TerraMosaic Daily Digest: October 3, 2026
Daily Summary
The October 3 issue is led by two studies that turn difficult hazard signals into inspectable evidence. A UAV highway-slope workflow fuses RGB, thermal and LiDAR decisions after foundation-model feature extraction, locating high-priority inspection zones across five slopes. In the Aegean, machine-learning detection and relocation of 66,109 earthquakes resolve 46 compact bursts and rapid migration along a 30-km structure, strongly supporting a deep magmatic intrusion during the 2025 Santorini–Amorgos crisis.
Measurement quality is as consequential as model choice. A Taiwan study corrects deformation and screens inconsistent GNSS/leveling observations before converting LiDAR heights, reducing holdout residual RMSE from 0.236 m to 0.041 m and showing that local coastal low-elevation classes can change near critical thresholds. Monthly shallow-groundwater mapping in the Weiku Oasis likewise combines optical, radar, topographic, meteorological, soil and human-activity variables, while keeping seasonal controls visible through SHAP analysis.
The engineering set emphasizes evolving material states and coupled response. Rapid-setting grout is represented from flowing slurry to solidified vein through time-dependent particle interactions calibrated against physical tests. A two-grid peridynamics–finite-volume formulation separates fine solid fracture resolution from a coarser fluid grid. Companion work addresses freeze–thaw subgrade deformation, dynamic rock strength, underground ground-motion intensity, soil–foundation interaction and excavation-induced tunnel response.
Several potentially useful records arrived with title and bibliographic metadata but without reliable abstracts. These include Loess Plateau erosion–connectivity contrasts, continental SWOT reservoir-operation modelling, variance-composite coastal monitoring and selected underground-engineering studies. They are retained only where their verified scope is relevant; no unavailable method, result or performance claim is inferred. Six records already covered in earlier Digests were excluded, along with materials, energy, logistics and ecological papers lacking a clear geohazard contribution.
Key Trends
The day’s strongest records connect observation quality, physical mechanism and operational interpretation across slopes, seismic crises, coasts and underground infrastructure.
- Multimodal slope inspection is moving from imagery to actionable zones: Foundation-model features and decision-level fusion combine RGB, thermal and LiDAR evidence while preserving modality-specific confidence and disagreement.
- Dense event catalogues can discriminate volcanic unrest mechanisms: Burst geometry, migration speed and cessation fronts make the Santorini–Amorgos sequence more consistent with propagating magmatic fluids than shallow hydrothermal activation.
- Vertical-reference quality now limits coastal classification: Deformation-aware GNSS/leveling control shows that centimetre- to decimetre-scale datum errors can move pixels across operational low-elevation thresholds.
- Engineering models are resolving state evolution rather than fixed properties: Grout solidification, fracture–flow coupling, freeze–thaw deformation and frequency-dependent foundation impedance are treated as changing processes.
- Sparse metadata requires bounded interpretation: Title-level records are useful discovery signals, but their methods and results remain unassessed until reliable abstracts or full texts are available.
Selected Papers
The October 3 selection is anchored by multimodal UAV inspection of highway slopes and a dense machine-learning earthquake catalogue that resolves the Santorini–Amorgos magmatic crisis. Supporting papers examine erosion–connectivity patterns, coastal elevation reference quality, time-dependent grouting, reservoir and shoreline monitoring, freeze–thaw deformation, dynamic rock strength, underground seismic demand and excavation-induced tunnel response. Records without reliable abstracts are retained only at title level and their unavailable results are not inferred.
1. A UAV-Based Highway Slope Inspection Method Based on Multimodal Sensors and Decision-Level Fusion Strategy
Core Problem: Highway-slope inspection must integrate heterogeneous RGB, thermal and LiDAR observations despite limited labelled imagery.
Key Innovation: Frozen DINOv2 encoders, modality-specific XGBoost classifiers and regularized decision-level fusion are evaluated on 2,007 aligned patch triplets from five highway slopes; the system reaches 90.48% accuracy, 85.13% macro recall and 82.90% F1 before aggregating predictions into priority inspection zones.
2. Episodic Seismicity Bursts Driven by Magma Intrusion During the 2025 Santorini–Amorgos Seismic Crisis
Core Problem: The exceptional productivity and migration of the Santorini–Amorgos earthquake sequence require discrimination among fault interaction, hydrothermal activation and magmatic intrusion.
Key Innovation: Machine-learning monitoring detects and relocates 66,109 events, resolving 46 hour-long bursts, migration up to about 3 km/h and a narrow 30-km structure; the geometry and back-fronts support a deep intrusion that stalled after propagating laterally.
3. Contrasting spatial patterns of soil erosion and sediment connectivity in a typical watershed of the Loess Plateau: Insights from the EC framework
Core Problem: Erosion intensity and sediment connectivity can form different spatial patterns, complicating identification of the parts of a watershed that actually deliver sediment.
Key Innovation: The verified title, authors and DOI establish an erosion–connectivity framework study in a typical Loess Plateau watershed; model construction, validation and quantitative findings cannot be assessed because a reliable abstract was unavailable.
