TerraMosaic Daily Digest: May 9, 2026
Daily Summary
May 9's papers centre on coupled cascade hazards, not isolated triggers. The Tut debris-flow study shows how earthquake-damaged slopes supplied coseismic landslide debris that rainfall later mobilized into 396 debris flows; the Boccassuolo landslide paper reconstructs a rapid transition from rotational slide to earth flow, infrastructure damage, river interaction, and small landslide-dammed lakes; and the paired Stromboli studies combine historical deposits, numerical reconstruction, and probabilistic invasion maps for deposit-derived pyroclastic density currents on inhabited volcanic slopes. Together, these papers treat runout, source conditioning, and downstream consequences as part of the same hazard system.
A second group turns uncertain forcing into operational thresholds, forecasts, or design constraints. Urban flood susceptibility in Guangzhou is built from social-media-derived flood inventories and GeoShapley interpretation; case-selective dynamical downscaling reproduces sub-daily extreme-rainfall statistics at a fraction of the cost of continuous convection-permitting simulation; the Kamchatka tsunami analysis tests source models against Pacific-wide buoy and tide-gauge records; and reservoir scheduling, streamflow prediction, Loess runoff-sediment response, and residual-soil water retention all make uncertainty explicit. Engineering papers extend the same logic into permafrost railways, cracked tunnel linings, saturated-site buried nuclear structures, CPT-based subsurface reconstruction, gassy clay, TBM rock fracturing, and automated shield-tunnel assembly.
Key Trends
The dominant movement is from single-hazard mapping toward coupled source-to-impact modelling with explicit uncertainty.
- Cascading hazards are being modelled as linked sequences: earthquake-preconditioned debris flows, Boccassuolo slide-to-earthflow evolution, Stromboli pyroclastic-density-current reconstruction, and Kamchatka tsunami propagation all trace how an initial disturbance moves through storage, runout, routing, and impact zones.
- Hazard forecasts are becoming cheaper without abandoning physical constraints: case-selective downscaling, physics-AI reservoir scheduling, graph-based streamflow forecasting, and runoff-sediment event classification use targeted simulation or network structure to reduce cost while preserving forcing and routing information.
- Risk evidence is diversifying beyond formal sensor inventories: Guangzhou flood susceptibility uses social media and news reports, northern Chile risk analysis uses institutional and community interviews, and Boccassuolo monitoring uses high-frequency satellite imagery to recover event evolution where conventional records are sparse.
- Subsurface and infrastructure papers emphasize uncertainty in material state: residual-soil SWCCs, overconsolidated gassy clay, permafrost subgrades, cracked tunnel linings, saturated buried nuclear structures, and CPT-based 3-D ground models all treat hidden material variability as central to safety rather than a secondary parameter.
- Remote-sensing methods are strongest when tied to hazard-relevant observables: glacier mapping, UAV coastal thermal retrieval, air-temperature estimation, saliency detection, optical-SAR fusion, wind assessment, and terrace or soil-landscape monitoring improve exposure, forcing, terrain, and environmental state estimation.
Selected Papers
This issue contains 33 selected papers from 1,141 papers analyzed. The leading papers connect earthquake-preconditioned debris-flow initiation, satellite reconstruction of the Boccassuolo landslide, probabilistic and historical Stromboli pyroclastic-flow hazard, explainable urban flood susceptibility, efficient extreme-precipitation downscaling, Pacific-wide tsunami propagation, volcanic tephra stratigraphy, and hydrological scheduling under future uncertainty. The broader set adds streamflow forecasting, residual-soil water-retention uncertainty, permafrost railway dynamics, Loess Plateau runoff-sediment response, tunnel lining crack mechanics, buried nuclear structure seismic response, deep-learning subsurface reconstruction, glacier mapping, graph contact-force surrogates, gassy clay constitutive modelling, TBM rock fracturing, coastal thermal mapping, optical-SAR fusion, and soil-landscape legacy analysis.
1. A hybrid approach to earthquake-preconditioned debris flow: regional susceptibility and event-based hazard assessments
Core Problem: The 2023 Kahramanmaras earthquake sequence left coseismic landslide debris that was later mobilized by intense rainfall, creating debris-flow hazards in a region without documented previous events.
Key Innovation: Field observations, remote sensing, logistic-regression susceptibility mapping, and RAMMS back-analysis are combined for 396 debris flows in the Tut region, linking post-seismic weakening to event-scale flow intensity and hazard maps.
