TerraMosaic Daily Digest: May 11, 2026
Daily Summary
The May 11 literature is unusually dense, but its strongest signal is specific rather than broad: hazard modelling is being pulled into fully chained workflows that start with observations and end in damage-relevant quantities. The leading debris-flow paper links radar precipitation, interception-infiltration-runoff modelling, and two-phase mass-flow simulation, demonstrating that tributary-specific rainfall and snowmelt-conditioned antecedent moisture can control initiation, erosion, and runout in inaccessible alpine terrain. The landslide cluster is similarly operational: railway embankment fragility is expressed through rainfall-conditioned landslide volume and soil-capacity uncertainty; rockfall-prone cliffs are monitored at sub-millimetre scale by combining GB-InSAR and infrared thermography; reservoir landslide-tsunami warning is reframed around the start of the high-energy wave window; and DTM, multispectral, and UAV-derived representations are tested for active, inactive, and physics-simulated landslide detection.
The same movement toward process-constrained prediction appears across flood, earthquake, cryosphere, wildfire, and infrastructure papers. Flood studies move beyond extent mapping: graph networks predict volume and edge flow under physical constraints, CNN surrogates estimate both depth and velocity, DEM-fused segmentation recovers flood depth from imagery, SWOT-constrained PINNs infer river bathymetry, and SAR benchmarks expose the persistent difficulty of flooded vegetation. Earthquake and ground-deformation studies emphasize mechanism-specific fragility, including liquefiable embankments under induced seismicity, salt-layer decoupling of active faults, oceanic transform rupture barriers, near-surface strike-slip rupture in soils, and urban microzonation. The cryosphere and wildfire papers contribute infrastructure-scale datasets and coupled feedback models: Alaska thaw inventories, glacierized-catchment isotope data, drought-induced Tibetan glacier mass loss, Arctic streamflow transitions, wildfire brown-carbon forcing, active-fire forecasting, and dense forest-disturbance labels. Across these domains, the scientific contribution is not a single dominant algorithm but a clearer coupling of observation, mechanism, uncertainty, and actionable damage state.
Key Trends
The day’s methodological center is the conversion of hazard observations into state variables that engineers, forecasters, and emergency managers can use: initiation timing, wave-arrival windows, embankment settlement, flood depth, bathymetry, fire-centre location, and deformation fields.
- Mass-movement papers now resolve event chains rather than isolated susceptibility. Debris-flow forecasting, rainfall-landslide fragility, rockfall monitoring, landslide-tsunami timing, and UAV scan-to-simulation each connect forcing, mechanics, and impact more tightly than a conventional map product.
- Flood models are moving from extent to hydraulic state. Depth, velocity, bathymetry, edge flow, sediment transport, and flooded vegetation all become explicit targets, supported by physics-informed learning and satellite-ground data fusion.
- Fragility is being tied to mechanism-specific damage states. Railway embankments, liquefiable soils, piles under scour, tunnels, converter valves, and heritage structures are assessed by settlement, deformation, scour, residual displacement, and vulnerability index.
- Remote sensing is being refined around known hazard-observation failures. The selected papers address vegetation-covered floods, patch-like SAR offset errors, large-gradient mining subsidence, thermally forced rock motion, domain-shifted damage assessment, and landslide feature redundancy.
- Climate-related hazards are becoming more data-rich and feedback-aware. Permafrost thaw, glacier melt, Arctic runoff, wildfire smoke forcing, active-fire prediction, and forest disturbance attribution are handled through larger datasets or coupled physical-atmospheric models.
Selected Papers
This issue contains 62 selected papers from 2,672 papers analyzed. The leading papers connect alpine debris-flow forecasting, rainfall-landslide railway fragility, thermo-mechanical rockfall monitoring, landslide-tsunami timing, AI/DTM landslide detection, UAV scan-to-simulation, salt-controlled active-fault seismic potential, oceanic transform rupture barriers, and liquefiable-soil embankment fragility. The broader set adds physics-informed flood graph networks, storm-surge emulation, geometric flood-depth retrieval, urban flood sediment dynamics, SWOT river bathymetry, flooded-vegetation SAR mapping, volcanic eruption forecasting, wildfire radiative forcing and active-fire prediction, permafrost thaw inventories, glacierized-catchment isotope data, drought-driven glacier mass loss, coastal erosion mitigation, tunnel deformation monitoring, scour, seismic microzonation, and cross-disaster building damage assessment.
