Ground-Penetrating Radar (GPR) technology, with its characteristics of being fast, non-destructive, and high-resolution, has become an important tool for detecting underground structures. However, GPR data inevitably suffer from environmental noise and electromagnetic interference during the acquisition process, leading to decreased data quality and increased complexity in data processing. Traditional filtering algorithms have limitations such as low discrimination between noise and signal, poor adaptability, and inability to process data in real time. This paper proposes a filtering model based on deep neural networks, called FilterNet. FilterNet combines Convolution Neural Networks (CNN) and recurrent neural networks (RNN) for processing multi-channel data. It can perform end-to-end filtering directly on the raw tunnel lining GPR data, achieving functions such as removing air reflection waves, denoising, and automatic gain. Using PSNR (Peak Signal-to-Noise Ratio) and SSIM (Structural Similarity Index) as statistical indicators, it is shown that the FilterNet model improves filtering precision. The SSIM of all three models is 0.997, and the PSNR of FilterNet1D and FilterNet are 19.06 and 19.41, respectively. Furthermore, tests on the model's processing efficiency indicate that FilterNet requires less memory and is more efficient than the UNet model. FilterNet's parameters are only 48 % of those of UNet. Its GFLOPS (Giga Floating Point Operations Per Second) is only one-third of UNet's, and it can process data in real time. Additionally, FilterNet performs exceptionally well in suppressing random noise.
This study investigates the seismogenic characteristics of the 2025 Dingri MS 6.8 earthquake through multi-parametric GNSS analyses of velocity field, strain rate evolution and displacement patterns across pre-seismic and co-seismic phases. Our findings demonstrate spatiotemporally heterogeneous crustal deformation exhibiting kinematic precursors correlating with subsequent rupture propagation. The epicentral region exhibited prolonged N-S compressional strain accumulation accompanied by accelerated E-W extensional deformation and progressive counterclockwise rotation of principal strain axes three years prior, indicating enhanced local normal fault activities. Co-seismic observations delineate significant displacement domains, with the XZSJ (∼95 mm) site documenting the largest northeastward motion, consistent with rupture propagation along secondary N-E trending structures. Co-seismic strain analysis identifies concentrated seismic moment release primarily west of the Xainza-Dinggye Fault and north of the Southern Qinghai-Xizang Detachment Fault system, displaying normal fault kinematics in agreement with the seismic source mechanism. The co-seismic strain partitioning pattern shows critical implications for regional N-S trending normal fault system, necessitating sustained geodetic monitoring to advance understanding of seismic cycle deformation in this area.
We conducted a rapid seismic intensity assessment of a M 6.8 earthquake in Dingri, Xizang, using a ground motion parameter attenuation model based on the shortest fault distance combined with either an empirical equation for the surface rupture length or data on the aftershocks that occurred within 1.5 hr after the earthquake. The assessment showed that the empirical equation for the relationship between the surface rupture length and magnitude established by Wells et al. yielded a surface rupture length that was closer to the actual value, while the seismic intensity determined using a combination of the ground motion parameter attenuation model and the empirical equation for the surface rupture length was relatively in line with the intensity from the actual investigation. This study also demonstrated that manual intervention and screening are needed for aftershocks within 1.5 hr after the earthquake if this information is to be employed in the intensity assessment. In addition, if the death assessment model does not consider the seismic vulnerability of local buildings, significant errors can occur in practice. Nevertheless, the disaster assessment results were obtained within 5 min after the earthquake, thus providing important data support for the government emergency command and decision-making associated with the emergency rescue response.
We proposes an AI-assisted framework for integrated natural disaster prevention and emergency response, leveraging the DeepSeek large language model (LLM) to advance intelligent decision-making in geohazard management. We systematically analyze the technical pathways for deploying LLMs in disaster scenarios, emphasizing three breakthrough directions: (1) knowledge graph-driven dynamic risk modeling, (2) reinforcement learning-optimized emergency decision systems, and (3) secure local deployment architectures. The DeepSeek model demonstrates unique advantages through its hybrid reasoning mechanism combining semantic analysis with geospatial pattern recognition, enabling cost-effective processing of multi-source data spanning historical disaster records, real-time IoT sensor feeds, and socio-environmental parameters. A modular system architecture is designed to achieve three critical objectives: (a) automated construction of domain-specific knowledge graphs through unsupervised learning of disaster physics relationships, (b) scenario-adaptive resource allocation using risk simulations, and (c) preserving emergency coordination via federated learning across distributed response nodes. The proposed local deployment paradigm addresses critical data security concerns in cross-border disaster management while complying with the FAIR principles (Findable, Accessible, Interoperable, Reusable) for geoscientific data governance. This work establishes a methodological foundation for next-generation AI-earth science convergence in disaster mitigation.
