TerraMosaic Daily Digest: May 4, 2026
Daily Summary
May 4's papers concentrate on where hazards become mechanically legible. Typhoon Gaemi landslides in Zixing are analysed as lithology-controlled clusters with field-validated mobility patterns; the Asadabad earthquake is resolved as a shallow, multi-stage rupture on an unmapped fault with pre-event deformation; earthquake-triggered landslide mapping is tested both as susceptibility modelling and as change detection under confounding terrain. Rockfall risk is linked directly to road-network function in Rome, while flood papers move from rapid Himalayan susceptibility mapping to low-fidelity inundation surrogates, WRF-driven damage forecasts, compound coastal-flood synthesis, transferable SAR-optical mapping, and flood segmentation from cross-polarized SAR.
The broader set treats cascading risk as a coupled physical and institutional problem. Tsunami ionospheric disturbances, Istanbul Natech exposure, wildfire hazard segmentation, tunnel-fire temperature reconstruction, liquefaction uplift, volcanic resistivity change, and snow-avalanche neural operators extend monitoring beyond conventional ground observations. Subsurface and underground studies resolve wetting collapse, shallow karst groundwater, tunnel settlement, shield-lining deformation, deep layered-rock support, mining-induced gas migration, rock damage, hydraulic-fracture microseismicity, repository creep, and cold-region soil mechanics. AI papers are strongest when constrained by physics, transfer, or explanation: InSAR unwrapping rejects unnecessary architectural complexity, disaster severity assessment adds preference-aligned rationales, and geospatial embeddings, building inspection, flood mapping, FWI, PINNs, and aerial detection are evaluated against deployment constraints.
Key Trends
The strongest contributions make the model answerable to a physical process, an exposed asset, or an operational warning problem.
- Landslide evidence is tied to triggering material and change history: lithology, rainfall intensity, pre-existing terrain, coseismic exposure, and post-event spectral change are treated as separable controls rather than folded into one susceptibility score.
- Flood modelling is judged by transfer and response time: Himalayan susceptibility mapping, WRF-hydraulic damage chains, alluvial-fan reconstruction, low-fidelity inundation surrogates, transferable SAR-optical U-Nets, and SAR fusion all target rapid use outside ideal calibration settings.
- Cascades are becoming asset-specific: roads, neighborhoods, industrial facilities, hospitals, metro systems, emergency supply chains, tunnels, buried structures, and shelters are modelled through their own failure states instead of as generic exposure layers.
- Physics-constrained AI is preferred where measurements are sparse: InSAR phase unwrapping, FWI, PINNs for tunnel fire and thermal fields, physics-informed groundwater flow, snow-avalanche operators, and permeability networks all embed governing structure to reduce brittle data dependence.
- Subsurface risk is represented as coupled multi-field behaviour: water, stress, heat, chemistry, salt, cyclic loading, and fracture networks jointly control wetting collapse, liquefaction uplift, karst flow, mine goafs, geothermal fractures, repository creep, and cold-region soil response.
Selected Papers
This issue contains 77 selected papers from 2,085 papers analyzed. The leading papers resolve lithology-controlled Typhoon Gaemi landslide clusters, shallow pre-event deformation and multi-stage rupture during the Asadabad earthquake, earthquake-induced landslide susceptibility and change detection, rockfall impacts on Rome's road network, and rapid flood and tsunami warning products. The broader set links compound flooding, wildfire hazard segmentation, Natech vulnerability, tunnel-fire reconstruction, liquefaction uplift, avalanche neural operators, karst groundwater, underground excavation and storage mechanics, and physics-constrained GeoAI for InSAR, flood mapping, FWI, building inspection, and aerial detection.
1. Lithological controls on clustered landslides: a case study of landslides triggered by Typhoon Gaemi (2024) in Zixing, Hunan Province, China
Core Problem: Extreme rainfall can trigger dense landslide clusters, but the lithological controls on clustering and post-failure mobility are rarely resolved with field validation.
Key Innovation: The Zixing study builds a remote-sensing and field-validated inventory for Typhoon Gaemi landslides and links clustering, mobility, slope gradient, residual soils, and lithology through statistics, machine learning, and material testing.
