TerraMosaic Daily Digest: June 14, 2026
Daily Summary
The June 14 issue is dominated by landslide work that treats deformation, triggering, and runout as linked components of one forecasting problem. Studies from the southeastern Tibetan Plateau, the Khodpe-Chainpur road corridor in Nepal, and a vegetation-rich volcanic terrain in eastern China map susceptibility or timing with machine-learning workflows tied to slope stability, rainfall memory, terrain position, and local structural controls. A Landslides paper adds a methodological caution: tree-ring reconstructions may recover landslide reactivation information even from apparently straight trees, expanding the observational archive beyond visibly tilted trunks.
The strongest process papers focus on failure after initiation. Mine overburden dump debris-flow experiments quantify how composition and slope govern runout regimes, while a hydro-mechanical model for compound slow-moving rockslides represents interacting rigid blocks, open fractures, basal hydraulic connections, and transition from pre-failure to post-failure motion. Alpine rockwall-talus analysis links climatic stress, lithology, deglaciation history, and fracture density to rockfall size and talus maturity. Together these papers move the field away from static hazard maps and toward mechanism-aware trajectories.
Remote sensing and AI papers broaden that trajectory into operational measurement. Three Remote Sensing papers fill gaps left by empty RSS abstracts: adaptive-sampling InSAR landslide susceptibility mapping, attention-weighted spatial ensemble susceptibility mapping, and UAV-LiDAR slip-surface reconstruction. A fourth landslide segmentation paper, DCA-UNet, points to boundary-aware feature learning, while flood segmentation, urban flood warning, InSAR atmospheric correction, seismic horizon tracking, DAS coverage, and geospatial foundation-model robustness show the same methodological pressure across hazards: models must be spatially adaptive, uncertainty-aware, and tied to physical observables rather than labels alone.
Key Trends
Five movements define the issue: spatiotemporal landslide prediction, post-failure mechanics, spatially adaptive remote-sensing AI, decision-ready earthquake and flood modeling, and infrastructure resilience.
- Landslide prediction is becoming explicitly spatiotemporal: Rainfall memory, CNN image recognition, slope-instability maps, random forests, adaptive InSAR sampling, and tree-ring evidence are being combined to recover not only where landslides occur but when they reactivate.
- Runout and post-failure mechanics are moving into the foreground: Mine overburden dump flumes, compound rockslide hydro-mechanics, alpine talus evolution, and UAV-LiDAR slip-surface reconstruction all target the geometry and mobility of failed mass after initiation.
- Remote-sensing AI is shifting from generic accuracy to spatial adaptivity: Hierarchical spatial ensembles, deformable-convolution segmentation, adaptive sampling, quasi-3D slip-surface inference, flood SAR explainability, and foundation-model robustness emphasize where model trust changes across terrain.
- Earthquake and flood papers are converging on usable decision variables: Flood-damage prioritization, real-time urban flood states, earthquake arrival picking, geotechnical damage attribution, bridge seismic response, and crisis-management studies translate hazard models into response or design choices.
- Infrastructure resilience is being modeled as a coupled system problem: Cascading infrastructure knowledge graphs, CAT bonds, disaster housing support, tunnel explosion risk, railway segmentation, and urban blast prediction extend geohazard thinking into recovery, finance, and network vulnerability.
Selected Papers
The selected papers emphasize landslide susceptibility, rainfall-triggered prediction, dendrogeomorphic reactivation records, alpine rockfall controls, mine-dump debris-flow runout, compound rockslide mechanics, UAV-LiDAR slip-surface reconstruction, flood and seismic monitoring, and infrastructure resilience. This issue contains 45 selected papers from 2642 papers analyzed.
1. Evaluation of landslide susceptibility in alpine canyon area based on random forest method: a case study of southeast Tibetan Plateau
Core Problem: The southeastern Tibetan Plateau contains dense historical landslides and complex terrain, but regional susceptibility mapping must handle interacting topographic, hydrologic, and infrastructure controls.
Key Innovation: Uses 20,514 historical landslide samples, 14 environmental predictors, and machine-learning comparison to evaluate high-resolution susceptibility in an alpine canyon setting.
