Initiated by Dr. Xin Wei, University of Michigan
Ongoing development by the community

TerraMosaic Daily Digest: May 20, 2026

Daily Summary

The May 20 papers are organized around observability: open datasets and coupled monitoring frameworks make slow or hidden geohazard states explicit. A northern-high-latitude permafrost-thaw ensemble and a 25-year quarterly Loess Plateau land-change record provide benchmark evidence for thaw vulnerability and erosion recovery, while the landslide papers add a hydro-mechanical storage metric for karst slope failure, a nonstationary rainfall-dependence model for future shallow-landslide probability, and a super-resolution multi-source detector for landslide mapping. Cold-region slope, glacier-crevasse, dam, and mining-deformation papers extend that theme from surface mapping to coupled thermo-hydro-mechanical response, acoustic precursors, seismic deformation, and three-dimensional ground movement.

The flood and infrastructure papers emphasize physical thresholds rather than black-box risk scores. Tropical-cyclone flood susceptibility is built from hydrodynamic simulations and explainable ensembles; storm-surge probabilities are constrained with thousands of synthetic cyclone simulations; the Tanzania corridor paper translates multi-sensor Earth-observation signals into engineering warning thresholds; darcyInterTransportFoam opens a fully coupled 3D surface-water and groundwater solver. Rockfill-dam, tunnel-face, microseismic, roadway-excavation, rough-fracture, and slope pile-group studies follow failure through measurable stress, discontinuity, gas, flow, and displacement states. Remote-sensing and foundation-model papers are most relevant when they strengthen disaster detection, deformation inversion, road-damage grounding, SAR generation, water-network mapping, and change detection.

Key Trends

The common thread is direct state reconstruction: permafrost thaw index, quarterly erosion-recovery maps, karst slope storage, nonstationary rainfall dependence, frozen-slope water-ice transition, crevasse acoustic precursors, cyclone flood susceptibility, corridor warning thresholds, and underground discontinuity or microseismic uncertainty.

  • Datasets increasingly encode process states, not only context: permafrost thaw, Loess Plateau erosion recovery, OpenSeisML, root-zone soil moisture, and composite water-network products provide reusable variables for thaw, erosion, subsurface structure, hydrologic forcing, and flood exposure.
  • Slope and cold-region studies resolve coupled water, ice, and mechanics: karst landslide storage, nonstationary shallow-landslide rainfall dependence, multi-source landslide detection, alpine turf-slope THM behaviour, glacier-crevasse acoustic staging, avalanche beacon geometry, and rockfill-dam seismic response treat instability as a coupled state rather than a static map label.
  • Flood and corridor studies convert forcing into thresholds: tropical-cyclone flood modelling, Bay of Bengal storm-surge probabilities, darcyInterTransportFoam, Tanzania corridor thresholds, GLOF warning experiences, isotope precipitation anomalies, IMERG vertical-error diagnosis, and flash-drought vegetation monitoring connect forcing to interpretable susceptibility, uncertainty, or response metrics.
  • Subsurface engineering papers quantify hidden degradation: mine support point clouds, tunnel-face discontinuities, microseismic uncertainty, THM rock damage, dust-methane airflow, rough-fracture flow, and slope pile groups turn inaccessible rock and infrastructure states into measured variables.
  • Remote-sensing and AI methods matter when they improve observable variables: multimodal disaster reasoning, ChangeVLM, GNSS-InSAR deformation inversion, WildRoadBench, GeoDiff-SAR II, POCA-Lite, DAAINet, DINO-MS, and GC2F-Net strengthen detection, deformation, change, retrieval, and representation tasks for hazard assessment.

Selected Papers

This issue contains 55 selected papers from 1,972 papers analyzed. The leading papers move from hazard labels to measurable states: high-latitude permafrost thaw, quarterly Loess Plateau erosion recovery, karst landslide storage, nonstationary shallow-landslide rainfall dependence, multi-source landslide detection, frozen-slope water-ice mechanics, tropical-cyclone flood susceptibility, storm-surge probability, corridor warning thresholds, and coupled surface-water and groundwater exchange. The broader set extends this state-based view to glacier-crevasse acoustic precursors, GLOF warning and evacuation, active-fault paleoseismic history, mining deformation, rockfill-dam seismic response, avalanche localization, tunnel-face discontinuities, deep-tunnel microseismic uncertainty, rough-fracture flow, infrastructure damage detection, and remote-sensing models that improve deformation, change, water-network, and SAR observables.

