dominantly left-lateral shear zone. Therefore, such fault stepovers are capable of localizing strain and could act as loci for seismic ruptures, bearing strong implications for the seismic hazard assessment of the region, as well as of other strike-slip fault zones. Full article

The North–South Seismic Belt is one of the major regions in China where strong earthquakes frequently occur. Predicting the monthly maximum magnitude is of significant importance for proactive seismic hazard defense. This paper uses seismic catalog data from the North–South Seismic Belt since 1970 to calculate and extract multiple seismic parameters. The monthly maximum magnitude is processed using Variational Mode Decomposition (VMD) with sample segmentation to avoid information leakage. The decomposed multiple modal data and seismic parameters together form a new dataset. Based on these datasets, this paper employs four deep learning models and four time windows to predict the monthly maximum magnitude, using prediction accuracy (PA), False Alarm Rate (FAR), and Missed Alarm Rate (MR) as evaluation metrics. It is found that a time window of 12 generally yields better prediction results, with the PA for Ms 5.0–6.0 earthquakes reaching 77.27% and for earthquakes above Ms 6.0 reaching 12.5%. Compared to data not decomposed using VMD, traditional error metrics show only a slight improvement, but the model can better predict short-term trends in magnitude changes. Full article

In this study, we aimed to assess the capacity of a physics-based earthquake simulator to improve our understanding of the seismogenic process. In this respect, we applied a previously tested earthquake simulator to two well-known and completely different seismogenic fault systems, namely the Italian Apennines and the Nankai subduction in Japan, for which long historical records of strong earthquakes are available. They are characterized by different fault mechanisms, fault sizes, and slip rates. Because of the difference in slip rates, the time scale of the seismicity patterns is different for the two systems (several hundreds of years for the Apennines and a few tens of years for the Nankai Fault). The results of simulations that produced synthetic catalogues of 100,000 years show these significant long-term seismicity patterns characterizing the seismic cycles for both seismogenic areas as follows: The average stress and the occurrence rate of earthquakes increase in the long term as the next major earthquake approaches; while the average stress increases uniformly, the occurrence rate stops increasing well in advance of the mainshocks; the b-value exhibits a long-term increase before major earthquakes and a fast decrease shortly before the mainshocks. Even if no specific statistical tool was applied for the quantification of the similarities between the seismicity patterns of the two seismic areas, such similarities are clearly justified by the large number of seismic cycles included in the 100,000-year synthetic catalogues. The paper includes a discussion on the capability of the simulation algorithm to reliably represent the real long-term seismogenic process. This question is difficult to answer because the available historical observations are of too short a duration to provide significant statistical results. In spite of the limitations characterizing the use of earthquake simulators for time-dependent earthquake hazard assessment, and the lack of convincing mechanistic explanations of the specific seismic patterns reproduced by our simulator algorithm, our results encourage further investigations into the application of simulators for the development of seismogenic models, including short-term features. Full article

Accurate seismic ground response analysis is crucial for the design and safety of civil infrastructure and establishing effective mitigation measures against seismic risks and hazards. This is a complex process due to the nonlinear soil properties and complicated underground geometries. As a simplified approach, the one-dimensional wave propagation model, which assumes that seismic waves travel vertically through a horizontally layered medium, is widely adopted for its reasonable performance in many practical applications. This study explores the potential of sequence deep learning models, specifically 1D convolutional neural networks (1D-CNNs), long short-term memory (LSTM) networks, and transformers, as an alternative for seismic ground response modeling. Utilizing ground motion data from the Kiban Kyoshin Network (KiK-net), we train these models to predict ground surface acceleration response spectra based on bedrock motions. The performance of the data-driven models is compared with the conventional equivalent-linear analysis model, SHAKE2000. The results demonstrate that the deep learning models outperform the physics-based model across various sites, with the transformer model exhibiting the smallest average prediction error due to its ability to capture long-range dependencies. The 1D-CNN model also shows a promising performance, albeit with occasional higher errors than the other models. All the data-driven models exhibit efficient computation times of less than 0.4 s for estimation. These findings highlight the potential of sequence deep learning approaches for seismic ground response modeling. Full article

A seismic hazard can be quantified by using probabilities. Modern seismic forecasting models (e.g., Operational Earthquake Forecasting systems) allow us to quantify the short-term variations in such probabilities. Indeed these probabilities change with time and space, in particular after strong seismic events. However, the short-term seismic hazard could also change during seismic swarms, i.e., a sequence with several small-/medium-sized events. The goal of this work is to quantify these changes, using the Italian Operational Earthquake Forecasting system, and also estimate the variations in the Gutenberg–Richter b-value. We focus our attention on three seismic swarms that occurred in Central Italy in October–November 2023. Our results indicate that short-term variations in seismic hazard are limited, less than an order of magnitude, and also that b-value variations are not significant. Placing our findings in a more general context, we can state that according to currently available models and catalogs, the occurrence of seismic swarms does not significantly affect the short-term seismic hazard. Full article

This paper is devoted to the features of seismological observations in the Arctic seas, which are complicated by harsh climatic conditions, the presence of ice cover, stamukhi and icebergs, and limited navigation. Despite the high risk of losing expensive equipment, the deployment of local networks of bottom seismographs or stations installed on ice is still necessary for studying the seismotectonic characteristics and geodynamic processes of the region under consideration, the deep structure of the crust and upper mantle, seismic hazards, and other marine geohazards. Various types of seismic stations used for long-term and short-term deployments in the Russian sector of the Arctic Ocean, as well as various schemes and workflows for their deployment/recovery, are described. The characteristics of seafloor seismic noise and their features are also considered. The results of deployments demonstrate that the characteristics of the stations make it possible to reliably record earthquake signals and seismic noise. Based on the experience gained, it was concluded that the preferred schemes for deploying ocean-bottom seismographs are those in which their subsequent recovery does not depend on their power resources. Usually, such schemes allow for the possibility of dismantling stations via trawling and are suitable for the shelf depths of the sea. The advantages of such schemes include the possibility of installing additional hydrophysical and hydrobiological equipment. When using pop-up ocean-bottom seismographs, special attention should be paid to the careful planning of the recovery because its success depends on the possibility of a passage to the deployment site, which is not always possible due to changing meteorological and ice conditions. Seismic records obtained on the seafloor are characterized by a high noise level, especially during periods of time when there is no ice cover. Therefore, it is recommended to install bottom stations for periods of time when ice cover is present. The frequency range of the prevailing noise significantly overlaps with the frequency range of earthquake signals that must be taken into account when processing bottom seismic records. Full article

Predicting the remaining time before the next earthquake based on seismic signals generated in a laboratory setting is a challenging research task that is of significant importance for earthquake hazard assessment. In this study, we employed a mel spectrogram and the mel frequency cepstral coefficient (MFCC) to extract relevant features from seismic signals. Furthermore, we proposed a deep learning model with a hierarchical structure. This model combines the characteristics of long short-term memory (LSTM), one-dimensional convolutional neural networks (1D-CNN), and two-dimensional convolutional neural networks (2D-CNN). Additionally, we applied a stacking model fusion strategy, combining gradient boosting trees with deep learning models to achieve optimal performance. We compared the performance of the aforementioned feature extraction methods and related models for earthquake prediction. The results revealed a significant improvement in predictive performance when the mel spectrogram and stacking were introduced. Additionally, we found that the combination of 1D-CNN and 2D-CNN has unique advantages in handling time-series problems. Full article