TerraMosaic Daily Digest: June 5, 2026
Daily Summary
The June 5 literature is anchored by a rare Landslides paper that reconstructs a post-seismic, rainfall-triggered debris flow from field evidence, UAV mapping, remote sensing, and numerical simulation. The event study is paired with two forecasting papers: one uses MT-InSAR time series, variational decomposition, PatchTST, learnable channel fusion, and TimeSHAP for interpretable landslide displacement prediction; the other converts sparse rainfall-debris flow records into probabilistic thresholds through Copula expansion, mixture modelling, and Bayesian updating. Together, these papers move early warning from single-threshold diagnosis toward monitored state reconstruction.
The engineering-geology papers sharpen the boundary between hydrology, material degradation, and infrastructure response. Phreatic-line inversion, soil-rock internal erosion, vegetated biopolymer stabilization, rainfall-infiltrated soil-nail slopes, grass-rooted lateritic soils, and shale bridge failure all examine how pore structure, seepage, reinforcement, and discontinuities govern slope stability. Tunnel and underground studies extend the same logic to mixed-face stability, groundwater perturbation during hydropower construction, immersed-tunnel seismic response, muck-volume measurement, tunnel-lining cracking, coal-burst precursors, and strainburst support. The remote-sensing and hydrology papers add scalable observation: palaeoflood reconstruction, national-scale flood data assimilation, permafrost runoff response, GeoAI wildfire synthesis, foundation-model point-cloud segmentation, and community flood image enhancement all target data gaps that limit operational hazard assessment.
Key Trends
Five movements define the issue: state-estimation warning, groundwater-linked slope stability, coupled underground sensing, sparse-record hazard hydrology, and remote-sensing methods for difficult geometry and limited labels.
- Debris-flow and landslide warning are becoming state-estimation problems: The Bujueyida debris-flow reconstruction, InSAR-PatchTST displacement model, and probabilistic rainfall-debris flow framework all infer evolving source conditions rather than relying on a single triggering index.
- Slope stability studies are integrating groundwater thresholds with material microstructure: Phreatic-line inversion, gap-graded soil-rock erosion, biopolymer vegetation reinforcement, lateritic-root pore structure, rainfall-infiltrated soil nails, and shale bridge tests connect hydraulic state to failure resistance.
- Underground hazards are being reframed as coupled sensing and control problems: Tunnel-face stability, hydropower groundwater dynamics, immersed-tunnel seismic response, muck-volume point clouds, lining-crack indices, coal-burst signal enhancement, and strainburst bolts all translate monitoring into actionable stability metrics.
- Hazard hydrology is expanding beyond instrumental records: Palaeoflood stratigraphy, real-time river data assimilation, permafrost runoff-response analysis, tidal-inlet storm sediment modelling, and island-flow PINNs extend flood and erosion assessment into sparse-data regimes.
- Remote-sensing methods are targeting sparse labels and difficult geometry: Zero-shot airborne point-cloud segmentation, GeoAI wildfire synthesis, flood-image super-resolution, GNSS/GRACE groundwater modelling, karst landform classification, and LiDAR/UAV methods address observational incompleteness rather than only benchmark accuracy.
Selected Papers
This issue contains 41 selected papers from 1,400 papers analyzed. The selected papers are ordered by relevance score and emphasize post-seismic debris-flow reconstruction, InSAR-based landslide forecasting, probabilistic rainfall thresholds, groundwater-linked slope stability, tunnel and mining hazards, sparse-record flood hydrology, and remote-sensing methods with direct transfer value for hazard observation.
1. Field investigation and numerical study of Bujueyida debris flow in Ganluo County (Sichuan, China) on 20 July 2024
Core Problem: The Bujueyida Gully event combined earthquake-weakened slopes, intense short-duration rainfall, loose source material, and railway exposure, requiring event-chain reconstruction rather than simple rainfall attribution.
Key Innovation: Combines field survey, UAV mapping, remote sensing, seismic history, rainfall analysis, and numerical simulation to identify seven unstable source slopes and reconstruct the post-seismic landslide-to-debris-flow process.