1. An Ensemble Dataset of Permafrost Thaw Conditions for Northern High Latitudes from Open Satellite Data

Source: Earth System Science Data Type: Open Permafrost Thaw Dataset Geohazard Type: Permafrost thaw, high-latitude ground instability, and cryosphere vulnerability Relevance: 8/10

Core Problem: Northern permafrost thaw is spatially heterogeneous and difficult to compare across products, limiting hazard and infrastructure assessments in thaw-sensitive terrain.

Key Innovation: An ensemble permafrost-thaw dataset built from 16 open satellite and environmental products uses XGBoost to map percent permafrost, mean annual ground temperature, and a Permafrost Thaw Index for northern high latitudes.

2. 25-year, quarterly land change maps of China's Loess Plateau reveal long-term and substantial water-induced soil erosion mitigation

Source: Earth System Science Data Type: Long-Term Erosion-Mitigation Dataset Geohazard Type: Water-induced soil erosion and land-change recovery on the Loess Plateau Relevance: 8/10

Core Problem: Erosion mitigation is often inferred from sparse snapshots rather than temporally dense land-change evidence over restoration timescales.

Key Innovation: A 25-year quarterly 10-30 m land-change record from Landsat and Sentinel-2 resolves one hundred seasonal steps of cropland, forest, grassland, and bare-soil dynamics across China's Loess Plateau.

3. Improving landslide susceptibility mapping with skeleton elastic storage coefficient and spatial learning models

Source: Geomatics, Natural Hazards and Risk Type: Karst Landslide Susceptibility Mapping Geohazard Type: Karst slope failure and hydro-mechanical landslide controls Relevance: 8/10

Core Problem: Karst landslide susceptibility models often miss subsurface storage and deformation processes that regulate slope instability under complex groundwater conditions.

Key Innovation: The study introduces skeleton elastic storage coefficient as a hydro-mechanical conditioning factor and combines it with spatial learning models for landslide susceptibility mapping in Bijie, China.

4. Probabilistic assessment of shallow landslides in a changing climate: insight from the rainfall dependence structure

Source: Landslides Type: Climate-Dependent Shallow-Landslide Probability Geohazard Type: Shallow landslide initiation under nonstationary rainfall dependence Relevance: 8/10

Core Problem: Future shallow-landslide risk depends not only on rainfall intensity or duration separately, but on how their dependence structure shifts under climate change.

Key Innovation: A nonparametric copula and Monte Carlo framework models climate-driven dependence between event rainfall and rainfall duration, then evaluates exceedance over a physically based shallow-landslide threshold.

5. Learning to detect landslides from multi-source remote sensing data with super-resolution reconstruction and deep feature fusion

Source: Natural Hazards Type: Multi-Source Landslide Detection Geohazard Type: Deep-learning landslide detection from optical and auxiliary remote-sensing data Relevance: 8/10

Core Problem: Remote-sensing landslide detection is limited by insufficient spatial detail and weak integration of optical imagery with auxiliary geospatial variables.

Key Innovation: A two-stage framework combines meta-learning-enhanced super-resolution reconstruction with multi-source deep feature fusion, improving landslide detection in benchmark and real-world case areas.

6. Thermo-hydro-mechanical coupling modelling and interface effects of transplanted turf slope soil in an alpine meadow

Source: Cold Regions Science and Technology Type: Cold-Region Slope Coupled Modelling Geohazard Type: Freeze-thaw deformation and alpine meadow slope stability Relevance: 8/10

Core Problem: Transplanted turf slopes in alpine meadows experience coupled water, heat, and mechanical changes that can weaken slope protection during freeze-thaw cycles.

Key Innovation: A thermo-hydro-mechanical model with water-ice phase change, frost heave, thaw settlement, and interface effects is developed for transplanted turf slope soil in cold-region engineering.

7. Integrating explainable ensemble machine learning with hydrodynamic modeling to assess tropical cyclone-induced flood susceptibility

Source: Geoscience Frontiers Type: Hydrodynamic Flood Susceptibility Geohazard Type: Tropical cyclone-induced flood hazard in coastal urban regions Relevance: 8/10

Core Problem: Tropical cyclone flood susceptibility requires both physically resolved inundation and interpretable controls, not only statistical flood labels.

Key Innovation: Hydrodynamic simulations for historical tropical cyclones are fused with explainable ensemble machine learning to produce 30 m flood-susceptibility maps and driver attribution for the Pearl River Delta.

