TerraMosaic Daily Digest: June 6, 2026
Daily Summary
The June 6 papers move landslide research toward explicit treatment of time, source conditions, and downstream consequences. A national-scale Italian study reframes landslide activity as a climate-sensitive signal that must be resolved through rainfall timing and regional dynamics rather than static susceptibility alone. A Natural Hazards paper advances co-seismic landslide mapping by replacing handcrafted change features with transfer-learned deep representations, while a reservoir-slope study couples landslide motion with surge impact to simulate a full failure-to-inundation chain. Debris-flow susceptibility work attacks a quieter but persistent weakness in machine learning inventories: the construction of defensible negative samples in surveyed terrain.
Cold-region and infrastructure papers converge on material-state monitoring. Ice-bearing glaciofluvial deposits, freeze-thawed red-bed sandstone, capillary barriers in mine dumps, Greenland and Antarctic surface melt, and High Mountain Asia GLOFs all show how phase change reorganizes hydrology, strength, and exposure. In parallel, rockfall barriers, root-reinforced liquefiable soils, groundwater-controlled marine-clay slopes, cataclastic weak rock, karst detection, tunnel digital twins, and tunnel-state PINNs convert difficult field conditions into measurable mechanical states. The strongest methodological thread is not generic AI; it is physically constrained inference from sparse, indirect, or heterogeneous observations.
Key Trends
Five movements define the issue: time-resolved landslide attribution, cryosphere phase-change hazards, hydro-mechanical geotechnics, digital-twin infrastructure monitoring, and remote sensing for missing state variables.
- Landslide assessment is shifting from static maps to time-resolved process attribution: The Italian national procedure links landslide dynamics to rainfall timing and climate stress, DRCD-TL targets rapid co-seismic mapping, the reservoir-slope study follows movement into surge inundation, and debris-flow susceptibility work improves the construction of non-event samples.
- Cryosphere hazards are being treated as coupled phase-change and exposure problems: Ice-bearing deposits, High Mountain Asia GLOFs, red-bed sandstone freeze-thaw damage, mine-dump capillary barriers, and GNSS-R melt retrievals all connect thermal state to slope, flood, and infrastructure risk.
- Geotechnical stability models are moving toward field-constrained hydro-mechanical states: Root-reinforced liquefaction tests, 2D-versus-3D groundwater models for marine-clay slopes, cataclastic weak-rock testing, karst LOTEM inversion, and Benggang susceptibility modelling emphasize mechanisms that can be constrained from site evidence.
- Infrastructure hazard analysis is becoming a digital-twin problem: Rockfall barrier modelling, tunnel lining neural operators, discontinuity-aware PINNs, wind-sand bridge-road simulations, and seismic surrogate demand models translate structural response into fast diagnostic or design tools.
- Remote sensing is being used to recover missing state variables, not only map footprints: Physics-guided flood models, InSAR coherence regression, GNSS-R ice-melt retrievals, InSAR-GNSS subsidence monitoring, and optical correction for wave-contaminated imagery all address states that are hard to observe directly.
Selected Papers
This issue contains 36 selected papers from 1,851 papers analyzed. The selected papers are ordered by relevance score and emphasize climate-sensitive landslide dynamics, co-seismic landslide mapping, reservoir landslide-surge chains, debris-flow susceptibility, freeze-thaw degradation, rockfall mitigation, GLOF risk, coastal subsidence, tunnel diagnostics, and physics-guided remote sensing.
1. Spatio-temporal variability of landslides as indicator of climate change impact: Towards an Italian national scale integrated procedure for the rainfall role analysis in landslide dynamics
Core Problem: National-scale landslide risk assessments often treat rainfall as a static trigger, even though climate change alters both rainfall timing and slope response across regions.
Key Innovation: Builds an Italian national-scale procedure to analyze the rainfall role in landslide dynamics through spatio-temporal variability, linking climate stress to observed ground instabilities.
2. DRCD-TL: a novel co-seismic landslide mapping approach by deep representation-driven unsupervised change detection based on transfer learning
Core Problem: Unsupervised co-seismic landslide mapping still relies heavily on handcrafted shallow change features that break down in complex mountain terrain.
Key Innovation: Introduces a transfer-learning change-detection framework that mines deep representations to highlight landslide information, validated in Mainling and Hokkaido against six baselines.
3. Prediction of movement trajectories and surge disaster-affected areas following instability and failure of a large, high-positioned, loess-mudstone interface landslide
Core Problem: High-position reservoir landslides can trigger surge waves, but rapid simulation of the full instability-to-inundation process remains difficult.
