TerraMosaic Daily Digest: June 18, 2026
Daily Summary
The June 18 literature is anchored by landslide mechanics rather than only mapping. Field evidence on landslide dams, physics-informed regional landslide probability, soil-moisture-aware susceptibility modelling, rainfall-driven slope-stability forecasting, and a unified elastoplastic-µ(I) formulation all point to the same shift: models are beginning to represent the transition from hydrological forcing to large-deformation failure more explicitly. Several remote-sensing papers extend this line into operational mapping, with InSAR susceptibility, earthquake-triggered landslide detection, UAV-LiDAR slip-surface estimation, and alpine deformation monitoring.
The second cluster expands geohazard intelligence into compound and cascading impacts. Hurricane Helene flood-landslide hotspots, liquefaction fragility mapping for Dhaka, flood susceptibility in Ethiopia, real-time urban flood forecasting, flood early-warning evaluation in Africa, and flood-damage drivers in Assam all connect physical hazard with exposure, service disruption, or loss. Permafrost and thaw-slump papers add a cryosphere dimension, treating deformation, climate forcing, and explainable machine learning as linked evidence streams.
A smaller but useful methods cluster addresses trust, data foundations, and agentic infrastructure. Burned-area mapping, wildfire detection, grass-curing prediction, tropical-cyclone precipitation extremes, and wildfire-weather attribution strengthen the event-scale climate-hazard record. BMAT, PANEL, semantic-entropy geotechnical design, and HydroCraft are less directly about landslides, but they matter because hazard workflows increasingly depend on reusable exposure data, domain-specific vision-language models, uncertainty-aware language-model support, and interactive hydrological modelling.
Key Trends
Five movements define this issue: process-aware landslide modelling, deformation-focused remote sensing, decision-oriented compound-risk analysis, climate-linked cryosphere and wildfire evidence, and uncertainty-aware AI infrastructure.
- Landslide modelling is moving from static susceptibility to process-aware failure prediction: Landslide-dam field evidence, rainfall-induced FOS forecasting, physics-informed regional landslide probability, unsaturated soil rheology, and soil-rock mixture CFD-DEM all represent hydrology, material degradation, or large deformation rather than only terrain correlation.
- Remote sensing is being used to infer hidden geometry and time-dependent deformation: InSAR susceptibility, UAV-LiDAR slip-surface estimation, mining deformation, alpine landslide monitoring, permafrost deformation, and 3D coseismic LiDAR displacement show a move from detection toward kinematic and structural interpretation.
- Compound risk studies are becoming more local and decision-oriented: Hurricane Helene hotspots, Dhaka liquefaction fragility, Wolaita flood susceptibility, urban flood forecasting, Assam damage drivers, and African flood-warning evaluation translate hazard outputs into damage, service access, or warning performance.
- Cold-region and wildfire hazards are treated as coupled climate-geomorphic systems: Thaw-slump susceptibility, alpine permafrost deformation, debris-flow impacts after wildfire, burned-area products, grass curing, wildfire detection, and wildfire-weather attribution all connect climate forcing with geomorphic or ecological response.
- AI infrastructure is shifting toward uncertainty and reusable domain tools: Semantic entropy for LLM-assisted geotechnics, HydroCraft, PANEL, and BMAT show that foundation-model and agentic workflows need uncertainty control, domain-specific interpretation, and exposure data before they become reliable hazard infrastructure.
Selected Papers
The selected papers emphasize landslide dams, rainfall-induced failure prediction, soil-moisture-informed susceptibility, unsaturated soil rheology, InSAR and UAV-based landslide monitoring, compound flood-landslide impact, liquefaction, thaw slumps, permafrost deformation, tunnel and pipeline response to faulting, volcanic and wildfire risk, remote-sensing datasets, and uncertainty-aware AI tools for geotechnical and hydrological workflows. This issue contains 43 selected papers.
