TerraMosaic Daily Digest: June 3, 2026
Daily Summary
The June 3 literature is organized around temporal state, not static hazard class. Landslide lakes are treated as pulsing systems controlled by hydro-climatology and landslide-toe morphology; alpine rock-slope instability is tracked with long-wavelength SAR to recover motion where C-band products decorrelate; and regional landslide prediction is recast from rainfall-window summaries to functional regression on continuous precipitation signals. The remote-sensing papers extend this temporal logic to annual landslide inventories, interpretable multi-temporal susceptibility, seasonally partitioned InSAR, and spatio-temporal slope-deformation graphs.
The surrounding geohazard papers sharpen the mechanics that connect observation to failure. Permafrost thaw slumps are mapped with heterogeneous sensor fusion; debris in dam-break waves is tracked as a moving impact load; liquefiable soils are shown to recover in ways that complicate post-earthquake hazard evaluation; and induced or near-fault seismic risk is coupled to deformation gradients, poroelastic stress, and exposure. Subsurface and infrastructure studies converge on coupled inversion and monitoring: tunnel water-bearing structures, rock-mass discontinuities, mine-roof water inrush, mountain-tunnel deformation, and moisture-weakened rock all require models that carry uncertainty through geology, hydrology, and mechanical response.
Key Trends
Five movements define the issue: temporal landslide state modelling, alpine InSAR monitoring, physics-to-consequence multi-hazard analysis, monitored subsurface inversion, and remote-sensing AI for incomplete data.
- Landslide analysis is becoming explicitly time-dependent: The landslide-lake, rainfall-functional-regression, annual Nujiang mapping, Tree-SHAP susceptibility, and dual-modal slope-deformation papers all replace one-time maps with histories of forcing, motion, and state change.
- Long-wavelength and seasonal InSAR are moving into alpine geohazard monitoring: L-band SAR and seasonally partitioned InSAR are used to recover deformation in vegetated, snow-affected, and high-relief terrain where conventional shorter-wavelength monitoring is fragile.
- Multi-hazard studies are coupling ground physics to consequences: Near-fault casualty modelling, liquefaction recovery, coastal urban multi-risk assessment, induced-fault rupture, and volcanic vent-opening work connect deformation or triggering mechanics with exposure and operational decision variables.
- Subsurface hazard models are shifting from deterministic geometry to monitored inversion: Tunnel, mining, and rock-mass papers use level-set inversion, data assimilation, microseismic response, temporal convolution, and coupled seepage-rheology tests to infer structures that cannot be directly observed.
- Remote-sensing AI is being rebuilt for incomplete, weak, and cross-modal data: Geospatial foundation-model benchmarking, missing-modality segmentation, shadow-aware image restoration, diffusion-based scene generation, and sparse-label Segment Anything adaptation address the failure modes that limit disaster mapping at scale.
Selected Papers
This issue contains 44 selected papers from 1,833 papers analyzed. The selected papers are ordered by relevance score and emphasize time-dependent landslide state, alpine deformation monitoring, hydroclimatic and seismic triggering, subsurface inversion, and remote-sensing methods with clear transfer value for hazard assessment.
1. Partial landslide occlusion of a valley river and the hydro-climatological drivers of landslide-lake ephemerality
Core Problem: Landslide lakes are often categorized as stable impoundments or unstable dams, leaving poorly constrained intermediate cases that repeatedly appear and drain over operationally relevant timescales.
Key Innovation: Uses remote-sensing chronology and hydrological classification at Te Horo, New Zealand, to define pulsing landslide-lake behavior driven by high-flow events, antecedent rainfall, and landslide-toe morphology.
2. Monitoring the displacement of large alpine rock slope instabilities with L-band SAR interferometric techniques
Core Problem: Large rock-slope instabilities can evolve from slow creep to catastrophic collapse, but vegetated and high-relief alpine terrain often causes decorrelation in conventional C-band InSAR monitoring.
Key Innovation: Demonstrates the value of L-band SAR interferometry for longer-term, area-wide displacement retrieval in alpine slope-instability settings where shorter wavelengths lose coherence.