1. From sky to ground: A multi-source observations-physical model cascade for debris flow forecasting in inaccessible alpine regions
Core Problem: Remote alpine debris flows are driven by coupled rainfall, snowmelt, infiltration, runoff, and two-phase sediment transport, but sparse gauges and limited field access make event-scale forecasting difficult.
Key Innovation: A sky-to-ground cascade links X-band radar precipitation, interception-infiltration-runoff modelling, and two-phase debris-flow simulation, reproducing the 2024 Ridi event at metre-scale and hourly resolution while showing how snowmelt-conditioned antecedent moisture amplifies erosion and runout.
2. Integrated method for fragility assessment of railway embankments under rainfall-induced landslides
Core Problem: Railway embankment vulnerability under rainfall-triggered landslides is hard to quantify because historical damage records are sparse and damage states are rarely linked to landslide volume.
Key Innovation: A hybrid fragility framework combines cumulative rainfall, landslide volume, maximum-likelihood regression, and Monte Carlo uncertainty in soil capacity to derive damage-state functions for Sri Lankan railway embankments.
3. Integration of Infrared Thermography and GB-InSAR for Dynamic Monitoring of Rock Face Movements: Case Study of La Cornalle Cliff (Switzerland)
Core Problem: Thermally driven sub-millimetre rock-face movement is a precursor-relevant process that conventional rockfall monitoring rarely resolves at high temporal resolution.
Key Innovation: GB-InSAR, infrared thermography, weather observations, Fourier-wavelet analysis, and analytical thermal-diffusion modelling are integrated to quantify daily thermo-mechanical cliff displacement at La Cornalle, Switzerland.
4. Study of maximum wave arrival time of landslide-tsunami in narrow reservoirs based on wavelet transform analysis
Core Problem: Reservoir landslide-tsunami early warning requires reliable estimates of when the high-energy wave window begins, not only maximum crest arrival.
Key Innovation: Eighty-one laboratory experiments, wavelet-transform energy extraction, 2D/3D regime separation, empirical formulas, SPH validation, and the Gongjiafang case are combined to estimate maximum-wave arrival time in narrow reservoirs.
5. Leveraging artificial intelligence and digital terrain models for active and inactive landslide detection
Core Problem: Landslide inventories need methods that can distinguish active and inactive failures while balancing terrain-resolution benefits against computational cost.
Key Innovation: AdaBoost-enhanced decision trees operating on hillshade-derived DTM features are tested across cell sizes in Cyprus, showing different optimal resolutions for relict and active landslide forms.
6. UAV-Assisted Scan-to-Simulation for Landslides Using Physics-Informed Gaussian Splatting
Core Problem: Landslide simulation pipelines often separate hazard mechanics from visually realistic site representation, limiting communication and interactive assessment.
Key Innovation: A UAV-based scan-to-simulation workflow converts 3D Gaussian Splatting reconstructions into volumetric domains for MPM landslide simulation, validated on a real Hong Kong landslide site.
7. Influence of salt-related mechanical layering on the seismic potential of active faults: Insights from the southwestern Valencia Trough (W Mediterranean)
Core Problem: Source-scaling assumptions for seismic hazard can fail where mechanically weak salt layers decouple supra- and subsalt fault systems.
Key Innovation: Structural and seismotectonic analysis of the Valencia Trough quantifies alternating tectonic and salt-withdrawal controls and proposes aspect-ratio-based magnitude estimates for partially decoupled ruptures.
8. Dynamics of Rupture Barriers on Oceanic Transform Faults: Insights From the Westernmost Gofar Transform Fault
Core Problem: Creeping barriers on oceanic transform faults arrest adjacent ruptures, but their internal structure and seismic-cycle role remain poorly constrained.
Key Innovation: Machine-learning-enabled ocean-bottom seismometer catalogs with more than 150,000 relocated events reveal migrating microearthquake swarms, fluid-creep coupling, and dilatancy strengthening at the Gofar transform barrier.
9. Seismic fragility functions for embankments on liquefiable soils affected by induced seismicity: application to the Groningen region
Core Problem: Embankments on liquefiable soils require fragility functions specific to ground motion, subsoil behavior, and settlement-based damage states.