2. Multi‐Stage Shallow Rupture and Pre‐Event Shallow Deformation During the 2025 Mw 5.9 Asadabad Earthquake (Afghanistan)
Core Problem: Moderate earthquakes can cause unexpectedly severe damage when shallow rupture occurs on immature or unmapped faults.
Key Innovation: Sentinel-1 InSAR and teleseismic waveforms reveal multi-stage shallow rupture, pre-event uplift, and secondary surface-approaching rupture during the 2025 Mw 5.9 Asadabad earthquake.
3. Rapid flood susceptibility mapping in the Indian Himalayan region using CNN‑U‑Net segmentation: insights from the 2025 monsoon events
Core Problem: Complex Himalayan catchments require rapid hazard information during the first hours of monsoon floods, when data and field access are limited.
Key Innovation: A CNN-U-Net framework combines 14 hydro-geomorphological predictors and a local convexity factor to map flood susceptibility after the 2025 Himachal Pradesh and Uttarakhand monsoon events.
4. Rockfall impacts on the road network of the Metropolitan Area of Rome: a risk prioritisation framework for decision-makers
Core Problem: Rockfall risk to road networks needs prioritization methods that connect slope source areas to network segments and emergency mobility.
Key Innovation: A Rome metropolitan framework integrates random-forest susceptibility, semi-empirical runout, impact probability, velocity, kinetic energy, and road-network exposure for decision-oriented risk ranking.
5. Assessing earthquake-induced landslide susceptibility: a comparative study with and without landslide inventory data in the Indian Himalayan region
Core Problem: Many seismically active mountain regions lack complete landslide inventories, limiting earthquake-induced susceptibility zonation.
Key Innovation: The Sikkim study compares frequency ratio, random forest, and Newmark displacement approaches with and without landslide inventory data across seismic, terrain, lithologic, hydrologic, road, land-use, and soil controls.
6. Change detection-based machine learning for earthquake-triggered landslide identification in complex mountain landscapes
Core Problem: New coseismic landslides can be confused with old landslide scars, bare ground, and human disturbance in satellite imagery.
Key Innovation: A change-detection machine-learning framework uses bi-temporal imagery and cross-zone tests after the 2017 Jiuzhaigou earthquake to separate newly exposed landslides from confounding terrain.
7. Near–real-time detection of tsunami-induced traveling ionospheric disturbances using multi-constellation GNSS and a temporal convolutional network
Core Problem: Tsunami warning systems need fast, physically consistent detection signals before coastal impact.
Key Innovation: Streaming multi-constellation GNSS dTEC from GUARDIAN is analysed with an unsupervised temporal convolutional network, phase-synchrony checks, and spatial consistency filters for near-real-time tsunami disturbance detection.
8. Improving flood and damage forecasting through WRF-based hydrological and hydraulic models in urban environments
Core Problem: Flood forecasts often stop at water levels and do not propagate meteorological uncertainty into direct damage estimates.
Key Innovation: WRF precipitation ensembles are bias-corrected and coupled with HEC-HMS, HEC-RAS, and empirical depth-damage curves to estimate flood hydrographs, hydraulics, and damage in Iran's Poldokhtar basin.
9. Identifying Natech vulnerability based on the Istanbul earthquake scenario at the neighborhood level
Core Problem: Industrial and commercial hazards can turn earthquake shaking into cascading technological risk, especially at neighborhood scale.
Key Innovation: The Istanbul scenario combines seismic structural damage estimates for 939 neighborhoods with more than 500,000 geocoded businesses and hazard-class information to derive Natech vulnerability indices.
10. Flash flooding in semiarid alluvial fan apron systems: a holistic approach to hazard assessment in the Menor Sea Basins (Spain)
Core Problem: Official hazard maps can miss active flood pathways in highly modified alluvial fan aprons.
Key Innovation: DEM, orthophoto, rainfall, gauge, geomorphological, risk-map, and Sentinel-2 turbidity evidence are combined to reconstruct the 2019 Menor Sea Basin flash flood and diagnose active fan sectors.