2. Interpretable Multivariate Landslide Displacement Forecasting Based on InSAR and Deep Learning: PatchTST with Learnable Channel Fusion
Core Problem: Remote-sensing-based landslide forecasting often lacks interpretability and must handle multivariate MT-InSAR deformation signals with trend, seasonal, and noise components.
Key Innovation: Adapts PatchTST to MT-InSAR landslide displacement series with learnable channel fusion, VMD decomposition, and TimeSHAP attribution for temporal and feature-level explanation.
3. A probabilistic modeling framework for data-driven forecasting of rainfall-induced debris flows
Core Problem: Rainfall-debris-flow warning models suffer from sparse event records, missed alarms, false alarms, and thresholds that often lack physical interpretation.
Key Innovation: Expands limited event data with Copula modelling, groups triggering mechanisms using mixture models, and updates rainfall thresholds with Bayesian inference.
4. Phreatic-line-based slope stability assessment and groundwater-level warning-threshold inversion
Core Problem: Groundwater warning thresholds for slopes are often empirical and disconnected from site-specific strength, suction, and geological conditions.
Key Innovation: Links field-measured groundwater levels with phreatic-line-based stability analysis and threshold inversion using unified saturated-unsaturated effective stress formulations.
5. Investigation into erosion of gap-graded soil-rock mixtures considering fine particle contents and seepage directions
Core Problem: Internal erosion in soil-rock mixtures depends on fine-particle content and seepage direction, but most tests do not control flow-gravity angle explicitly.
Key Innovation: Develops a steerable seepage testing device and couples laboratory measurements with PNM-CFD-DEM simulations to resolve multiscale particle migration.
6. Soil stabilization and erosion resistance capabilities of vegetative biomaterials based on biopolymer
Core Problem: Cement-based ecological slope protection can compact soil and reduce compatibility with vegetation, limiting sustainable erosion control.
Key Innovation: Evaluates xanthan-gum and gellan-gum biopolymers with mechanical tests, permeability tests, SEM, NMR, plant-growth experiments, and simulated erosion tests.
7. Seismic critical stability analysis of purely cohesive non-homogeneous soil slopes using stress characteristics method
Core Problem: Seismic stability of non-homogeneous cohesive slopes is usually evaluated through assumed geometry and slip surfaces rather than critical equilibrium profiles.
Key Innovation: Uses the stress characteristics method to derive critical slope profiles with factor of safety equal to one under pseudo-static seismic loading.
8. Shallow stability analysis of rainfall-infiltrated slopes considering soil nail reinforcement
Core Problem: Rainfall infiltration alters pore-water pressure and shallow sliding potential, but existing reinforced-slope mechanisms can violate velocity compatibility or flow rules.
Key Innovation: Combines rainfall infiltration simulation with an improved rotational-translational upper-bound mechanism and explicit soil-nail reinforcement effects.
9. Pore structure and hydro-mechanical behavior of ryegrass-planted lateritic soil: insights from CT scanning and numerical simulation
Core Problem: Lateritic soils crack in dry environments, accelerating rainfall infiltration and reducing strength; root density can either improve or disturb the pore network.
Key Innovation: Uses CT reconstruction, single-phase flow simulation, and finite elements to quantify how ryegrass roots modify pore geometry and hydro-mechanical response.
10. Shear Progressive Failure Mechanisms of Shale with Non-persistent Joints
Core Problem: Rock-bridge failure in non-persistent jointed shale controls slope instability and crack coalescence, but progressive shear mechanisms remain difficult to quantify.
Key Innovation: Combines compression-shear tests, acoustic emission, digital image correlation, and 3D scanning to resolve bridge failure surfaces and crack propagation.
11. Seismic response of the immersed tunnel considering spatially random soils and water-stratum-structure coupling effect
Core Problem: Immersed-tunnel seismic models often simplify soil heterogeneity and water-soil-structure coupling, which can bias dynamic response estimates.
