TerraMosaic Daily Digest: April 30, 2026
Daily Summary
April 30 is led by landslide papers that treat failure as a time-resolved cascade rather than a static scar. The Matai'an study is the central result: a 308 million m3, two-stage landslide in Taiwan is reconstructed from precursor geomorphology and seismic source inversion, linking earthquake-damaged cracking, landslide-dam formation, breach flooding, and downstream seismic warning. The same logic appears in rainfall-conditioned shaking-table tests on sandy slopes, Typhoon Wipha road-corridor failures, Quaternary mesa-fringe landslides in Patagonia, Nu River old-landslide geometry, and saturated loess instability experiments: the decisive information is not only where slopes fail, but how hydrology, shaking, lithology, river incision, and inherited structure sequence the failure.
The broader geohazard literature extends this process view into infrastructure and prediction systems. Rockburst experiments, coastal cavity-collapse tests, earthquake-cycle simulations, Kamchatka slip partitioning, and tunnel or pipeline reliability papers connect material state to cascading exposure. Remote sensing and AI contributions are most useful where they become operationally constrained: high-frequency EO foundation modelling for rapid hazards, uncertainty-aware AI weather evaluation, EuroMineNet for mining footprints, SAR-optical flood-region image generation, causal wildfire forecasting, and physically regularized geomechanical neural models all move beyond generic accuracy toward time, uncertainty, and mechanism.
Key Trends
The selected papers share a common technical direction: hazard assessment is becoming event-sequenced, uncertainty-aware, and tied to physical observables.
- Landslide analysis is moving from inventory to cascade reconstruction: Matai'an, Typhoon Wipha, the Patagonian mesa, the Nu River, loess instability tests, and fractured rock-mass experiments all emphasize precursor development, staged failure, and downstream consequences.
- Multi-trigger mechanics are replacing single-trigger explanations: rainfall plus shaking, seepage plus excavation, unloading plus loading rate, waves plus rainfall plus overtopping, and vibration plus suffusion show that hazard thresholds emerge from coupled forcing histories.
- Forecasting papers are being judged by uncertainty and extrapolation: AI weather ensembles, record-breaking extreme evaluation, signature-kernel scoring, spatial-extreme neural networks, and probabilistic geolocalization all test whether models remain useful at the edge cases that matter for disasters.
- Earth observation is becoming denser in time and more benchmarked: HighFM, EuroMineNet, Sentinel-1 snowmelt timing, radar-gauge precipitation, SAR-optical cloud-free image generation, and mountain air-temperature fusion prioritize repeatable products and operational update cycles.
- Infrastructure geohazards are becoming instrumented digital systems: tunnel MWD, LiDAR odometry, cutterhead torque prediction, support timing, pipeline fault interaction, and copula-based tunnel reliability translate geologic uncertainty into monitored engineering decisions.
Selected Papers
This digest features 72 selected papers from 1,722 papers analyzed. The sequence opens with the giant two-stage Matai'an landslide in Taiwan, then moves through rainfall-conditioned seismic slope failure, arid mesa landslide evolution, Typhoon Wipha road-corridor disruption, Nu River old-landslide geometry, saturated loess instability, fractured rock-mass failure, coastal cavity collapse, tunnel rockburst mechanics, Kamchatka megathrust slip, earthquake-cycle simulation, high-frequency Earth-observation foundation models, uncertainty-aware AI weather forecasting, EuroMineNet mining footprints, multi-hazard infrastructure risk, and operational remote-sensing tools for floods, snowmelt, wildfire, tunnels, and mountain climate.
1. A giant two-stage landslide at Matai'an, Taiwan: insights from seismic source inversion and remote-sensing analysis
Core Problem: The Matai'an event links slope collapse, landslide-dam formation, and breach flooding, but such cascades are difficult to reconstruct fast enough for downstream warning.
Key Innovation: Multitemporal remote sensing is integrated with seismic source inversion to resolve a two-stage, 308 million m3 failure and to show how geomorphic precursors, near-real-time seismic detection, and riverine seismic stations can support dam-breach warning.
