TerraMosaic Daily Digest: June 19, 2026
Daily Summary
The June 19 literature is dominated by landslide intelligence that is becoming more mechanistic and more updateable. The strongest papers pair transfer learning with physical constraints, Bayesian uncertainty with InSAR-augmented sampling, crack mapping with SPH kinematics, discrete-fracture networks with rockfall block detection, and nonlocal rheology with basal sliding. Together, they move landslide assessment away from static terrain correlation and toward models that can carry physical state, deformation evidence, and uncertainty into prediction.
A second cluster treats geohazard mechanics as coupled systems rather than isolated triggers. Anti-dip rock-slope damage, debris-flow barrier overflow, seepage erosion in layered soil-rock mixtures, rock-ice avalanche rheology, liquefiable sloping ground, fault-crossing tunnels, rockburst, and tunnelling gushing all emphasize the same problem: failure often emerges from interacting structure, water, stress path, and engineered boundary conditions. Laboratory testing, physical simulation, and numerical modelling remain essential companions to remote sensing and machine learning.
The broader hazard set extends this logic to cities, rivers, coasts, and climate extremes. ICOPS-CNN subsidence mapping, Alpine high-impact precipitation, Shapley-based compound-flood attribution, road-speed flood resilience, wildfire forecasting and cause prediction, drought-heat modelling, river-flow intermittency, and disaster-data temporality all focus on making hazard evidence operational. The methods papers on InSAR phase filtering, geological model construction, ground-motion generation, and uncertainty-aware geotechnical AI are useful because they strengthen the data and modelling infrastructure behind future geohazard workflows.
Key Trends
Five movements define this issue: physical landslide AI, structure-aware slope and debris-flow mechanics, coupled seismic ground-infrastructure response, attributable hydroclimatic extremes, and uncertainty-aware hazard data infrastructure.
- Landslide susceptibility is becoming dynamic, physical, and uncertainty-aware: Deep transfer learning, Bayesian updating, InSAR augmentation, crack-informed SPH simulation, and nonlocal 3D-SPH show that landslide models are beginning to encode process, deformation, and epistemic uncertainty rather than only map static correlations.
- Rock-slope and debris-flow mechanics are being tied back to measurable structure: Deterministic fracture networks, anti-dip slope damage analysis, debris-flow overflow experiments, soil-rock erosion tests, and rock-ice avalanche scaling all connect field or laboratory observables to failure mechanisms.
- Seismic geohazard studies are converging on coupled infrastructure-ground response: Fault-crossing tunnels, liquefiable sloping deposits, piled filling slopes, ground-motion generation, induced seismicity, and earthquake hydrological response papers all examine how earthquake loading moves through ground, fluid, and engineered systems.
- Compound hydroclimatic hazards are becoming more attributable and operational: Alpine high-impact precipitation, Shapley-based compound floods, urban road-speed flood resilience, drought-heat prediction, river intermittency, future rainfall, and storm-cluster loss studies quantify not only where hazards occur but which drivers control their impacts.
- Reliable hazard AI depends on better data infrastructure, not only larger models: InSAR phase filtering, UAV-to-volume geological modelling, gully extraction, disaster-data temporality, Bayesian constitutive uncertainty, and vision-foundation geotechnical monitoring point toward auditable, uncertainty-aware, domain-grounded AI pipelines.
Selected Papers
The selected papers emphasize landslide susceptibility and kinematics, rockfall and debris-flow mechanics, urban subsidence, earthquake deformation, liquefaction, tunnel and slope response under seismic loading, compound flood attribution, wildfire and drought forecasting, river-flow intermittency, gully mapping, geological model construction, disaster-data quality, and uncertainty-aware AI tools for geotechnical and hydrological workflows. This issue contains 47 selected papers.
1. Landslide susceptibility assessment integrating deep transfer learning and physical models in the Baihetan reservoir area, China
Core Problem: Regional landslide susceptibility around large reservoirs remains difficult to transfer across terrain, hydrological, and engineering settings when models rely only on empirical correlations.
Key Innovation: Integrates deep transfer learning with physical model information to make susceptibility assessment more portable and process-aware in the Baihetan reservoir area.
