TerraMosaic Daily Digest: August 25, 2026
Daily Summary
Direct slope-hazard studies emphasize internal state evolution over simple trigger counts. Rock-avalanche forecasting is systematically validated across 22 case histories with GPU-accelerated Bayesian calibration, while debris-flow prediction improves by learning spatially and temporally varying parameters inside a two-phase model. Landslide case studies resolve delayed failure through confluence-fed groundwater recharge at Jure, staged avalanche-creep-runout behavior under mining and rainfall at Chupitian, and clustered shallow failures driven by rapid transition from antecedent drought to extreme rainfall; companion studies on loess, residual clay, soil-rock mixtures, and soil-water characteristic curves sharpen the material mechanics behind wetting-induced weakening and flow-like transition.
Seismic, coastal, hydroclimatic, and climate-exposure papers similarly replace simplified hazard descriptors with structure-aware and compound formulations. Single-station state-space learning, Bayesian inversions of fault stress and oceanic transform sources, sub-meter offshore seismic imaging, empirical site-amplification synthesis, and stochastic tunnel fragility analyses all improve how earthquakes are detected, characterized, or translated into infrastructure demand. Beyond seismic hazards, recent analyses define worst-case moraine-breach volumes for glacial lake outburst floods, show how extreme tsunamis can erase reef shielding and generate harbor vortices, classify high-frequency North Sea sea-level extremes, link U.S. springtime precipitation extremes to BAM-modulated MCS-cyclone coupling, project climate-conditioned probable maximum precipitation and rainfall-driven portfolio risk, and extend exposure analysis to wildfire-prone settlements, urban and humid heat, glacier retreat, and East Greenland Atlantification.
A distinct but important methods stream expands analytical capacity without itself constituting geohazard validation unless directly tested. Wide-area potential geohazard screening fuses InSAR, optical imagery, terrain, and vector data; landslide susceptibility mapping adds bootstrap uncertainty and township holdouts; and remote-sensing studies advance cloud-mask super-resolution, onboard cloud removal, knowledge-guided change synthesis, weakly supervised sonar segmentation, multimodal diffusion reconstruction, adaptive land-cover fusion, and SAR-based wave or 3D target retrieval. Across these papers, the common direction is toward physics-aware modeling, explicit uncertainty, computational efficiency, and stricter evaluation of generalization, data leakage, and reproducibility.
Key Trends
Five scientific and methodological trajectories define the August 25 literature: state-dependent slope mechanics, uncertainty-aware hybrid forecasting, region- and structure-aware seismic assessment, compound coastal-hydroclimatic hazard framing, and a distinct stream of transferable scientific-AI methods built around physics and evaluation rigor.
- Hydrological memory is emerging as a first-order slope-failure control: The strongest landslide papers move beyond event-total rainfall toward persistent internal water-state evolution. The Jure study attributes delayed failure to continued recharge from a confluence area after peak rainfall, the clustered shallow-movement study identifies rapid drought-to-deluge transition as the dominant trigger, and the loess and NMR-based soil-water studies resolve how moisture, shrink-swell behavior, and thixotropy reorganize strength before failure or remobilization.
- Probabilistic and hybrid runout forecasting is becoming more defensible: Several papers improve forecast credibility by embedding learning inside physically structured models and by quantifying uncertainty explicitly. Rock-avalanche runout is benchmarked through Bayesian calibration and leave-one-out validation over 400,000 simulations, debris-flow mobility is improved by machine-learned parameter fields within a two-phase model, landslide displacement forecasting uses sequential Bayesian dictionary learning, and Bayi susceptibility mapping evaluates both predictive skill and spatial uncertainty with bootstrap and township holdouts.