Key Innovation: More than 3,000 fully coupled nonlinear PLAXIS2D analyses with PM4Sand generate Groningen-specific fragility functions and a general model for embankment settlement under induced seismicity.
10. Sequential Feature Selection for Efficient Landslide Segmentation from Multi-Spectral Data
Core Problem: Deep landslide segmentation models often ingest many correlated bands without knowing which spectral or topographic cues are physically useful.
Key Innovation: Sequential Forward Floating Selection with a lightweight U-Net++ proxy identifies an eight-channel Landslide4Sense input subset that matches or exceeds larger configurations while improving interpretability.
11. A High-Precision Monitoring Method for Surface Subsidence in Western Chinese Mining Areas by Fusing InSAR and LiDAR
Core Problem: Mining subsidence monitoring is split between InSAR, which fails in large-gradient zones, and LiDAR, which is stronger for large deformation but weaker for subtle changes.
Key Innovation: LiDAR-based stable-feature extraction, InSAR gradient/coherence zoning, and inverse-MSE weighted fusion reduce full-gradient deformation error to 39 mm at the Sihe Coal Mine.
12. Probabilistic stability and runout dynamics of open-pit mine active dumps: A multi-parameter numerical simulation framework integrating shear strength reduction and risk assessment
Core Problem: Active mine dumps require simultaneous mapping of critical zones, failure probability, and runout distances under geo-mining and seismic conditions.
Key Innovation: UAV photogrammetry, 3D topographic modelling, shear-strength reduction, and hybrid numerical simulations are combined to assess dump stability and safe working limits in the Raniganj coalfield.
13. Convolutional neural network model for rapid prediction of urban flash flood water depth and velocity maps
Core Problem: Real-time flash-flood prediction is constrained by the cost of 2D shallow-water simulations, especially when both depth and velocity maps are needed.
Key Innovation: A CNN surrogate trained on hydrodynamic simulations uses rainfall hyetographs and hydro-morphological descriptors to predict urban flood depth and velocity envelopes for Matera, Italy.
14. DUALFloodGNN: Physics-informed Graph Neural Network for Operational Flood Modeling
Core Problem: Fast flood surrogates must preserve hydrodynamic consistency while remaining efficient enough for operational autoregressive forecasts.
Key Innovation: DUALFloodGNN embeds global and local physical constraints in a graph neural network that jointly predicts node water volume and edge flow using multi-step curriculum learning.
15. PACT: Peak-Aware Cross-Attention Graph Transformers for Efficient Storm-Surge Emulation
Core Problem: Coastal hazard planning needs fast storm-surge emulators that retain peak behaviour across heterogeneous atmospheric forcing histories.
Key Innovation: PACT combines graph-encoded forcing patches, station-query cross-attention, temporal transformers, and peak-aware losses to emulate storm-surge forecasts at US Northeast tide gauges.
16. Geometric Flood Depth Estimation: Fusing Transformer-Based Segmentation with Digital Elevation Models
Core Problem: Post-disaster flood masks do not provide the vertical water-depth information needed for navigability, exposure, and structural-risk assessment.
Key Innovation: Mask2Former flood segmentation is fused with DEM-derived water-land boundaries to estimate water-surface elevation and per-pixel flood depth without running a hydrodynamic model.
17. An Integrated Modeling Framework for Sediment Dynamics During Urban Flooding: Application to Hurricane Harvey in Houston
Core Problem: Urban flood impacts include sediment redistribution, yet many flood models focus only on water depth and extent.
Key Innovation: An integrated modelling framework for Houston links flood hydraulics with sediment dynamics to evaluate how extreme urban inundation mobilizes and deposits sediment.
18. Semantic Segmentation of Flooded Vegetation in SAR Imagery: A Benchmark Dataset and Comparative Study of Deep Learning Models
Core Problem: Flooded vegetation remains a difficult class for SAR-based flood delineation because canopy structure obscures open water signatures.
Key Innovation: VegFlood provides 1,707 Sentinel-1 image-mask pairs, DEM and hydrological context, and model baselines showing how vegetation density, water level, and geomorphology control segmentation accuracy.
19. Advancing river bathymetry mapping through physics-informed neural networks and SWOT satellite observations
Core Problem: Most river beds remain unsurveyed, limiting flood models even as SWOT now observes water-surface elevations globally.