2. Satellite-based investigation of the spring 2025 Boccassuolo Landslide, Northern Apennines (Italy)
Core Problem: The spring 2025 Boccassuolo landslide evolved from rotational sliding to earth flow and earth slide, damaging roads, a bridge, buildings, and the Dragone Stream corridor.
Key Innovation: PlanetScope and Sentinel-2 imagery reconstruct the event chronology, displacement style, infrastructure impact, and formation of two small landslide-dammed lakes with potential flood concern.
3. Assessing Deposit‐Derived Pyroclastic Flow Hazard at Stromboli (Italy): 2. Probabilistic Invasion Maps
Core Problem: Stromboli's paroxysms can trigger hot deposit-derived flows that escape the Sciara del Fuoco and threaten inhabited basins, but future invasion probabilities remain uncertain.
Key Innovation: Topographic basin analysis, a tested numerical flow model, and a temporal paroxysm-to-PDC occurrence model produce probabilistic invasion maps over 10- and 50-year horizons.
4. Assessing Deposit‐Derived Pyroclastic Flow Hazard at Stromboli (Italy): 1. Reconstruction of the Dynamics of the 11 September 1930 Event
Core Problem: The 11 September 1930 Stromboli event remains a key analogue for deposit-derived pyroclastic flows that can reach populated valleys.
Key Innovation: New field deposits, historical records, eyewitness constraints, and numerical modelling reconstruct flow thickness, runout, and source conditions for the 1930 event.
5. An Explainable Ensemble Machine Learning Framework for Flood Susceptibility Mapping Using Social Media Data: A Case Study of Guangzhou, China
Core Problem: Urban flood inventories are often incomplete, and non-flood samples can bias machine-learning susceptibility models.
Key Innovation: Flood reports from social media and news are combined with similarity- and diversity-based negative sampling, heterogeneous bagging, and GeoShapley interpretation for Guangzhou.
6. Case‐Selective Dynamical Downscaling Enables a Tenfold Cost Reduction for Extreme Precipitation Statistics
Core Problem: Convection-permitting climate simulations are too expensive for long-period extreme-precipitation statistics over large regions.
Key Innovation: Case-selective dynamical downscaling simulates only windows likely to contain extremes and reproduces 1-6 h precipitation statistics at roughly one-tenth of full continuous-simulation cost.
7. Transoceanic propagation of the tsunami from 2025 Mw 8.8 Kamchatka Earthquake across the Pacific Ocean
Core Problem: The 2025 Mw 8.8 Kamchatka tsunami requires source-model testing against basin-scale buoy and tide-gauge observations.
Key Innovation: Forty deep-ocean buoys, ten coastal gauges, multilayer elastic-Earth source calculations, and spectral-energy analysis constrain transoceanic propagation and coastal oscillation patterns.
8. Stratigraphy, eruptive dynamics and hazard implications of the 1.47 Ma Amaranto tephra: an atypical plinian eruption in the Michoacán-Guanajuato volcanic field (México)
Core Problem: The Amaranto tephra is a regionally distributed marker near Morelia, but its eruptive source, dynamics, and hazard implications have remained uncertain.
Key Innovation: Stratigraphy, componentry, vesicularity, density, geochronology, and dispersal interpretation reconstruct an atypical 1.47 Ma plinian eruption and its hazard relevance.
9. Physics‐AI Synergized Optimization‐Learning‐Simulation Framework for Robust Cascade Reservoir Scheduling Under Future Hydrological Uncertainty
Core Problem: Future runoff uncertainty can produce abrupt wet-dry shifts that challenge hydropower, flood-control, and ecological scheduling rules.
Key Innovation: An optimization-learning-simulation framework combines INSGA-III, VIKOR, physics-informed reinforcement LSTM, SARIMA, bootstrap, and Cholesky simulation while enforcing water-balance constraints.
10. Sociopolitical Construction of Disaster Risk in Informal Settlements (SoPoCoDRIS): Evidence from Extractive Cities in Northern Chile
Core Problem: Risk in informal settlements is shaped not only by physical exposure to earthquakes and tsunamis but also by housing policy, relocation, institutional fragmentation, and community agency.
Key Innovation: Forty-five interviews in northern Chile identify mechanisms by which risk is politically produced and sustained across formal and informal urban space.
11. Role of river network information in streamflow prediction using graph wavenet and entity-aware long short-term memory models
Core Problem: Flood response in river basins depends on upstream-downstream connectivity that sequence models may not represent explicitly.