- Seismic assessment is shifting toward sparse-sensor inference and coupled infrastructure response: The earthquake cohort emphasizes both faster source inference and more realistic translation to damage. SeisMamba shows that single-station waveform modeling can retain useful cross-region performance for early warning, Bayesian studies infer evolving fault stress and transform-fault source properties under limited observations, offshore imaging revises marine fault slip hazard upward, and tunnel, bridge, site-amplification, and damage-classification papers show that heterogeneous soils, pulse-like motions, interface mechanics, and empirical amplification structure materially change infrastructure assessment.
- Coastal and hydroclimatic extremes are being reframed as compound and upper-bound problems: Multiple studies treat hazard not as a single peak variable but as the interaction of timing, geometry, and scenario severity. Tsunami modeling shows reef shielding can collapse under stronger Manila Trench scenarios and create harbor vortices, the GLOF brief derives worst-case breach depths and flood volumes, North Sea analysis isolates high-frequency sea-level event types and their compound amplification with residual extremes, and precipitation studies connect U.S. MCS extremes to BAM-cyclone coupling while projecting larger-area probable maximum precipitation under climate change.
- Transferable Earth-observation and scientific-AI methods now prioritize physics, uncertainty, and evaluation discipline: A large secondary stream advances methods that are relevant to hazard science without always being hazard-validated. This includes correlated aleatoric-epistemic uncertainty for dense prediction, solver-emulator comparisons for PDE learning, operator learning with embedded physics, diffusion-based structural or multimodal geospatial reconstruction, CUDA morphology, cloud-mask downscaling, onboard cloud removal, and explicit warnings about data leakage and replicability. The common trajectory is stronger physical structure and more careful claims about generalization.
Selected Papers
The selected papers combine validated studies of landslides, debris flows, earthquakes, tsunamis, coastal flooding, glacial lake outburst floods, wildfire exposure, subsidence, heat, and cryospheric change with a parallel cohort of remote-sensing, simulation, and scientific-AI methods. Read the first set as domain-specific hazard evidence, and the second as transferable analytical advances whose relevance to geohazards depends on explicit application and testing.
1. How well can rock avalanche runout be forecasted? GPU-Accelerated validation of a method for forecasting rock avalanche motion using a numerical model
Core Problem: Rock-avalanche forecast skill across rheologies and case types had not been comprehensively validated because of data and compute limits.
Key Innovation: Uses GPU-accelerated Bayesian calibration and leave-one-out validation of ORIN-3D across 22 case histories and about 400,000 simulations.
2. SeisMamba: Low-Latency Single-Station Seismic Magnitude Estimation for Spatially Distributed Earthquake Early Warning
Core Problem: Fast, accurate earthquake magnitude estimation is hard in regions lacking dense regional sensor networks.
Key Innovation: A lightweight single-station Mamba waveform model that improves accuracy-latency tradeoffs and cross-region robustness.
3. Mechanisms of reef shielding breakdown and harbor vortex formation under extreme tsunamis: A case study of Yongxing Island, the South China Sea
Core Problem: Reef shielding can fail during extreme tsunamis, creating hazardous harbor vortices and amplified impacts.
Key Innovation: Identifies breakdown mechanisms linking reef shielding loss to vortex formation in a real island setting.
4. Brief communication: Towards defining the worst-case breach scenarios and potential flood volumes for moraine-dammed lake outbursts
Core Problem: GLOF modeling needs defensible upper-bound breach depths and flood volumes for moraine-dammed lakes.
Key Innovation: Derives a methodology for estimating maximum breach depth and potential outburst flood volume from past failures.
5. Post-LGM intensification of marine faulting: resolution-dependent hazard assessment
Core Problem: Existing offshore seismic-hazard assessments may miss elevated fault slip rates and episodic activity.
Key Innovation: Uses sub-meter seismic imaging to reveal much higher marine fault slip rates and glacial-cycle-linked accelerations.
6. Enhancing debris-flow mobility and hazard forecasting using machine-learned spatial-temporal parameters in a two-phase flow model
Core Problem: Fixed parameters in two-phase debris-flow models fail to capture spatial-temporal variability and limit forecast skill.