Key Innovation: A dual-network PINN assimilates multi-temporal SWOT water-surface observations with gradually varied flow physics to infer riverbed elevation and quantify sensitivity to hydraulic parameter uncertainty.
20. Forecasting volcanic eruptions across scales
Core Problem: Volcano monitoring can often detect unrest but still struggles to translate signals across eruption scales, durations, and impact-relevant forecast horizons.
Key Innovation: The Science perspective frames eruption forecasting as a cross-scale problem, emphasizing the need to connect short-term monitoring with longer-term constraints on volcanic systems.
21. Strong global radiative effects from wildfire dark brown carbon
Core Problem: The warming contribution of wildfire dark brown carbon is poorly constrained despite its prevalence in smoke plumes.
Key Innovation: Aircraft, ground, satellite, and aerosol-climate modelling show that strongly absorbing wildfire brown carbon is globally widespread and can exert radiative effects comparable to or exceeding black carbon under upper-bound assumptions.
22. A comprehensive database of thawing permafrost locations across Alaska: version 2.0.0
Core Problem: Abrupt permafrost thaw damages ecosystems and infrastructure, but regional inventories remain fragmented across field and remote-sensing sources.
Key Innovation: The Alaska Permafrost Thaw Database compiles 19,540 thaw and thermokarst locations from 44 sources, spanning 1950-present observations across multiple spatial resolutions.
23. Large-scale isotopic fingerprinting of cryosphere and hydrological components in glacierized catchments
Core Problem: Hydrological partitioning in glacierized catchments is limited by the lack of globally harmonized stable-isotope measurements across cryospheric endmembers.
Key Innovation: A global georeferenced database compiles 12,348 δ18O and δ2H measurements from 63 publications, covering precipitation, snow, glacier ice, meltwater, rock glaciers, permafrost thaw waters, lakes, streams, and groundwater.
24. The processes and impacts of drought-induced extreme glacier mass loss on the south‐central Tibetan Plateau
Core Problem: Drought can intensify glacier mass loss, but its process chain through albedo, snowfall, radiation, and lake water budgets remains under-resolved.
Key Innovation: Energy-mass balance reconstruction for 853 glaciers links ablation-season drought to exceptional shortwave radiation, reduced snowfall, enhanced melt, and downstream lake-budget effects.
25. Climate change-driven runoff variations in alpine catchments: Quantitative attribution using three innovative methods coupled with cryospheric processes
Core Problem: Runoff attribution in high mountains is uncertain when glacier melt and evapotranspiration are not directly observed at basin scale.
Key Innovation: A coupled VIC-OGGM framework and three attribution methods separate climate, glacier meltwater, and land-surface-change contributions to Qilian Mountain runoff variability.
26. Study on the mechanical and hydrological properties of vegetation bag root-soil-bag composites under heavy rainfall conditions
Core Problem: Heavy rainfall threatens earthen slopes, and nature-based slope reinforcement must quantify mechanical and hydrological effects together.
Key Innovation: Root morphology, tensile strength, shear testing, permeability, rainfall interception, and slope models show how vegetated geobags stabilize slopes through canopy interception, root reinforcement, and bag retention.
27. Topographical Cues as Determinants of Landslide Risk Perception
Core Problem: Risk communication can fail when perceived landslide danger diverges from terrain-driven exposure and susceptibility.
Key Innovation: The study analyzes topographic cues as determinants of landslide risk perception, linking physical terrain attributes to how people interpret slope hazard.
28. Advancing the BRAMS wildfire–atmosphere modelling system: application to an extreme wildfire event
Core Problem: Wildfire smoke perturbs radiation, stability, and weather, but coupled fire-atmosphere simulations remain difficult at local to mesoscale.
Key Innovation: BRAMS v6.0 integrates crown-fire spread, dynamic smoke emissions, aerosol-radiation interactions, and atmospheric feedbacks, reproducing optical-depth structure during the 2017 Portugal wildfire.
29. Set Prediction for Next-Day Active Fire Forecasting
Core Problem: Wildfire forecasts often predict broad fire danger rather than localized active-fire events.
Key Innovation: WISP reframes next-day fire prediction as ranked point-set forecasting on a 375 m grid using meteorology, vegetation, land, and fire-history covariates with Hungarian matching.