2. Dynamic response and failure evolution of sandy slopes under ground motion before and after rainfall infiltration
Core Problem: Sandy slopes can be sequentially weakened by rainfall and then destabilized by ground motion, yet the nonlinear failure evolution is rarely observed stage by stage.
Key Innovation: Shaking-table experiments before and after rainfall infiltration quantify changes in natural frequency, damping, crack coalescence, and the four-stage transition from early deformation to overall sliding.
3. Landslides on an arid mesa: From lava flow to landslide fringe in the Patagonian volcanic tableland
Core Problem: Large landslide fringes around volcanic tablelands are regionally known, but their kinematics and chronology remain poorly resolved in arid, non-glaciated terrain.
Key Innovation: Remote-sensing mapping, morphometrics, field survey, and cosmogenic 36Cl dating reconstruct Quaternary retrogressive sliding, spreading, toppling, rockfall, and flow-like reactivation around Mesa Vizcaina.
4. From clustered landslides to road network disruption: evidence from Typhoon Wipha (2025) in Ningde, Fujian, China
Core Problem: Typhoon-triggered landslide clusters can disrupt road corridors even when event rainfall is not exceptional by historical comparison.
Key Innovation: Sentinel-2 change detection, field mapping, UAV imagery, rainfall records, and slope-stability analysis show that antecedent rainfall, coastal orographic forcing, and wind-root-soil damage drove the Wipha 2025 highway failure.
5. Geomorphic and fluvial controls on the distribution and geometry of old landslides in the Nu River Basin, China
Core Problem: Old landslides record long-term hillslope instability, but their geometry is often analyzed without quantitative links to river incision and fault-controlled landscape adjustment.
Key Innovation: A 258-landslide inventory in the Nu River Basin relates landslide clustering and allometry to lithology, faults, slope position, steepness index, knickpoints, and erosional modification.
6. Investigating instability mechanism of loess: experimental insights from two constant shear modes
Core Problem: Rainfall and groundwater fluctuations destabilize loess slopes, but the initiation point of saturated loess instability is hard to detect experimentally.
Key Innovation: Constant-shear drained and reduced-pressure drained triaxial tests introduce a practical instability-detection method and isolate how pore-pressure rise, confining-pressure reduction, and loading rate control failure.
7. Progressive failure process of rock mass with intermittent fractures: Experimental and numerical investigations
Core Problem: Intermittent fractures control rock-slope strength through progressive bridge failure, yet load transfer and crack propagation are difficult to observe directly.
Key Innovation: Digital image correlation and DEM simulations track shear and tensile strain fields, crack penetration, deflection, and bottom-up rock-bridge failure in coplanar fractured rock masses.
8. Effect of Grain Shape and Porosity on Coastal Internal Erosion and Cavity Collapse under Continued Waves, Rainfalls and Overtopping
Core Problem: Coastal ground collapse under waves, rainfall, and overtopping depends on grain fabric, but collapse thresholds are often treated as material-independent.
Key Innovation: Prototype-scale experiments show that grain shape and porosity strongly control suction, internal erosion, cavity expansion, and collapse resistance under continued marine and rainfall forcing.
9. Combined effects of internal unloading and external loading rate on hard rock tunnel rockburst: Evidences from reduced-scale model tests and numerical simulations
Core Problem: Deep hard-rock tunnels experience both excavation unloading and external stress loading, but their combined effects on rockburst severity remain poorly constrained.
Key Innovation: Reduced-scale model tests and PFC3D simulations reveal how internal unloading lowers buckling thresholds while loading rate changes fragmentation and surrounding-rock strength response.
10. Crack Energy Evolution Mechanism and Rockburst Characteristics of Granite Under Varying Loading Rates: An Experimental and Numerical Investigation
Core Problem: Rockburst proneness in deep granite is governed by loading-rate-dependent crack growth and energy release, which are difficult to separate in field observations.
Key Innovation: Experiments, high-speed imaging, and numerical simulations connect loading rate to crack-induced energy evolution, fragmentation intensity, and rockburst proneness indices.
