TerraMosaic Daily Digest: June 26, 2026
Daily Summary
The June 26 literature is unusually concentrated around landslide intelligence as an end-to-end system: records, triggers, motion, and response. A UNDRR-linked paper on landslide inventory and loss-and-damage reporting attacks a persistent blind spot in disaster databases, while the Turkiye and Hindu Kush studies treat landslides as part of cascading multi-hazard systems rather than isolated slope failures. The strongest technical papers then move from inventory to prediction: differentiable landslide simulation, Riemann-based SPH for rainfall-induced slope failure, probabilistic rainfall-debris-flow forecasting, dynamic early warning, and time-series rainfall regression all aim to preserve the forcing-to-motion chain.
Process mechanics and operational mapping reinforce each other. Several papers target the transition from rainfall, thaw, seismic shaking, or erosion into movement: aspect-controlled shallow landslides, retrogressive thaw-slump activation windows, sensitive-clay progressive failure, riverbank collapse, postfire debris-flow vulnerability, rockfall statistics, and GLOF hydraulics. The methods are diverse, but the shared contribution is sharper state representation: pore pressure, deformation, entrainment, thermal-hydrologic windows, GNSS-network dependencies, and non-stationary strength are treated as observable or inferable variables rather than hidden background conditions.
The remote-sensing and AI layer is also more applied than ornamental. LandslideSegFormer, ASA-DETR, UAV YOLO detection, pixel-based earthquake-triggered landslide mapping, multi-agent rapid response, thaw-slump mapping, InSAR-based mining hazard prediction, levee-piping time-series analysis, and Prithvi robustness testing all ask whether models can survive scale, domain shift, weak labels, sparse observations, and operational reporting. The day therefore reads less like a list of algorithms and more like a blueprint for geohazard systems that connect detection, mechanics, uncertainty, exposure, and accountable records.
Key Trends
Five movements define this issue: record-to-impact integration, physically constrained landslide forecasting, activation thresholds in cold and mountain terrain, remote sensing tied to response workflows, and deformation-memory approaches to infrastructure hazard.
- Landslide science is tightening the record-to-impact chain: Inventory mapping, loss-and-damage reporting, multi-hazard Turkiye scenarios, and Hindu Kush risk assessment all expose the same gap: hazard products matter most when they connect physical events to exposure, vulnerability, and accountable consequences.
- Forecasting models are becoming physically constrained: Differentiable landslide simulation, Riemann-SPH slope failure, probabilistic debris-flow forecasting, functional rainfall regression, and dynamic warning systems all move beyond black-box labels toward time-dependent forcing, material state, and uncertainty.
- Cold-region and high-mountain hazards are treated as activation problems: Retrogressive thaw-slump windows, hemispheric thaw-slump mapping, GLOF reconstruction, freeze-thaw-seepage rock degradation, and Himalayan multi-hazard assessment frame hazard as the timing of thermal, hydrologic, and geomorphic thresholds.
- Remote sensing is being coupled to decision workflows: LandslideSegFormer, ASA-DETR, UAV YOLO deposits, MAS-LAND, InSAR riverbank collapse, UATEM landslide imaging, and mining-zone deformation models convert imagery or deformation data into response-facing information.
- Infrastructure hazards are being read through deformation memory: Levee piping, earthen-dam breach evolution, mining subsidence, liquefaction recovery, slope-tunnel coupling, and GB-RAR early warning show that slow or transient deformation histories can be as important as instantaneous loads.
Selected Papers
The selected papers span landslide loss-and-damage reporting, earthquake-triggered cascading hazards, rainfall-induced landslide motion, SPH slope-failure simulation, debris-flow forecasting and entrainment, dynamic landslide early warning, landslide segmentation, multi-agent rapid response, shallow-landslide controls, thaw-slump activation, riverbank collapse, GLOF reconstruction, levee piping, earthen-dam breach evolution, liquefaction recovery, mining deformation, rock-failure warning, and geospatial foundation-model robustness. This issue contains 46 selected papers from 1829 analyzed.
1. Multi-hazards assessment in Türkiye following large earthquakes
Core Problem: Future risk in southeastern Turkiye cannot be represented by shaking alone because the 2023 earthquake sequence linked landslides, liquefaction, infrastructure disruption, and flood amplification.
Key Innovation: Builds a scenario-based multi-hazard framework that combines deterministic earthquake ruptures, co-seismic landslide susceptibility, InSAR deformation, flood simulation, and climate-adjusted rainfall over a 50-year horizon.