30. Rapid Forest Fuel Load Estimation via Virtual Remote Sensing and Metric-Scale Feed-Forward 3D Reconstruction
Core Problem: Wildfire risk assessment needs affordable vertical fuel-load information beyond what routine satellite imagery provides.
Key Innovation: Virtual remote sensing, feed-forward 3D reconstruction, metric-scale alignment, BEV projection, and watershed segmentation estimate forest structure, LAI, and combustible fuel load.
31. Dense labelled time-series for mapping European forest disturbance agents
Core Problem: Disturbance-agent mapping is limited by sparse, inconsistent labels and weak temporal segmentation.
Key Innovation: DISFOR provides dense 2015-2024 European forest disturbance-agent labels, Sentinel-2 time series, image chips, interpreter confidence, and hierarchical agent classes including wildfire.
32. Shifting streamflow regimes in the arctic: a multi-basin analysis using the VIC model
Core Problem: Arctic warming is altering freshwater export, but basin-specific hydrological regime shifts remain unevenly constrained.
Key Innovation: VIC simulations and observations across six major Arctic rivers quantify transitions from snowmelt-dominated to increasingly rainfall-dominated streamflow regimes.
33. Projected changes in Arctic river streamflow and shifting climatic drivers: A linear ensemble deep learning approach
Core Problem: High-latitude runoff projections must represent nonlinear interactions among warming, precipitation, evapotranspiration, and permafrost degradation.
Key Innovation: A linear ensemble deep-learning approach projects six Arctic river basins and identifies shifting climatic drivers as permafrost extent declines.
34. Varying the Combination of Hydrological Models in Time and Space: Toward a More Accurate Representation of Streamflow in Large‐Sample Hydrology
Core Problem: Large-sample hydrology often relies on fixed model structures that cannot adapt to changing high- and low-flow conditions.
Key Innovation: An analog-based dynamic combination of 156 FUSE model simulations across 559 US catchments varies model weights in space and time to reduce high-flow and low-flow trade-offs.
35. Hundred-year changes in water and sediment in the Inner Mongolia reach of the Yellow River: trend transitions, driving mechanisms, and attribution quantification
Core Problem: Flood-season water and sediment changes over a century reflect intertwined reservoir operation, land use, vegetation, and climate drivers.
Key Innovation: Hydrological and sediment records from 1919-2024 are segmented by trends, abrupt changes, cycles, and attribution to quantify phase-specific runoff and sediment declines.
36. Scaling relationships between catchment area and peak discharge across the Fitzroy Basin, eastern Australia
Core Problem: Extreme-discharge estimation in large basins is complicated by uneven gauge records, nonstationary cross sections, and nonuniform rainfall.
Key Innovation: Historical peak flows from 136 gauges in Australia's Fitzroy Basin are used to quantify area-discharge scaling and identify gauges that disproportionately influence flood-frequency relationships.
37. Geo-spatial assessment of human-induced soil erosion and environmental hazard using the RUSLE model: A path toward sustainable land management in the Duhok Watershed
Core Problem: Land-use change can concentrate soil loss and sediment delivery, but priority zones need spatially explicit erosion-hazard mapping.
Key Innovation: RUSLE, GIS, remote sensing, and land-use overlays map erosion hazard in the Duhok watershed, identifying very high and extremely high soil-loss zones linked to farming, urban expansion, and damming.
38. Projected effects of land use change on future soil erosion in southern Italy
Core Problem: Planning for erosion resilience requires linking future land-use scenarios with rainfall erosivity and sediment-delivery services.
Key Innovation: Dyna-CLUE and InVEST are coupled with dense field data to reconstruct and project soil erosion regulation in southern Italy from 1960 to 2060.
39. Submergence controls on wave transmission for concrete-based oyster mats and oyster shell bags
Core Problem: Living-shoreline structures must balance oyster survival with wave attenuation, but performance declines as submergence increases.
Key Innovation: Hydrodynamic testing of oyster mats and shell bags quantifies how structural width and water depth control wave-energy reduction for shoreline erosion mitigation.
40. Experimental analysis of trunk scour on the seaward side of a detached low-crested rubble-mound breakwater under irregular waves
Core Problem: Scour at low-crested breakwaters threatens structural stability, but ripple-scale morphology can obscure scour-hole measurements.
Key Innovation: Mobile-bed experiments and inverse continuous wavelet transform post-processing isolate wave-number, period, height, and crest-submergence controls on seaward trunk scour.