11. Interseismic, Coseismic, and Early Postseismic Slip Associated With the 2025 Mw 8.8 Kamchatka Earthquake
Core Problem: Large subduction earthquakes alter coupling, afterslip, and nearby volcanic stress fields, but the full seismic-cycle slip partition is hard to resolve.
Key Innovation: Geodetic and seismic observations quantify interseismic coupling, coseismic rupture, postseismic afterslip, dynamic overshoot, cascading triggering, and extensional strain near Kamchatka volcanoes.
12. Seismicity Burst Evolution Suggests Systematic Fault Zone Condition Changes in Japan After the 2011 M9 Tohoku‐Oki Earthquake
Core Problem: Earthquake clustering can reveal hidden fault-zone damage and stress evolution, but conventional catalogs rarely isolate compact burst behavior over decades.
Key Innovation: A statistical analysis of more than 4,000 seismicity bursts in Japan shows persistent post-Tohoku changes near the volcanic front, lower b-values, altered stress ratios, and reduced effective stress drops.
13. Comprehensive Earthquake Cycle Modeling With Multiscale Patches: Unified Simulation Reproducing Observed Laws and Characteristics
Core Problem: Earthquake-cycle models must reproduce magnitude-frequency scaling, Omori aftershocks, repeating earthquakes, and rupture initiation within one long-duration framework.
Key Innovation: A quasi-dynamic rate-and-state model with multiscale frictional patches in the northern Japan Trench reproduces observed seismicity laws and earthquake-cycle characteristics.
14. HighFM: Towards a Foundation Model for Learning Representations from High-Frequency Earth Observation Data
Core Problem: Most Earth-observation foundation models use low-revisit imagery, limiting their value for fast-moving hazards and real-time response.
Key Innovation: HighFM adapts masked autoencoding to more than 2 TB of geostationary SEVIRI imagery, adding fine-grained temporal encodings and improving cloud-masking and active-fire detection tasks.
15. On the Predictive Skill of Artificial Intelligence-based Weather Models for Extreme Events using Uncertainty Quantification
Core Problem: AI weather models can match deterministic benchmarks but still struggle to represent uncertainty and extremes that drive disaster impact.
Key Innovation: Perturbed ensembles for Pakistan floods, China heatwave, and global threshold events compare FuXi, GraphCast, SFNO, IFS ENS, and AIFS ENS to expose where probabilistic AI weather forecasts miss extremes.
16. EuroMineNet: A multitemporal Sentinel-2 benchmark for spatiotemporal mining footprint analysis in the European Union (2015–2024)
Core Problem: Mining-induced land degradation requires temporally consistent monitoring, but existing datasets lack decade-long expert annotations and change-aware evaluation.
Key Innovation: EuroMineNet provides annual Sentinel-2 mining footprint labels for 133 EU sites from 2015 to 2024, benchmarks 31 deep-learning models, and releases a change-aware temporal IoU protocol.
17. Climate-resilient systems prioritisation index: a spatial framework for multi-hazard mitigation via nature-based solutions
Core Problem: Climate-resilience planning needs spatial prioritization that links multi-hazard exposure to nature-based mitigation rather than treating hazards independently.
Key Innovation: The CRESPI framework provides a GIS-based spatial index for prioritizing nature-based solutions across multi-hazard climate resilience planning contexts.
18. Inoperability Assessment of Interdependent Critical Infrastructures exposed to Natural Hazards considering Climate Change
Core Problem: Critical infrastructure failures cascade across sectors, and climate-change projections must preserve spatial hazard correlation to estimate inoperability.
Key Innovation: Spatially coherent stochastic hazard fields are coupled with a dynamic inoperability input-output model to evaluate cascading effects in interdependent power and water systems.
19. Korean Peninsula—Updated Sea-Level Rise Assessment
Core Problem: National sea-level assessments become outdated as tide gauges, satellite altimetry, and vertical land motion estimates evolve.