Key Innovation: Applies a two-phase flow framework to the Likan Highway landslide group to predict movement trajectories and surge-affected areas for a potential landslide-surge disaster chain.
4. Enhancing debris flow susceptibility models using negative samples from hazard-bearing body data
Core Problem: Machine-learning debris-flow susceptibility maps are highly sensitive to negative samples, especially where surveyed and unsurveyed catchments are not clearly separated.
Key Innovation: Uses road and building-roof information from hazard-bearing watersheds to construct alternative negative samples and test their effect on susceptibility modelling in Ya'an, China.
5. Phase change-driven structural degradation of ice-bearing glaciofluvial deposits from initial thawing to freeze-thaw cycling
Core Problem: Ice-bearing glaciofluvial deposits in southeastern Xizang are destabilized by warming and engineering activity, but their phase-change degradation remains poorly constrained.
Key Innovation: Examines the transition from initial thawing to freeze-thaw cycling to resolve how ice content reorganizes the soil-ice skeleton and increases landslide, debris-flow, and tunnel-safety risk.
6. Analysis of large-scale guiding flexible rockfall barriers of steel wire-ring nets using an improved truss equivalent method
Core Problem: Large guiding flexible rockfall barriers are difficult to optimize because full-scale wire-ring net models are computationally expensive.
Key Innovation: Develops an improved truss equivalent method with a mechanically informed hardening law, validated against panel and full-scale field tests for barrier response.
7. Static and cyclic liquefaction instability of reinforced soil with plant roots
Core Problem: Root reinforcement is widely invoked for slope stabilization, but static and cyclic instability under extreme rainfall-like shear conditions is still poorly linked.
Key Innovation: Uses constant-shear-stress tests to quantify static instability onset and its relation to cyclic liquefaction behavior in root-reinforced geomaterials.
8. Glacial Lake Outburst Floods in High Mountain Asia: Historical Evidence, Future Changes, and Risk-Reduction Strategies from a Remote-Sensing Perspective
Core Problem: High Mountain Asia GLOF knowledge remains fragmented as glacier retreat changes lake-dam configurations and downstream exposure.
Key Innovation: Synthesizes historical evidence, future change pathways, and risk-reduction strategies from a remote-sensing perspective across High Mountain Asia.
9. Land subsidence rates and their implications for potential land loss in coastal Semarang-Demak
Core Problem: Coastal Semarang-Demak faces persistent subsidence and land-loss risk, but deformation rates and their spatial implications require updated geodetic evidence.
Key Innovation: Combines SBAS-InSAR and GNSS monitoring from Sentinel-1A imagery to quantify subsidence rates and evaluate potential land loss in northern Java.
10. Advanced Flood Prediction with Physics-Guided Deep Learning: Combining UNet, FNO, and SAR/Optical Imagery
Core Problem: Flood mapping must generalize across sparse observations and heterogeneous terrain while preserving hydrodynamic consistency.
Key Innovation: Combines UNet, Fourier Neural Operator components, SAR/optical imagery, DEM features, and shallow-water-equation constraints in a physics-guided flood prediction framework.
11. Influence of index contribution rate and machine learning models on Benggang susceptibility at the slope unit scale
Core Problem: Benggang susceptibility prediction depends on both environmental indicator selection and model choice, yet contribution thresholds are rarely tested systematically.
Key Innovation: Uses slope units, GeoDetector contribution rates, multi-scale segmentation, and machine-learning comparison to evaluate susceptibility in Huichang County.
12. Late Quaternary tectonic activity characteristics of the western Qiulitage thrust fault, southwestern Tianshan, China
Core Problem: Seismic hazard in the southwestern Tianshan depends on poorly constrained Late Quaternary thrust-fault activity and surface rupture history.
Key Innovation: Combines remote-sensing interpretation, field investigation, UAV surveying, and OSL dating to quantify recent fault activity and earthquake rupture evidence.
13. Unraveling soil erosion severity in the endorheic basin of Hodna, Algeria: a spatiotemporal responses framework under LULCC dynamics
Core Problem: Water erosion in endorheic basins changes with land-use and land-cover dynamics, but spatio-temporal hazard responses remain difficult to quantify.
Key Innovation: Builds a spatio-temporal erosion-severity framework under LULCC dynamics for the Hodna basin to support water and soil conservation planning.
14. Tsunami evacuation planning for Camana, Peru: insights from agent-based modeling
Core Problem: Coastal tourist communities in the Ring of Fire need evacuation strategies that account for pedestrian behavior, density, routes, and vertical shelters.