2. Spatiotemporal prediction of rainfall-induced landslides using CNN-based image recognition and slope instability identification
Core Problem: Factor-of-safety thresholds from slope-stability models often fail to match the timing of observed rainfall-induced landslides.
Key Innovation: Integrates physically based hydrological and stability simulations with CNN, DBSCAN, and LSTM components to identify slope instability and forecast landslide timing.
3. Can straight tree sampling in landslide areas improve spatio-temporal reconstruction of landslide reactivations?
Core Problem: Dendrogeomorphic reconstructions usually sample visibly tilted trees, potentially discarding trees that recorded smaller or older landslide movement.
Key Innovation: Compares tilted and straight Picea abies trees on a landslide to test whether apparently undisturbed stems preserve usable growth reactions for reactivation histories.
4. Climatic, lithologic and topographic control on alpine rock fracturing and talus evolution
Core Problem: Rockfall rates vary strongly across alpine rockwall-talus systems, but the relative influence of climate history, lithology, and topography remains difficult to separate.
Key Innovation: Combines rock-mass measurements, fracture-density surveys, talus morphology, clast properties, glacier-stress estimates, and frost-cracking context to explain rockfall size and talus maturity.
5. An Optimized Heterogeneous Ensemble Learning Algorithm for InSAR Landslide Susceptibility Mapping Based on the Adaptive Sampling Strategy
Core Problem: Landslide susceptibility models can be limited by expensive sampling strategies and weak adaptability across heterogeneous reservoir terrain.
Key Innovation: Combines hotspot-based adaptive sampling, Sentinel-1 InSAR, Monte Carlo frequency-ratio screening, and an SVM-RF-XGBoost stacking ensemble for landslide susceptibility in the Baihetan Reservoir area.
6. Composition and slope controls on runout hazard in mine overburden dump debris flows: flume experiments and dimensionless flow regime analysis
Core Problem: Failures of mine overburden dumps can transform into debris flows with mechanics that differ from natural mixtures because blasted waste is angular and poly-disperse.
Key Innovation: Runs 81 controlled flume experiments varying mass ratio, size ratio, and slope, then uses dimensionless regime analysis to quantify runout and flow controls.
7. Slope Stability assessment of a roadside landslide along Khodpe-Chainpur Road Section (NH-64), Sudurpashchim Province, Nepal
Core Problem: Roadside landslides in Himalayan corridors threaten monsoon-season connectivity, yet many mitigation designs lack integrated geophysical and stability evidence.
Key Innovation: Applies geophysical characterization, slope-stability analysis, and mitigation evaluation to the Baluwatar landslide along Nepal's NH-64 corridor.
8. A hydro-mechanical model for compound slow-moving rockslides with pre-existing open fractures
Core Problem: Compound rockslides can move as interacting blocks connected by open fractures and basal pathways, complicating pre-failure and post-failure prediction.
Key Innovation: Models rigid blocks, vertical and basal fractures, water exchange, and hydro-mechanical coupling in a unified framework for slow-moving rockslide evolution.
9. Displacement-Based Estimation of Quasi-Three-Dimensional Landslide Slip Surfaces Using UAV LiDAR Data
Core Problem: Buried slip surfaces remain a major uncertainty where borehole data are sparse, limiting reliable landslide hazard assessment.
Key Innovation: Uses multi-temporal UAV-LiDAR displacement gradients, longitudinal and transverse profile fitting, and borehole validation to estimate quasi-3D slip surfaces for two Japanese landslides.
10. Attention-Driven Hierarchical Spatial Adaptive Ensemble for Landslide Susceptibility Mapping
Core Problem: Conventional landslide susceptibility ensembles often apply fixed global weights or kernel-constrained local averaging despite spatially varying model reliability.
Key Innovation: Builds a two-stage hierarchical spatial adaptive ensemble that fuses GWR, geographically optimal similarity, and DNN base learners through attention-derived spatial weights.
11. DCA-UNet for Landslide Segmentation with Deformable Convolution and Aggregated Attention
Core Problem: Landslide segmentation in remote-sensing imagery is affected by irregular shapes, scale variation, and weak boundaries.
Key Innovation: Proposes a UNet variant combining deformable convolution and aggregated attention to improve landslide boundary and shape representation.