2. Bayesian deep learning framework for updating landslide susceptibility assessment based on epistemic uncertainty with InSAR augmented samples
Core Problem: Conventional data-driven susceptibility maps rarely quantify epistemic uncertainty or update themselves when deformation evidence becomes available.
Key Innovation: Uses Bayesian deep learning and InSAR-augmented samples to update landslide susceptibility while exposing uncertainty from data sampling and model knowledge gaps.
3. A new unstable block identification method using deterministic discrete fracture networks for rockfall hazard assessment in open-pit mines
Core Problem: Rockfall hazard assessment in fractured open-pit slopes depends on identifying unstable blocks, but manual mapping is slow and exposes field teams to risk.
Key Innovation: Builds a deterministic discrete-fracture-network method to identify potentially unstable rock blocks and support more systematic rockfall hazard assessment.
4. Deep learning-augmented crack mapping and SPH-based dynamic simulation for landslide kinematic prediction
Core Problem: Early crack patterns contain information about landslide motion, yet they are often disconnected from dynamic runout simulation.
Key Innovation: Combines deep-learning crack extraction with SPH-based simulation so mapped surface damage can inform landslide kinematic prediction.
5. Nonlocal rheology-based 3D-SPH model incorporating basal sliding quantification for granular landslide dynamics
Core Problem: Granular landslide mobility is controlled by nonlocal rheology and basal sliding, both of which are hard to represent in practical three-dimensional runout models.
Key Innovation: Introduces a nonlocal rheology-based 3D-SPH formulation that explicitly quantifies basal sliding effects in granular landslide dynamics.
6. Deep-seated zonal damage in anti-dip layered rock slope: A case study in southwest China
Core Problem: Anti-dip layered rock slopes can develop deep zonal damage before visible failure, but internal damage organization is difficult to infer from surface evidence alone.
Key Innovation: Uses a southwest China case to characterize deep-seated zonal damage and link layered structural control to progressive rock-slope weakening.
7. Debris flow overflowing rigid barrier: Theoretical models and experimental verification
Core Problem: Rigid barriers can reduce debris-flow impact, but overflow behavior remains a key failure mode for barrier sizing and downstream risk.
Key Innovation: Combines theoretical models with experiments to describe overflow conditions and improve the mechanics basis of debris-flow barrier design.
8. Land subsidence through ICOPS time-series coupled with metaheuristic optimized CNN-based susceptibility mapping: a case study in Bandung, Indonesia
Core Problem: Urban subsidence requires both deformation monitoring and forward susceptibility mapping, especially where groundwater extraction and urban growth interact.
Key Innovation: Couples Sentinel-1 ICOPS time-series deformation with CNN susceptibility models optimized by metaheuristic search for Bandung, Indonesia.
9. Experimental investigation on erosion characteristics in double-layer soil-rock mixtures considering seepage direction and layer-specific fine particle contents
Core Problem: Internal erosion in layered soil-rock mixtures depends on seepage direction and fine-particle distribution, but these controls are rarely isolated experimentally.
Key Innovation: Tests double-layer mixtures under controlled seepage to clarify how layer-specific fines and hydraulic direction govern erosion susceptibility.
10. Rheological scaling for granular systems with varying size, density, and friction: implications for rock-ice avalanches
Core Problem: Rock-ice avalanche mobility is governed by mixed particles with contrasting size, density, and friction, complicating scaling from experiments to natural events.
Key Innovation: Develops rheological scaling for heterogeneous granular systems and frames its implications for warming-driven rock-ice avalanche dynamics.
11. Shaking table study of dynamic response of biotreated liquefiable sandy ground with varying slopes
Core Problem: Inclined saturated sand deposits are vulnerable to seismic liquefaction, but the performance of biotreated ground under slope-dependent shaking remains uncertain.
Key Innovation: Uses shaking-table tests to evaluate how MICP biotreatment changes acceleration, pore-pressure, and deformation response of liquefiable slopes.
12. Spatio-Temporal Patterns and Drivers of High-Impact Precipitation Events in the European Alps (1961-2022)
Core Problem: High-impact precipitation in the Alps drives floods and landslides, yet long-term spatial patterns and atmospheric controls remain unevenly characterized.
Key Innovation: Analyzes 1961-2022 Alpine high-impact precipitation to relate event clustering and drivers to mountain flood and landslide exposure.