Key Innovation: Learns dynamic internal parameter fields from terrain, rainfall, and vibration data while preserving a physically interpretable two-phase flow model.
7. Predictive Performance and Resampling-Based Prediction Uncertainty of a Stacking Ensemble for Landslide Susceptibility Assessment in Bayi District, China
Core Problem: Ensemble landslide susceptibility studies often emphasize accuracy while underexamining uncertainty and spatial stability.
Key Innovation: Uses bootstrap and township holdout evaluation to show a Stacking ensemble can improve both predictive skill and uncertainty characterization.
8. Automatic Identification and Assessment of Potential Geohazards in a Wide Area Based on Multisource Remote Sensing and Deep Learning
Core Problem: Potential geohazard identification across wide areas still depends heavily on manual interpretation and fragmented workflows.
Key Innovation: Automates deformation-zone delineation, threatened-object detection, geohazard screening, and risk assessment by fusing InSAR, optical imagery, terrain, and vector data.
9. Hydro-mechanical triggering mechanism of the delayed Jure landslide: Critical role of progressive infiltration from the confluence area in Southern Himalaya
Core Problem: Delayed post-rainfall landslides are poorly explained by conventional rainfall-threshold logic.
Key Innovation: Quantifies how confluence-area groundwater recharge progressively elevates pore pressure and drives delayed failure in the Jure landslide.
10. Mechanism and kinetics of catastrophic landslide triggered by mining and rainfall at Chupitian Village, Guizhou, China, June 10, 2021
Core Problem: The failure sequence, creep behavior, and kinetic controls of a mining- and rainfall-triggered catastrophic landslide were unclear.
Key Innovation: Integrates UAV mapping, SBAS-InSAR, Massflow simulation, and GBRT-SHAP analysis within one event reconstruction.
11. Extreme rainfall-induced clustered shallow mass movements: surface disturbances, prediction, and cascading hazards
Core Problem: How extreme rainfall triggers clustered shallow mass movements, surface disturbance, and cascading impacts.
Key Innovation: An integrated treatment of clustered shallow failures spanning disturbance patterns, prediction, and cascading hazards.
12. Physics-guided sequential Bayesian dictionary learning with increment-domain modeling for landslide displacement prediction
Core Problem: How to improve landslide displacement prediction under complex, evolving temporal dynamics.
Key Innovation: A physics-guided sequential Bayesian dictionary-learning approach in increment space for adaptive landslide forecasting.
13. Rainstorm Erosion Hotspots and Traits Across Loess Plateau Management Zones
Core Problem: Which erosion processes and hotspot landforms dominate small catchments across Loess Plateau management zones after extreme rainstorms?
Key Innovation: Combines 30 post-event surveys with UAV orthophotos and DTMs to quantify process-specific erosion intensity, event density, and landslide hotspot settings.
14. Bayesian Inference of Complex Stress Evolution in Rate-and-State Governed Faults Constrained by Seismicity Rate Observations
Core Problem: How can time-varying fault stress evolution and uncertainty be inferred from seismicity rates without dense geodetic coverage?
Key Innovation: Introduces a physics-informed Bayesian inversion with flexible trapezoidal stress sources and uncertainty estimation, validated on the Reno-Mogul swarm.
15. Bayesian Source Characterization of Oceanic Transform Fault Earthquakes Using Relative Teleseismic Rayleigh Wave Spectra
Core Problem: How to recover source location, depth, and rupture complexity for oceanic transform earthquakes despite path effects and limited local observations?
Key Innovation: Uses relative teleseismic Rayleigh-wave spectra with inferred data errors to suppress 3-D path bias and improve Bayesian source estimates.
16. Enhanced Occurrence of Springtime Extreme Precipitation Associated With Mesoscale Convective Systems Over the US by Baroclinic Annular Mode
Core Problem: Why do U.S. springtime MCS extremes intensify under certain large-scale circulation states?