Key Innovation: The Korean Peninsula assessment updates relative and geocentric sea-level rise using additional gauge and satellite data, identifying where coastal hazard acceleration remains most pronounced.
20. Towards generative location awareness for disaster response: A probabilistic cross-view geolocalization approach
Core Problem: Emergency response depends on locating disaster imagery quickly, but cross-view geolocalization often lacks uncertainty and explainability.
Key Innovation: ProbGLC combines probabilistic and deterministic cross-view geolocalization to support uncertainty-aware, explainable location awareness for rapid disaster response.
21. Physics-based models outperform AI weather forecasts of record-breaking extremes
Core Problem: AI weather forecasts may perform well on average but can underpredict unprecedented extremes that dominate hazard losses.
Key Innovation: A Science Advances evaluation shows physics-based HRES forecasts still outperform leading AI models for record heat, cold, and wind, especially where events exceed the training distribution.
22. SURFLI: mapping social and flood vulnerabilities across urban–rural low-income/low-access areas in North Carolina
Core Problem: Flood risk in low-income and low-access communities requires spatially explicit vulnerability mapping rather than flood exposure alone.
Key Innovation: SURFLI maps social and flood vulnerability across urban-rural low-income/low-access areas in North Carolina to support targeted risk reduction.
23. Using Environmental Tracers to Reduce Uncertainty in Natural Flood Management Modeling
Core Problem: Nature-based flood management models are uncertain when calibrated only on streamflow because they may misrepresent surface and subsurface flow partitioning.
Key Innovation: Environmental tracers are added to model calibration, reducing uncertainty in woodland-planting flood scenarios and improving process representation in an upland catchment.
24. High-resolution dynamic surface water mapping in the Hindu Kush Himalaya region by integrating SDGSAT-1 and Sentinel-2 observations
Core Problem: Surface water in the Hindu Kush Himalaya is dynamic and hazard-relevant, but cloud cover and complex terrain limit high-resolution monitoring.
Key Innovation: SDGSAT-1 and Sentinel-2 observations are integrated for high-resolution dynamic surface-water mapping across the Hindu Kush Himalaya.
25. RRBF-KMA: High-resolution radar–gauge merged precipitation dataset for South Korea (2016–2024)
Core Problem: Radar-gauge precipitation fields are essential for flood and landslide applications, but many regions lack high-resolution merged datasets.
Key Innovation: RRBF-KMA delivers a high-resolution radar-gauge merged precipitation dataset for South Korea from 2016 to 2024.
26. Integrating machine learning with a novel karst hydrological model to enhance extreme streamflow simulation in karst regions
Core Problem: Karst basins produce nonlinear flood and drought responses that conventional hydrologic models often underfit.
Key Innovation: Machine learning is integrated with a karst hydrological model to improve extreme-streamflow simulation.
27. Utilization of traffic images for accurate and realtime flood depth measurement in complex urban environments
Core Problem: Real-time urban flood depth is difficult to measure in complex streets where fixed gauges are sparse.
Key Innovation: Traffic images are used to estimate flood depth accurately and in real time, turning existing cameras into operational flood sensors.
28. An innovative urban drainage network pipe diameter optimization method based on an enhanced genetic algorithm and multi-stage disaster mitigation model
Core Problem: Urban drainage upgrades need pipe-diameter optimization that accounts for system cost and staged disaster mitigation.
Key Innovation: An enhanced genetic algorithm supports multi-stage drainage network pipe-diameter optimization for urban flood mitigation.
29. Hyperspectral constraints reduce bias in ECOSTRESS evapotranspiration and drought indicators
Core Problem: Thermal drought indicators can be biased when hyperspectral constraints are absent from evapotranspiration estimates.
Key Innovation: Hyperspectral information is used to reduce bias in ECOSTRESS evapotranspiration and drought indicators.
30. Snow Water Storage Within Eight Pacific Coastal Watersheds in British Columbia (Canada) Inferred From Four Years of Airborne Lidar Data
Core Problem: Snow storage in Pacific coastal watersheds is difficult to quantify at the spatial detail needed for runoff and flood forecasting.