12. Explainable Flood Segmentation on Sentinel-1 SAR Imagery: A Comparative Study of CNN and Transformer Architectures
Core Problem: SAR-based flood mapping must distinguish floodwater from permanent water under cloud-independent but noisy observation conditions.
Key Innovation: Compares CNN and Transformer architectures for Sentinel-1 flood segmentation and emphasizes explainability for emergency-response use.
13. Beyond hazard: leveraging flood damage mapping in the prioritization of flood-prone areas
Core Problem: Urban flood planning often emphasizes hazard depth or extent while underusing spatially explicit damage estimates for prioritization.
Key Innovation: Shows how flood-damage mapping can rank flood-prone areas for decision support rather than serving only as a post-hoc loss estimate.
14. AI-assisted framework using physically informed rainfall-drainage features for real-time urban flood risk forecasting
Core Problem: Urban flood warnings need actionable risk states rather than raw sensor forecasts or unstructured rainfall predictions.
Key Innovation: Derives physically guided rainfall and drainage features and uses AI classification to forecast low-, medium-, and high-risk urban flood states in real time.
15. Contrastive Learning for Seismic Horizon Tracking with Domain-Specific Priors
Core Problem: Unsupervised 3D seismic horizon tracking fails near faults when signal-based trace alignment and texture-based deep models are used separately.
Key Innovation: Uses signal-derived local horizon correspondences as priors for self-supervised contrastive training of a texture-based seismic horizon tracker.
16. WaveDINO: Learning-Based Atmospheric Correction of Unwrapped InSAR Interferograms Validated by GNSS: Results at Laguna del Maule and Campi Flegrei Volcanoes
Core Problem: InSAR deformation signals can be distorted by atmospheric phase delays, seasonal effects, and decorrelation, limiting volcanic and ground-deformation interpretation.
Key Innovation: Introduces a learning-based atmospheric-correction approach validated against GNSS observations at Laguna del Maule and Campi Flegrei.
17. Learning Earthquake Wave Arrival Time Picking from Labels with Inaccuracies
Core Problem: Earthquake phase-picking models are trained on labels that can contain systematic timing inaccuracies, which limits warning and monitoring reliability.
Key Innovation: Studies learning strategies for arrival-time picking under inaccurate labels, with direct relevance to automated seismic monitoring pipelines.
18. Domain-Guided Prompting of the Segment Anything Model for Seismic Interpretation: The Role of Attributes, Visualization, and Hybrid Prompts
Core Problem: General segmentation foundation models need domain guidance before they can support geologic target delineation in seismic volumes.
Key Innovation: Tests how seismic attributes, visualization choices, and hybrid prompts shape Segment Anything Model outputs for seismic interpretation.
19. Detection of the Earth Tides by Diamagnetic Levitation
Core Problem: Small gravity variations from mass redistribution are difficult to detect with compact instrumentation, limiting dense geophysical monitoring.
Key Innovation: Demonstrates diamagnetic levitation as a sensitive platform for detecting Earth tides, with potential transfer to gravity-based mass-change monitoring.
20. A local terrain smoothing approach for stabilizing microscale and high-resolution mesoscale simulations: a case study using FastEddy (v3.0) and WRF (v4.6.0)
Core Problem: High-resolution terrain-following simulations can become numerically unstable over steep slopes, forcing excessive terrain smoothing.
Key Innovation: Introduces localized terrain smoothing to preserve resolution while stabilizing FastEddy and WRF simulations in steep-slope domains.
21. Influence of fabric anisotropy on stability and failure mechanism of slopes: insights from an anisotropic Mohr-Coulomb model
Core Problem: Slope stability assessments often ignore fabric anisotropy even though anisotropy can change failure geometry and mobilized shear strength.
Key Innovation: Uses an anisotropic Mohr-Coulomb model to evaluate how fabric direction and strength anisotropy alter slope stability and failure mechanisms.
22. Analytical framework for lateral bearing capacity of two-pile near clay slope: Group and slope effects
Core Problem: Pile foundations near clay slopes must account for coupled group effects and slope boundary effects on lateral capacity.
Key Innovation: Develops an analytical framework for two-pile lateral bearing capacity near clay slopes, separating group interaction from slope influence.