Key Innovation: Links positive BAM phases to slower extratropical cyclones and a near doubling of extreme MCS precipitation occurrence.
17. UHI-Bench: Benchmarking Dual-Source Urban Heat Island Modeling Across Cities in Diverse Climate Regimes
Core Problem: How can land-surface and near-surface air-temperature UHI signals be modeled jointly across cities despite missing data and spatiotemporal mismatch?
Key Innovation: Introduces the first dual-source UHI benchmark with unified cross-city transfer tasks, environmental covariates, and multi-family baseline comparisons.
18. Offshore turbidity currents forecasting (Part II): Anomaly detection using deep learning
Core Problem: Offshore turbidity currents need anomaly-detection tools for forecasting hazardous events.
Key Innovation: Applies deep learning anomaly detection to forecast turbidity-current behavior.
19. Exposure of settlements to wildfires in a transboundary wildland-urban interface region in Central Europe
Core Problem: Settlements in a transboundary wildland-urban interface need spatially explicit wildfire exposure assessment.
Key Innovation: Combines satellite data, local information, simulations, and validation into actionable cross-border exposure maps.
20. Investigation of North Sea high-frequency sea-level extremes using long-term sea-level measurements
Core Problem: Standard hourly sea-level analyses miss short-period extremes that can still amplify hazardous coastal water levels.
Key Innovation: Combines long-term tide-gauge decomposition and clustering to classify high-frequency sea-level extremes and compound events.
21. Influence of water saturation on sandstone strainburst failure under true-triaxial unloading with different intermediate principal stresses
Core Problem: The coupled effects of water saturation and intermediate stress on strainburst intensity were poorly resolved.
Key Innovation: Shows through true-triaxial unloading tests that saturation suppresses dynamic strainburst even as failure severity can remain high.
22. Stochastic Seismic Fragility Analysis for Tunnel Linings with Rubber Concrete Under Near-Fault Ground Motions
Core Problem: How near-fault motion variability affects tunnel lining damage and whether a rubber concrete interlayer reduces fragility.
Key Innovation: A stochastic IDA-based fragility framework calibrated with lab-tested rubber-sand-concrete damping properties.
23. Investigation of tunnel-soil-structure interaction under dynamic loading in non-homogeneous soils: case study of Catania, Sicily, Italy
Core Problem: How soil heterogeneity, tunnel depth, building position, and seismic input shape coupled tunnel-soil-structure response.
Key Innovation: A fully coupled finite-element case study that tests configuration-dependent seismic interaction in non-homogeneous soils.
24. On the consistency of site amplification estimates from empirical and numerical methods in Italy
Core Problem: Whether empirical site-amplification estimates are mutually consistent and how they compare with 1D numerical simulations.
Key Innovation: A national-scale Italian comparison showing empirical amplification estimates are robust and outperform simple 1D models.
25. Improving the accuracy of EXSIM-estimated horizontal PGA and 5% damped PSA for South Korea using two-stage correction models
Core Problem: How to reduce EXSIM bias in PGA and PSA estimation for South Korea using limited observations.
Key Innovation: A two-stage residual-correction workflow combining regression and LightGBM with source, site, and topographic predictors.
26. Rapid post-earthquake damage assessment using patch-level CNNs and VLM
Core Problem: How to accelerate building damage assessment from post-earthquake imagery beyond manual inspection.
Key Innovation: A large patch-level indoor damage dataset and CNN-versus-VLM benchmark for rapid binary damage classification.
27. A machine learning-based hazard assessment framework for urban road subsidence induced by pipeline damage
Core Problem: How to assess urban road subsidence hazard caused by buried pipeline damage.
Key Innovation: A machine-learning hazard-assessment framework targeted specifically at pipeline-induced subsidence risk.