Key Innovation: Graph WaveNet and entity-aware LSTM models are compared using hydrological and Euclidean adjacency matrices for hourly streamflow forecasts across South Korean river networks.
12. Probabilistic soil-water characteristic curve of residual soils: A case history from Korea
Core Problem: Residual soils in Korea are commonly unsaturated, but incomplete soil-water characteristic curve data limit probabilistic stability assessment.
Key Innovation: A copula-based framework simulates bivariate van Genuchten SWCC parameters from 52 measured curves, quantifying uncertainty in residual-soil water retention.
13. Numerical study on dynamic responses of a train-track-subgrade system in permafrost regions under seasonal variation
Core Problem: Railway subgrades in permafrost regions undergo seasonal hydrothermal changes that alter dynamic stiffness, damping, and long-term operational safety.
Key Innovation: A 3-D train-track-permafrost subgrade model, validated against Qinghai-Tibet Railway monitoring, simulates seasonal dynamic response under coupled freeze-thaw conditions.
14. Event-based runoff-sediment response patterns and their driving mechanisms: a comparative study of typical watersheds on the Chinese Loess Plateau
Core Problem: Loess Plateau watersheds show sharply different runoff-sediment regimes, increasing uncertainty in event-scale sediment forecasting.
Key Innovation: The MESH framework identifies flood events and combines multidimensional descriptors, Spearman correlation, random forests, rating curves, and hysteresis analysis across three watersheds.
15. Effect of lining crack location on the bearing capacity of tunnels in metamorphic layered soft rock
Core Problem: Operating tunnels in high mountain canyon regions can develop lining cracks whose structural risk depends on bedding dip and crack position.
Key Innovation: Model tests and XFEM simulations quantify crack development, internal-force redistribution, deformation, and ultimate bearing capacity for different crack locations.
16. Centrifuge shaking table tests on the bidirectional seismic response of buried nuclear structures in saturated sites
Core Problem: Buried small modular reactor structures lack experimental evidence for seismic performance under saturated clay and liquefiable interlayer conditions.
Key Innovation: Centrifuge shaking-table tests at 50g compare site-structure kinematics and internal response under bidirectional Kobe motions of increasing intensity.
17. Multi-scale reconstruction of 3D subsurface model using deep learning techniques based on cone penetration testing data
Core Problem: Underground construction safety depends on subsurface models, but interpolation struggles with complex strata and sparse CPT coverage.
Key Innovation: GeoView uses positional encoding, neighbouring CPT information, and staged multi-scale MLP prediction to reconstruct 3-D soil-type distributions across open-source and metroline datasets.
18. From Spectral Indices to Artificial Intelligence: A Review of Remote Sensing Methodologies for Glacier Mapping
Core Problem: Glacier retreat monitoring needs transferable, validated mapping methods, yet training data and benchmark standards remain uneven.
Key Innovation: The review traces glacier mapping from spectral indices to machine learning and U-Net-style segmentation, highlighting dataset, transferability, and uncertainty gaps.
19. Autocorrelation Seismic Imaging of Northern Taiwan Using Ambient Noise Data
Core Problem: Northern Taiwan's transition from subduction to collision and post-collisional collapse requires dense constraints on crustal reflectors and Moho geometry.
Key Innovation: FormosaArray ambient-noise autocorrelation images Moho depth, lower-crustal reflectors, and shallow intracrustal interfaces that bound crustal seismicity.
20. Encoding Interactions between Polygonal Blocks into Graph Neural Network for Fast and Accurate Contact Forces
Core Problem: Implicit DDA contact-force calculations are accurate but slowed by open-close iteration at each time step.
Key Innovation: A dynamic graph neural network learns polygonal block contact-state transitions and bypasses iterative contact solving while preserving geometric and kinematic interaction information.
21. Constitutive modeling of overconsolidated gassy clay: A unified hardening approach
Core Problem: Gas bubbles can weaken or strengthen overconsolidated clay through competing cavity disruption and bubble-flooding drainage effects.
Key Innovation: A unified hardening elastoplastic model introduces a parameter for bubble effects and captures coupled overconsolidation and gas-bubble influence on clay response.
22. Influence of cutterhead cooling water on microwave-induced fracturing of hard rocks during microwave-TBM synchronous operation
Core Problem: Microwave-assisted TBM excavation must account for simultaneous cooling-water spray, temperature reduction, and thermal-crack evolution.