4. Quality Control of GNSS/Leveling-Derived Geoid Heights and Implications for LiDAR-Based Coastal Low-Elevation Classification
Core Problem: Vertical deformation, benchmark instability and epoch mismatch can contaminate GNSS/leveling control and propagate into LiDAR-based coastal elevation classes.
Key Innovation: Vertical-velocity correction, leave-one-out validation and iterative outlier screening reduce holdout residual RMSE from 0.236 m to 0.041 m; broad lowland patterns persist, but local pixels near critical elevation thresholds change classification.
5. Modeling fracture grouting diffusion of cementitious rapid-setting grouts with continuous time-dependent viscosity and phase evolution
Core Problem: Rapid-setting grout changes continuously from a flowing suspension to a solid vein, a transition that split fluid–solid schemes do not represent consistently.
Key Innovation: A discrete-element updating scheme links viscosity-dependent particle contacts to solidification, is calibrated with viscosity, flowability and discharge tests, and reproduces experimental filling patterns and pressure distributions.
6. Towards continental-scale SWOT-based modeling of reservoir operations
Core Problem: Continental water-risk analysis needs reservoir-operation information where operational records and in-situ monitoring are incomplete.
Key Innovation: The verified title, authors and DOI establish continental-scale modelling based on Surface Water and Ocean Topography observations; reservoir inference, validation and operational skill cannot be assessed because a reliable abstract was unavailable.
7. Satellite-Based Coastal Monitoring through Variance Image Composites (SatVar)
Core Problem: Repeated satellite scenes need a compact representation that makes persistent and episodic coastal change visible for regional monitoring.
Key Innovation: The verified title, authors and DOI establish a variance-image-composite method named SatVar; sensor inputs, temporal compositing, validation and detectable change classes cannot be assessed because a reliable abstract was unavailable.
8. Thermal and Deformation Behavior of Subgrade with Different Surface Coverings under Freeze–Thaw Cycles: A Model Test and Comprehensive Evaluation
Core Problem: Surface covering can alter the coupled thermal and deformation response of cold-region subgrade during repeated freezing and thawing.
Key Innovation: The verified title, authors and DOI establish a comparative model-test evaluation across surface coverings; boundary conditions, deformation magnitudes and performance ranking cannot be assessed because a reliable abstract was unavailable.
9. A Decoupling-Learning-Based Unified Predictor for Dynamic Rock Strength under Sparse and Heterogeneous Data
Core Problem: Dynamic rock-strength prediction must reconcile small datasets collected across different lithologies, loading conditions and experimental protocols.
Key Innovation: The verified title, authors and DOI establish a decoupling-learning unified predictor for sparse heterogeneous data; representation strategy, baselines and out-of-domain performance cannot be assessed because a reliable abstract was unavailable.
10. Information-Theoretic Development of a Finite-Wavelength-Corrected Ground-Motion Intensity Measure for Underground Structures
Core Problem: Conventional ground-motion measures may not represent the finite-wavelength effects that control underground structural demand.
Key Innovation: The verified title, authors and DOI establish an information-theoretic intensity measure with finite-wavelength correction; formulation, selected demand measures and predictive gains cannot be assessed because a reliable abstract was unavailable.
11. Analysis of Soil–Foundation–Structure Interaction Based on a Novel Lumped-Parameter Model with Improved Legendre Polynomials
Core Problem: Time-domain seismic analysis needs a compact representation of frequency-dependent foundation impedance without introducing artificial inertial forces.
Key Innovation: Ratios of improved Legendre polynomials are decomposed into low-order spring–damper–inerter elements, producing a mass-free lumped model that reproduces elastic half-space impedance across a broad frequency band and is tested in nonlinear structural response examples.
12. Simplified Analysis Method for Transverse Response of a Tunnel Induced by Overlying Foundation-Pit Excavation
Core Problem: Excavation above an existing tunnel changes unloading and soil restraint in the transverse direction, requiring a tractable screening method for deformation and internal force.
Key Innovation: The verified title, authors and DOI establish a simplified transverse-response analysis for an overlying foundation pit; governing assumptions, calibration and case-study accuracy cannot be assessed because a reliable abstract was unavailable.
13. Accurate Monthly Estimation of Shallow Groundwater Depth in Arid Oasis Regions Driven by Multi-Source Data and Machine Learning
Core Problem: Sparse wells and strong seasonal heterogeneity limit monthly mapping of shallow groundwater in intensively managed arid oases.
Key Innovation: Optical, radar, terrain, meteorological, soil and human-activity predictors are compared across RF, SVM and XGBoost strategies; the best monthly models average R² = 0.790 and SHAP reveals seasonally shifting controls while valley depth remains consistently influential.
14. A Two-Grid Strategy for Coupling Peridynamics and Finite-Volume Methods in Hydraulic-Fracturing Simulations
Core Problem: Engineering-scale fracture–flow simulation is expensive when fluid pressure and solid damage must share the finest discretization.
Key Innovation: A coarse finite-volume fluid grid is coupled to fine peridynamic particles through strain-rate interpolation, pressure reconstruction and damage-based permeability upscaling, improving efficiency while preserving the main fracture and pore-pressure patterns in benchmarks.