2. How to better link landslide inventory mapping with loss and damage reporting
Core Problem: Landslide impacts are systematically undercounted in global disaster databases because small, frequent events often fall below reporting thresholds and inventories are poorly linked to loss records.
Key Innovation: Develops a maturity framework for connecting landslide inventory mapping with national loss-and-damage reporting, based on literature review, expert questionnaires, workshops, and UNDRR/EW4ALL needs.
3. Physics-embedded differentiable simulation of rainfall-induced landslide motion
Core Problem: Slow-moving landslides are difficult to forecast from sparse satellite and rainfall observations without drifting into physically unrealistic model states.
Key Innovation: Embeds hydrological diffusion and viscoplastic flow in a differentiable simulator, allowing automatic parameter inversion and physically constrained prediction for California Coast Range landslides.
4. A two-phase Riemann-based SPH model for large deformation and rainfall-induced slope failure
Core Problem: Hydro-mechanical coupling in unsaturated slopes becomes numerically fragile once deformation turns into fluid-like runout.
Key Innovation: Adds a low-dissipation Riemann solver, controlled dissipation, and unified transport velocity to a two-phase SPH framework, validating it against soil-column tests, rainfall flumes, and the Fei Tsui Road landslide.
5. Integrated multi-hazard vulnerability and risk assessment in the Hindu Kush Mountains of Northern Pakistan: Challenges and recommendations
Core Problem: The Hindu Kush combines earthquakes, floods, landslides, debris flows, and GLOFs, yet risk studies often treat these hazards separately.
Key Innovation: Integrates hazard-specific assessments, open remote-sensing data, field observations, GIS, and AHP-based multi-criteria weighting to map composite vulnerability and risk in Chitral, northern Pakistan.
6. A probabilistic modeling framework for data-driven forecasting of rainfall-induced debris flows
Core Problem: Rainfall-threshold debris-flow warning systems are limited by sparse monitoring data, false alarms, and binary classifications.
Key Innovation: Combines Copula data expansion, mixture-model grouping of triggering mechanisms, and Bayesian rainfall thresholds to produce probabilistic debris-flow forecasts.
7. A Variable-Mass Shallow Water SPH framework for debris flow simulation with erosion and entrainment
Core Problem: Depth-integrated SPH models are efficient for large debris flows but commonly treat flow mass as fixed, which conflicts with erosion and entrainment physics.
Key Innovation: Introduces a variable-mass shallow-water SPH framework with physically based entrainment, particle-grid coupling, adaptive refinement, and mass-conserving bed-flow exchange.
8. Displacement prediction model for rainfall-induced landslides based on PDL-VMD-SVR: A case study of the southern slope landslide at Fushun West Open-pit Mine, China
Core Problem: Rainfall-landslide displacement relationships are lagged and nonlinear, making static rainfall-displacement models unreliable for warning.
Key Innovation: Combines polynomial distributed lag modeling, variational mode decomposition, and support vector regression to represent rainfall hysteresis and improve Fushun mine-slope displacement prediction.
9. Dynamic landslide early warning based on frequency-domain enhanced deep learning
Core Problem: Warning models often miss the joint influence of displacement trends, seasonal cycles, reservoir level, rainfall, and slope structure.
Key Innovation: Introduces DPFEB-Net with frequency-domain enhancement, environmental drivers, geological variables, and MCMC uncertainty bounds for dynamic early warning in the Three Gorges Reservoir area.
10. LandslideSegFormer: A non-local context-aware dual-branch network for accurate small-scale landslide segmentation from remote sensing imagery
Core Problem: Transformer landslide segmentation often misses small failures and loses boundary detail.
Key Innovation: Redesigns SegFormer into LandslideSegFormer with transformer-convolution fusion, non-local feature aggregation, and a gated hybrid-attention decoder for small-scale landslide mapping.
11. MAS-LAND: A multi-agent system for landslide detection and rapid response
Core Problem: Post-event landslide mapping is rarely connected to infrastructure exposure and operational reporting.
Key Innovation: Builds MAS-LAND, an LLM-enhanced multi-agent workflow that couples landslide segmentation, building and road exposure assessment, and automated civil-protection reporting.
12. Tectonic and microtopographic controls on aspect-dependent distribution of rainfall-induced shallow landslides
Core Problem: Aspect-dependent shallow landslide initiation is often attributed to vegetation or soil hydraulics, but tectonic and microtopographic controls can dominate.
Key Innovation: Analyzes more than 700 rainstorm-triggered landslides in eastern China, linking bimodal aspect distribution to fault orientation, slope form, weathering, and stability modeling.