41. Response of Laterally Loaded Piles Under Combined Scour and Vertical Loading Conditions
Core Problem: Coastal pile foundations can lose lateral capacity when scour reduces embedment and shifts bending demand.
Key Innovation: Laboratory tests and PIV capture soil displacement fields, showing how general and local scour alter rotation centers, shear zones, bending moments, and P-delta effects.
42. A nonlinear phenomenological model for beach evolution
Core Problem: Existing shoreline-instability models can obscure dynamic controls behind quasi-periodic beach undulations.
Key Innovation: A nonlinear one-line-model formulation represents wave attack, alongshore currents, and scale-selective damping, reproducing sinuous shoreline evolution and chaotic planform behavior.
43. Coupled coastal modelling system and particle tracking model for shoreline morphodynamics and management at the Kitchener Drain Outlet, Northern Nile Delta, Egypt
Core Problem: Sediment imbalance and engineering interventions at coastal outlets require scenario-based assessment of erosion and accretion trade-offs.
Key Innovation: A coupled Coastal Modelling System and Particle Tracking Model evaluates nourishment, sediment traps, jetties, groins, spur dikes, and detached breakwaters at the Kitchener Drain outlet.
44. A method for predicting the rock strata subsidence boundary at deep-buried mining areas with thin bedrock and Thick soil layer
Core Problem: Mining subsidence boundaries are difficult to predict where thick soil layers overlie thin bedrock and key strata control deformation transfer.
Key Innovation: A boundary-prediction method characterizes rock-strata movement and surface subsidence evolution for deep-buried coal mining under complex overburden conditions.
45. Full-field deformation quantification of underground tunnels using SLAM LiDAR point cloud based on unsupervised neural implicit learning
Core Problem: Tunnel risk management needs robust full-field deformation quantification on irregular point-cloud surfaces without manual parameter tuning.
Key Innovation: NIFCyl uses unsupervised neural implicit learning and cylinder-based spatial averaging on SLAM LiDAR point clouds, outperforming M3C2 in synthetic and active mine-tunnel tests.
46. Numerical modeling and validation of backfill grouting in shield tunneling using the material point method
Core Problem: Backfill grouting quality depends on grout filling patterns and pressure evolution in the shield tail gap, which are hard to measure directly.
Key Innovation: GPU-accelerated v-p WCGIMP material point modelling is validated against grouting tests and applied to engineering-scale shield tunnel simulations.
47. A Span–Strain Framework for Determining Sequential Excavation Strategies in Tunnelling
Core Problem: Tunnel excavation strategies often require site parameters that are expensive to determine for small and medium projects.
Key Innovation: A span-strain decision graph built from 672 numerical simulations predicts SEM choices with substantial agreement against 96 tunnel projects.
48. Vlasov beam-based modeling of multimodal deformations in rectangular tunnels under 3D soil-tunnel interactions
Core Problem: Large rectangular tunnels can experience torsion, distortion, warping, flattening, bending, and shear that simplified models omit.
Key Innovation: A Vlasov beam-based 3D soil-tunnel interaction model derives finite-element solutions for six deformation modes and validates them across three case studies.
49. Discrete Element Investigation of the Influence of Shallow Soil Density on the Manifestations of Strike-Slip Surface Fault Rupture
Core Problem: Surface fault rupture patterns depend on shallow soil density, affecting fault-displacement hazard at the ground surface.
Key Innovation: Three-dimensional DEM simulations with tens of millions of grains show how dense soils localize multiple shear bands while loose soils develop narrower diffuse deformation zones.
50. Role of Seismic Microzonation Studies in Defining Design Response Spectra for Urban Areas
Core Problem: Urban earthquake design needs local response spectra that reflect site effects rather than generic code values.
Key Innovation: The paper reviews and applies microzonation principles for defining design response spectra in urban areas.
51. Geographic Bias Analysis and Cross-Domain Generalization in Deep Learning-Based Building Damage Assessment
Core Problem: Satellite damage-assessment models can fail when deployed outside the geographies represented in training data.
Key Innovation: Systematic xView2 evaluation across 17 disaster locations shows strong geographic bias and tests fusion augmentation, supervised fine-tuning, and CORAL domain adaptation for unseen regions.