3. Landslide Mapping and Susceptibility Assessment in the Middle and Lower Reaches of the Nujiang River (2017-2025) Using Satellite Embedding and Multidimensional Environmental Factors
Core Problem: Rapid terrain change and fragile geomorphic conditions along the Nujiang River make annual landslide mapping and susceptibility updating difficult with conventional inventory workflows.
Key Innovation: Builds a multi-source Earth-observation workflow using satellite embedding-change intensity and environmental conditioning factors to construct annual landslide maps and 10 m susceptibility estimates.
4. On the use of rainfall time series for regional landslide prediction by means of functional regression
Core Problem: Regional landslide prediction usually compresses rainfall histories into fixed cumulative windows, discarding temporal structure that may control slope response.
Key Innovation: Tests functional generalized additive models that use continuous rainfall time series directly, shifting prediction from scalar thresholds toward time-series signal modelling.
5. Interpretable Multi-Temporal Landslide Susceptibility Assessment Using Random Forest and Tree-SHAP in the Eastern Himalayan Syntaxis
Core Problem: Long-term landslide susceptibility evolution and the changing importance of conditioning factors remain poorly constrained in the deeply incised Eastern Himalayan Syntaxis.
Key Innovation: Compiles a 30-year, 1350-landslide inventory, builds period-specific Random Forest susceptibility models, and uses Tree-SHAP to diagnose temporal shifts in controlling factors.
6. Multi-source heterogeneous feature fusion framework for identifying retrogressive thaw slumps on the Qinghai-Tibet plateau
Core Problem: Retrogressive thaw slumps have subtle spectral signatures and complex morphology, making consistent detection difficult across permafrost landscapes.
Key Innovation: Integrates topographic, environmental, spectral, thermal, and segmentation-model information in a heterogeneous feature-fusion framework for RTS inventory construction on the Qinghai-Tibet Plateau.
7. Slope deformation prediction via geometric-temporal dual-modal fusion: An interpretable spatio-temporal model
Core Problem: Slope-deformation forecasting models often rely on single graph structures and provide limited explanation of which spatial nodes and temporal features drive predictions.
Key Innovation: Introduces geometric-temporal dual-modal fusion for graph convolutional prediction and pairs it with local interpretability and spatial smoothing constraints.
8. The Effect of Hydraulic Fracturing on the 2021 Ms 6.0 Luxian Earthquake as Revealed by Deformation and Numerical Simulations
Core Problem: The contribution of hydraulic fracturing to moderate earthquakes remains difficult to separate from background tectonic loading and pre-existing fault conditions.
Key Innovation: Uses deformation observations and numerical simulation to evaluate how hydraulic fracturing may have affected the 2021 Luxian earthquake sequence.
9. In-situ evidence of soil recovery after earthquake sequences redefines liquefaction hazard evaluation paradigms
Core Problem: Post-earthquake liquefaction assessments rely on in-situ tests, but the measured soil state may evolve during recovery after earthquake sequences.
Key Innovation: Analyzes strong-motion data and more than 4600 CPT records from Canterbury to show time-dependent soil strengthening and fabric evolution that complicate reliquefaction evaluation.
10. A deep learning-based algorithm for spatiotemporal monitoring of debris in dam-break waves
Core Problem: Mobilized debris in dam-break, tsunami, and flash-flood waves creates moving impact loads that are difficult to detect and track when object counts and velocities are high.
Key Innovation: Develops a hybrid deep-learning debris identification and tracking algorithm for spatiotemporal monitoring under extreme-flow conditions.
11. 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: Casualty estimation near active faults can miss deformation-gradient and hanging-wall effects when based only on intensity or building proxies.
Key Innovation: Constructs a comprehensive risk index coupling InSAR deformation, ground-motion parameters, surface susceptibility, fault distance, and structural lethality for the Jishishan earthquake.
12. Multi-hazard risk assessment in coastal urban settlements: application to the RETURNVILLE virtual testbed
Core Problem: Dense coastal settlements face interacting natural hazards, but single-hazard assessments miss compound and cascading risk patterns.
Key Innovation: Uses the RETURNVILLE virtual testbed to evaluate spatially heterogeneous multi-hazard risk under coupled coastal urban exposure conditions.
13. Geospatial Foundation Models to Enable Progress on Sustainable Development Goals
Core Problem: Geospatial foundation models are proliferating, but their practical value across sustainability and environmental hazard tasks remains weakly benchmarked.