Key Innovation: Four years of airborne lidar are used to infer snow water storage across eight British Columbia coastal watersheds.
31. A global high-resolution dataset of snowmelt runoff onset timing from Sentinel-1 SAR, 2015–2024
Core Problem: Snowmelt onset controls seasonal runoff and water hazards, but global high-resolution timing data remain limited.
Key Innovation: A Sentinel-1 SAR dataset maps global snowmelt runoff onset timing from 2015 to 2024.
32. Causal Graph Neural Networks for robust wildfire forecasting across geographic shifts
Core Problem: Wildfire forecasting models often transfer poorly across geographic shifts because they learn correlations without fire-mechanism structure.
Key Innovation: A causal graph neural network uses learned graph structure and backdoor-adjusted pooling to improve burned-area forecasting robustness across regions.
33. High-quality cloud-free optical image generation using multi-temporal SAR and contaminated optical data
Core Problem: Clouds remove optical observations during floods and other rapidly changing events, limiting post-event mapping.
Key Innovation: CRGenNet generates cloud-free optical images from multi-temporal SAR and contaminated optical data and introduces the TCSEN12 flood-region benchmark.
34. Bayesian polarimetric detector for improved forest loss monitoring using dual-polarization Sentinel-1 data
Core Problem: Near-real-time forest loss monitoring with Sentinel-1 must handle mixed land covers and different scattering responses after clearing.
Key Innovation: An unsupervised Bayesian polarimetric change detector jointly processes VV and VH Sentinel-1 time series to improve sensitivity across disturbance conditions.
35. AGI-LO: Adaptive Geometric-Intensity LiDAR odometry for tunnel-like degraded environments
Core Problem: LiDAR odometry loses accuracy in tunnels and sharp turns because motion assumptions, frame overlap, and geometric features degrade simultaneously.
Key Innovation: AGI-LO adaptively combines geometric and intensity features to improve odometry in high-speed or feature-poor tunnel-like environments.
36. MmSAM: multimodal meets SAM2 for efficient remote sensing semantic segmentation
Core Problem: Foundation-model segmentation is hard to adapt to multimodal remote sensing without high computation or modality mismatch.
Key Innovation: MmSAM fine-tunes SAM2 for multimodal remote sensing by using extra modalities as hard- and soft-MoE prompts rather than equal image inputs.
37. A Sequential Cooperative Inversion Framework of DC Resistivity and Frequency-Domain Electromagnetic Data to Enhance Subsurface Imaging in Geoscience and Engineering
Core Problem: Separate DC resistivity and frequency-domain electromagnetic inversions can produce incomplete or inconsistent subsurface images.
Key Innovation: A sequential cooperative inversion framework fuses both methods to better delineate hydrogeologically important resistive structures in synthetic and field cases.
38. Cross-Sections and Dimensions: A LiDAR-Based GIS Tool for Bankfull Channel Mapping
Core Problem: Bankfull channel dimensions are difficult to map consistently across landscapes from point-cloud data.
Key Innovation: A LiDAR-based GIS tool extracts cross-sections and dimensions for bankfull channel mapping.
39. Development of a method for estimating soil moisture using obliquely reflected light in two-dimensional soil experiments
Core Problem: Two-dimensional slope-failure experiments need low-cost, noninvasive moisture measurements that work across dark and light soils.
Key Innovation: Obliquely reflected light is used to relate dark-pixel proportion to volumetric water content, improving soil-moisture visualization in 2D models.
40. Application of the multimodal model for predicting rock mechanical parameters via MWD feature splicing and fusion
Core Problem: Deep tunnel construction needs real-time rock mechanical parameters, but laboratory testing lags behind excavation decisions.
Key Innovation: A multimodal CNN-XL framework fuses MWD statistics and tunnel-face imagery, with SHAP interpretation, to predict cohesion, modulus, friction angle, and rock strength.
41. Dust storms: Hidden drivers of extreme rainfall and global precipitation shifts
Core Problem: Dust storms are usually treated as hazards themselves, but their influence on heavy rainfall and precipitation redistribution is underused in prediction.