28. Physics-Integrated Operator Learning via Gaussian Splatting Representations
Core Problem: Data-driven neural operators drift during long rollouts and underuse known physics.
Key Innovation: Gaussian-splat continuous fields inject governing operators directly into evolution maps.
29. It depends: Incorporating correlations for joint aleatoric and epistemic uncertainties of high-dimensional output spaces
Core Problem: High-dimensional outputs need joint aleatoric and epistemic uncertainty without full-covariance cost.
Key Innovation: A unified low-rank-plus-diagonal covariance approximation for correlated joint uncertainty in dense prediction tasks.
30. From Numerical Simulators of PDEs to Neural Emulators and Back
Core Problem: Neural PDE emulators are often treated as black-box replacements rather than analyzed against numerical solvers.
Key Innovation: A unified solver-emulator view using Fourier analysis, benchmarking, and differentiable-physics studies.
31. Data Leakage Inflates Generalizability of Power Outage Prediction Models
Core Problem: Power-outage prediction studies overstate generalization because random splits leak spatial and temporal structure.
Key Innovation: A systematic comparison of random, leave-one-state-out, and leave-one-event-out tests showing major performance collapse.
32. Deep Learning Super Resolution for Satellite Cloud Mask Downscaling
Core Problem: High-frequency geostationary cloud masks are too coarse for many downstream monitoring applications.
Key Innovation: Cross-sensor deep super-resolution and a matched SEVIRI-MODIS cloud-mask dataset for 4x downscaling.
33. TorchMorph: CUDA-accelerated Morphological Transforms
Core Problem: Morphological transforms and distance transforms are bottlenecked by CPU-only tooling outside GPU training loops.
Key Innovation: A CUDA PyTorch extension covering high-dimensional morphology, distance transforms, and related operators with SciPy-like APIs.
34. Weakly Supervised Seafloor Segmentation for Seagrass Habitat Mapping in Side-Scan Sonar Imagery
Core Problem: Dense manual annotation limits side-scan-sonar seafloor mapping in deep or turbid environments.
Key Innovation: Weakly supervised ViT segmentation with CAM refinement, CRF tuning, and self-training for acoustic imagery.
35. ORBITALIF: An Efficient Spiking Federated Learning Framework for Onboard Cloud Removal
Core Problem: Cloud cover and downlink constraints delay disaster-relevant Earth observation
Key Innovation: Combines spiking onboard inference with inter-satellite federated learning for energy-efficient cloud removal
36. Real-World Knowledge-Guided Change Data Synthesis for Remote Sensing
Core Problem: Expanding change-detection training data without relying on rigid handcrafted change rules.
Key Innovation: Uses pretrained vision-language models to reason about plausible change locations and class transitions inside a unified change-synthesis framework.
37. Experimental study on wave attenuation through living shorelines of submerged breakwater and marsh vegetation
Core Problem: Living shorelines need quantified performance for reducing incoming wave energy.
Key Innovation: Experimentally evaluates wave attenuation from combined submerged breakwater and marsh-vegetation shorelines.
38. Quantifying the Wave Attenuation of Hybrid Seagrass-Reef Structures - An Experimental Study
Core Problem: Coral-reef monitoring needs scalable observations across habitat, environment, threat, and ecosystem-service metrics.
Key Innovation: Synthesizes modern reef remote-sensing capabilities across mapping, environmental monitoring, and management-relevant products.
39. Assessing financial risk to property portfolios from physical rainfall extremes
Core Problem: Property portfolios need practical ways to quantify losses from extreme-rainfall-driven flooding.
Key Innovation: Links open hazard and asset data to portfolio-scale financial risk estimates and adaptation payback.
40. LFD (v1.0): latent-compression-free generative diffusion with geological priors and geophysical regularization for implicit structural modeling
Core Problem: Implicit structural modeling from seismic data needs faster generation while honoring geological constraints.
Key Innovation: Applies diffusion generation directly in model space with geological priors and geophysical regularization.