Key Innovation: Introduces SustainFM, a benchmark spanning the Sustainable Development Goals and including environmental hazard detection, to evaluate when foundation models outperform conventional approaches.
14. Early Detection of Geohazards in Alpine Regions Using Seasonally Partitioned InSAR: A Case Study of the Eastern Himalayan Syntaxis
Core Problem: Seasonal surface change in alpine regions reduces InSAR coherence over long intervals, masking early deformation signals from unstable slopes and related geohazards.
Key Innovation: Partitions Sentinel-1 interferograms by seasonal coherence migration and screens transition-season pairs to improve early geohazard detection in the Eastern Himalayan Syntaxis.
15. Multi-Model Machine Learning Mapping of Gully Erosion Susceptibility in the Heihe Region of the Xiaoxingan Mountains, China
Core Problem: Gully erosion in Northeast China's Mollisol belt threatens a major grain-producing region, but susceptibility models rarely test human-footprint indicators with spatially explicit validation.
Key Innovation: Combines a 4020-gully inventory with multi-model machine learning, cross-validation, and anthropogenic conditioning factors to map susceptibility in the Heihe region.
16. Experimental study on bio-grouting in metal tailings: injectability, diffusion heterogeneity and strength-porosity relationship
Core Problem: MICP-based tailings reinforcement is promising, but injectability, heterogeneous cementation, and the strength-porosity relationship remain poorly quantified at engineering scale.
Key Innovation: Uses a large-scale pressurized column model to test bio-grouting transport, spatial cementation heterogeneity, and mechanical improvement in metal tailings.
17. Dynamic response of offshore wind turbine supported by suction bucket in sands under multi-hazard loads considering the scour-hole dimensions
Core Problem: Offshore wind turbine foundations are affected by simultaneous wind, wave, seismic, and scour loading, but combined dynamic failure mechanisms remain insufficiently characterized.
Key Innovation: Uses 3D numerical analyses to evaluate pore pressure, displacement, and acceleration responses of suction-bucket foundations under coupled environmental and seismic loads.
18. Moisture-Dependent Evolution of Creep and Permeability in Argillaceous Siltstone During Shear-Compression Transition
Core Problem: Long-term slope and tunnel stability depends on how moisture alters coupled creep and permeability during shear-compression transitions in weak rock.
Key Innovation: Quantifies strength, stiffness, creep, and permeability evolution of argillaceous siltstone across controlled moisture contents and confining pressures.
19. A geology-informed graph neural network solution for addressing the data sparsity challenge in 3D geological modeling
Core Problem: Engineering-geology models often must reconstruct discontinuous subsurface bodies from sparse boreholes and interpreted sections.
Key Innovation: Embeds geological priors, stratigraphic connectivity, and lens-distribution scalar fields into a topology-aware graph neural network for 3D geological reconstruction.
20. Interface-based inversion of complex water-bearing structures in tunnels using level-set methods
Core Problem: Water-bearing faults, fracture zones, and karst bodies ahead of tunnel faces have irregular boundaries that are blurred by conventional resistivity inversion.
Key Innovation: Reformulates tunnel resistivity inversion as level-set boundary recovery, enabling more stable delineation of complex water-bearing structures.
21. Self-correcting prediction of rock mass discontinuity ahead of tunnel face: A model-driven and observation-corrected loop
Core Problem: Deep-learning forecasts of discontinuities ahead of tunnel faces are vulnerable to sparse, noisy, and incomplete field observations.
Key Innovation: Combines drilling data, ConvLSTM prediction, uncertainty estimation, and sequential data assimilation into a self-correcting tunnel-face forecasting loop.
22. Shear creep characteristics of carbonaceous shale under coupled hydro-mechanical-chemical coupling: Macroscopic-microscopic behavior
Core Problem: Weak carbonaceous shale interlayers in mine slopes degrade under acidic water, rainfall-induced wetting-drying, seepage, and sustained shear stress.
Key Innovation: Uses coupled THMC shear rheology, SEM, and fracture-statistics analysis to link microstructural damage to long-term strength loss and creep acceleration.
23. Numerical and field investigations of seepage-damage-microseismic response induced by mining in the weakly cemented sandstone roof
Core Problem: Weakly cemented sandstone roofs in mines are prone to coupled seepage, damage, and water inrush, but these processes are difficult to observe together.