1. Analysis of Dam Characteristics and Failure Trend of Landslide Dams With Different Materials: Field Investigation
Core Problem: Material-dependent landslide-dam stability and breach behavior remain difficult to generalize from laboratory tests or single-event case studies.
Key Innovation: Uses field investigations of landslide dams with different material compositions to relate dam geometry and internal material structure to failure tendencies.
2. Physics-informed neural networks for efficient probabilistic analysis of regional rainfall-induced landslides
Core Problem: Regional rainfall-induced shallow-landslide probability requires physically consistent infiltration modeling, but full Richards-equation solutions are computationally costly at scale.
Key Innovation: Embeds hydrological physics into a neural surrogate for efficient probabilistic landslide analysis across regional soil and rainfall conditions.
3. Enhancing landslide susceptibility mapping with the integration of soil moisture and machine learning models in Minxian County, China
Core Problem: Static terrain and lithology predictors miss transient hydrological states that can materially alter landslide susceptibility.
Key Innovation: Integrates soil-moisture information with machine-learning susceptibility models to test hydrological controls on landslide prediction in Minxian County.
4. Prediction of Rainfall-Induced Slope Stability Spatiotemporal Evolution Based on a Hybrid Transformer-LSTM Deep Learning Framework
Core Problem: Slope factor-of-safety prediction under rainfall must capture nonlinear temporal infiltration and spatial stability evolution without the cost of full numerical simulation.
Key Innovation: Combines Transformer and LSTM components to learn spatiotemporal FOS evolution for rainfall-induced slope-stability assessment.
5. A unified elastoplastic-µ(I) rheology model for unsaturated soils: formulations and two-point MPM implementation
Core Problem: Models of geomaterial solid-fluid transition often separate elastoplastic deformation from flow-like landslide motion.
Key Innovation: Couples Modified Cam-Clay plasticity with µ(I) rheology and two-point MPM to represent progressive cohesion loss, pore-pressure effects, and large deformation.
6. An Optimized Heterogeneous Ensemble Learning Algorithm for InSAR Landslide Susceptibility Mapping Based on the Adaptive Sampling Strategy
Core Problem: Class imbalance and spatial heterogeneity limit the reliability of machine-learning landslide susceptibility maps derived from deformation and terrain predictors.
Key Innovation: Uses adaptive sampling with heterogeneous ensemble learning to improve InSAR-informed susceptibility mapping.
7. Toward an Automatic Pixel-Based Detection of Earthquake-Triggered Landslides in Arid Environments Using Optical Imagery
Core Problem: Post-earthquake landslide mapping in arid regions is hindered by subtle spectral contrasts and limited automated pixel-level workflows.
Key Innovation: Develops an optical-imagery workflow for automatic pixel-based detection of earthquake-triggered landslides in arid terrain.
8. Geotechnical Characterization, Risk Analysis, and Design of Stabilization Measures for a Landslide Along the RN16 Coastal Highway in Morocco
Core Problem: Coastal highway landslides require engineering designs that connect geotechnical characterization, risk assessment, and stabilization measures.
Key Innovation: Combines site investigation, stability analysis, and mitigation design for a landslide-prone segment of Morocco’s RN16 coastal highway.
9. Rockfall Volume-Cumulative Frequency Relationships for Rockfall Hazard Quantification Using Historical and Change Detection Data
Core Problem: Rockfall hazard estimates depend strongly on volume-frequency relationships, yet historical records and change-detection inventories often capture different event ranges.
Key Innovation: Combines historical and change-detection data to estimate rockfall volume-cumulative frequency relationships for hazard quantification.
10. Modeling of rainfall-induced failure of unsaturated soil-rock mixture slopes: CFD-DEM coupling method
Core Problem: Rainfall-induced failure of unsaturated soil-rock mixture slopes involves coupled hydro-mechanical processes and particle-scale soil-rock interactions.
Key Innovation: Uses a CFD-DEM coupling approach to simulate infiltration-driven failure in unsaturated soil-rock mixture slopes.