8. Early-Warning Engineering for Corridor Vulnerability: A Remote Sensing Threshold Model for Disaster Risk Reduction in Tanzania

Source: International Journal of Disaster Risk Reduction Type: Remote-Sensing Corridor Early Warning Geohazard Type: Climate-sensitive transport-corridor degradation and disaster-risk reduction Relevance: 8/10

Core Problem: Transport corridors exposed to rainfall, vegetation stress, and terrain degradation need warning thresholds that are physically interpretable for engineering action.

Key Innovation: Multi-sensor remote-sensing indicators are translated into corridor-vulnerability thresholds for Tanzania, linking vegetation dynamics and terrain response to operational early-warning logic.

9. darcyInterTransportFoam v1.0: an open-source, fully-coupled 3D solver for simulating surface water – saturated groundwater processes and exchanges

Source: Geoscientific Model Development Type: Open Surface-Water Groundwater Solver Geohazard Type: Floodplain, hyporheic, and saturated groundwater exchange processes Relevance: 8/10

Core Problem: Surface-water and saturated-groundwater interactions are often simplified, limiting simulation of exchange, turbulence, and sediment-water interface processes during hydrologic extremes.

Key Innovation: darcyInterTransportFoam provides an open-source fully coupled 3D OpenFOAM solver for surface-water and saturated-groundwater processes and their exchanges.

10. Multi-stage Acoustic characteristics in glacier crevasse development under hydraulic action

Source: Cold Regions Science and Technology Type: Glacier Crevasse Acoustic Monitoring Geohazard Type: Crevasse growth, ice-dam failure, and glacial outburst hazard precursors Relevance: 8/10

Core Problem: Hydraulically driven glacier crevasse development can precede ice-dam failure, but its progressive damage stages are difficult to monitor directly.

Key Innovation: Acoustic-emission observations identify multi-stage crack development and a sharp signal increase before penetration, supporting precursor monitoring for glacier-dam instability.

11. Numerical simulation of a rockfill dam subjected to consecutive earthquakes: total and effective stress analysis versus recorded field data

Source: Soil Dynamics and Earthquake Engineering Type: Rockfill Dam Seismic Response Geohazard Type: Consecutive-earthquake deformation and dam safety Relevance: 8/10

Core Problem: Rockfill dams can accumulate deformation during earthquake sequences, but total-stress and effective-stress simulations need validation against field records.

Key Innovation: Numerical simulations of El Infiernillo Dam compare constitutive approaches against recorded consecutive earthquakes, linking model choice to crest settlement and dynamic response.

12. Three-Dimensional Deformation Field Inversion Based on Fused Monitoring Data of GNSS and InSAR: A Case Study of Jinchuan No. 2 Mining Area

Source: Remote Sensing Type: GNSS-InSAR Deformation Inversion Geohazard Type: Mining subsidence, fissures, and three-dimensional ground deformation Relevance: 8/10

Core Problem: Mining areas need three-dimensional deformation fields because single-sensor line-of-sight monitoring cannot resolve the full displacement vector.

Key Innovation: GNSS and InSAR observations are fused to invert 3D deformation in the Jinchuan No. 2 mining area, improving interpretation of subsidence, fissure formation, and engineering impacts.

13. Probabilistic assessment of tropical cyclone-induced storm surge hazards along the Bay of Bengal coast

Source: Natural Hazards Type: Probabilistic Storm-Surge Hazard Assessment Geohazard Type: Tropical cyclone storm surge and coastal flood hazard Relevance: 7/10

Core Problem: Bay of Bengal storm-surge risk is difficult to quantify because rare cyclone extremes interact nonlinearly with tides, waves, bathymetry, and coastal geometry.

Key Innovation: A fully coupled ADCIRC+SWAN system forced by 2,570 synthetic tropical cyclones estimates present-day storm-surge extremes and bivariate hazard probabilities along the Bay of Bengal coast.

14. Public warning and evacuation experiences during recent GLOF events (2019 and 2023) and recommendations for future preparedness: Insights from Lunana, Bhutan

Source: International Journal of Disaster Risk Reduction Type: GLOF Warning and Evacuation Analysis Geohazard Type: Glacial lake outburst flood early warning and community evacuation Relevance: 7/10

Core Problem: Glacial lake outburst flood early-warning systems depend on warning dissemination and evacuation behaviour, but event-based community evidence remains limited.

Key Innovation: Surveys, interviews, and focus groups from 2019 and 2023 Lunana GLOF events identify environmental cues, informal warning channels, evacuation constraints, and preparedness improvements.