On January 7, 2025, a M 6.8 earthquake occurred at Dingri County, Shigatse City, Xizang Autonomous Region, which is the most largest earthquake in this region in the most recent five years. The maximum recorded peak ground acceleration in this earthquake is 0.43 g, which is significantly higher than the local design basis earthquake intensity level. This report focuses on the post-earthquake seismic damage investigation of rural self-built houses and local public buildings, which shows quite different seismic performance. Collapse and heavy damage of rural self-built houses without seismic resistant measures were observed, while the masonry houses with seismic resistant measures show good seismic performance against collapse. The structural systems of public buildings with proper seismic design suffered slight seismic damage. Different form slight structural damage, the seismic damages of non-structural components, which resulted in the interruption of public building functionality were highlighted in this report. With the investigation results of seismic damage suffered by local houses and buildings, the casualty distribution and causes were analyzed in detail.
This paper reports the recorded structural responses of four 170 m-320 m tall buildings in China to the mainshock of the M 7.9 Myanmar earthquake on March 28, 2025. The buildings are located approximately 1 200 km-2 000 km away from the epicenter. The following observations are made by preliminary analysis of the data: (1) the base motion of the buildings exhibited significant long-period components in the range of 2 s-10 s; (2) the identified fundamental periods were much larger than the empirical equations in the design codes, suggesting that the empirical equations may be overly conservative; (3) the amplification of floor accelerations was much more significant than code provisions for determining the seismic demands on non-structural elements, possibly attributing to the overly high damping ratios assumed in the design codes; (4) the buildings exhibited large enough equivalent lateral stiffnesses to satisfy the drift limit under frequent earthquakes by the Chinese seismic provisions, and (5) the significant durations of the shaking of the upper floors of the buildings were comparable to those of the base motions.
On March 28, 2025, a moment magnitude (MW) 7.9 earthquake struck Myanmar, marking it as the most powerful seismic event of the year. The earthquake occurred along the Sagaing Fault, a prominent right-lateral strike-slip fault and the most significant active tectonic structure in Myanmar. In response to this seismic emergency, China Earthquake Networks Center (CENC) promptly published several reports on seismic source parameters. The focal mechanism solution indicates strike-slip faulting as the causative mechanism of the earthquake. Analysis of the rupture process indicates that it predominantly propagates from north to south. The rupture extended over a length exceeding 200 km and persisted for approximately 95 s. According to the estimated seismic intensity map, the meizoseismal region experienced shaking intensity reaching up to X(10), while the area with an intensity of more than VI exceeds 443 487 km2. This earthquake inflicted substantial casualties and extensive property damage, underscoring the long-standing seismic hazard posed by the Sagaing Fault and highlighting the need for enhanced seismic preparedness and risk mitigation strategies in the region.
A devastating Mw 7.7 earthquake struck near Mandalay, Myanmar, on March 28, 2025, causing extensive damage and casualties across Myanmar and neighboring regions. The 2025 event occurred in a well-recognized seismic gap along the Sagaing Fault. Here we focus on the mainshock rupture properties based on back-projection of teleseismic P waves and early aftershock locations, analysis of near-field seismic recordings for the mainshock initiation, and remotely triggered seismicity following the Mw7.7 mainshock. We find that the ∼500 km mainshock rupture can be revealed by both rapid back-projection of teleseismic P waves from multiple broadband arrays and early aftershock locations within about 3 h from the Thai Meteorological Department (TMD) catalog. The rupture speed went supershear in the southward propagation after the initial bilateral subshear ruptures, as expected for large strike-slip earthquakes of such sizes. Clear fault zone head waves that are reflected along a bimaterial fault interface are observed at the only near-fault station GE.NPW on the slower side about 2.6 km away from the Sagaing fault, consistent with the preferred direction of a supershear rupture propagating to the south. In addition, aftershocks from the regional TMD catalog appear to be located mostly to the east of the mainshock rupture. While we cannot completely rule out mis-locations from the one-sided station distribution, these off-fault seismicity could also be explained by reactivations of subsidiary faults within the Shan Plateau, or an eastward dipping of the mainshock rupture plane. Although no immediate foreshocks were found from several nearby stations, we identify one sub-event with magnitude ∼6 at the beginning of the mainshock with a slightly different focal mechanism about 20-30 km south of the hypocenter determined by the United States Geological Survey (USGS). The mainshock also occurred when the tidal stresses reached its maximum on the right-lateral strike-slip fault, likely indicating that the timing of the mainshock is modulated by the solid earth tides. We find a significant increase of seismic activity near the Thailand/Myanmar border, in multiple (geothermally active) regions of Yunnan province in Southwest China, as well as the Xingfengjian reservoir in the Guangdong province in South China. Because static stress changes from the mainshock are small but negative near the Thailand/Myanmar border, the occurrence of microseismicity in this and other regions can be mainly explained by remote triggering from dynamic stress changes of the mainshock rupture. Our analyses demonstrate the importance of rapid analysis on openly available seismic data and catalog to better understand the rupture properties and triggered seismicity following large earthquakes.