11. When Less Is More: Simplicity Beats Complexity for Physics-Constrained InSAR Phase Unwrapping
Core Problem: Operational InSAR unwrapping is slowed by complex computer-vision models that may violate smooth elastic deformation physics.
Key Innovation: A large LiCSAR ablation study shows that a simpler U-Net outperforms attention-heavy models and avoids unphysical high-frequency artefacts in geophysical regression.
12. Compound flooding risks in coastal cities: a review of causes, mechanisms, and assessment methodologies
Core Problem: Coastal-city flood risk is increasingly compound, but assessment methods remain fragmented across drivers and scales.
Key Innovation: A systematic review of 2015-2025 literature synthesizes mechanisms, definitions, assessment frameworks, climate-change effects, and the emerging role of AI in compound-flood prediction.
13. A review of tsunami vulnerability assessment: focus on human indicator
Core Problem: Tsunami vulnerability assessments often overemphasize physical exposure and underrepresent social capacity, warning, evacuation, and recovery.
Key Innovation: A systematic review organizes human vulnerability indicators into exposure, early warning, evacuation-response, and recovery domains for more locally adaptable tsunami vulnerability models.
14. A transferable deep learning framework for flood mapping: Spatial generalization across hydro-climatic regimes using satellite imagery
Core Problem: Deep flood-mapping models often report in-domain accuracy without testing spatial transfer or operational robustness.
Key Innovation: Separate Sentinel-1 and Sentinel-2 U-Net models are trained on globally sourced flood data and evaluated for spatial generalization beyond the training regions.
15. Strategies for Predicting Flood Inundation in a Large and Complex Floodplain Based on Low‐Fidelity Hydrodynamic Models
Core Problem: High-fidelity hydrodynamic models are too slow for ensemble or real-time flood applications in large urbanized floodplains.
Key Innovation: A low-fidelity spatial analysis and Gaussian-process learning surrogate is tested on the Lower Brisbane River floodplain with tributary interactions, tide effects, and dam releases.
16. Wildfire hazard segmentation from space-sensed data using deep learning models: preserving the spatial geographical properties of the data
Core Problem: Wildfire hazard segmentation must preserve geographic structure rather than treating pixels as independent image fragments.
Key Innovation: Deep segmentation models are designed and evaluated for wildfire hazard mapping from space-sensed data while retaining spatial geographic properties.
17. Cross-Polarization Fusion of VV AND VH SAR Observations for Improved Flood Mapping
Core Problem: Single-polarization SAR flood maps can fail under complex surface scattering and land-cover ambiguity.
Key Innovation: VV and VH SAR observations are fused to improve flood-water delineation, strengthening all-weather flood mapping for emergency response.
18. FLoRA: Fusion-Latent for Optical Reconstruction and Flood Area Segmentation via Cross-Modal Multi-Task Distillation Network
Core Problem: Optical imagery is interpretable but weather-limited, whereas SAR is all-weather but visually ambiguous.
Key Innovation: FLoRA distils optical RGB and NDVI priors into Sentinel-1 SAR representations through cross-modal multi-task learning that jointly reconstructs optical imagery and segments flood water.
19. An analytical model for liquefaction-induced underground buried structure uplift with experimental validation
Core Problem: Underground structures can uplift during liquefaction, but analytical criteria often lack experimental validation.
Key Innovation: An analytical uplift model is developed and checked against experiments to quantify buried-structure response under liquefaction.
20. Temporal Change in Resistivity Structure Before and After the 2018 Small Phreatic Eruption Iwo‐Yama Volcano, Kirishima Volcanic Complex, Japan
Core Problem: Small phreatic eruptions can occur with limited surface warning, making subsurface hydrothermal monitoring critical.
Key Innovation: Time-varying resistivity structure around Iwo-yama volcano is reconstructed before and after the 2018 eruption to identify changes in hydrothermal state.
21. A physics-informed multi-scale fourier neural operator framework for snow avalanche dynamics simulation
Core Problem: Avalanche simulation needs fast models that retain the governing multi-scale physics of moving snow masses.
Key Innovation: A physics-informed multi-scale Fourier neural operator framework is proposed for snow-avalanche dynamics simulation.