23. Effect of a shallow subway tunnel and adjacent foundation on the seismic response of ground surface
Core Problem: Urban seismic response can be modified by shallow tunnels and nearby foundations, but their combined effect on ground motion is rarely isolated.
Key Innovation: Evaluates how a shallow subway tunnel and adjacent foundation alter ground-surface seismic response for underground-infrastructure risk assessment.
24. Role of soil type in geotechnical damage observed in Golbasi during the 2023 Turkiye earthquake sequence
Core Problem: The 2023 Turkiye earthquake sequence produced spatially variable geotechnical damage whose relationship to soil type must be clarified for microzonation.
Key Innovation: Analyzes observed Golbasi damage through the lens of soil-type controls to improve interpretation of earthquake-induced ground failures.
25. Acoustic Emission-Based Investigation of Freeze-Thaw-Induced Fracture Propagation and Damage Evolution in Water-Saturated Rock
Core Problem: Freeze-thaw cycling weakens saturated rock masses and can promote fracture growth relevant to alpine rockfall and cold-region slope instability.
Key Innovation: Uses acoustic emission monitoring to track fracture propagation and damage evolution in water-saturated rock under freeze-thaw conditions.
26. GeoHazards, Vol. 7, Pages 71: A New Catalogue of Historical Eruptions in Santorini Volcano: Documentation and Completeness Analysis
Core Problem: Historical eruption catalogs contain uneven documentary evidence, uncertain eruption size, and completeness biases that affect volcanic hazard interpretation.
Key Innovation: Builds a revised Santorini eruption catalogue with reliability scoring, bimodal VEI assignment, and completeness analysis using statistical and Monte Carlo tests.
27. Future projections of heat extreme modulation by subseasonal oscillations over India as simulated by the CMIP6 models
Core Problem: Future heat extremes over India depend partly on subseasonal oscillations, but model skill differs across oscillation modes.
Key Innovation: Evaluates CMIP6 representation of tropical subseasonal oscillations and uses best-performing ensembles to project heat-extreme modulation.
28. Distributed Acoustic Sensing for Urban Monitoring: Coverage Thresholds and Percolation
Core Problem: Urban DAS monitoring depends on fiber-network coverage, but spatial thresholds for reliable city-scale observability remain underdefined.
Key Innovation: Frames DAS coverage as a percolation problem to analyze thresholds for urban monitoring network connectivity.
29. Calibrated uncertainty for continental design flood estimation using conformal probabilistic machine learning
Core Problem: Continental design-flood estimates need calibrated uncertainty for ungauged or sparsely gauged basins.
Key Innovation: Applies conformal probabilistic machine learning to produce calibrated uncertainty intervals for large-scale flood estimation.
30. Evolution of earthquake crisis management and recovery in Italy after 2009: An expert commentary
Core Problem: Earthquake recovery practice has evolved after repeated Italian disasters, but transferable institutional lessons remain dispersed.
Key Innovation: Synthesizes crisis-management and recovery lessons from L'Aquila, Emilia-Romagna, Central Italy, and Ischia with comparison to French practice.
31. Cascading failure analysis of critical infrastructure in conflict zones: a large language model-driven knowledge graph framework
Core Problem: Infrastructure failure in conflict zones can cascade across interdependent systems, but event knowledge is fragmented across text sources.
Key Innovation: Uses a large-language-model-driven knowledge graph to extract and analyze cascading infrastructure failure pathways.
32. RFDTM: A national-scale and wall-to-wall 30 m resolution mangrove sub-canopy topography dataset for New Zealand derived from ICESat-2 ATLAS and multi-band SAR
Core Problem: Coastal inundation and ecosystem-risk modeling are limited by poor sub-canopy elevation data in mangrove environments.
Key Innovation: Produces a national 30 m mangrove sub-canopy topography dataset using ICESat-2 and multi-band SAR for hydrologic and vulnerability applications.
33. Assessing the Robustness of Prithvi Geospatial Foundation Model for Coastal Habitat Mapping Under Data Availability and Domain Shift Scenarios
Core Problem: Operational habitat and coastal mapping requires models that remain reliable under label scarcity and domain shift.
Key Innovation: Stress-tests the Prithvi geospatial foundation model for coastal habitat mapping across data availability and domain-shift scenarios.