Key Innovation: Couples numerical seepage-damage simulation with field microseismic monitoring to identify water-inrush hazard zones above the mining face.
24. An Attention-Enhanced Temporal Convolutional Network for Surrounding Rock Deformation Prediction in Mountain Tunnels
Core Problem: Mountain-tunnel deformation prediction must combine geological and construction conditions with evolving monitoring sequences.
Key Innovation: Introduces an attention-enhanced temporal convolutional network that fuses static tunnel attributes with historical deformation sequences across 290 monitoring sections.
25. Study on Three-Dimensional Deformation Inversion in Mining Areas Based on PIM Optimized by CMA-ES and Multi-Source InSAR
Core Problem: Mining deformation is three-dimensional, while conventional InSAR is one-dimensional and can lose coherence in high-deformation areas; PIM estimates also degrade near subsidence edges.
Key Innovation: Fuses multi-source InSAR with a CMA-ES-optimized probability integral model to improve three-dimensional deformation inversion in mining areas.
26. Injection-induced poroelastic controls on the transient process of dynamic fault rupture in geo-energy engineering
Core Problem: Elastostatic fault-stability criteria capture slip initiation but not the nucleation and transient rupture phases of injection-induced seismicity.
Key Innovation: Incorporates analytical poroelastic stress into an interfacial slip model to examine dynamic rupture regimes under geo-energy injection.
27. Brief communication: Vent opening at Campi Flegrei - clues from dyke propagation patterns
Core Problem: Forecasting future vent-opening locations in calderas is central to volcanic hazard management but often relies mainly on past vent density.
Key Innovation: Uses empirical dyke-propagation patterns to produce vent-opening probability maps for Campi Flegrei that align with caldera structure and recent seismicity.
28. Liquid limit and rheology of clay: insight from diffuse double layer theory
Core Problem: Liquid limit and yield stress are basic descriptors of clay behavior, yet their quantitative link across mineralogy and pore-water chemistry remains uncertain.
Key Innovation: Derives liquid limit and yield stress from diffuse double layer theory, explicitly accounting for surface area, cation exchange capacity, temperature, salinity, and pH.
29. Failure Mechanisms and Mechanical Behavior of Layered Sandstone Under Complex Stress Paths: An Experimental Study
Core Problem: Layered sandstone in slopes and underground openings experiences anisotropic failure under cyclic and confining stress paths.
Key Innovation: Uses systematic experiments to relate bedding angle, confining pressure, cyclic loading, energy dissipation, and failure mode through a three-dimensional strength surface.
30. Assessment of Dynamic Soil Stiffness of Lump-Sand Mixtures: Insights from Bender Element Tests
Core Problem: Small-strain stiffness of heterogeneous lump-sand mixtures varies with moisture, confining pressure, and lump content, affecting dynamic ground-response estimates.
Key Innovation: Uses bender-element testing to quantify shear-wave velocity and small-strain modulus across dry and saturated lump-sand mixtures.
31. Analytical investigation of steady-state seepage flow around shallow tunnels with arbitrary cross-sections
Core Problem: Non-circular shallow tunnels lack unified analytical seepage solutions despite increasing engineering use.
Key Innovation: Uses conformal mapping to derive steady-state seepage solutions for arbitrary tunnel cross-sections in shallow strata.
32. Erosional behavior of sandy soil stabilized with alkali-activated municipal solid waste incineration bottom ash and industrial by-products
Core Problem: Sandy soils are vulnerable to rainfall and wind erosion, while conventional stabilizers can carry environmental and material costs.
Key Innovation: Tests alkali-activated binders from bottom ash, waste kaolin, and red mud, showing large gains in strength and dual-mode erosion resistance.
33. Efficient Coupled FEM-SBFEM Framework for Dynamic Slope Stability Analysis of Stepped Pile-Reinforced Slopes: Adaptive Frobenius Norm-Based Termination and Reduced-Order Modeling
Core Problem: Three-dimensional dynamic analysis of reinforced slopes is computationally expensive when unbounded soil domains and soil-structure interaction are represented explicitly.
Key Innovation: Combines FEM-SBFEM coupling, reduced-order modelling, recursive convolution, and adaptive termination criteria for efficient dynamic slope-stability analysis.