22. Towards reliable multimodal disaster severity assessment through preference optimization and explainable vision-language reasoning
Core Problem: Damage-assessment models need both reliable classification and transparent reasoning, but annotated rationales are scarce.
Key Innovation: A human-in-the-loop pipeline builds reasoning and preference datasets, then combines supervised fine-tuning and direct preference optimization for explainable vision-language disaster assessment.
23. Integrating initial fractal dimension with LightGBM for enhanced wildfire burned-area prediction
Core Problem: Environmental predictors alone often miss the geometric complexity of early fire spread.
Key Innovation: Initial fractal dimension from MODIS active-fire masks during the first 72 hours is added to LightGBM, improving burned-area prediction for large US wildfires.
24. Multi-scale Assessment of Wetting-Induced Collapse in a Lateritic Soil: From Microstructure to Foundation Performance
Core Problem: Collapsible tropical soils can lose strength abruptly when wetted, threatening foundations and earth structures.
Key Innovation: Oedometer, shear, plate-load, microstructural, and suction tests quantify wetting-induced collapse from pore-scale fabric to foundation bearing capacity.
25. Assessment of multiple predictors to the psychological effects of flooding for residential and business areas in Peninsular Malaysia
Core Problem: Flood risk studies often underrepresent intangible psychological damage despite its role in wellbeing and recovery.
Key Innovation: A Peninsular Malaysia survey uses willingness-to-pay and predictor analysis to quantify household and business psychological impacts from flooding.
26. Mapping flood memory: How risk perception and social vulnerability drive flood insurance patterns in the U.S
Core Problem: Flood-insurance behaviour reflects memory, vulnerability, and perceived risk, not only physical hazard exposure.
Key Innovation: US flood-insurance patterns are mapped against flood memory, risk perception, and social vulnerability to interpret protection gaps.
27. Leveraging Past Recovery Strategy for Future Policy Implementation: Post-2018 Earthquake Residential Housing Reconstruction in Bayan Timur Hamlet, Lombok, Indonesia
Core Problem: Recovery policies often fail to learn systematically from household reconstruction strategies after earthquakes.
Key Innovation: The Lombok study extracts past recovery strategies to inform future housing-reconstruction policy implementation.
28. Post-earthquake functionality and patient waiting times of fully equipped hospital critical rooms using 3D virtual reality damage scenarios
Core Problem: Hospital functionality can collapse through nonstructural and content damage even when structural damage is limited.
Key Innovation: 3D virtual-reality damage scenarios and probabilistic room-level simulations estimate residual functionality and waiting times for emergency, ICU, and operating rooms under earthquake levels.
29. A hybrid complex network-ML approach for critical state identification and mitigation in emergency supply chain cascading failures
Core Problem: Emergency supply chains can fail through nonlinear cascades before collapse is obvious.
Key Innovation: Complex-network cascading models, Markov chains, and ensemble machine learning identify phase-transition thresholds and mitigation points in emergency supply networks.
30. Dynamic evolution of metro network vulnerability: shifting mechanisms under cascading failures
Core Problem: Static network snapshots miss how metro vulnerability changes as networks densify and passenger flows grow.
Key Innovation: A 21-year Beijing Metro analysis combines unsupervised stage detection, multi-scale critical-station ranking, and agent-based cascade simulation.
31. SPAMoE: Spectrum-Aware Hybrid Operator Framework for Full-Waveform Inversion
Core Problem: Full-waveform inversion remains ill-posed and computationally expensive, especially for multi-scale geological features.
Key Innovation: SPAMoE uses spectral-preserving encoders and frequency-band routing across neural operators to stabilize multi-scale velocity reconstruction.
32. A damage-coupled elastoplastic constitutive framework for ductile-like failure of rocks with stress-induced stiffness anisotropy and damage-dependent asymmetric yielding: numerical implementation and applications
Core Problem: Rock failure models often underrepresent stress-induced anisotropy and asymmetric damage during progressive failure.
Key Innovation: A thermodynamic elastoplastic framework couples stiffness degradation, damage-dependent yielding, Hoek-Brown and Drucker-Prager behaviour, and finite-element implementation.