41. Age-specific exposure to human-induced increases in humid heat
Core Problem: Measuring how human-caused warming changes age-specific exposure to dangerous humid heat
Key Innovation: Combines demographic data and grid-scale attribution to quantify disproportionate child exposure under current and future warming
42. Near-term, geospatial opportunity for biomass carbon storage to address the wildfire and climate crises
Core Problem: Whether anoxic burial of wildfire-treatment biomass can scale as a near-term carbon-storage pathway
Key Innovation: Combines western US geospatial water-balance and life-cycle modeling to map where biomass burial is most feasible and carbon efficient
43. Macroscopic Shear Behavior and Microstructural Evolution of Intact Loess from the Dongzhi Tableland
Core Problem: How wetting-driven microstructural change degrades loess shear strength and alters failure mode was insufficiently resolved.
Key Innovation: Identifies water-content thresholds and a three-stage strength-decay mechanism linking SEM and MIP observations to macroscopic shear behavior.
44. What Drives the Glacier Retreat, and How Do We See It? A Study of Measurement Methods and Environmental Drivers of Retreat in the Amundsenisen Glacial System, Svalbard
Core Problem: Glacier retreat estimates vary with measurement method, complicating interpretation of environmental controls.
Key Innovation: Systematically compares five terminus-measurement methods while analyzing multi-decadal retreat drivers for four Svalbard outlet glaciers.
45. Study on the influence of moisture content on the thixotropy and phase transition of loess
Core Problem: The rheological transition governing redeposited loess reactivation and flow initiation was insufficiently quantified.
Key Innovation: Uses three-interval thixotropy testing to define moisture-sensitive thixotropy and a dual stress-strain criterion for fluidization.
46. NMR-based characterization of soil-water characteristic curves considering volume change during drying-wetting cycles
Core Problem: Conventional soil-water characteristic curve measurement is slow and often ignores volume change during drying-wetting cycles.
Key Innovation: Introduces a rapid NMR-based route to derive SWCCs while explicitly accounting for shrink-swell effects.
47. Experimentally calibrated CFRP arch barrel interface strengthening for seismic retrofit of an earthquake-damaged historical masonry bridge: field damage assessment, interface characterization, and 3D nonlinear seismic analysis
Core Problem: How to retrofit an earthquake-damaged historical masonry bridge by stabilizing interface-controlled failure mechanisms.
Key Innovation: An experimentally calibrated CFRP interface-strengthening strategy embedded in 3D nonlinear seismic analysis.
48. SMS-Diff: Physics-aware spectral-spatial diffusion for unified geospatial multi-modal reconstruction
Core Problem: How to reconstruct heterogeneous geospatial modalities in a unified physics-aware framework.
Key Innovation: A spectral-spatial diffusion approach designed for unified multi-modal geospatial reconstruction.
49. Climate change projections of probable maximum precipitation over large areas through depth-area-duration analysis: A case study in Brazil
Core Problem: How climate change alters probable maximum precipitation over large areas in Brazil.
Key Innovation: A depth-area-duration framework for projecting climate-conditioned probable maximum precipitation.
50. CRISP: Calibration-Aware Visual State Space Duality for Remote Sensing Semantic Segmentation
Core Problem: How can state-space remote-sensing segmenters recover sharp boundaries and intra-class variation without losing linear-time efficiency?
Key Innovation: Pairs a duality calibration operator with a multi-prototype head to restore local contrast and preserve detail in VSSD backbones.
51. DDMS: Discriminative Distillation of Multi-view Foundational Features into Single-view Models
Core Problem: How can single-view foundation features inherit 3D consistency and local distinctiveness from stronger multi-view geometry models?
Key Innovation: Distills fused foundation and multi-view teacher features with a discriminative ranking objective to preserve both semantics and geometry.