Key Innovation: Uses agent-based modelling with reinforcement learning to evaluate tsunami evacuation scenarios and infrastructure interventions for Camana, Peru.
15. Experimental methods and creep testing study of undisturbed cataclastic weak rock under lateral confinement
Core Problem: Cataclastic weak rock masses in tectonically disturbed mountain regions are hard to sample undisturbed, limiting laboratory characterization for slope and tunnel projects.
Key Innovation: Develops experimental methods and lateral-confinement creep tests for undisturbed cataclastic weak rock to better constrain engineering mechanical behavior.
16. Evolution of capillary barrier effect in cover layers of mine dumps in seasonally frozen regions
Core Problem: Freeze-thaw cycles can degrade capillary barrier cover layers on mine dumps, affecting percolation control and slope stability.
Key Innovation: Examines soil-rock mixture cover layers under seasonally frozen conditions to track how capillary barrier performance evolves through freeze-thaw cycling.
17. Effects of model dimensionality (2D vs. 3D) on simulated groundwater flow conditions in post-glacial marine clay deposits in the context of slope stability analysis
Core Problem: Traditional 2D slope-stability groundwater models may misrepresent flow conditions in meandering post-glacial marine clay settings.
Key Innovation: Compares 2D and 3D hydrogeological simulations to identify when simplified river-as-divide assumptions affect groundwater conditions and factors of safety.
18. Numerical simulation of wind sand flow fields and deposition-erosion characteristics in the bridge-road transition section under different wind angles
Core Problem: Bridge-road transitions in desert railways suffer severe windblown sand deposition and erosion, but wind-angle effects on flow fields are insufficiently resolved.
Key Innovation: Uses numerical simulations to quantify wind-sand flow and deposition-erosion patterns around bridge-road transitions under different wind directions.
19. Physics-Data Bi-directional neural operator for fast forward and inverse analysis of tunnel lining towards practical digital twins
Core Problem: Real-time tunnel digital twins need fast forward and inverse mappings between loads and lining response under non-uniform conditions.
Key Innovation: Develops a physics-data bidirectional neural operator for efficient forward and inverse tunnel lining analysis suitable for practical digital twin updates.
20. Discontinuity-embedded reduced-order PINNs for intelligent diagnosis of longitudinal structural mechanical state in jointed segmental tunnels
Core Problem: Tunnel-state inverse problems must handle stiffness discontinuities at joints that conventional PINNs often smooth away.
Key Innovation: Embeds joint discontinuities in reduced-order physics-informed neural networks to diagnose longitudinal mechanical states in jointed segmental tunnels.
21. 3D LOTEM inversion based on initial-model optimization-a case study of deep karst detection
Core Problem: Deep karst targets occur under complex surface and subsurface interfaces where long-offset transient electromagnetic inversion is sensitive to initial-model choice.
Key Innovation: Uses optimized initial models in 3D LOTEM inversion to improve resistivity interpretation for deep karst detection in engineering investigations.
22. Multiscale mineral-structural-mechanical analyses of red-bed sandstone subjected to freeze-thaw cycles
Core Problem: High-altitude red-bed sandstone can weaken under freeze-thaw cycling, but links between mineral change, pore structure, and mechanical damage are still incomplete.
Key Innovation: Combines multiscale mineral, structural, and mechanical analyses to resolve freeze-thaw damage in northeastern Qinghai-Tibet Plateau sandstone.
23. Geology-guided machine learning models for geotechnical characterisation of Bangkok subsoil with applications to pile foundations
Core Problem: Urban geotechnical design is constrained by spatially variable subsoil conditions and sparse borehole data.
Key Innovation: Combines geological guidance with machine-learning models to characterize Bangkok subsoil and predict parameters for pile-foundation applications.
24. A Multidecadal Catalog of Normal-Faulting Earthquakes Across the Tibetan Plateau Derived From InSAR and Body-Waveform Modeling: Insights Into the Plateau's Extensional Tectonics and Dynamics
Core Problem: Extensional seismicity across the Tibetan Plateau is incompletely resolved, limiting understanding of regional fault geometry and seismic strain release.
Key Innovation: Compiles 40 Mw >= 5 normal-faulting events since 1976 using InSAR and body-waveform modelling to quantify rupture geometry and extension partitioning.
25. Beyond Backscatter: InSAR coherence from detected SAR images
Core Problem: Coherence is valuable for deformation and damage analysis but normally requires precisely coregistered SAR complex data.