13. InSAR-derived surface deformation and numerical simulation: constraints on the seismogenic fault and rupture process of the 2021 MS 5.4 Baicheng earthquake
Core Problem: The 2021 Baicheng earthquake produced damaging deformation in a fold-thrust belt where salt structures and shallow ruptures complicate source interpretation.
Key Innovation: Combines InSAR coseismic deformation with numerical simulation to constrain seismogenic fault geometry and rupture behavior.
14. Multimethod geophysical investigation for cavity detection in the Northwestern part of Jebel Hafeet area, Al-Ain, United Arab Emirates
Core Problem: Subsurface cavities can destabilize infrastructure, but single geophysical methods may miss geometry or ambiguity in complex carbonate terrain.
Key Innovation: Integrates electrical resistivity and gravity surveys to detect cavities and assess associated ground-instability risk at Jebel Hafeet.
15. Spatiotemporal quantification of relative contributions of compound flood drivers using Shapley value-based approach
Core Problem: Compound flood severity depends on nonlinear interactions among drivers, making it difficult to assign responsibility to rainfall, surge, river flow, or other controls.
Key Innovation: Applies a Shapley-value framework to quantify spatiotemporal driver contributions across combinations, strengthening interpretability of compound-flood analysis.
16. Dynamic centrifuge tests of arching effect in embankment undergoing basal settlement and water infiltration
Core Problem: Basal settlement and infiltration can change arching in embankments, affecting deformation and stability during coupled hydromechanical loading.
Key Innovation: Uses dynamic centrifuge testing with water supply to observe arching, settlement, and deformation evolution under controlled basal subsidence and infiltration.
17. Dynamic responses and damping damage of aseismic joints for strike-slip fault-crossing tunnels subjected to excitation
Core Problem: Tunnels crossing strike-slip faults are highly vulnerable during earthquakes, and the damage mechanisms of aseismic joints require physical verification.
Key Innovation: Uses shaking-table models to examine mitigation performance and damage accumulation in aseismic joints for fault-crossing tunnels.
18. Frictional behavior of dolomite-rich fault gouges: Effects of mineralogical heterogeneity and implications for induced seismic risk in carbonate reservoirs
Core Problem: Induced seismic risk in carbonate reservoirs depends on fault-gouge friction, which can vary with mineralogical heterogeneity.
Key Innovation: Measures dolomite-rich gouge frictional behavior to connect mineral composition with fault stability and induced seismic potential.
19. Root reinforcement and riverbank stability: Biomechanical contribution of Alnus glutinosa across contrasting temperate environments
Core Problem: Vegetation reinforcement is central to riverbank stability, but root biomechanical properties may vary across environmental settings.
Key Innovation: Compares Alnus glutinosa root systems across contrasting temperate environments to quantify biomechanical contributions to bank stability.
20. Operational Diagnosis of Urban Road Flood Resilience Using Vehicle Speed
Core Problem: Urban flood resilience is often measured with static indicators that do not capture real-time transport disruption during inundation.
Key Innovation: Uses vehicle-speed data to diagnose operational road resilience at high spatiotemporal resolution during urban flooding.
21. CanadaFireSat: Towards high-resolution wildfire forecasting with multiple modalities
Core Problem: Wildfire forecasting requires high-resolution multimodal data that can represent weather, fuels, and landscape conditions in rapidly changing fire seasons.
Key Innovation: Introduces CanadaFireSat as a multimodal forecasting direction for boreal wildfire monitoring and management.
22. Integrating physical modeling and data-driven learning: A DCNN for InSAR phase filtering and applications
Core Problem: InSAR deformation products can fail under low coherence or dense fringes when filters oversmooth phase or preserve noise.
Key Innovation: Combines physical modeling and deep convolutional learning to improve phase filtering for deformation retrieval and topographic applications.
23. A decoupled generative framework for physics-constrained non-stationary ground motion simulation
Core Problem: Ground-motion simulation must reproduce non-stationary intensity, frequency, and amplitude patterns despite scarce data for extreme scenarios.
Key Innovation: Decouples physics-constrained simulation into hierarchical components to generate non-stationary earthquake ground motions more flexibly.
24. Zoning method for seismic response of piled filling slopes based on post-test wave velocity from large-scale shaking table tests
Core Problem: High filling slopes show spatially variable seismic response, making uniform aseismic design insufficient.