Key Innovation: Global observations and simulations show that dust acts as ice nuclei, enhancing 7-day precipitation and shifting aerosol-rainfall patterns toward heavier rainfall.
42. The Role of Tropical Cyclone—Ocean Interactions in Future Changes in Hurricane Katrina
Core Problem: Future hurricane projections differ depending on whether atmosphere-ocean coupling is represented.
Key Innovation: Convection-permitting atmosphere-only and atmosphere-ocean regional ensembles for Hurricane Katrina show that coupling damps pressure decreases but amplifies precipitation scaling.
43. Modeling Spatial Extremal Dependence of Precipitation Using Distributional Neural Networks
Core Problem: Spatial dependence of rainfall maxima is hard to estimate where likelihood-based max-stable models become intractable.
Key Innovation: Distributional neural networks estimate spatio-temporal extremal dependence and uncertainty for Western German rainfall maxima, including the July 2021 flood event.
44. Signature Kernel Scoring Rule: A Spatio-Temporal Diagnostic for Probabilistic Weather Forecasting
Core Problem: Weather forecast verification often ignores path-dependent space-time structure, weakening evaluation of probabilistic ML forecasts.
Key Innovation: A signature-kernel scoring rule treats weather variables as continuous paths and provides a strictly proper diagnostic for WeatherBench 2 models.
45. A Time-Dependent Analytical Model for Shield Tunnel Uplift Incorporating Low-Carbon Grout Dynamics
Core Problem: Shield tunnel uplift during synchronous grouting depends on time-evolving grout rheology that static models omit.
Key Innovation: A Pasternak foundation beam model incorporates low-carbon grout strength and viscosity evolution and is validated against Shanghai Metro monitoring data.
46. A multiscale computational framework for material-interface-structure integrated design of UHPC-based strengthening systems for shield tunnel linings
Core Problem: UHPC strengthening systems for shield tunnel linings require compatibility across material, interface, and structural scales.
Key Innovation: A multiscale computational framework links material-interface damage evolution to full lining performance and design choices for prefabricated UHPC panels.
47. Impact characteristics of shield disc cutters under composite strata: insights from full-scale cutting tests and theoretical studies
Core Problem: Shield disc cutters are damaged by hard-soft stratum interfaces, especially as tunnel diameters increase.
Key Innovation: Full-scale cutting tests and theory derive an impact-load model that accounts for rock damage, strength contrast, and interface bonding.
48. Quantitative analysis of surrounding rock response considering the dynamic coupling evolution of 3D strength criterion parameters and seepage parameters
Core Problem: Deep excavations under high ground stress and pore pressure require coupled strength and seepage parameters rather than static rock-mass assumptions.
Key Innovation: A differential-iteration elastoplastic model couples 3D strength criteria, hydro-mechanical effects, and dynamic parameter evolution for surrounding-rock response analysis.
49. Real-time cutterhead torque prediction via cluster-based geological identification in karst shield tunneling
Core Problem: Karst shield tunnelling faces abrupt geological transitions that can overload cutterheads if torque is not predicted in real time.
Key Innovation: K-means and DBSCAN formation identification are embedded in a CNN-LSTM-MHA model for cutterhead torque prediction and adaptive control.
50. A multi-borehole drilling-rate based method for spatial characterization of internal rock mass discontinuities in tunnelling
Core Problem: Internal discontinuities control tunnelling risk, but advance prediction from sparse boreholes is uncertain.
Key Innovation: Measurement-while-drilling rate anomalies from multiple boreholes are clustered and fitted with sample-consensus planes to reconstruct discontinuity geometry.
51. Conical strain wedge model for nonlinear analysis of buried continuous pipelines under strike-slip fault movement
Core Problem: Fault displacement can buckle or rupture buried pipelines, while beam-on-foundation models underrepresent 3D soil plasticity.
Key Innovation: A conical strain wedge model refines pipeline-soil interaction under strike-slip movement by incorporating soil flow-around failure.