15. Unraveling Holocene surface-faulting history of the Çay segment, Sultandağı Fault (SSW Türkiye): linking paleoseismic evidence to segment-scale seismic behavior

Source: Natural Hazards Type: Paleoseismic Surface-Faulting History Geohazard Type: Active fault rupture history and segment-scale seismic hazard Relevance: 7/10

Core Problem: Segment-scale seismic behaviour on active normal faults is poorly constrained where paleoseismic evidence is sparse.

Key Innovation: Trenching, stratigraphic analysis, radiocarbon dating, and Bayesian age modelling document five Holocene surface-rupturing earthquakes on the Çay segment of the Sultandağı Fault.

16. Cooperative avalanche beacon localization using machine-learned geometric constraints and sequential regression-plane intersection

Source: Cold Regions Science and Technology Type: Avalanche Beacon Localization Geohazard Type: Avalanche search-and-rescue localization under uncertain geometry Relevance: 7/10

Core Problem: Avalanche rescues require rapid and accurate beacon localization under noisy signal geometry and limited rescuer coordination.

Key Innovation: A cooperative three-rescuer framework uses machine-learned geometric constraints and sequential regression-plane intersection to improve avalanche beacon localization.

17. Towards multimodal geospatial reasoning: a foundation model approach for disaster detection from social media, news, and weather data

Source: Natural Hazards Type: Multimodal Disaster Reasoning Geohazard Type: Disaster detection from social media, news, weather, and satellite reference data Relevance: 7/10

Core Problem: Disaster monitoring across social, meteorological, and Earth-observation streams lacks a shared spatial reasoning framework.

Key Innovation: A grid-based multimodal foundation-model approach aligns social media, news, weather, and satellite reference signals for disaster detection and geospatial reasoning.

18. A multi-chain surrogate-assisted hybrid optimization framework for joint identification of groundwater contaminant sources and hydrogeological parameters

Source: Hydrology and Earth System Sciences Type: Groundwater Source Inversion Geohazard Type: Groundwater contamination source and hydrogeological parameter uncertainty Relevance: 7/10

Core Problem: Groundwater contamination source identification is underdetermined when source locations and aquifer parameters are both unknown.

Key Innovation: A multi-chain surrogate-assisted hybrid optimization framework combines simulation, Tabu Search, and surrogate modelling to jointly identify contaminant sources and hydrogeological parameters.

19. Integrating GRACE and InSAR Data to Assess Land Surface Displacement From Managed Aquifer Recharge

Source: Water Resources Research Type: GRACE-InSAR Displacement Fusion Geohazard Type: Aquifer storage change and managed-recharge surface deformation Relevance: 7/10

Core Problem: Managed aquifer recharge can deform land surfaces, but GRACE storage signals and InSAR displacement observations operate at incompatible scales.

Key Innovation: Machine-learning integration of GRACE and InSAR links groundwater storage change to land-surface displacement during managed aquifer recharge.

20. Prediction of Hydroclimatic Anomalies Using a New Isotope Precipitation Index

Source: Geophysical Research Letters Type: Isotope Hydroclimate Anomaly Index Geohazard Type: Drought, moisture recycling, and hydroclimatic anomaly prediction Relevance: 7/10

Core Problem: Hydroclimatic anomalies depend on atmospheric moisture pathways that conventional precipitation indices do not directly observe.

Key Innovation: A stable-isotope precipitation index captures moisture-source and recycling information to improve prediction of hydroclimatic anomalies.

21. Drivers of Basal Melt Variability for Pine Island Glacier Ice Shelf: Ocean Forcing Versus Geometric Feedback

Source: Geophysical Research Letters Type: Ice-Shelf Melt Variability Analysis Geohazard Type: Ice-shelf basal melt and marine cryosphere instability Relevance: 7/10

Core Problem: Basal melt of the Pine Island Glacier Ice Shelf reflects both ocean forcing and geometric feedback, complicating attribution of ice-shelf instability.

Key Innovation: The study separates ocean-forcing and geometry-driven controls on basal melt variability, clarifying feedbacks relevant to ice-shelf retreat.

22. Cloud‐Rain Vertical Inconsistency Increases IMERG Precipitation Uncertainty

Source: Geophysical Research Letters Type: Satellite Precipitation Error Diagnosis Geohazard Type: Precipitation uncertainty for flood and landslide forcing Relevance: 7/10

Core Problem: IMERG precipitation uncertainty can increase when cloud and rain vertical structures are inconsistent, degrading hazard-forcing estimates.