33. Real-time full-field reconstruction and prediction of temperature field evolution stimulated by dynamic tunnel fires using a physics-informed neural network (PINN) assisted by online-updated sparse monitoring data
Core Problem: Tunnel-fire decisions require full-field temperature information from sparse sensors under changing heat release.
Key Innovation: An online-updated physics-informed neural network reconstructs and predicts full-field tunnel-fire temperature without pretraining.
34. A complex frequency-domain physical neural network method for calculating the quasi-steady-state temperature field of tunnels
Core Problem: Cold-region tunnel frost-damage mitigation needs efficient temperature-field solutions under periodic boundaries and multiple materials.
Key Innovation: A multi-domain frequency-domain physics-informed neural network solves quasi-steady multilayer heat conduction in tunnels.
35. Simulating and characterising concurrent lateral–vertical saltwater intrusion in low-lying coastal aquifers under pumping and storm-surge forcing
Core Problem: Coastal aquifer studies often separate pumping-driven lateral intrusion from storm-surge-driven vertical intrusion.
Key Innovation: A coupled density-dependent surface-subsurface model decomposes lateral, vertical, and cross-effect salt-mass contributions over pumping and surge scenarios.
36. Physics-informed extreme learning machine with time-marching and random features for unsaturated groundwater flow
Core Problem: Physics-informed groundwater solvers can be too costly for strongly nonlinear unsaturated flow.
Key Innovation: TR-PIELM combines time marching, random features, governing-equation constraints, and boundary enforcement for efficient unsaturated groundwater flow simulation.
37. A semi-distributed model with physically-derived HRUs and time-dependent parameters for event-based flood simulation
Core Problem: Flood forecasting is weakened by nonstationary runoff generation and static parameterization.
Key Innovation: PHAST uses physically derived hydrological response units and time-adaptive parameters to represent infiltration, overland flow, subsurface flow, and routing during flood events.
38. Predicting the probability of concentrated flow occurrence using a dynamic approach on high-resolution soil surface data
Core Problem: Concentrated flow controls accelerated soil erosion, but its spatial probability is rarely derived from 3D surface structure.
Key Innovation: Virtual rainfall, flow routing, and computer-vision channel analysis estimate concentrated-flow probability from high-resolution soil-surface data.
39. Optimizing the penetration of a quasi-RTBM under an ultra shallow cover to enhance tunneling performance and minimize ground settlement via discrete-continuum coupling analysis
Core Problem: Large-section rectangular tunnelling under ultra-shallow cover can cause ground deformation controlled by operational penetration.
Key Innovation: DEM-FEM coupling links RTBM penetration to localized soil arching, tunnelling performance, and surface settlement.
40. Thermo-mechanical damage characteristics and fracture behavior of backfill bodies in thermal energy storage cavities: An experimental and numerical study
Core Problem: Backfilled thermal-storage cavities may lose stability through thermal fatigue and stress redistribution.
Key Innovation: Heated backfill compression tests, acoustic emission, and numerical modelling resolve microcrack initiation, propagation, and coalescence under thermo-mechanical loading.
41. Deformation evolution mechanism and intelligent prediction of red clay stratum under subway cyclic loading using a hybrid deep learning model
Core Problem: Subway cyclic loading can accumulate deformation in red clay, but the mechanism and prediction tools remain incomplete.
Key Innovation: Dynamic triaxial tests, SEM-EDS, a modified shakedown limit, and hybrid deep learning quantify and predict red-clay deformation under subway loads.
42. Zoning criteria and applications for pressure-relief gas migration-storage areas in stope based on the “analogous hyperbola” model of mining-induced strata movement
Core Problem: Gas extraction in mining faces requires zoning criteria tied to strata movement and gas-enriched channels.
Key Innovation: An analogous-hyperbola model of mining-induced strata movement defines gas migration-storage zones and supports an integrated extraction strategy.
43. 3D geological model-based coupled modeling method for braced excavation integrated with adjacent buildings and its application to monitoring scheme optimization
Core Problem: Borehole-defined geology is difficult to transfer into excavation simulations with adjacent buildings.
Key Innovation: A 3D geological model maps stratigraphic geometry and properties into numerical grids for excavation-building coupling and monitoring-scheme optimization.
