In this research, a new framework of multi-task deep learning is proposed for earthquake magnitude, epicentral distance, and focal depth prediction based on the seismic waveforms simultaneously. The study relies on 3 channel seismic recordings from the Stanford Earthquake Dataset (STEAD) that contain a large amount of waveform data and exhibit various characteristics, providing a robust training set for the model. The first step in pre-processing the raw seismic signals is to apply a Butterworth band-pass filter to remove noise and improve the quality of the signals. Then, a Wavelet-Lifting Fourier Transformer (WL-FT) is used to identify meaningful features from the seismic data through multi-resolution characteristics, frequency-domain information, and crosschannel dependency. This allows the model to acquire powerful and noise-tolerant representations. Then the extracted feature vectors are fed to the Capsule-based Temporal Deformable Convolution embedded multi-decoder (CT-DCMD) module. The capsule network is able to capture the hierarchical relationships between features, and the temporal deformable convolution dynamically learns the important time regions with adaptive changes of the receptive field. This has been beneficial for the model to capture complex temporal and structural features of seismic signals. The refined features are then passed through batch normalization, GELU activation, and dropout to stabilize training, enhance non-linearity, and generalize. Last, a multi-decoder structure is employed to simultaneously predict magnitude, epicentral distance, and focal depth. The proposed model is assessed with the help of a number of performance metrics like MAE, MSE, RMSE, and R 2 . The proposed model scored MAE and MSE values of 0.126 and 0.025, respectively. The experimental results show that the model can predict the seismic data with high accuracy and reliability, which is suitable for prediction and early warning of seismic data in the real world.
Rapid focal mechanism estimation plays a critical role in earthquake and tsunami hazard assessment, although it is not the sole parameter, particularly in time-sensitive early warning scenarios. Conventional approaches based on full waveform inversion generally require substantial processing time, limiting their applicability for early warning when timely information is essential for decisionmaking considerations. In the initial stage following an earthquake, rapid indication of tsunami potential is critically needed; therefore, complementary approaches based on P-wave polarity can help fill this information gap. Although Indonesia operates an extensive Strong-Motion Accelerograph (SMA) network, its utilization for focal mechanism determination remains limited. To address this challenge, we propose InaFocMec, a novel method integrating deep learning-based polarity detection with the SKHASH algorithm for focal mechanism inversion. The model was trained using optimized P-wave time windows ranging from 0.20 to 0.05 seconds and evaluated on a comprehensive dataset representing Indonesian tectonic settings, achieving 99.76% classification accuracy. The resulting focal mechanism solutions for significant earthquakes show strong agreement with global and national reference catalogs, demonstrating robustness. These findings highlight the potential of InaFocMec as a rapid tool to enhance earthquake source characterization and support more effective tsunami early warning systems in Indonesia.
The conservation and safeguarding of immovable cultural heritage sites is an urgent issue of international concern especially in China, which has geological activities and climate sensitive areas. Conventional monitoring tools are known to be expensive, low temporal resolution, and small spatial coverage, which makes them unable to offer real-time and comprehensive risk-assessment on large site networks. In this research an apt risk identification model of the immovable heritage sites under the influence of heterogeneous satellite measurements will be presented and this shall incorporate the multi-source remote sensing imagery such as Synthetic Aperture Radar (SAR), multispectral optical data, and thermal infrared data. The suggested framework is based on an object-oriented analytical paradigm which conceptualizes individual monitored locations as Digital Site Object (DSO), and considers individual monitored sites to be a first-class data unit that can be accessed and operated in a single geospatial information infrastructure. The model finds a hierarchical feature extraction pipeline, which is integrated with convolutional neural networks (CNNs) and transformer-based attention patterns to detect structural deformation, surface subsidence, vegetation encroachment, and water infiltration abnormalities. Multi-temporal change detection algorithm is used on time-series satellite measurements which provide the ability to quantify the change of risk on the four major dimensions: geological instability, climatic exposure, anthropogenic disturbance, and structural vulnerability. Data on 186 nationally registered immovable heritage sites within the Loess Plateau region in Shaanxi Province and karst landscape in Guizhou Province in China were used to validate the model. Experimental data show that the proposed model has a total risk classification rate of 91.4% and a kappa coefficient of 0.886, surpassing those of baseline methods such as single-source assessment and traditional machine learning methods. The experiment also reveals that heterogeneous data fusion could greatly improve the latent risk detection that would not have been detected by any of the sensors. Such results offer an empirical and extensively generalizable approach to the monitoring of heritage sites on the national level and a theoretical approach to the inclusion of intelligent remote sensing systems as part of the cultural heritage management technologies.
Closely spaced adjacent structures are vulnerable to dynamic interaction during earthquakes, which may amplify structural responses and increase seismic damage. Although various passive control devices have been proposed for adjacent structures, nonlinear friction-based control under simultaneous bidirectional horizontal ground motions remains comparatively less explored, and efficient analytical tuning methodologies for such devices are limited. This study proposes a Nonlinear Dynamic Amplifier Friction Vibration Absorber (NDAFVA) for mitigating the earthquake-induced response of adjacent structures subjected to bidirectional horizontal ground motions. Nonlinear dynamic models are developed for Single-Degree-of-Freedom (SDOF) and Multi-Degree-of-Freedom (MDOF) systems by combining dynamic amplification with nonlinear frictional energy dissipation. The NDAFVA is evaluated under cyclic pulse excitations and 20 recorded near-field and far-field earthquake ground motions. Optimal absorber parameters are determined using hybrid H2/H[Formula: see text] optimization, and closed-form tuning relationships are derived using Taylor-series-based analytical formulations. The NDAFVA provides substantial response mitigation across the considered seismic excitations. Across the 20 ground motions, the maximum reductions in peak acceleration and displacement are 69% and 72.5%, respectively, while the mean and standard-deviation values are also reported in Tables 3 and 4 . The closed-form tuning relationships closely reproduce the optimized parameters and provide greater response reduction than conventional H2 and H[Formula: see text] tuning for the benchmark systems considered. These results demonstrate that combining dynamic amplification with nonlinear frictional energy dissipation provides an effective passive control strategy, while the proposed closed-form relationships offer a computationally efficient approach for preliminary seismic-control design of adjacent structures.
Preparations to minimize deaths and damage to coastal infrastructure, livelihoods, and resources have been realized by the entire country after the occurrence of several tsunami incidents. The scientific community at the international level is trying to develop and diversify methodologies for tsunami vulnerability assessment. Usually, the vulnerability of coastal communities is assessed based on reading previous literature, post-tsunami reports recorded in various places, and some are also based only on available data. Numerous studies have attempted to use various methods for vulnerability assessment focus on physical vulnerability assessment, but there is less research on socio-economic vulnerability assessment. Although many researchers try to focus on the socio-economic aspect, they do not reach the level they should, especially on human indicators. Therefore, the aim of this research is to validate the indicators related to the human dimension based on previous studies and tsunami events that have occurred in different countries. The objective of this study is to summarize and assess the socio-economic indicators for tsunami vulnerability assessment and to scrutinize aspects of vulnerability that are less explored and can be improved based on previous findings. This review is the first to consolidate and validate human-focused socio-economic indicators for tsunami vulnerability assessment through a PRISMA-based systematic synthesis, providing a standardized classification of human indicators into four key dimensions: exposure, warning capacity, evacuation and emergency capacity, and recovery capacity. This paper discusses the importance of these indicators based on previous research and suggests potential improvements for future studies.