Key Innovation: Builds 2D and 3D seawater-seabed-tunnel models that integrate acoustic-structure coupling with random-field soil heterogeneity.
12. Groundwater dynamics induced by tunneling and reservoir filling at the site of Kala hydropower project
Core Problem: Hydropower excavation and reservoir filling can reorganize groundwater systems, affecting tunnel leakage, phreatic surfaces, and site safety.
Key Innovation: Combines hydrochemistry, packer tests, tunnel inflow observations, and transient groundwater modelling to separate leakage sources and predict recharge evolution.
13. Modeling of the largest snow avalanche tragedy in the Western Carpathians (Velka Fatra Mountains - Slovakia)
Core Problem: The 1924 Rybie avalanche remains the deadliest avalanche in Slovakia, but historical records provide limited quantitative constraints on release zones, snow depth, and destructive pressure.
Key Innovation: Uses LiDAR terrain data, GIS potential-release-area mapping, AvaFrame, and RAMMS back-calculation to reconstruct the likely avalanche scenario.
14. 3D point cloud segmentation-based volume measurement of muck in earth pressure balance shield tunneling
Core Problem: Inaccurate muck-volume measurement in EPB shield tunnelling can cause over-excavation or under-excavation, leading to settlement or heave hazards.
Key Innovation: Uses LiDAR point-cloud segmentation and a surface-based volume algorithm validated against engineering records and laboratory error analysis.
15. Cross-project graph convolutional network model for just-in-time rock mass classification in TBM tunneling
Core Problem: TBM rock-mass classification models often fail to generalize across projects because geology, machine scale, and class imbalance change simultaneously.
Key Innovation: Integrates mechanics-driven input features, graph convolution, and KMeans-SMOTE to transfer rock-mass classification across hard-rock tunnel projects.
16. Limit equilibrium models for tunnel face stability in mixed ground
Core Problem: Mixed soft-hard ground creates complex shield-tunnel face failures, yet criteria for upper-soft/lower-hard and upper-hard/lower-soft strata remain unclear.
Key Innovation: Defines strength-ratio-based mixed-ground classes and derives limit support-pressure models from numerical failure mechanisms.
17. Multi-parameter quantitative analysis of tunnel lining cracks based on an improved tunnel crack index considering fractal dimension
Core Problem: Tunnel lining crack assessment is limited by vague diagnostic scales, single-index evaluation, and inconsistent extraction of complex crack parameters.
Key Innovation: Combines decomposable crack geometry, K-means, Bayesian decision boundaries, fractal dimension, and partial least squares regression into an improved tunnel crack index.
18. Signal Enhancement and Identification Methodology for Coal Burst Early Forecasting Induced by High-Intensity Mining
Core Problem: Coal-burst monitoring under high-intensity mining is hindered by low signal-to-noise ratios and poorly defined fracture precursors.
Key Innovation: Uses phase-space reconstruction, dual-domain features, and physics-informed signal enhancement to isolate deterministic fracture precursors from mechanical noise.
19. Mechanism of Pre-tensioned Bolt for Strainburst Mitigation: Insights from the Regulation of Stress/Energy Fields and Fracture Pattern
Core Problem: Pre-tensioned bolts are widely used against strainburst, but their effect on stress fields, energy release, fracture modes, and ejected debris is not fully explained.
Key Innovation: Combines theory, experiments, numerical modelling, and moment-tensor inversion to show how pre-tension alters energy fields and suppresses tensile fracture propagation.
20. Study on the Fracture Characteristics and Bolting Effect of Deep Tunnels Based on in-situ Mechanical Behavior of Rock Mass
Core Problem: Deep tunnel failure depends on lateral stress state and energy localization, while support effects are often evaluated without tracking fracture-energy transfer.
Key Innovation: Tracks plastic energy evolution to classify unloading-induced failure patterns and quantify how bolting redirects energy into deeper stable rock.
21. Entropy-based probabilistic constitutive model for rocks: Coupled damage-permeability evolution and Bayesian calibration
Core Problem: Subsurface hazard and engineering systems require coupled damage-permeability models that account for uncertainty without assuming predefined damage distributions.