11. DCA-UNet for Landslide Segmentation with Deformable Convolution and Aggregated Attention
Core Problem: Landslide segmentation from remote sensing imagery must resolve irregular boundaries and multi-scale morphology under heterogeneous backgrounds.
Key Innovation: Introduces a U-Net variant using deformable convolution and aggregated attention for landslide semantic segmentation.
12. Displacement-Based Estimation of Quasi-Three-Dimensional Landslide Slip Surfaces Using UAV LiDAR Data
Core Problem: Slip-surface geometry is difficult to infer where direct subsurface data are sparse, yet it controls landslide volume and stabilization design.
Key Innovation: Uses UAV LiDAR-derived displacement patterns to estimate quasi-3D landslide slip surfaces.
13. Identifying Local Hotspots of Compounded Physical and Social Impacts from Hurricane Helene-Induced Flood and Landslides
Core Problem: Post-disaster prioritization can miss communities where physical damage, service isolation, and pre-existing vulnerability compound at local scales.
Key Innovation: Builds a block-group-scale composite index combining flood and landslide building damage, road-based service isolation, and vulnerability for Hurricane Helene impacts.
14. Probabilistic liquefaction hazard mapping in the DMDP area of Bangladesh: Assessing surficial manifestation and shallow foundation damage using fragility functions
Core Problem: Urban liquefaction risk needs site-specific probabilities for both surficial manifestation and shallow foundation damage.
Key Innovation: Uses SCPT data, site-specific PGA, and fragility functions to map probabilistic liquefaction and foundation-damage risk in the Dhaka region.
15. Attention-Driven Hierarchical Spatial Adaptive Ensemble for Landslide Susceptibility Mapping
Core Problem: Susceptibility mapping can lose spatial structure when model ensembles treat conditioning factors as independent tabular predictors.
Key Innovation: Uses hierarchical spatial adaptation and attention-driven ensemble learning to refine landslide susceptibility estimation.
16. Mining-Induced Deformation and Slope Stability in Steep Mountainous Areas Based on InSAR Monitoring and Rock Movement Theory
Core Problem: Mining deformation in steep mountainous terrain can propagate into slope-instability risk that is difficult to monitor with ground observations alone.
Key Innovation: Combines InSAR monitoring with rock movement theory to assess deformation and slope stability in a southwestern China mining area.
17. Landslide Deformation Remote Monitoring in Alpine Mountains Using UAV Photogrammetry and Infrared Thermography
Core Problem: Alpine landslide monitoring needs high-resolution deformation and thermal information in steep terrain where field access is limited.
Key Innovation: Integrates UAV photogrammetry with infrared thermography for remote monitoring of landslide deformation in mountainous terrain.
18. Influence of topsoil depth and rock block content on the stability of soil-rock mixture slopes
Core Problem: The stability of soil-rock mixture slopes depends on topsoil depth and block content, but these controls are often simplified in slope analyses.
Key Innovation: Evaluates how topsoil thickness and rock-block fraction affect stability responses in soil-rock mixture slopes.
19. GIS and remote sensing-based hybrid AHP-machine learning framework for flood susceptibility and risk assessment in the Wolaita Zone, Southern Ethiopia
Core Problem: Flood risk assessment in data-scarce regions must combine sparse observations, terrain controls, and interpretable decision criteria.
Key Innovation: Integrates GIS, remote sensing, AHP, and ensemble machine-learning models for flood susceptibility and risk mapping in southern Ethiopia.
20. AI-assisted framework using physically informed rainfall-drainage features for real-time urban flood risk forecasting
Core Problem: Real-time urban flood warning requires actionable hazard classes rather than raw rainfall or water-level forecasts.
Key Innovation: Uses physically informed rainfall-drainage features in an AI-assisted framework for real-time multi-class urban flood risk forecasting.
21. Thaw slump susceptibility assessment in the central Qinghai-Tibet Plateau permafrost region based on an interpretable 2D-CNN framework with SBAS-InSAR deformation
Core Problem: Permafrost thaw-slump susceptibility reflects coupled thermal, hydrological, terrain, and deformation conditions.