34. Hyperbolic Geometry-Guided Unsupervised Domain Adaptive Semantic Segmentation in Remote Sensing
Core Problem: Remote-sensing segmentation models often drift under cross-region and cross-sensor distribution shifts.
Key Innovation: Uses hyperbolic geometry to preserve global semantic structure during unsupervised domain adaptation for remote-sensing segmentation.
35. Gaussian Mixture Distribution stereo matching network for satellite images
Core Problem: Terrain and infrastructure mapping from satellite stereo imagery requires robust disparity estimation under complex surface conditions.
Key Innovation: Introduces a Gaussian mixture distribution stereo-matching network for satellite images, improving dense 3D reconstruction inputs for geomorphic mapping.
36. EchoNet: Efficient railway segmentation from satellite imagery
Core Problem: Railway exposure mapping is needed for infrastructure risk analysis but remains labor-intensive at broad scale.
Key Innovation: Develops an efficient satellite-image railway segmentation model that can support transport-network exposure inventories.
37. BlastContextNet: Fast blast loading prediction in large-scale urban environments via spatial perception
Core Problem: Large-scale urban blast loading prediction is computationally expensive because geometric preprocessing dominates runtime.
Key Innovation: Uses a spatial perception module to learn local geometric context directly from raw patches for fast urban shock-wave prediction.
38. Disaster Aid and Rental Assistance: Examining Federal Housing Support for Homeowners and Renters After Disasters
Core Problem: Federal disaster housing support may affect homeowners and renters unevenly after major hazards.
Key Innovation: Examines patterns in disaster aid and rental assistance to clarify equity gaps in post-disaster housing recovery.
39. Lowering premiums and increasing investor returns by issuing multi-country CAT bonds
Core Problem: Single-country catastrophe bonds may face high premiums and limited diversification for low-frequency extreme hazards.
Key Innovation: Analyzes how multi-country catastrophe bonds can reduce premiums while improving investor returns through pooled disaster risk.
40. Efficiency of the prescriptive recommendations of the Iranian NBC-8 code for RC tie elements in confined masonry buildings
Core Problem: Prescriptive seismic detailing rules for confined masonry may be too general across building heights and demand levels.
Key Innovation: Evaluates the effectiveness of Iranian NBC-8 recommendations for reinforced-concrete tie elements in confined masonry buildings.
41. Seismic performance of integral abutment bridges with prestressed ultra-high-performance concrete (UHPC) girders, considering soil-structure interaction (SSI): a comparative analysis
Core Problem: Integral abutment bridge seismic behavior depends on continuous superstructure-foundation interaction and girder material choice.
Key Innovation: Compares UHPC and conventional concrete integral abutment bridges while explicitly considering soil-structure interaction.
42. Numerically predicted seismic performance of post-tensioned cross-laminated timber shear wall structures after 50 years of service life
Core Problem: Post-tensioning losses and timber creep can alter the seismic performance of cross-laminated timber structures over decades.
Key Innovation: Predicts 50-year seismic behavior of post-tensioned CLT shear walls using validated loss estimates and calibrated lateral-performance models.
43. Study on the diffusion and explosion characteristics of hydrogen leakage incident in tunnel under longitudinal ventilation
Core Problem: Hydrogen leakage in semi-confined road or rail tunnels can create fire and explosion hazards that depend on ventilation and ignition conditions.
Key Innovation: Simulates hydrogen diffusion and explosion characteristics in tunnel environments under different ignition locations, times, and wind speeds.
44. Robust Spatial Georeferencing for UAV-UGV Mobile Mapping Platforms in Urban Canyons via Asymmetric GNSS/UWB Fusion
Core Problem: Urban canyon mapping degrades GNSS positioning for UAV-UGV systems, limiting reliable mobile sensing in obstructed hazard-response environments.
Key Innovation: Fuses asymmetric GNSS and UWB observations to improve spatial georeferencing for UAV-UGV mobile mapping platforms.
45. A Method for Propagating Uncertainty of LiDAR Measurements to QSM-Derived Tree Metrics
Core Problem: LiDAR-derived structural metrics are often used in terrain and vegetation analysis without explicit propagation of measurement uncertainty.
Key Innovation: Develops a method for propagating LiDAR measurement uncertainty into quantitative structure model outputs.