Key Innovation: Integrates maximum-entropy damage modelling, sigmoid permeability evolution, hybrid optimization, and Bayesian calibration for probabilistic rock hydro-mechanics.
22. Reconstructing palaeoflood magnitude and frequency for flood hazard assessment in arid regions: The northern Judea Desert Streams, Israel
Core Problem: Arid-region flood-frequency analysis is constrained by scarce gauges and high event variability, even as infrastructure exposure expands.
Key Innovation: Uses palaeoflood sediment evidence, geomorphic analysis, and hydraulic reconstruction to estimate flood magnitude and frequency in Judean Desert streams.
23. Application of real-time data assimilation system to improve streamflow forecasts in Japan
Core Problem: Operational flood forecasting at national scale must update hydrodynamic states from unevenly distributed real-time observations.
Key Innovation: Implements an online LETKF data-assimilation system using hourly water-surface observations from 1800 gauges to improve streamflow and inundation forecasts in Japan.
24. Declining Runoff Sensitivity to Precipitation Following Permafrost Degradation: Insights From Event-Scale Runoff Response in the Yellow River Source Region
Core Problem: Frozen-ground degradation may alter flood generation and water-resource resilience, but event-scale runoff sensitivity remains debated.
Key Innovation: Uses observation-only ensemble rainfall-runoff analysis to quantify four decades of changing precipitation response across frozen-ground zones in the Yellow River source region.
25. Three Decades of GeoAI for Wildfire Science: A Systematic and Meta-Analysis Review
Core Problem: GeoAI wildfire studies are fragmented by task, data source, and method, limiting synthesis across the full wildfire management cycle.
Key Innovation: Meta-analyzes 449 peer-reviewed studies from 1994-2024 to evaluate how supervised AI, remote sensing, and GIS are used across wildfire science tasks.
26. Slip rate partitioning among fault segments in the central part of North Anatolian Fault Zone
Core Problem: Seismic hazard along the North Anatolian Fault depends on how deformation is partitioned between the master fault and splay faults.
Key Innovation: Uses 105 GNSS sites to build a block model that estimates slip rates across the central NAFZ and its branching fault segments.
27. OBS Data Mining Reveals Seismic Structure and Dynamics of the Oceanic Blanco Transform Fault, Northeast Pacific
Core Problem: Oceanic transform-fault deformation is partitioned between seismic and aseismic slip, but event depths and segment dynamics are poorly constrained offshore.
Key Innovation: Uses machine-learning phase pickers and high-precision locations from a 54-station OBS array to build a 30,000-event Blanco Transform Fault catalogue.
28. Parameterized physics-informed neural networks for flow reconstruction around idealized islands under data-sparse conditions
Core Problem: Coastal island management often lacks dense measurements of near-bed flow fields that control erosion and benthic-habitat stress.
Key Innovation: Combines LES pretraining, tomographic PIV refinement, and parameterized PINNs to reconstruct 3D flow fields around islands under sparse observations.
29. Storm-Induced Sediment Transport at a Small Tidal Inlet
Core Problem: Small tidal inlets are common but understudied, and storm-driven sediment transport controls shoreline, backbarrier, and navigation hazards.
Key Innovation: Uses Delft3D, field water-level and current measurements, and storm-normal comparisons to model deposition patterns at Allens Pond Inlet.
30. Enhancing community-based participatory flood imagery using an AI-based super-resolution framework
Core Problem: Community flood photos and videos often contain motion blur, compression artifacts, low resolution, and inconsistent metadata, reducing their value for damage assessment.
Key Innovation: Fine-tunes Real-ESRGAN on participatory flood imagery to improve perceptual quality and analytical usability of ground-level disaster observations.
31. Consistency-Guided Distillation from Vision Foundation Models for Zero-Shot Airborne Point Cloud Segmentation
Core Problem: Large airborne point-cloud segmentation still depends on expensive 3D labels, while 2D foundation-model distillation suffers from semantic bleeding in complex scenes.