52. Seismic response of underground station–soil–structure cluster interaction systems: Insights from shaking table tests
Core Problem: Underground stations, tunnels, and nearby buildings interact through foundation soils during earthquakes.
Key Innovation: Shaking-table tests compare soil-station, soil-structure, and combined station-soil-structure cluster systems under different ground-motion inputs.
53. Out-of-plane response of lined tunnels with joints in time space to seismic excitations induced by SH waves in visco-elastic half-space
Core Problem: Jointed tunnel linings require time-domain out-of-plane response analysis under inclined seismic waves.
Key Innovation: A Fourier-space soil-structure interaction formulation with joint springs is transformed to time space by FFT for lined tunnels in visco-elastic half-space.
54. Copula-based system reliability analysis of shallow-buried rectangular tunnels considering groundwater and multiple failure modes
Core Problem: Shallow rectangular tunnels can fail through vault, sidewall, or invert mechanisms under uncertain pore pressure and correlated soil parameters.
Key Innovation: Upper-bound limit analysis is combined with copula-based probabilistic modelling to quantify tunnel system reliability under groundwater effects.
55. Permeability evolution driven by the suffusion process in silica sands under cyclic loading: Mathematical model and experimental verification
Core Problem: Granular soils undergoing suffusion change permeability as particles migrate under seepage and dynamic loading.
Key Innovation: A Kozeny-Carman extension using erosion ratio is validated with cyclic-load sand-column tests to predict permeability evolution.
56. Vibration-exacerbated suffusion in soils with different gradations: a coupled DEM–PNM study
Core Problem: Train vibration can intensify soil suffusion around shield tunnels, but gradation controls are poorly understood.
Key Innovation: Coupled DEM-PNM simulations show how particle-size ratio and fines content modulate vibration-driven pore-structure evolution and fine-particle loss.
57. GPU-accelerated semi-implicit two-phase double-layer MPM with adaptive projection-error compensation for accurate soil–water coupling
Core Problem: Two-phase MPM simulations of soil-water hazards can accumulate nonphysical water-volume errors.
Key Innovation: A GPU-accelerated double-layer MPM adds adaptive projection-error compensation to preserve incompressible water volume in porous-flow and free-surface coupling.
58. Geomechanics-motivated discovery of three-dimensional coupled governing equations via physics-informed symbolic genetic algorithms
Core Problem: Prescribed governing equations can bias geomechanical models when multiphysics coupling is unknown or misspecified.
Key Innovation: A physics-informed symbolic genetic algorithm discovers explicit 3D coupled PDEs from sparse observations using tensor-consistent operators and field separation.
59. Inverse physics-informed neural network framework for physics-consistent estimation of dispersion coefficient and mechanistic understanding of solute transport
Core Problem: Conventional breakthrough-curve fitting can estimate dispersion coefficients that reproduce data but violate transport physics.
Key Innovation: An inverse PINN embeds the advection-dispersion equation to estimate physically consistent dispersion coefficients for sand-illite solute tests.
60. Physics-informed GRU encoder-decoder model for predicting cross-scale mechanical behavior of rock
Core Problem: Microscale nanoindentation measurements do not directly translate into macroscopic rock stress-strain behavior.
Key Innovation: A physics-informed GRU encoder-decoder maps nanoindentation-derived micromechanical parameters to macroscopic rock response with plastic-evolution constraints.
61. A critical state-constrained neural network for modeling the constitutive behavior of sand and clay under monotonic loading
Core Problem: Purely data-driven soil constitutive models may become unstable or nonphysical outside narrow training conditions.
Key Innovation: A critical-state-constrained neural network integrates strain decomposition, elastic updating, and a learnable critical state line for sand and clay.
62. Shear strength and durability enhancement of fly ash and bottom ash mixed soil using microbial-induced calcite precipitation: A comparative study
Core Problem: Fly ash and bottom ash can be reused in geotechnical fills but need sufficient strength and durability.
Key Innovation: MICP treatment with Sporosarcina pasteurii improves ash-mixed soil shear strength and durability under wetting-drying and freeze-thaw cycles.