Key Innovation: Combines SBAS-InSAR deformation with an interpretable 2D-CNN framework to assess thaw-slump susceptibility on the Qinghai-Tibet Plateau.
22. Climate-driven thaw slump susceptibility on the Qinghai-Tibet plateau using geographically explainable machine learning
Core Problem: Regional thaw-slump susceptibility needs models that can explain how climate and terrain controls vary geographically.
Key Innovation: Applies geographically explainable machine learning to map and interpret climate-driven thaw-slump susceptibility on the Qinghai-Tibet Plateau.
23. Psychometric and cultural characterization of relative volcanic risk perception levels of individuals highly exposed to proximal activity from Villarrica volcano, Chile
Core Problem: Volcanic risk communication depends on how exposed residents and visitors perceive proximal volcanic activity.
Key Innovation: Combines psychometric and cultural-theory approaches to characterize volcanic risk perception near Villarrica volcano.
24. Quantifying the current and future likelihood of the 2022 extreme wildfire weather conditions in France with anthropogenic climate change
Core Problem: Extreme wildfire weather requires attribution frameworks that distinguish present-day likelihood from future climate-conditioned risk.
Key Innovation: Quantifies current and future likelihood of the 2022 French wildfire-weather conditions under anthropogenic climate change.
25. A Streamlined Flood-Specific Evaluation Framework: Assessing African Riverine Early Warning Systems
Core Problem: Riverine flood early warning systems need evaluation criteria tailored to flood-specific lead time, reliability, and user actionability.
Key Innovation: Develops a streamlined framework for assessing African riverine early warning systems.
26. Hotspots and Drivers of Flood Damage in Assam, India: A Spatio-Temporal Assessment of Flood-Induced Damage and Their Driving Factors
Core Problem: Flood-loss mitigation in Assam requires spatially and temporally explicit understanding of damage hotspots and their drivers.
Key Innovation: Maps flood-induced damage patterns and analyzes driving factors across Assam, India.
27. Land Surface Deformation of Alpine Permafrost in the Earthquake-Impacted Source Area of the Yellow River During 2017-2024
Core Problem: Earthquake-impacted alpine permafrost can deform over multi-year periods, affecting slope and hydrological stability.
Key Innovation: Uses remote-sensing deformation time series to characterize land-surface deformation in the Yellow River source area from 2017 to 2024.
28. Extraction of Detailed 3D Coseismic Displacements in the 2024 Noto Peninsula Earthquake from Airborne LiDAR Data
Core Problem: Earthquake deformation fields are difficult to resolve in three dimensions at fine spatial detail using conventional two-dimensional displacement products.
Key Innovation: Derives detailed 3D coseismic displacements from airborne LiDAR data for the 2024 Noto Peninsula earthquake.
29. Response of Tunnel-track system to fault dislocation considering rock mass rupture patterns
Core Problem: Tunnel-track systems crossing faults can experience complex deformation governed by rock-mass rupture patterns.
Key Innovation: Models tunnel-track response to fault dislocation while explicitly considering rock-mass rupture geometry.
30. Performance of large deep tunnel complex considering strike-slip fault dislocation
Core Problem: Large deep tunnel complexes crossing strike-slip faults require deformation and damage assessment beyond single-tunnel simplifications.
Key Innovation: Evaluates the performance of a deep tunnel complex under strike-slip fault dislocation scenarios.
31. Efficient Mitigation Measures for Reducing the Kinematic Distress of Offshore Pipelines Due to Seismic Fault Rupture
Core Problem: Offshore pipelines crossing active faults can suffer large kinematic distress during seismic rupture.
Key Innovation: Assesses mitigation measures for reducing fault-rupture-induced pipeline deformation and distress.
32. Contribution of Tropical Cyclones to Hourly Precipitation Extremes in the Contiguous United States
Core Problem: Hourly precipitation extremes are a key driver of flash flooding, yet the tropical-cyclone contribution varies regionally and temporally.