Key Innovation: Distills vision-foundation-model knowledge through geometry-constrained pseudo-label purification and consistency guidance for zero-shot 3D segmentation.
32. Fluvial geomorphology and GIS-based assessment of water resource vulnerability in oil-affected river systems of Venezuela
Core Problem: Oil infrastructure, deforestation, and land-use change threaten hydrological and geomorphic stability in Venezuela's major river systems.
Key Innovation: Applies integrated GIS, remote sensing, proximity analysis, and geomorphic sensitivity assessment to map water-resource vulnerability.
33. Long-term hydro-sedimentary dynamics of the Ucayali River (Amazon Basin) revealed through combined observations, remote sensing, and SWAT-Amazon modelling
Core Problem: Hydro-sedimentary budgets in large Amazonian rivers are difficult to quantify because conventional monitoring lacks adequate spatial and temporal coverage.
Key Innovation: Combines observations, remote sensing, and SWAT-Amazon hydrological-hydraulic modelling to quantify floodplain-controlled sediment dynamics in the Ucayali River.
34. Physics-guided phase-specific machine learning for vibration-driven droplet migration in constricted pores
Core Problem: Dynamic multiphase transport in fractured or constricted geomaterials is nonlinear and regime-dependent, reducing robustness of single global machine-learning models.
Key Innovation: Separates pore-scale droplet migration into physically distinct oscillatory and propagation phases before training phase-specific predictors.
35. Texture-terrain integration for classification of karst Fenglin-Fengcong landforms in Southwest China
Core Problem: Fenglin-Fengcong karst landforms have heterogeneous morphology and weak spectral contrast, complicating terrain classification for karst hazard and geomorphic analysis.
Key Innovation: Integrates GF-1 imagery, ALOS PALSAR DEM derivatives, texture descriptors, and object-based Random Forest classification across three karst regions.
36. Modeling groundwater depletion in Hungary through GRACE and GLDAS observations analyzed with ensemble machine learning models
Core Problem: Groundwater storage modelling can be spatially biased where monitoring networks are sparse and unevenly distributed.
Key Innovation: Combines GRACE, GLDAS, and ensemble machine-learning models to estimate regional groundwater depletion under limited in-situ monitoring.
37. Granite fracturing under multipoint sequential and mobile laser irradiation using a grain model
Core Problem: Laser-assisted hard-rock excavation requires understanding of thermal damage, crater formation, and crack propagation under moving and multipoint irradiation.
Key Innovation: Combines laser tests, microstructural analysis, and a GBM-CZM thermal-mechanical damage model to simulate granite fracturing.
38. Study on the pull-out mechanical behavior of expansive mortar grouted anchors and the mechanical response of fractures at different angles
Core Problem: Conventional cementitious grouts can shrink and weaken stress transfer in fractured rock masses, reducing anchorage reliability.
Key Innovation: Combines pull-out tests and numerical simulations to evaluate expansive mortar grouted anchors and fracture-angle-dependent mechanical response.
39. Analysis of TBM disc cutter performance based on the pressure distribution to optimize operational efficiency
Core Problem: TBM cutter performance depends on disc-rock pressure distribution, which controls rock fracture and specific energy consumption.
Key Innovation: Extends CSM-style cutter-force modelling with alternative contact-pressure profiles and laboratory-informed numerical models across rock types.
40. Dynamic vulnerability assessment of urban rail transit systems using spatiotemporal deep learning
Core Problem: Urban rail vulnerability changes with dynamic passenger flow and network topology, which static assessments cannot capture.
Key Innovation: Uses a graph convolutional temporal Kolmogorov-Arnold network and node-importance metrics to assess structural and functional vulnerability under disruptions.
41. A Fusion Method of GNSS-IR Sea Surface Height Using the Carrier and SNR
Core Problem: GNSS-IR sea-surface-height retrieval can use SNR and carrier phase, but combining them robustly remains difficult for coastal water-level monitoring.
Key Innovation: Develops a Bayesian multi-source fusion framework combining SNR and dual-frequency carrier phase for improved GNSS-IR sea-level inversion.