13. Weekly thermal-hydrologic activation windows of retrogressive thaw slumps: Implications for early warning of slope instability in degrading permafrost regions
Core Problem: The weekly timing of thaw-slump activation is poorly constrained despite accelerating permafrost degradation.
Key Innovation: Develops a thermal-hydrologic activation-window framework using air temperature, ground-surface temperature, soil moisture, snow-memory indicators, ROC thresholds, and bootstrap uncertainty.
14. Domain knowledge constrained dynamic graph learning of GNSS sensor networks for spatio-temporal slope deformation monitoring
Core Problem: GNSS slope-monitoring networks contain physical dependencies that generic temporal models often ignore.
Key Innovation: Uses domain-knowledge-constrained dynamic graph learning to model spatiotemporal slope deformation, embedding sensor-network structure and rainfall-related controls into deformation prediction.
15. Hemispheric-scale mapping of thaw slumps using a cloud-native and transferable deep learning framework
Core Problem: Retrogressive thaw slumps are expanding across permafrost regions, but consistent hemispheric inventories remain difficult.
Key Innovation: Creates a cloud-native transferable deep-learning framework for hemispheric-scale thaw-slump mapping, targeting cross-region generalization in Arctic and high-mountain terrain.
16. GeoHazards, Vol. 7, Pages 66: Toward an Automatic Pixel-Based Detection of Earthquake-Triggered Landslides in Arid Environments Using Optical Imagery
Core Problem: Earthquake-triggered landslides in arid terrain can be spectrally subtle and spatially fragmented in optical imagery.
Key Innovation: Tests automatic pixel-based optical-image detection for earthquake-triggered landslides in arid environments, addressing rapid post-seismic mapping where vegetation cues are weak.
17. Effect of a hybrid baffle-flexible barrier system for debris flow mitigation: A real-scale investigation in a natural gully
Core Problem: Rigid baffles and flexible barriers are commonly studied separately, but field mitigation often requires coupled energy dissipation.
Key Innovation: Uses a real-scale natural-gully investigation to test a hybrid baffle-flexible barrier system and evaluate debris-flow impact reduction under field conditions.
18. Seismically induced rockfall modelling using a 3D integrated approach: Ischia Island (Italy) case study
Core Problem: Volcanic islands with steep fractured slopes require rockfall simulations that preserve three-dimensional topographic and source-zone complexity.
Key Innovation: Applies an integrated 3D approach to model earthquake-induced rockfall on Ischia Island, connecting seismic triggering, block motion, and runout patterns.
19. 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 landslide risk frameworks need a clearer procedure for separating rainfall as predisposing versus triggering forcing under climate change.
Key Innovation: Reviews Italian landslide variability and develops a national-scale integrated procedure for analyzing the rainfall role in landslide dynamics.
20. Dynamic risk assessment of riverbank collapse: A cascading framework integrating explainable AI and InSAR
Core Problem: Riverbank collapse evolves through interacting deformation, hydrology, and exposure, making static susceptibility maps inadequate.
Key Innovation: Builds a cascading dynamic-risk framework that integrates InSAR deformation and explainable AI to predict riverbank-collapse risk.
21. A multi-hazard framework to assess co-seismic ground instabilities scenarios: Application to a virtual test bed
Core Problem: Earthquakes can trigger several ground-instability modes, but scenario tools often handle landslides, liquefaction, and related failures separately.
Key Innovation: Develops a multi-hazard virtual-testbed framework for co-seismic ground instabilities, supporting scenario-based comparison of spatially compounding earthquake effects.
22. On the use of rainfall time series for regional landslide prediction by means of functional regression
Core Problem: Standard rainfall-landslide models compress precipitation time series into scalar thresholds, discarding temporal information.
Key Innovation: Uses functional generalized additive models to preserve full rainfall time series while combining them with terrain predictors for regional landslide prediction.
23. Probabilistic modeling of failure mechanisms of sensitive clay slopes with non-stationary heterogeneity
Core Problem: Sensitive clay landslides involve progressive failure and large runout, yet non-stationary shear-strength heterogeneity is rarely included probabilistically.
Key Innovation: Combines CPT-informed Bayesian updating with GPU-accelerated GIMP large-deformation modeling to simulate Saint-Jude-type sensitive-clay slope failure.
24. Rapid investigation of a complex-terrain landslide using the unmanned aerial transient electromagnetic method: Validation against electrical resistivity tomography
Core Problem: Slip-surface mapping in steep complex terrain is slow and hazardous when based only on contact geophysics.