Key Innovation: FY-3G precipitation radar and FY-4B cloud observations are used to diagnose how cloud-rain vertical inconsistency amplifies satellite rainfall errors.

23. WildRoadBench: A Wild Aerial Road-Damage Grounding Benchmark for Vision-Language Models and Autonomous Agents

Source: arXiv Type: Aerial Road-Damage Benchmark Geohazard Type: Road-infrastructure damage detection after hazards Relevance: 7/10

Core Problem: Vision-language models and autonomous agents lack realistic aerial benchmarks for grounding road damage in complex post-disaster scenes.

Key Innovation: WildRoadBench provides a wild UAV road-damage grounding benchmark with tracks for vision-language models and autonomous agents.

24. WaveGraphNet: Physics-Consistent Guided-Wave Damage Localization through Coupled Inverse-Forward Graph Learning

Source: arXiv Type: Guided-Wave Damage Localization Geohazard Type: Structural-health monitoring for infrastructure damage Relevance: 7/10

Core Problem: Guided-wave damage localization must respect wave physics while solving inverse localization from sparse measurements.

Key Innovation: WaveGraphNet couples inverse and forward graph learning with physics-consistent guided-wave constraints for structural damage localization.

25. OpenSeisML: Open Large-Scale Real Seismic and well-log Dataset for Generative AI

Source: arXiv Type: Open Seismic Generative-AI Dataset Geohazard Type: Subsurface seismic and well-log data for geologic modelling Relevance: 7/10

Core Problem: Generative AI for subsurface interpretation is limited by the scarcity of large open seismic and well-log datasets.

Key Innovation: OpenSeisML releases a large-scale real seismic and well-log dataset to support generative modelling, stratigraphic interpretation, and geologic data synthesis.

26. Towards Integrated Rock Support Visualisation in 3D Point Cloud of Underground Mines

Source: arXiv Type: Underground Mine Support Point Clouds Geohazard Type: Rock support visualization and underground excavation safety Relevance: 7/10

Core Problem: Underground support assessment requires joint interpretation of discontinuities, rock bolts, and excavation geometry in three dimensions.

Key Innovation: Automated structure mapping and rock-bolt identification are integrated in 3D point clouds to visualize rock-support geometry and discontinuity relationships in underground mines.

27. GeoDiff-SAR II: 3D-Driven Foundation Diffusion Models for SAR Generation via Decoupled Control

Source: arXiv Type: SAR Foundation Diffusion Model Geohazard Type: Synthetic SAR generation for Earth-observation hazard analysis Relevance: 7/10

Core Problem: SAR-based hazard mapping is constrained by limited labelled scenes and hard-to-control image geometry.

Key Innovation: GeoDiff-SAR II uses 3D-driven decoupled control in a foundation diffusion model to generate SAR imagery with improved structural and geometric consistency.

28. Neural Negative Binomial Regression for Weekly Seismicity Forecasting: Per-Cell Dispersion Estimation and Tail Risk Assessment

Source: arXiv Type: Weekly Seismicity Forecasting Geohazard Type: Earthquake-rate forecasting and seismic tail-risk estimation Relevance: 7/10

Core Problem: Seismicity forecasts must capture overdispersed event counts and rare tail behaviour across spatial cells.

Key Innovation: A neural negative-binomial regression estimates cell-specific dispersion for weekly seismicity forecasting and tail-risk assessment in Central Asia.

29. ChangeVLM: A Language-Guided Semantic Alignment Framework for Binary Remote Sensing Change Detection

Source: Remote Sensing Type: Language-Guided Change Detection Geohazard Type: Remote-sensing change detection for disaster and land-surface monitoring Relevance: 7/10

Core Problem: Binary remote-sensing change detection often lacks semantic alignment, which limits interpretability and transfer across scenes.

Key Innovation: ChangeVLM introduces language-guided semantic alignment for prompt-free, interpretable bi-temporal change detection.

30. Evaluating Multi-Source Soil Moisture Products for Root-Zone Soil Moisture Representation in Yunnan, China

Source: Remote Sensing Type: Root-Zone Soil-Moisture Evaluation Geohazard Type: Soil-moisture forcing for drought, flood, and slope-hydrology assessment Relevance: 7/10

Core Problem: Root-zone soil moisture products differ substantially, creating uncertainty for hydroclimatic and slope-hydrology applications.

Key Innovation: Multi-source soil-moisture products are evaluated against root-zone representation in Yunnan, identifying product behaviour relevant to hydrologic hazard analysis.