Key Innovation: Uses post-test wave-velocity changes from large-scale shaking-table tests to zone seismic response and damage in piled filling slopes.
25. True triaxial physical simulation and characteristic analysis of rockburst in deep tunnels triggered by fatigue disturbance
Core Problem: Rockburst under high in-situ stress and fatigue disturbance is difficult to reproduce with conventional loading paths.
Key Innovation: Applies true triaxial physical simulation to analyze immediate and delayed rockburst behavior in deep tunnels.
26. Explainable spatiotemporal GCN for detecting gushing during EPB shield tunnelling
Core Problem: Gushing events in EPB shield tunnelling arise from coupled geology and machine-control mismatches, requiring real-time interpretable detection.
Key Innovation: Builds an explainable spatiotemporal graph convolutional network for ring-by-ring gushing risk classification.
27. Seismic response of tunnels with isolation layer crossing fault zones: large-scale shaking table tests and numerical verification
Core Problem: Isolation layers for tunnels crossing fault zones need experimental validation under large-scale seismic loading.
Key Innovation: Combines shaking-table tests and numerical verification to evaluate tunnel response and isolation-layer effectiveness across fault zones.
28. Wave overtopping detection and intertidal zone delineation using semantic segmentation in coastal scenes
Core Problem: Coastal overtopping and intertidal boundaries are hard to monitor continuously from imagery under changing shoreline and foam conditions.
Key Innovation: Uses semantic segmentation to classify coastal scene pixels, detect overtopping, and delineate intertidal zones from camera data.
29. Influence of roughness on overtopping flow at a dike using deep learning-based video measurement techniques
Core Problem: Thin, unsteady overtopping flows over dikes are difficult to measure, limiting flood-defense design and roughness evaluation.
Key Innovation: Uses deep-learning video measurement to quantify overtopping velocity and water-layer thickness under different roughness conditions.
30. Aufeis in a warming world: Global patterns, processes, and environmental implications
Core Problem: Aufeis affects cold-region hydrology, infrastructure, and geomorphology, but global patterns and warming-driven changes are fragmented across studies.
Key Innovation: Synthesizes global aufeis processes, distributions, and environmental implications under warming conditions.
31. Magnetotelluric Imaging Suggests Minimal Downward Saline Fluid Migration in the Region of the Largest-Known Injection-Induced Earthquake, Oklahoma
Core Problem: The fluid pathways involved in basement-hosted induced earthquakes remain difficult to infer after wastewater injection events.
Key Innovation: Uses rapid-response magnetotelluric imaging around the Pawnee earthquake to test whether saline fluids migrated downward into the seismogenic region.
32. Hydrological Impact of Earthquakes on Reverse and Normal Faults: Results From Numerical Models
Core Problem: Earthquake-induced hydrological signals reflect coupled deformation, permeability, thermal pressurization, and fault-zone fluid flow.
Key Innovation: Models reverse and normal fault ruptures with rate-state friction and coupled fluid flow to separate mechanisms of hydrological response.
33. Seismogenic Structure of the 1975 Haicheng Ms 7.3 Earthquake (NE China) Inferred from 3D Magnetotelluric Imaging
Core Problem: The structural controls on the 1975 Haicheng earthquake remain important for understanding seismogenic zones in northeastern China.
Key Innovation: Uses 3D magnetotelluric imaging to infer conductive structures associated with the Haicheng seismogenic system.
34. Analysing impact of future emission scenarios on rainfall in rainfed agricultural zones in Northern Iraq using satellite data and CMIP6 models
Core Problem: Fine-temporal future rainfall shifts in northern Iraq remain uncertain for rainfed agricultural zones and water-harvesting infrastructure.
Key Innovation: Combines satellite data, CMIP6 scenarios, and weather generation to project rainfall changes under SSP245 and SSP585.
35. Unravelling wind-driven impact of storm clusters, a case study for the insurer Generali France
Core Problem: Winter windstorm losses in Europe often arise from clusters, complicating attribution of damage to individual events for insurance and reinsurance.
Key Innovation: Analyzes storm-cluster impacts in France to connect meteorological clustering with loss attribution and insured damage aggregation.