52. Joint-Embedding Prediction of Masked Point Tubes for Self-Supervised Learning on 4D Point Cloud Videos
Core Problem: 4D point-cloud pretraining overemphasizes reconstruction instead of semantic dynamics.
Key Innovation: JEPA-style masked point-tube latent prediction for unlabeled 4D point clouds.
53. A mesh-free multiresolution deep energy method with phase-field modeling of brittle fracture
Core Problem: Phase-field brittle fracture needs dense discretization for unknown crack paths.
Key Innovation: Mesh-free deep energy method with multiresolution spline encoding and Monte Carlo integration.
54. Equivariant Covariance Tensors: Guaranteed SPD Uncertainty for Tensor-Valued Geometric Learning
Core Problem: Tensor-valued geometric learning lacks symmetry-consistent full uncertainty estimation with guaranteed SPD covariance.
Key Innovation: E(3)-equivariant full-covariance prediction on the SPD manifold with a robust Log-Euclidean scoring objective.
55. Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
Core Problem: Improving accuracy and stability of neural approximation for smooth functions and PDE solutions.
Key Innovation: Augments neural networks with polynomial components, weak orthogonality constraints, basis pruning, and polynomial preconditioning.
56. Voronoi-Assisted Optimization for Diffusing Unsigned Distance Fields from Unoriented Points
Core Problem: Recovering stable unsigned distance fields directly from unoriented point clouds without heavy neural models.
Key Innovation: Uses Voronoi-guided bidirectional normal alignment and diffusion to build UDFs robustly for complex open and non-manifold geometries.
57. An AI-driven framework for state-dependent constitutive modeling of soils integrating particle swarm optimization and machine learning
Core Problem: Soil stress-strain behavior is nonlinear and state-dependent, making adaptive constitutive prediction difficult.
Key Innovation: Combines PSO, machine learning, and adaptive integration to update constitutive parameters in real time.
58. Advancing pressure-flow scour estimation through particle swarm optimized ensemble learning and shapley additive exPlanations-based physical interpretation
Core Problem: Pressure-flow scour estimation remains difficult to predict accurately and interpret physically.
Key Innovation: Pairs particle-swarm-optimized ensemble learning with SHAP-based interpretation for scour estimation.
59. Multi-objective optimization of an arc-shaped wave run-up suppression structure for a rounded square column
Core Problem: Wave run-up on marine columns needs suppression structures with balanced hydrodynamic performance.
Key Innovation: Uses multi-objective optimization to design an arc-shaped wave run-up suppression structure.
60. Computational study on wave attenuation by an inclined porous-plate wave barrier
Core Problem: Inclined porous-plate barriers need quantified effectiveness for attenuating waves.
Key Innovation: Computationally evaluates wave-reduction performance of an inclined porous-plate barrier.
61. Narwhals document atlantification of East Greenland
Core Problem: Quantifying how Atlantification is changing remote East Greenland coastal waters and glacier-front exposure
Key Innovation: Uses instrumented narwhals plus historical ocean data to document deep warming reaching marine-terminating glaciers
62. Adaptive Fusion of Multiple Land-Cover Products for Improved Spatial Representation of Key Land Classes in Central Asia
Core Problem: Existing land-cover products disagree spatially and can share the same classification errors.
Key Innovation: Introduces class- and pixel-specific reliability-adaptive fusion with a residual expert for disagreement cases.
63. Phenology-Aware Compound Heat and Drought Events and Potential Exposure for Summer Maize in the Huang-Huai-Hai Plain, China
Core Problem: Uniform thresholds miss stage-specific crop exposure to concurrent heat and drought stress.
Key Innovation: Builds a phenology-aware daily framework with stage-specific heat thresholds and crop-adjusted SPEI.
64. SAR-Oriented and Physics-Guided Ocean Wave Spectrum Retrieval
Core Problem: SAR-based wave retrieval methods often recover scalar metrics better than physically consistent spectra.