31. Real-Time Early Warning of Incipient Fire in Multiple Urban Scenarios: A Deep Learning-Based Monitoring Method

Source: Remote Sensing Type: Urban Fire Early Warning Geohazard Type: Incipient fire detection in multiple urban remote-sensing scenarios Relevance: 7/10

Core Problem: Urban fire warning needs fast detection of small incipient fire signatures across heterogeneous monitoring scenes.

Key Innovation: A deep-learning monitoring method is designed for real-time incipient fire warning across multiple urban scenarios.

32. Airflow organization for coordinated control of multiple airborne hazards: A case of dust and methane in mechanized roadway excavation

Source: Tunnelling and Underground Space Technology Type: Roadway Airborne-Hazard Control Geohazard Type: Dust and methane hazards in mechanized tunnel or mine-roadway excavation Relevance: 7/10

Core Problem: Dust and methane can require competing airflow controls in mechanized roadway excavation, creating coupled occupational and explosion hazards.

Key Innovation: Simulation, experiments, and field tests evaluate airflow organization for coordinated control of dust and methane using a radial air-distribution strategy.

33. Identification of well logging data lithology based on physics-informed graph attention network

Source: International Journal of Rock Mechanics and Mining Sciences Type: Physics-Informed Lithology GNN Geohazard Type: Well-log lithology interpretation for underground and reservoir engineering Relevance: 7/10

Core Problem: Lithology identification from well logs can violate physical relationships when treated as a purely statistical classification task.

Key Innovation: A physics-informed graph attention network embeds rock-physics constraints into well-log lithology classification.

34. Experimental investigation of inertial flow in 3D printed intersected rough fractures: Effects of JRC and intersection angle

Source: International Journal of Rock Mechanics and Mining Sciences Type: Rough-Fracture Inertial Flow Geohazard Type: Hydraulic flow through intersecting rock fractures Relevance: 7/10

Core Problem: Flow in intersecting rough fractures becomes inertial and geometry-dependent, affecting seepage, grouting, and fractured-rock transport predictions.

Key Innovation: 3D-printed fracture experiments quantify the effects of joint roughness coefficient and intersection angle on inertial flow behaviour.

35. Pore-fracture evolution and fractal characteristics of coal under triaxial compression: Insights from in-situ nuclear magnetic resonance

Source: International Journal of Rock Mechanics and Mining Sciences Type: Coal Pore-Fracture Evolution Geohazard Type: Coal deformation, fracture growth, and mining geohazard precursors Relevance: 7/10

Core Problem: Coal failure under triaxial compression involves coupled pore and fracture evolution that is difficult to observe continuously.

Key Innovation: In-situ nuclear magnetic resonance tracks pore-fracture evolution and fractal characteristics during triaxial coal compression.

36. Theoretical feasibility assessment of hydraulic fracturing with microwave assistance: Insights from a coupled electromagnetic-thermo-hydro-mechanical (ETHM) phase-field simulation

Source: International Journal of Rock Mechanics and Mining Sciences Type: Microwave-Assisted Fracturing Model Geohazard Type: Coupled hydraulic fracturing and thermal rock damage Relevance: 7/10

Core Problem: Hydraulic fracturing in hard rock may be enhanced by microwave-induced weakening, but the coupled feasibility remains unclear.

Key Innovation: A coupled electromagnetic-thermo-hydro-mechanical phase-field model tests microwave-assisted hydraulic fracturing mechanisms.

37. Enhancing understanding of vegetation responses to flash droughts: the need for high-temporal-resolution monitoring data

Source: Journal of Hydrology Type: Flash-Drought Vegetation Monitoring Geohazard Type: Rapid drought impacts on vegetation and ecological vulnerability Relevance: 7/10

Core Problem: Flash-drought impacts can be missed by coarse temporal vegetation monitoring.

Key Innovation: The paper argues for high-temporal-resolution monitoring to resolve vegetation response timing and mechanisms during flash droughts.

38. CWN-Net: a precise segmentation model for composite water networks using multi-source and multi-temporal remote sensing imagery

Source: Journal of Hydrology Type: Composite Water-Network Segmentation Geohazard Type: River, lake, and water-network mapping for flood and hydrologic monitoring Relevance: 7/10

Core Problem: Composite water networks are difficult to segment where channels, wetlands, shadows, and seasonal water interact.

Key Innovation: CWN-Net combines multi-source and multi-temporal remote-sensing imagery with a segmentation architecture tailored to composite water networks.