Key Innovation: Applies unmanned aerial transient electromagnetics to the Huadian landslide and validates interpretations against ERT, field observations, and meteorological context.
25. Multi-source heterogeneous feature fusion framework for identifying retrogressive thaw slumps on the Qinghai-Tibet plateau
Core Problem: Thaw slumps have heterogeneous spectral, thermal, and geomorphic signatures that challenge single-source mapping.
Key Innovation: Fuses multi-source heterogeneous features to identify retrogressive thaw slumps on the Qinghai-Tibet Plateau, improving permafrost-hazard detection.
26. Hydraulic reconstruction and geomorphological impacts of surge-induced GLOFs in the Upper Indus Basin
Core Problem: Surge-induced glacial lake outburst floods in the Upper Indus Basin require reconstruction of both flood hydraulics and downstream geomorphic change.
Key Innovation: Combines hydraulic reconstruction with geomorphic impact analysis to diagnose GLOF routing, erosion, and landscape response in a high-mountain basin.
27. ASA-DETR: Adaptive sparse attention enhanced RT-DETR for remote sensing landslide detection with multi-scale frequency-aware feature learning
Core Problem: Operational landslide mapping needs detectors that remain sensitive to scale and frequency content in complex mountain imagery.
Key Innovation: Introduces ASA-DETR, an adaptive sparse-attention RT-DETR variant with multi-scale frequency-aware learning for remote-sensing landslide detection.
28. An improved YOLO-based model for detecting small landslide deposits along mountainous rural roads via UAV imagery
Core Problem: Small landslide deposits along mountain roads are easily missed in UAV imagery but matter for rural-road exposure.
Key Innovation: Improves YOLO-based detection for small landslide deposits along mountainous roads using UAV imagery and task-specific model adaptation.
29. Assessment of the vulnerability of buildings destroyed during postfire debris flow events in Kule village, Yajiang County, China
Core Problem: Postfire debris flows damage buildings in ways that depend on flow depth, impact, and local vulnerability, yet empirical damage models remain sparse.
Key Innovation: Assesses building vulnerability in Kule village after postfire debris flows and derives damage-probability relationships for debris-flow impact assessment.
30. GeoHazards, Vol. 7, Pages 68: Geotechnical Characterization, Risk Analysis, and Design of Stabilization Measures for a Landslide Along the RN16 Coastal Highway in Morocco: A Case Study at KP 178+000
Core Problem: A coastal highway landslide requires geotechnical characterization that connects site conditions, risk, and engineering stabilization.
Key Innovation: Combines geotechnical characterization, risk analysis, and stabilization design for the RN16 coastal highway landslide in Morocco.
31. GeoHazards, Vol. 7, Pages 69: Rockfall Volume-Cumulative Frequency Relationships for Rockfall Hazard Quantification Using Historical and Change Detection Data
Core Problem: Rockfall hazard estimates depend on frequency-volume relationships that are often poorly constrained by short inventories.
Key Innovation: Uses historical records and change-detection data to build rockfall volume-cumulative frequency relationships for hazard quantification.
32. Remote Sensing, Vol. 18, Pages 1999: Attention-Driven Hierarchical Spatial Adaptive Ensemble for Landslide Susceptibility Mapping
Core Problem: Susceptibility models can miss spatially variable controls and local feature interactions in heterogeneous terrain.
Key Innovation: Introduces an attention-driven hierarchical spatial adaptive ensemble for landslide susceptibility mapping.
33. Multi-scale time-series framework for piping-induced levee failure using a piping potential index
Core Problem: Levee piping develops through multi-scale hydrologic memory rather than instantaneous river-stage forcing alone.
Key Innovation: Defines a piping potential index from river level, groundwater, rainfall, and SMAP soil moisture over multiple accumulation windows to anticipate levee failures.
34. Mechanisms of seepage and overtopping-induced failure in earthen dams: A coupled two-layer SPH investigation
Core Problem: Overtopping simulations of earthen dams can be unrealistic when initial seepage fields are not physically established.
Key Innovation: Uses a coupled two-layer SPH framework with a two-stage seepage-to-overtopping strategy to simulate phreatic-line development, erosion, and breach evolution.
35. In-situ evidence of soil recovery after earthquake sequences redefines liquefaction hazard evaluation paradigms
Core Problem: Post-earthquake CPT investigations may capture transient soil recovery rather than the state that controlled liquefaction during shaking.
Key Innovation: Analyzes Canterbury earthquake records and more than 4600 CPTs to show time-dependent strengthening and fabric effects that complicate reliquefaction assessment.