34. Posterior distribution estimation of parameters related to unsaturated soil hydraulic properties by combining parameterized physics-informed neural networks and Bayesian inference
Core Problem: Uncertain unsaturated hydraulic parameters are central to infiltration and slope-hydrology modelling, but Bayesian estimation can be computationally prohibitive.
Key Innovation: Combines parameterized physics-informed neural networks with Bayesian inference to estimate posterior distributions for unsaturated soil hydraulic properties.
35. A two-stage data fusion framework for multi-source precipitation estimates integrating gauge, radar, and satellite observations via deep learning and Bayesian model averaging
Core Problem: Reliable precipitation inputs for hazard modelling require fusion of gauges, radar, and satellites across inconsistent scales and uncertainties.
Key Innovation: Combines 3D-CNN/ConvLSTM correction with Bayesian model averaging to produce calibrated multi-source precipitation estimates over complex terrain.
36. STARS: Shared-specific Translation and Alignment for missing-modality Remote Sensing Semantic Segmentation
Core Problem: Disaster mapping often faces missing optical, SAR, or elevation modalities, which can cause collapse in multimodal segmentation models.
Key Innovation: Introduces shared-specific translation and alignment to preserve segmentation performance when one or more remote-sensing modalities are absent.
37. SARU: A Shadow-Aware and Removal Unified Framework for remote sensing images with new benchmarks
Core Problem: Shadows in optical remote-sensing images degrade object detection and segmentation, especially in high-relief or urban terrain.
Key Innovation: Unifies shadow detection and removal through semantic-color fusion and a training-free physical restoration algorithm, supported by new benchmarks.
38. Controllable Remote Sensing Image Generation With Semantic Consistency via Diffusion Models
Core Problem: Remote-sensing diffusion models need semantically controlled training data and outputs to support reliable downstream mapping.
Key Innovation: Builds image-text-segmentation triplets and uses semantic masks with text prompts to generate controllable, semantically consistent remote-sensing images.
39. Wetland mapping from sparse annotations with satellite image time series and temporal-aware segment anything model
Core Problem: Dense pixel labels are expensive, and static image foundation models struggle with seasonal satellite time series.
Key Innovation: Adapts Segment Anything into a temporal-aware weak-supervision framework that uses sparse point labels and satellite time series.
40. Robust urban SLAM via resilient GNSS-IMU-LiDAR-camera-loop fusion
Core Problem: Urban and infrastructure mapping systems degrade when GNSS, LiDAR, camera, or loop-closure cues fail under difficult scene conditions.
Key Innovation: Fuses GNSS, IMU, LiDAR, camera, and loop constraints in a resilient SLAM framework applicable to repeated infrastructure and hazard-scene mapping.
41. Prediction models for significant duration of offshore ground motions in subduction zone using the S-Net database
Core Problem: Offshore significant-duration models remain sparse despite their importance for subduction-zone earthquake and tsunami-source characterization.
Key Innovation: Uses 19,058 S-Net ocean-bottom strong-motion records to build significant-duration prediction models for interface and slab earthquakes.
42. Probabilistic framework for seismic risk analysis of process plants
Core Problem: Process-plant seismic risk depends on component damage, process-network dependencies, redundancy, deterioration, and economic interruption.
Key Innovation: Builds a probabilistic hazard-risk framework that propagates component failures through process networks while accounting for correlated damage and standby equipment.
43. Cajon Pass and the Southern San Andreas Fault System: Earthquake Cycle Stress Accumulation and Present-Day Loading
Core Problem: The Cajon Pass junction may control whether large ruptures propagate between the San Andreas and San Jacinto fault systems, but present-day stress loading is uncertain.
Key Innovation: Uses 4D earthquake-cycle simulations informed by a 1000-year rupture history to estimate stress accumulation and fault interaction near Los Angeles.
44. Grain-Scale Heterogeneity in Fault Veins Influences Shear Rupture, Damage Distribution, and Microseismicity
Core Problem: Fault veins are common in fault zones, but the role of grain-scale heterogeneity in rupture style and off-fault damage is poorly constrained.
Key Innovation: Uses discrete-element simulations to show how heterogeneous quartz veins alter shear strength, distributed cracking, and microseismicity patterns.