39. An equal-stiffness-based method for analysis and design of pile groups on slopes

Source: Computers and Geotechnics Type: Slope Pile-Group Design Geohazard Type: Pile foundations on slopes and valley-crossing bridge infrastructure Relevance: 7/10

Core Problem: Pile groups installed on slopes have nonuniform stiffness and loading conditions that conventional level-ground design does not capture.

Key Innovation: An equal-stiffness method and slope-adapted interaction model support analysis and design of pile groups on sloping ground.

40. Automatic identification and quantitative integrity evaluation of rockmass discontinuities on tunnel face based on 3D laser scanning point cloud data

Source: Transportation Geotechnics Type: Tunnel-Face Discontinuity Evaluation Geohazard Type: Rockmass integrity and tunnel-face instability Relevance: 7/10

Core Problem: Tunnel-face stability depends on discontinuity networks that are laborious to map and quantify manually.

Key Innovation: 3D laser-scanning point clouds are used to automatically identify discontinuities and evaluate rockmass integrity on tunnel faces.

41. Energy damage characteristics of rocks under coupled thermo-hydro-mechanical conditions: An insight from laboratory to engineering perspectives

Source: Journal of Rock Mechanics and Geotechnical Engineering Type: Thermo-Hydro-Mechanical Rock Damage Geohazard Type: Rock damage in deep tunnels under coupled heat, water, and stress Relevance: 7/10

Core Problem: Deep rock engineering exposes rocks to coupled thermal, hydraulic, and mechanical conditions that alter energy dissipation and failure.

Key Innovation: Laboratory-to-engineering analysis quantifies energy damage characteristics of rocks under coupled thermo-hydro-mechanical loading.

42. Analysis of microseismicity localization uncertainty in deep tunnels

Source: Journal of Rock Mechanics and Geotechnical Engineering Type: Deep-Tunnel Microseismic Uncertainty Geohazard Type: Microseismic localization and tunnel safety monitoring Relevance: 7/10

Core Problem: Microseismic monitoring in deep tunnels is limited by uncertainty in source location and wave-velocity structure.

Key Innovation: NGBoost and Bayesian modelling are used to quantify microseismic localization uncertainty and improve interpretation of deep-tunnel events.

43. Leveraging the TabPFN Algorithm for High‐Resolution Mapping of Groundwater Bicarbonate and Its Scaling Risk Across China

Source: Water Resources Research Type: Groundwater Scaling-Risk Mapping Geohazard Type: Groundwater bicarbonate and scaling risk across China Relevance: 6/10

Core Problem: Groundwater scaling risk requires high-resolution mapping of bicarbonate concentrations over large heterogeneous regions.

Key Innovation: The TabPFN algorithm is applied to map groundwater bicarbonate and scaling risk across China at high spatial resolution.

44. Gene‐Informed Modeling of Denitrification Process in Groundwater Through Dynamic Flux Balance Analysis and Deep Learning

Source: Water Resources Research Type: Gene-Informed Groundwater Denitrification Geohazard Type: Groundwater biogeochemistry and nitrate attenuation Relevance: 6/10

Core Problem: Groundwater denitrification depends on microbial function that is difficult to represent in reactive-transport models.

Key Innovation: Dynamic flux balance analysis is combined with deep learning and gene-informed constraints to model groundwater denitrification processes.

45. Miller-Index-Based Latent Crystallographic Fracture Plane Reasoning with Vision-Language Models

Source: arXiv Type: VLM Fracture-Plane Reasoning Geohazard Type: Crystallographic fracture reasoning with vision-language models Relevance: 6/10

Core Problem: Vision-language models do not inherently encode crystallographic constraints needed for fracture-plane reasoning.

Key Innovation: A Miller-index-based latent reasoning framework tests whether VLMs can infer crystallographic fracture planes from structured visual and symbolic information.

46. Learning Stratigraphically Consistent Relative Geologic Time from 3D Seismic Data via Sinusoidal Mapping

Source: arXiv Type: Stratigraphic-Time Learning Geohazard Type: 3D seismic interpretation and stratigraphic consistency Relevance: 6/10

Core Problem: Relative geologic time inferred from 3D seismic volumes can violate stratigraphic ordering and continuity.

Key Innovation: A sinusoidal mapping framework learns stratigraphically consistent relative geologic time from 3D seismic data.

47. POCA-Lite: A Lightweight Change-Detection Architecture with Geometry-Aware Auxiliary Supervision and Feedback Fusion

Source: Remote Sensing Type: Lightweight Change Detection Geohazard Type: Efficient remote-sensing change detection for land-surface monitoring Relevance: 6/10

Core Problem: Operational change detection needs lightweight models that still preserve geometry and temporal feedback.