36. Three-dimensional coupled discontinuous smoothed particle hydrodynamics (3CDSPH) method for simulating rockslide-generated surge waves considering rock damage-contact and fluid-solid interaction
Core Problem: Rockslide-tsunami simulation requires coupled damage-contact and fluid-solid interaction rather than purely kinematic slide forcing.
Key Innovation: Develops a 3D coupled discontinuous SPH method for simulating rockslide-generated surge waves with rock damage, contact, and water interaction.
37. An efficient non-hydrostatic numerical model for subaerial granular landslide-generated waves
Core Problem: Subaerial granular landslides can generate dispersive waves that are expensive to model in full 3D.
Key Innovation: Presents SH-NH2L, coupling Savage-Hutter granular flow with a two-layer non-hydrostatic water model to reproduce landslide-generated waves at lower computational cost.
38. Hazard prediction of backfill and dumping zones of open-pit mining areas using InSAR and multi-factor ensemble modeling
Core Problem: Open-pit backfill and dumping zones combine consolidation settlement with possible failure, requiring deformation-aware hazard prediction.
Key Innovation: Fuses DEM reconstruction, multi-track 3D InSAR, nonlinear deformation metrics, and ensemble learning to predict hazards in backfill and dumping zones.
39. Remote Sensing, Vol. 18, Pages 2028: A Variational Data Assimilation Framework for Mining Subsidence Reconstruction from Heterogeneous D-InSAR and TLS Observations
Core Problem: Mining subsidence observations from D-InSAR and terrestrial laser scanning differ in geometry, precision, and sampling.
Key Innovation: Develops a variational data-assimilation framework to reconstruct mining subsidence from heterogeneous D-InSAR and TLS observations.
40. Predicting the Failure of Rock Using Critical Slowing Down Theory on Infrared Radiation Temperature Characteristics: Insight from Noise Correction
Core Problem: Catastrophic rock failure lacks robust precursor indicators that can tolerate noisy infrared measurements.
Key Innovation: Combines infrared radiation temperature, acoustic emission, denoised crack-evolution thermograms, and critical-slowing-down theory to detect rock-failure precursors.
41. Study on Deformation and Failure of Interlayered Rock Masses Under Microscale Freeze-Thaw-Seepage Coupled Effects
Core Problem: Cold-region rock masses degrade through coupled freeze-thaw cracking and seepage, weakening slopes and underground structures.
Key Innovation: Builds a particle-flow freeze-thaw-seepage model to quantify crack evolution, contact-force change, and strength loss under cycles, seepage rates, and interlayer dip angles.
42. Improving peak ground acceleration prediction and seismic intensity estimation using ensemble learning for earthquake early warning in the Taiwan region
Core Problem: Rapid earthquake warning needs accurate PGA and intensity estimates that generalize across Taiwan source and site conditions.
Key Innovation: Uses ensemble learning to improve PGA prediction and seismic-intensity estimation for earthquake early warning in Taiwan.
43. Near-fault ground motion effects and casualty patterns revealed by InSAR deformation and multi-factor risk coupling: A case study of the 2023 Jishishan earthquake
Core Problem: Near-fault deformation and casualty patterns require coupling observed ground deformation with exposure and vulnerability factors.
Key Innovation: Combines InSAR deformation and multi-factor risk coupling to analyze near-fault ground-motion effects in the 2023 Jishishan earthquake.
44. Remote Sensing, Vol. 18, Pages 2056: A GB-RAR Deformation Early Warning Method Based on a Hybrid Algorithm for Optimizing Prediction Models
Core Problem: Ground-based radar deformation monitoring needs robust prediction-model optimization for warning under noisy time-series conditions.
Key Innovation: Develops a hybrid-algorithm approach for optimizing GB-RAR deformation early-warning models.
45. Dynamic response of high-steep rock slope-tunnel system with cataclastic rock belt under seismic-rainfall coupling: Shaking table test
Core Problem: High-steep rock slopes with tunnels can respond differently when seismic loading is combined with rainfall weakening.
Key Innovation: Uses shaking-table testing to study dynamic response of a high-steep rock slope-tunnel system with a cataclastic rock belt under seismic-rainfall coupling.
46. Assessing the Robustness of Prithvi Geospatial Foundation Model for Coastal Habitat Mapping Under Data Availability and Domain Shift Scenarios
Core Problem: Geospatial foundation models must be tested under data scarcity and domain shift before being trusted for environmental monitoring.
Key Innovation: Benchmarks the Prithvi geospatial foundation model against supervised segmentation for coastal habitat mapping across sample-size and domain-shift scenarios.