Sea fog is a precarious weather disaster affecting transportation on the sea. The accuracy of the threshold method for sea fog detection is limited by time and region. In comparison, the deep learning method learns features of objects through different network layers and can therefore accurately extract fog data and is less affected by temporal and spatial factors. This study proposes a scSE-LinkNet model for daytime sea fog detection that leverages residual blocks to encoder feature maps and attention module to learn the features of sea fog data by considering spectral and spatial information of nodes. With the help of satellite radar data from Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP), a ground sample database was extracted from Moderate Resolution Imaging Spectroradiometer (MODIS) L1B data. The scSE-LinkNet was trained on the training set, and quantitative evaluation was performed on the test set. Results showed the probability of detection (POD), false alarm rate (FAR), critical success index (CSI), and Heidke skill scores (HSS) were 0.924, 0.143, 0.800, and 0.864, respectively. Compared with other neural networks (FCN, U-Net, and LinkNet), the CSI of scSE-LinkNet was improved, with a maximum increase of nearly 8%. Moreover, the sea fog detection results were consistent with the measured data and CALIOP products.

This study develops a MATLAB-based (MathWorks, Natick, MA, USA; version R2024b) first-order probabilistic screening framework for the Shenhu gas hydrate production area, northern South China Sea, to separate the roles of bottom-water warming, geothermal and fluid perturbation, hydrate saturation, excess pore pressure, and slope geometry. The framework integrates methane hydrate phase equilibrium, geothermal profiles, transient heat diffusion, base of gas hydrate stability zone (BGHSZ) migration, conceptual reservoir phase layering, and infinite-slope factor-of-safety (FoS) analysis. The modeled present-day BGHSZ (~271 mbsf) is broadly consistent with reported bottom-simulating-reflector depths. At the baseline thermal diffusivity, conductive potential dissociation thickness stays below 4 m even for +3.0 °C over 1000 years; across the full tested diffusivity–warming–time matrix it ranges 0–19.5 m; geological tests, reported as equilibrium BGHSZ offsets, give larger responses. Monte Carlo analysis under three parallel slope-angle assumptions shows that absolute FoS depends strongly on the slope distribution. Conductive warming reduces the median FoS by about 5.5%, with about 9% of realizations exceeding a 10% reduction, yet low-FoS outcomes remain absent within the tested ranges. Within the adopted parameterization, modeled FoS variability is controlled primarily by slope angle, friction angle, and the prescribed excess pore-pressure ratio, while conductive thermal forcing produces a comparatively modest reduction in relative stability margin.

Active faults exhibit complex geometric features, with rupture processes controlled by the coupled effects of tectonic loading, fault geometry, and internal displacement heterogeneity. To elucidate the interrelationship between long‐term deformation, stress redistribution, and rupture evolution, we constructed a three‐dimensional numerical model of the Longitudinal Valley Fault (LVF) in eastern Taiwan using the discrete element method (DEM). The model simulates heterogeneous displacement, rupture sequences, and energy release under long term quasi‐static tectonic loading. Model results show spatially heterogeneous rupture and cumulative deformation. Rupture in the late stage concentrates mainly on the eastern side and in the central and northern sections of the fault. Cumulative displacement follows a non‐monotonic pattern with depth, characterized by shallow discrete adjustment, a continuous intermediate depth deformation band, and overall weakening with localized high displacement patches at greater depths. Along strike, high displacement zones and rupture clusters are spatially associated with local bends and rapid strike variations, suggesting that geometric complexity may influence the spatial organization of heterogeneous deformation. Results of stress partitioning further indicate that stronger normal confinement at depth may limits the development of a continuous deep high displacement band. These results provide mechanical constraints for identifying candidate regions with enhanced deformation concentration and rupture susceptibility, guiding further investigation of potential rupture prone zones.

Understanding the shallow rupture mechanisms on coseismic faults and assessing the influence of fault area propagation is essential for disaster prevention. Since 2000, Hualien and nearby areas in eastern Taiwan have experienced frequent earthquakes, making it a good area to study the evolution of fault rupture. This study proposes a two-dimensional dynamic discrete element model to simulate the shallow rupture behavior of the Milun Fault. Results indicate that the rupture process proceeds through multiple evolutionary stages, with fractures propagating upward from depth but failing to fully break through to the surface, resulting instead in surface cracking without complete rupture. The second deviatoric stress invariant serves as an effective indicator of stress accumulation and release during rupture progression. For the preferred model, the modeled vertical uplift near the fault reached 0.6 m, consistent with field observations reporting a maximum coseismic uplift of approximately 0.585 m along the Milun Fault. Given the scarcity of near-fault observational constraints, the simulation represents a physically plausible scenario rather than a unique reconstruction. The integration of stress evolution, crack propagation, and near-field displacement provides new insight into the mechanical processes governing shallow thrust fault rupture and can be applied to similar fault systems exhibiting near-surface deformation.