63. Bayesian parametric calibration of Shanghai deep clayey soil with CPTu data under partial drainage conditions
Core Problem: Shanghai deep clay behaves between drained and undrained states, making uniform parameter choices unreliable for deep excavations.
Key Innovation: An enhanced Modified Cam-Clay model is linked to CPTu cavity expansion and calibrated with Bayesian inference for partially drained clay parameters.
64. Using 2D electrical resistivity imaging and borehole data to estimate N60-value of soils with k-means clustering for subsurface geomaterials categorization
Core Problem: Drilling alone is costly for site investigation, while geophysical surveys need calibration to engineering soil indices.
Key Innovation: 2D electrical resistivity imaging, SPT borehole data, and k-means clustering are integrated to estimate N60 values and categorize subsurface geomaterials.
65. Comprehensive performance tests and evaluations of the slope protection effect of lignin and guar gum-modified loess
Core Problem: Loess slopes require eco-friendly treatments that improve shear strength while reducing erosion, disintegration, and permeability.
Key Innovation: Direct shear, disintegration, permeability, SEM, and slope-model tests evaluate lignin and guar gum treatments and identify effective dosage ranges.
66. Time-dependent optimization of support application timing in soft rock tunnels considering creep deformation and structural degradation
Core Problem: Soft-rock tunnel support timing is often empirical despite creep deformation, moisture weakening, and structural degradation.
Key Innovation: A time-resolved framework couples Burgers-Drucker-Prager creep failure with support-capacity degradation to optimize primary support and secondary lining timing.
67. ZAYAN: Disentangled Contrastive Transformer for Tabular Remote Sensing Data
Core Problem: Remote-sensing tabular data are heterogeneous, label-scarce, and redundant, limiting generalization of deep tabular models.
Key Innovation: ZAYAN uses self-supervised feature-centric contrastive learning and a Transformer to improve remote-sensing tabular and flood-prediction benchmarks under label scarcity.
68. A generalised pre-training strategy for deep learning networks in semantic segmentation of remotely sensed images
Core Problem: ImageNet pretraining creates domain gaps for remote-sensing segmentation, while domain-specific pretraining datasets can generalize poorly.
Key Innovation: A generalized pretraining strategy discourages domain-specific feature dependence during pretraining to improve transfer across remote-sensing segmentation scenarios.
69. Noise2Map: End-to-End Diffusion Model for Semantic Segmentation and Change Detection
Core Problem: Remote-sensing segmentation and change detection models often struggle with temporal inconsistency and fine-grained structure.
Key Innovation: Noise2Map repurposes diffusion denoising for fast end-to-end semantic and change-map prediction without costly sampling.
70. A Small Object Detection Transformer for UAV Remote Sensing Imagery via Multi-Scale Perception and Cross-Spatial-Frequency Domain Fusion
Core Problem: Small objects in UAV imagery are hard to detect because sparse pixels and complex backgrounds degrade Transformer features.
Key Innovation: MSF-DETR combines multi-scale perception, contextual anchor attention, and spatial-frequency enhancement for UAV small-object detection.
71. A 3D Gaussian Splatting Method with Deterministic Structure-Sensitive Adaptive Density Control for UAV Orthophoto Generation
Core Problem: UAV orthophoto generation with 3D Gaussian Splatting is limited by weak textures, occlusions, memory cost, and unstable densification.
Key Innovation: A single-GPU 3DGS workflow uses adaptive spatial blocks, deterministic structure-sensitive density control, and tiled orthographic rendering with weighted blending.
72. A global mountain long-term monthly mean air temperature dataset from priority-based weighted fusion of multi-source 1-km satellite LST products and in-situ observations
Core Problem: Mountain air-temperature gradients are poorly captured by coarse reanalyses, biasing climate and hazard analyses in complex terrain.
Key Innovation: A 1-km global mountain monthly air-temperature dataset fuses six satellite LST products with 6,842 station records and topographic auxiliaries for 1997-2020.