44. Performance assessment of offshore wind turbines with pile–wheel composite foundations considering combined effects of wind–wave–current–earthquake
Core Problem: Offshore wind foundations in seismic regions experience combined environmental and earthquake loading.
Key Innovation: A soil-water-structure numerical model evaluates pile-wheel composite foundation performance under coupled wind, wave, current, and seismic excitations.
45. A Novel Elastoplastic Dynamic Damage Constitutive Model and Numerical Implementation of Rock Under True Triaxial Multistage Disturbance
Core Problem: Rock around excavations experiences coupled static stress and micro-dynamic disturbances not captured by simple constitutive laws.
Key Innovation: True-triaxial static-dynamic tests and a new elastoplastic dynamic damage model reproduce deformation and strength response under multistage disturbance.
46. Experimental Study of the Effect of Shale Fracture Shear-Slip Behavior Under Hydraulic Pulses
Core Problem: Deep shale stimulation needs fracture-network complexity without uncontrolled induced seismic response.
Key Innovation: Hydraulic pulse injection experiments compare monotonic and pulsed fracture shear-slip behaviour in Southern Sichuan Basin shale.
47. Investigation on the Grouting Diffusion in Granite with Rough Fracture: Insights from Physical Model Tests and Numerical Simulation
Core Problem: Rough fractures strongly control grout diffusion, pressure evolution, and sealing performance.
Key Innovation: Physical fracture tests, 3D scanning, NMR, and numerical simulation quantify slurry diffusion and sealing under roughness-controlled conditions.
48. Study on Microfracture Mechanism and Constitutive Model of Deep Stratified Sandstone Under Impact Load
Core Problem: Deep stratified sandstone fails under coupled in-situ stress, impact disturbance, and bedding anisotropy.
Key Innovation: Improved SHPB and microscopic tests quantify strength, energy evolution, crack propagation, and constitutive response across bedding angles.
49. Geomechanical Modeling and Critically Stressed Fracture Analysis for Geothermal Resource Management – A Study from the Menderes Metamorphic Basement, Gediz Graben, Türkiye
Core Problem: Geothermal production requires understanding stress regime, caprock integrity, and critically stressed fracture behaviour.
Key Innovation: A geomechanical model for the Gediz Graben reservoir integrates breakouts, stress ratios, normal faulting, and fracture stability for resource management.
50. Efficient Prediction of THM Multi-field Coupled Mechanical Properties of CO2-Filled Storage Based on Machine Learning
Core Problem: CO2 mineralization storage in mine goafs needs rapid prediction of coupled mechanical responses.
Key Innovation: Machine-learning surrogate models are trained for THM-coupled storage units to predict key stress and deformation responses under varying storage parameters.
51. Investigation on performance-based deformation control criteria for shield tunnel linings under typical adjacent construction disturbances
Core Problem: Existing shield-lining deformation criteria are often empirical and case-specific.
Key Innovation: Numerical analysis of upper loading, lateral unloading, and upper unloading supports performance-based deformation-control criteria and safety classification.
52. Study on the TCNN-augmented adaptive smoke control in tunnel fires: theory, technical framework, and numerical demonstration
Core Problem: Tunnel smoke-control systems must adapt to dynamic fire states with sparse sensing and limited communication.
Key Innovation: A TCNN-augmented framework reconstructs temperature slices and adapts ventilation using backlayering length, IoT sensing, and AI control.
53. Failure characteristics and structural ring properties of deep tunnels with rockbolts in horizontal layered rock
Core Problem: Weak bedding planes in layered rock complicate rockbolt support design under high geostress.
Key Innovation: Physical model tests and DEM simulations compare unsupported, system-bolted, and invert-rockbolt cases to resolve stress transfer and structural ring behaviour.
54. MWD experimental exploration and case studies for elastic constant and in-situ stress estimation based on the impact-rotary rock-breaking mechanism
Core Problem: High in-situ stress in plateau tunnels requires rapid prediction during excavation.
Key Innovation: Measurement-while-drilling parameters and borehole wave velocity are integrated in laboratory and case-study workflows to estimate elastic constants and in-situ stress.