Key Innovation: POCA-Lite combines geometry-aware auxiliary supervision and feedback fusion for efficient bi-temporal remote-sensing change detection.

48. DAAINet: Domain Adversarial Anti-Interference Network for Bi-Temporal Change Detection

Source: Remote Sensing Type: Domain-Adversarial Change Detection Geohazard Type: Remote-sensing change detection under domain interference Relevance: 6/10

Core Problem: Bi-temporal change detection can fail when domain shifts and nuisance interference dominate the change signal.

Key Innovation: DAAINet uses domain-adversarial anti-interference learning to improve change detection robustness.

49. GeoHybridGNN: A Hybrid Intelligent Mapping Framework for Porphyry Copper Prospectivity Mapping Integrating Remote Sensing, Geology, and Geochemistry

Source: Remote Sensing Type: Hybrid Geologic Prospectivity GNN Geohazard Type: Mineral prospectivity mapping from remote sensing, geology, and geochemistry Relevance: 6/10

Core Problem: Prospectivity mapping must integrate spatially structured geologic, geochemical, and remote-sensing evidence.

Key Innovation: GeoHybridGNN combines graph learning with multi-source geoscience evidence for porphyry copper prospectivity mapping.

50. An Efficient Remote Sensing Cross-Modal Retrieval Method Based on Hashing Contrastive Learning

Source: Remote Sensing Type: Cross-Modal Remote-Sensing Retrieval Geohazard Type: Efficient retrieval across remote-sensing modalities Relevance: 6/10

Core Problem: Cross-modal Earth-observation retrieval needs compact representations that preserve semantic correspondence between different sensors.

Key Innovation: A hashing contrastive-learning method improves efficient retrieval across remote-sensing modalities.

51. Learning Scale-Consistent Representations via Multi-Scale Local Consistency for Remote Sensing Imagery

Source: Remote Sensing Type: Scale-Consistent Remote-Sensing Representation Geohazard Type: Self-supervised remote-sensing features for multiscale terrain analysis Relevance: 6/10

Core Problem: Remote-sensing representations often change with image scale, reducing transfer across resolutions and scene sizes.

Key Innovation: DINO-MS learns scale-consistent representations through multi-scale local consistency for remote-sensing imagery.

52. GC2F-Net: A Global Category-Center Prior-Guided Spatial-Frequency Collaborative Network for Remote Sensing Semantic Segmentation

Source: Remote Sensing Type: Spatial-Frequency Segmentation Network Geohazard Type: Remote-sensing semantic segmentation for land-surface mapping Relevance: 6/10

Core Problem: Semantic segmentation of remote-sensing imagery must represent both category-level priors and fine spatial-frequency detail.

Key Innovation: GC2F-Net combines global category-center priors with spatial-frequency collaboration for remote-sensing semantic segmentation.

53. Numerical solution of steady-state turbulent flow of newtonian fluid in a two-dimensional rough fracture

Source: Computers and Geotechnics Type: Rough-Fracture Turbulent Flow Geohazard Type: Fluid transport in rough rock fractures Relevance: 6/10

Core Problem: Turbulent flow in rough fractures affects seepage, contaminant transport, and subsurface engineering but is difficult to solve accurately.

Key Innovation: A numerical solution is developed for steady-state turbulent Newtonian flow in two-dimensional rough fractures.

54. Field and laboratory discrepancies in shear wave velocity across different rock types

Source: Journal of Rock Mechanics and Geotechnical Engineering Type: Rock Shear-Wave Velocity Discrepancy Geohazard Type: Geotechnical site characterization and rock dynamic properties Relevance: 6/10

Core Problem: Laboratory and field shear-wave velocities can diverge across rock types, complicating dynamic-property assignment for engineering analysis.

Key Innovation: The study compares field and laboratory shear-wave velocity measurements to diagnose scale, fabric, and testing discrepancies across rock types.

55. Improved phase field simulation of 3D multiple cracks propagation in rock materials via unified shear energy criterion

Source: Journal of Rock Mechanics and Geotechnical Engineering Type: Three-Dimensional Crack Phase Field Geohazard Type: Rock crack propagation and brittle failure modelling Relevance: 6/10

Core Problem: Three-dimensional multiple-crack propagation in rock requires phase-field criteria that better capture shear-driven fracture growth.

Key Innovation: A unified shear-energy criterion improves phase-field simulation of 3D multiple-crack propagation in rock materials.