Key Innovation: True triaxial tests on basalt, granite, and sandstone quantify cooling-water effects on surface temperature, thermal cracking, and cutting debris during microwave-induced fracturing.
23. Automatic lining fabrication for shield tunnels: theoretical framework and full-scale test verification
Core Problem: Automatic segment assembly remains a bottleneck for self-driving and synchronous shield tunnelling.
Key Innovation: A framework combining kinematic calculation, machine vision, gap and height-difference extraction, and control strategies is verified through full-scale shield-tunnel assembly testing.
24. Buckling stability assessment of a 15 MW offshore floating wind turbine substructure with a fully coupled approach
Core Problem: Large floating wind turbine substructures can experience local buckling under sub-yield extreme loads.
Key Innovation: A fully coupled 15 MW FOWT analysis identifies heading-dependent high-risk panel regions and evaluates buckling utilization under IEC-defined extreme loading.
25. A Spatio-Temporal Machine Learning Framework for High-Accuracy Thermal Mapping of Dynamic Coastal Waters Using UAV Imagery
Core Problem: UAV thermal imagery over dynamic coastal waters is biased by nonlinear environmental and tidal effects.
Key Innovation: A spatio-temporal MLP uses geographic coordinates and tidal status to correct UAV temperatures and map a nuclear-power-plant thermal plume at high accuracy.
26. HyTC-TaNet: A Hybrid Deep Learning Model Capturing Multiday Temporal Dependencies for Daily Mean Air Temperature Estimation With Spatial Applicability Analysis
Core Problem: Daily mean air-temperature estimation from LST often ignores multiday thermal dependence and spatial applicability.
Key Innovation: HyTC-TaNet combines transformer temporal attention, CNN feature extraction, and an area-of-applicability confidence metric for satellite-based air-temperature estimation.
27. S2AM: Dynamic Center–Surround Mechanism for Remote Sensing Salient Object Detection
Core Problem: Remote-sensing saliency models often lack explicit centre-surround contrast and unstable localization in complex scenes.
Key Innovation: S2AM adapts SAM2 features with saliency-aware domain adaptation, centroid-guided localization, and structure-constrained decoding.
28. MSFE-Net: A Task-Oriented Optical–SAR Fusion Framework for Robust Industrial Object Detection
Core Problem: Image-level fusion metrics may not predict downstream object-detection performance in optical-SAR remote sensing.
Key Innovation: MSFE-Net integrates pixel-level optical-SAR fusion with YOLOv11 and demonstrates task-level robustness for oil-tank and photovoltaic-array detection.
29. Fully Automated Wind Site Assessment in Complex Terrain Using Satellite Data and Global Circulation Models
Core Problem: Wind assessment in complex terrain needs automated high-resolution roughness, terrain, canopy, and flow modelling without local calibration data.
Key Innovation: Satellite-derived DSM/DTM, forest canopy, ERA5 forcing, deep-learning roughness maps, OpenFOAM IDDES, and dynamic downscaling are combined for automated wind-site assessment.
30. Legacies of past agriculture in forest areas – soils of former field systems in northeastern Bavaria, Germany
Core Problem: Historical agriculture leaves geomorphic and soil-stratigraphic legacies that affect present forest soils and erosion interpretation.
Key Innovation: Geopedological profiles and transects across Bavarian lynchet systems identify truncation, upslope sediment accumulation, and multi-phase tillage-induced landform formation.
31. A sea–land gradient perspective on ecosystem services: unravelling spatiotemporal evolution and interaction mechanisms
Core Problem: Coastal ecosystem services vary along sea-land gradients, but their interaction mechanisms and spatial-temporal evolution remain poorly resolved.
Key Innovation: A sea-land gradient framework is used to analyse ecosystem-service changes and interaction mechanisms in a coastal natural-hazards and risk context.
32. Climate-dominated, LUCC-enhanced water–ecosystem responses in the humid temperate Tumen River basin
Core Problem: Climate and land-use/cover change jointly alter basin water and ecosystem behaviour, complicating attribution of environmental risk.
Key Innovation: The Tumen River study separates climate-dominated and LUCC-enhanced responses in a humid temperate basin.
33. Whole-life evolution of soil aggregate structure and nutrient (C-N-P) retention function in terraced citrus orchards
Core Problem: Long-term terrace construction and orchard management alter soil aggregate stability and nutrient retention on transformed slopes.
Key Innovation: A chronosequence of citrus terraces quantifies aggregate-size distribution, C-N-P stoichiometry, and depth-dependent structural evolution over 5-45 years.