55. Study on the morphology features and formation mechanism of shear fracture zone (SFZ) of flawed sandstone under static and dynamic coupled compressive-shear loading
Core Problem: Deep rock masses experience combined shear, compression, dynamic disturbance, and flaw-controlled fracture coalescence.
Key Innovation: SHPB-based static-dynamic coupled tests resolve shear-fracture-zone morphology, crack sequence, and preloading effects in flawed sandstone.
56. Fluid injection protocols influence microseismic source mechanisms: Insights from fracture-vein interactions
Core Problem: Injection protocols can change fracture geometry and microseismic source mechanisms in veined rock.
Key Innovation: A hydro-mechanical DEM model with moment-tensor decomposition tests how injection schedule and vein orientation govern fracture propagation and microseismicity.
57. Two-phase hydromechanical modeling using zero-thickness interface elements for fluid flow simulation in faulted hydrocarbon reservoirs
Core Problem: Fluid flow in faulted reservoirs depends on coupled deformation and multiphase dynamics along faults.
Key Innovation: Zero-thickness interface elements represent faults inside a two-phase hydromechanical finite-element model.
58. Creep behaviour under triaxial direct shear tests and coupled thermo-mechanical constitutive modelling of granite host rock for deep geological disposal
Core Problem: Long-term repository stability depends on granite creep under high temperature and pressure.
Key Innovation: A high-temperature, high-pressure triaxial direct-shear apparatus and coupled thermo-mechanical constitutive model quantify creep and long-term strength of Beishan granite.
59. Rotated-envelope kinematics for roof collapse in Hoek–Brown rock masses: tension cutoff, axisymmetry, and elliptical openings
Core Problem: Roof-collapse prediction must account for opening geometry, axisymmetry, and tensile cutoff in nonlinear rock masses.
Key Innovation: A rotated-envelope kinematic limit-analysis framework provides geometry-portable collapse bounds for rectangular, circular, and elliptical openings.
60. Seismic data-driven 4D-LSM for progressive failure of rock
Core Problem: Rock-fracture prediction is limited by unresolved micro-damage heterogeneity.
Key Innovation: A seismic data-driven 4D lattice spring model maps detected seismic events into virtual damage zones that progressively degrade local bonds during simulation.
61. Influence of effective pressure on the hydraulic properties of intact, micro-fractured and macro-fractured Westerly Granite under hydrostatic conditions
Core Problem: Fracture transmissivity at depth depends on pressure, scale, and surface roughness that remain partly constrained.
Key Innovation: Hydrostatic stress cycling to 150 MPa compares intact, thermal micro-fractured, fabricated macro-fractured, and tensile-fractured Westerly granite samples.
62. Impact of fractal dimensions of rainfall time series on mountain flood forecasting
Core Problem: Mountain rainfall is heterogeneous, but fractal time-series properties are rarely used in flood forecasts.
Key Innovation: Rainfall and water-level fractal dimensions from a 339 km2 watershed are linked to minute-scale forecasting performance.
63. Characterizing the distribution and flow dynamics of shallow karst fissure-type groundwater in urban environments using multi-geophysical time-lapse monitoring combined with saline tracer tests
Core Problem: Urban karst groundwater flow is hard to map because fissures are heterogeneous, infilled, and difficult to access.
Key Innovation: Time-lapse borehole-surface ERT, saline tracers, and multi-geophysical monitoring characterize groundwater distribution and dynamics for hazard mitigation.
64. Machine learning models unravel contrasting enrichment of uranium and arsenic in groundwater from the Songnen Basin, China
Core Problem: Contrasting uranium and arsenic behaviour complicates groundwater safety assessment.
Key Innovation: Interpretable machine learning and hydrogeochemistry identify different controls on uranium depletion and arsenic enrichment along a groundwater flow path.
65. Hydrobiogeochemical impacts of water sealing on acid mine drainage in karst regions
Core Problem: Karst connectivity makes acid mine drainage control uncertain after water sealing.
Key Innovation: A paired sealed-unsealed site comparison quantifies hydrological, sulfate, iron, and microbial responses to grouting-based water sealing.