Key Innovation: Quantifies the contribution of tropical cyclones to hourly precipitation extremes across the contiguous United States.
33. Debris Flows Suppressed Riverine Productivity and Respiration Following High-Severity Wildfire on the Klamath River, California
Core Problem: High-severity wildfire can trigger debris flows whose downstream ecological effects are rarely quantified at river-system scale.
Key Innovation: Links post-wildfire debris flows to suppressed riverine productivity and respiration on the Klamath River.
34. Spatiotemporal prediction of grass curing in Victoria, Australia using multi-source remote sensing and deep learning models
Core Problem: Grassland dryness and curing are critical wildfire-risk variables that need spatially continuous prediction.
Key Innovation: Uses multi-source remote sensing and deep-learning models to predict monthly grass curing in Victoria, Australia.
35. Developing 30 m resolution monthly burned area products in Africa (2017-2025)
Core Problem: Africa lacks long-term high-resolution monthly burned-area products despite its dominant contribution to global burned area.
Key Innovation: Combines Landsat, Sentinel-2, and Sentinel-1 features to generate 30 m monthly burned-area products for Africa from 2017 to 2025.
36. Immediate remote sensing: Dynamic context-adaptive fusion for Himawari-8/9 10-minute wildfire detection
Core Problem: Wildfire detection from geostationary satellites must adapt to changing context while preserving 10-minute operational latency.
Key Innovation: Introduces a dynamic context-adaptive fusion approach for Himawari-8/9 wildfire detection.
37. Assessment of the Relationship Between Seismic Vulnerability and Seismic Risk Perception: A Case Study of Peshawar, Pakistan
Core Problem: Physical seismic vulnerability and public risk perception may diverge, complicating preparedness planning.
Key Innovation: Compares structural seismic vulnerability with risk-perception patterns in Peshawar, Pakistan.
38. A Variational Data Assimilation Framework for Mining Subsidence Reconstruction from Heterogeneous D-InSAR and TLS Observations
Core Problem: Mining subsidence monitoring must integrate heterogeneous deformation observations with different resolutions and error structures.
Key Innovation: Uses variational data assimilation to reconstruct subsidence from D-InSAR and terrestrial laser-scanning observations.
39. BMAT: A footprint-level building facade material dataset for 73 major cities worldwide
Core Problem: Urban hazard and damage models often lack footprint-level building facade material data across cities.
Key Innovation: Provides a global city-scale building facade material dataset that can support exposure, vulnerability, and damage modelling.
40. PANEL: A photovoltaic-specific vision-language model for zero-shot and few-shot PV interpretation tasks in remote sensing
Core Problem: Remote-sensing interpretation often requires domain-specific reasoning under limited labeled samples.
Key Innovation: Builds a photovoltaic-specific vision-language model demonstrating zero-shot and few-shot remote-sensing interpretation capacity transferable to infrastructure mapping.
41. A semantic entropy framework for quantifying and mitigating uncertainty in LLM-assisted geotechnical design
Core Problem: LLM-assisted geotechnical design needs explicit uncertainty quantification before outputs can inform engineering workflows.
Key Innovation: Uses semantic entropy to quantify and reduce uncertainty in LLM-assisted geotechnical design tasks.
42. HydroCraft: an efficient online platform for hydrological modeling integrated with a large language model agent
Core Problem: Hydrological modelling platforms remain difficult for non-specialists to configure and interrogate across datasets and modelling choices.
Key Innovation: Integrates hydrological modelling with a large-language-model agent to support interactive online model setup and analysis.
43. Influential factors on cyclic resistance of non-plastic and low-plastic silts
Core Problem: Low-plastic and non-plastic silts can undergo cyclic strength loss, but their resistance depends on interacting material and loading factors.
Key Innovation: Analyzes influential factors controlling cyclic resistance in non-plastic and low-plastic silts relevant to seismic ground-failure assessment.