36. ML-based fire cause prediction: Integrating population mobility and meteorological data
Core Problem: Wildfire cause records are often incomplete, limiting prevention and risk-management strategies.
Key Innovation: Combines meteorological data, geography, and population mobility in a machine-learning framework to predict fire causes and audit unknown records.
37. Copula-based risk identification and HydroFusionNet-driven prediction with potential causal-linkage interpretation for compound drought-heat events
Core Problem: Compound drought-heat events require joint-risk estimation and interpretable prediction because their impacts emerge from coupled water and heat stress.
Key Innovation: Combines copula-based risk identification with HydroFusionNet prediction and causal-linkage interpretation for semi-arid basin hazards.
38. Hydrologically guided transfer learning for ungauged basin prediction: similarity-based adaptation and interpretation of runoff generation mechanisms
Core Problem: Ungauged-basin prediction suffers from distribution shift and weak interpretability when deep models are transferred across catchments.
Key Innovation: Introduces hydrologically guided transfer learning with similarity-based adaptation and interpretation of runoff-generation mechanisms.
39. Wavelet-Kalman-LSTM fusion: enhancing meteorological drought prediction with robust feature selection and uncertainty quantification
Core Problem: Drought prediction must handle non-stationary climate signals and quantify uncertainty across multiple time scales.
Key Innovation: Combines wavelet decomposition, Kalman filtering, LSTM modeling, feature selection, and uncertainty quantification for SPI drought forecasts.
40. Spatial Patterns and Controlling Factors of River Flow Intermittency in Africa
Core Problem: River intermittency in data-scarce African basins is poorly mapped despite its importance for drought, ecosystem, and water-security risk.
Key Innovation: Produces a high-resolution four-class intermittency map using sequential random-forest modeling across millions of river reaches.
41. Spatiotemporal evolution, physical drivers, and future prediction of global terrestrial moisture imbalance: a multi-expert AI-agent approach based on SVBI
Core Problem: Existing indices can miss vapor supply-demand mismatch and flash anomalies in global moisture imbalance.
Key Innovation: Defines SVBI and applies a multi-expert AI-agent framework to analyze drivers and future evolution of terrestrial moisture imbalance.
42. Uncertainty-informed large vision model for automated recognition of excavated rock chips in tunnel boring machines
Core Problem: Rock-chip recognition in tunnel boring is sensitive to field lighting and image variability, limiting real-time ground-condition inference.
Key Innovation: Integrates U-Net with the Segment Anything vision foundation model and uncertainty handling for automated TBM muck-image interpretation.
43. A Bayesian-based framework for quantifying model-inherent uncertainties in soil constitutive models
Core Problem: Soil constitutive models carry model-form and parameter uncertainty that propagate into slope, foundation, and ground-response simulations.
Key Innovation: Develops a Bayesian framework for quantifying model-inherent uncertainties in constitutive behavior.
44. Extraction of permanent gullies integrating a thalweg-guided dataset and morphological structures with deep learning in the Mollisol region of Northeast China
Core Problem: Large-area gully mapping is limited by complex backgrounds and scarce annotated data.
Key Innovation: Uses thalweg-guided datasets and morphology-aware deep learning to extract permanent gullies in the Mollisol region.
45. Bridging the visual-computational gap: transforming UAV reality meshes into volumetric geological models via explicit-implicit integration
Core Problem: UAV reality meshes capture outcrop geometry but are not directly usable as volumetric geological models for simulation or hazard analysis.
Key Innovation: Combines explicit and implicit modeling to transform UAV reality meshes into volumetric geological representations.
46. Between Hope and Hype: Civil Society Organisations in Disaster Risk Reduction - A Systematic Global Review and Future Research Agenda
Core Problem: Civil society organizations are widely invoked in disaster risk reduction, but evidence on their roles and limits is dispersed across regions and hazard contexts.
Key Innovation: Synthesizes global literature to clarify where civil society organizations strengthen disaster-risk governance and where claims remain overextended.
47. Temporal characteristics of disaster data: how to avoid misconceptions, misinterpretation, or misuse
Core Problem: Temporal properties of disaster datasets can be misunderstood, leading to flawed trend analysis, misleading comparisons, or inappropriate operational use.
Key Innovation: Examines disaster-data temporality and offers guidance for avoiding common misinterpretations in hazard and risk analysis.