66. Out of bounds: the use of model ensembles to explore model structural uncertainty under extreme events
Core Problem: Urban drainage simulations under extremes depend strongly on structural assumptions and boundary conditions.
Key Innovation: Four model configurations are compared to evaluate how model structure changes drainage response during out-of-bounds events.
67. Intensification of extreme climatic events in Central Asia’s arid regions amid climate change
Core Problem: Arid oasis systems face coupled climate extremes, groundwater depletion, and ecological decline.
Key Innovation: A 1 km flash-drought workflow, deep-learning groundwater inversion, structural equation modelling, and lag analysis quantify extreme-event intensification in Xinjiang.
68. Non-stationary drought assessment through integration of rainfall and GRACE-based terrestrial water storage data
Core Problem: Rainfall-only stationary drought indices can miss hydrological depletion under climate variability.
Key Innovation: A nonstationary combined TWS-rainfall index integrates IMD rainfall and GRACE terrestrial water storage for drought assessment.
69. A coupled hydro-mechanical-chemical simulation of osmotic consolidation and swelling in expansive soils
Core Problem: Expansive bentonite barriers require coupled hydraulic, mechanical, and chemical simulation.
Key Innovation: A u-p formulation is coupled with advection-dispersion and electrochemical clay constitutive laws to model osmotic consolidation and swelling.
70. Laboratory investigation on the static and dynamic mechanical behaviors of rubber-sand mixture
Core Problem: Rubber-sand mixtures need mechanical characterization before safe use in dynamic geotechnical applications.
Key Innovation: Drained monotonic and cyclic triaxial tests quantify rubber-content and confining-pressure effects on stiffness, dilation, and accumulated strain.
71. Mechanical Properties and Energy Evolution of Frozen Sodium Carbonate Clay under Uniaxial Compression Conditions
Core Problem: Frozen saline soils have salt- and temperature-dependent failure modes relevant to cold-region transport infrastructure.
Key Innovation: Uniaxial compression, DIC, and energy analysis identify shear-band patterns and damage evolution in frozen sodium carbonate clay.
72. Permeability coefficient prediction of coarse-grained soils through a Bayesian-optimized physics-constrained neural network
Core Problem: Permeability prediction needs nonlinear accuracy without losing physical interpretability.
Key Innovation: A Bayesian-optimized physics-constrained neural network embeds the Kozeny-Carman relation and residual learning to estimate permeability and uncertainty.
73. An Innovative Multi-Model Ensemble Strategy for Subgrade Compaction Quality Assessment Incorporating Moisture Content Effects and Underlying Soil Layer Properties
Core Problem: Specification-based compaction control often neglects moisture content and underlying layer effects.
Key Innovation: A stacking ensemble model uses compaction control values, moisture effects, and layer properties to improve subgrade compaction quality assessment.
74. Experimental investigation of heavy haul railway embankment deformation under traffic-induced bi-directional principal stress rotation
Core Problem: Heavy-haul rail foundations undergo complex cyclic stress paths that standard tests can underestimate.
Key Innovation: Finite-element stress-path extraction and bidirectional dynamic simple shear tests quantify principal-stress-rotation effects on embankment deformation.
75. Geospatial foundation-model embeddings improve population estimation unevenly across space and scale
Core Problem: Disaster exposure mapping needs subnational population estimates where census data are sparse or outdated.
Key Innovation: Population Dynamics Foundation Model embeddings are benchmarked against harmonized covariates in Brazil, Nigeria, and the United States, showing scale- and geography-dependent gains.
76. Synergistic Perception and Generative Recomposition: A Multi-Agent Orchestration for Expert-Level Building Inspection
Core Problem: Building inspection is difficult because cracks, spalling, low contrast, and limited pixel labels degrade segmentation.
Key Innovation: FacadeFixer orchestrates detection and segmentation agents with generative recomposition to improve multi-defect facade inspection under scarce annotations.
77. ADR-YOLO: An Adaptive Dual-Refinement Network for Object Detection in Aerial Images
Core Problem: Aerial imagery contains scale variation and complex backgrounds that weaken rapid object detection.
Key Innovation: ADR-YOLO adds stage-specific denoising and selective attention with scale-aware fusion for robust object detection in aerial images.