52. DA-SegFormer: Damage-Aware Semantic Segmentation for Fine-Grained Disaster Assessment
Core Problem: Emergency response requires fine-grained damage classes, but minor and major damage are underrepresented and texture cues are lost during resizing.
Key Innovation: DA-SegFormer adds class-aware sampling, online hard-example mining, Dice loss, and resolution-preserving inference to improve UAV damage segmentation on RescueNet.
53. Robust Building Damage Detection in Cross-Disaster Settings Using Domain Adaptation
Core Problem: Damage classifiers trained on multi-disaster benchmarks often fail under unseen geographic and disaster-domain shifts.
Key Innovation: A two-stage supervised domain-adaptation ensemble adapts the xView2 first-place pipeline to Ida-BD and shows that domain adaptation is essential for trustworthy disaster-response deployment.
54. Suppressing the Patch-like Errors of SAR Intensity Offset Tracking Based on Z-Score Standardization and INFLO Structural Density Analysis
Core Problem: SAR offset tracking can generate patch-like errors from isolated high-intensity scatterers, reducing deformation-field reliability.
Key Innovation: Z-score screening and INFLO structural-density analysis remove outlier scatterers before NCC, increasing valid offset coverage and deformation-estimation quality in Amnye Machen and Central Tianshan tests.
55. Shear behaviour of foam-conditioned gravelly clay: insights from unconsolidated undrained triaxial compression tests
Core Problem: Shield tunnelling in gravelly clay requires reliable conditioning to control excavated muck behavior under undrained loading.
Key Innovation: Unconsolidated undrained triaxial compression tests quantify shear behavior of foam-conditioned gravelly clay for EPB shield applications.
56. Seismic performance and failure mechanisms of minarets: lessons from the 2023 Kahramanmaraş Earthquakes for structural design and mitigation
Core Problem: Slender heritage structures remain vulnerable to earthquake failure mechanisms that are not fully captured by generic building guidance.
Key Innovation: Observed damage from the 2023 Kahramanmaraş earthquakes is used to identify minaret failure mechanisms and implications for structural design and mitigation.
57. Assessment of the seismic vulnerability index of three Moorish houses in the Albayzín, Granada (Spain)
Core Problem: Heritage earthen buildings need vulnerability assessment that respects geometry, materials, construction, and rehabilitation state.
Key Innovation: The SVIVA index is applied to three listed Moorish houses to identify structural factors controlling seismic vulnerability.
58. Visco-elastically connected shear-flexural structure: elastoplastic seismic response property and a simplified analysis procedure
Core Problem: Seismic design needs simplified procedures that retain essential elastoplastic response of coupled shear-flexural systems.
Key Innovation: A visco-elastically connected shear-flexural model is developed to approximate nonlinear seismic response for structural assessment.
59. Inelastic seismic response of low-rise multistory realistic buildings under bidirectional interaction due to ground excitation
Core Problem: Low-rise buildings can experience bidirectional seismic interaction effects that alter inelastic response.
Key Innovation: Numerical analysis evaluates bidirectional interaction under ground excitation for realistic low-rise multistory buildings.
60. Shaking table test of a novel slope-type sliding pendulum isolation system for seismic protection of electrical equipment
Core Problem: Electrical equipment needs isolation systems that reduce seismic acceleration while maintaining re-centering capacity.
Key Innovation: Shaking-table tests of a slope-type sliding pendulum isolation system demonstrate reduced seismic response and residual displacement for scaled converter-valve equipment.
61. Inherited Structure of a Paleo‐Subduction Zone Beneath the Northern Tibetan Plateau: Insights From 3‐D Anisotropic Magnetotelluric Imaging
Core Problem: Ancient sutures may remain weak zones that localize deformation and seismicity, but their deep structure is poorly constrained.
Key Innovation: Three-dimensional anisotropic magnetotelluric imaging identifies a northeast-dipping low-resistivity paleo-subduction structure interpreted as a fluid-bearing inherited lithospheric scar.
62. Real-time phase-resolved wave prediction over planar coastal bathymetries using U-Net convolutional neural networks
Core Problem: Coastal warning and operations require short-lead wave prediction under nonlinear shoaling and breaking.
Key Innovation: A compact U-Net predicts nearshore surface-elevation time series from a single offshore gauge over planar bathymetries, outperforming linear and second-order wave-theory baselines.