Key Innovation: Trains a Residual U-Net to regress coherence from detected Sentinel-1 SAR images, potentially lowering the data barrier for change and instability monitoring.
26. Antarctic surface melt mapping using Tianmu-1 GNSS-R observations
Core Problem: Antarctic surface melt is a key cryospheric state variable, but existing microwave approaches have coarse spatial resolution or limited sensitivity.
Key Innovation: Uses Tianmu-1 GNSS-R observations with cross-calibrated multi-satellite data and an empirical algorithm to retrieve Antarctic surface melt maps.
27. Measuring Greenland Ice Sheet Melt Based on GNSS-R and Random Forest Using FengYun-3E GNOS-R
Core Problem: Greenland melt monitoring requires higher spatial detail than many microwave products provide under accelerating climate warming.
Key Innovation: Combines FengYun-3E GNOS-R GNSS-R observations with Random Forest modelling to detect Greenland Ice Sheet melt state.
28. Explainable physics-informed Bayesian graph neural networks (PI-BSTGNN) for groundwater surrogate modeling and sensitivity analysis
Core Problem: Groundwater hazard modelling needs fast surrogates that retain physics and quantify sensitivity under stochastic hydrogeological conditions.
Key Innovation: Uses MODFLOW6 realizations to train a physics-informed Bayesian spatio-temporal graph neural network for explainable groundwater surrogate modelling.
29. Wave modulation of flood-driven jet dynamics, suspended-sediment dispersion and event-scale morphodynamics in the Modaomen Estuary
Core Problem: Flood jets in wave-influenced estuaries control sediment dispersion and event-scale erosion-deposition, but wave modulation remains hard to quantify.
Key Innovation: Uses a coupled TELEMAC-MASCARET wave-current-sediment model to resolve how waves alter flood-driven jets and suspended-sediment redistribution.
30. Role of Unsaturated Zone on Aquifer Leakage-Assessed With Tidal Response
Core Problem: Traditional tidal-response analyses can underestimate groundwater leakage by neglecting unsaturated-zone effects.
Key Innovation: Develops an analytical tidal-response model that incorporates unsaturated-zone behavior and capillary effects to better resolve aquifer leakage.
31. Numerical investigation on hydrodynamic and wave attenuation performance of a stepped-type floating breakwater
Core Problem: Floating breakwaters offer flexible coastal protection, but conventional forms perform poorly against long-period waves.
Key Innovation: Uses two-dimensional RANS simulations to evaluate a stepped trapezoidal floating breakwater and its wave attenuation performance.
32. ARMOR: Adaptive meshing with reinforcement optimization of implicit fields for real-time 3D monitoring in unexposed scenes
Core Problem: Mines, tunnels, and lava tubes require real-time 3D mapping, but unexposed scenes challenge conventional meshing and structural assessment workflows.
Key Innovation: Introduces adaptive meshing with reinforcement optimization of implicit fields for real-time 3D monitoring in complex unexposed environments.
33. Correcting Wave-Induced Reflectance Anomalies in Pan-Sharpened Very High-Resolution Satellite Data
Core Problem: Sub-meter aquatic satellite data are vulnerable to wave glint and whitecap contamination, reducing reliability for bathymetry and habitat mapping.
Key Innovation: Develops a correction approach for wave-induced reflectance anomalies in pan-sharpened very high-resolution imagery.
34. Cave system evolution in dedolomite (central Slovenia)
Core Problem: Dedolomite cave systems can precondition karst landscapes, but their structural, sedimentary, and chronological development remains difficult to reconstruct.
Key Innovation: Combines geomorphological mapping, structural analysis, LiDAR morphometry, sedimentology, U-Th dating, and OSL geochronology to reconstruct cave-system evolution.
35. Physics-informed deep surrogate demand modeling for framed core-tube structures under seismic loadings
Core Problem: Nonlinear time-history analysis for high-rise seismic demand is too expensive for repeated performance evaluation.
Key Innovation: Develops a physics-informed deep surrogate to rapidly predict multi-component nonlinear seismic demands for framed core-tube structures.
36. Seismic response and influencing factors of rectangular groove aqueduct with pile foundations in seasonally frozen regions
Core Problem: Aqueducts with pile foundations in seasonally frozen regions experience coupled fluid-structure, pile-soil, frozen-soil, and topographic effects during earthquakes.
Key Innovation: Enhances spring-mass fluid-structure interaction modelling and finite-element analysis to quantify frozen-soil and topographic controls on aqueduct seismic response.