Key Innovation: Develops a physics-guided Swin Transformer that reconstructs wave frequency spectra while preserving spectral energy consistency.
65. Enhanced 3D Lightning Localization for Low-Frequency Radio Observations over the Tibetan Plateau
Core Problem: Aging sensors and electromagnetic noise degrade 3D lightning localization accuracy in low-frequency networks.
Key Innovation: Introduces phase-preserving denoising that improves time-of-arrival consistency without changing hardware.
66. A Closed-Form Statistical Expression for Evaluating Wind Speed and Direction Prediction Intervals from Doppler Lidar Arc Scans
Core Problem: Operational arc-scan Doppler lidar needs robust prediction intervals for wind retrieval under limited scans.
Key Innovation: Derives a closed-form uncertainty expression that avoids unstable cross-covariance estimation.
67. Hydrodynamic characteristics of wetting water-front evolution in an air-saturated sandstone fracture considering wall roughness and matrix imbibition
Core Problem: Unsaturated seepage exchange between rock fractures and matrix is difficult to quantify.
Key Innovation: Develops and validates a fractal piecewise model linking wall roughness and matrix imbibition to wetting-front evolution.
68. Experimental and Numerical Studies on Dynamic Response and Fracture Characteristic of Sandstone with Multiple Defects
Core Problem: Multiple interacting discontinuities complicate dynamic rock failure and energy dissipation mechanisms.
Key Innovation: Uses validated experiments and coupled simulations to resolve crack evolution, force-chain disturbance, and energy partitioning in defective sandstone.
69. Structure and Shear Behavior of a High Liquid Limit Residual Mottled Clay in Cambodia
Core Problem: How the structure and shear behavior of high-liquid-limit residual mottled clay relate to bank-slope instability.
Key Innovation: A multi-method material characterization tying iron-cemented structure and disturbance sensitivity to failure-prone behavior.
70. A novel seismic strengthening method for masonry walls using 3D-Printed polymer mesh: out-of-plane experiments and in-plane numerical analysis
Core Problem: How 3D-printed polymer mesh strengthening compares with conventional textile-reinforced mortar for masonry walls.
Key Innovation: An experimental-numerical benchmark showing customizable polymer meshes can deliver meaningful seismic strengthening.
71. Optimal soil dielectric model map for global soil moisture retrieval from SMAP
Core Problem: How to improve global SMAP soil-moisture retrieval by choosing regionally optimal soil dielectric models.
Key Innovation: A global map of optimal dielectric models for soil-moisture inversion.
72. Data-centric probabilistic machine learning framework for sandstone strength prediction under freeze-thaw conditions
Core Problem: How to predict sandstone dynamic strength under freeze-thaw cycles when data are scarce.
Key Innovation: A data-centric probabilistic ML pipeline combining augmentation, Gaussian processes, uncertainty, and interpretability.
73. Physics-guided surrogate modeling for full-process stress-strain curves of brittle rock
Core Problem: How to emulate full stress-strain curves of brittle rock efficiently without losing physical realism.
Key Innovation: A physics-guided surrogate model aimed at full-process brittle-rock response rather than single strength targets.
74. Critical-state-surface-based modelling of triaxial behaviour in soil-rock mixtures with varying rock block contents
Core Problem: How varying rock block content changes the triaxial and critical-state behavior of soil-rock mixtures.
Key Innovation: A critical-state-surface-based constitutive treatment spanning different block-content regimes.
75. Widespread Divergence Between Precipitation and Discharge Trends Across Global Rivers
Core Problem: How often do long-term precipitation and river discharge trends diverge globally, and what drives that mismatch?
Key Innovation: Shows widespread decoupling at 10,179 sites and identifies precipitation-extreme frequency and runoff ratio as key drivers.
76. Sensitivity of Spatial Snowmelt Simulations to Radiative Forcing and Model Top Layer Thickness
Core Problem: How do top-layer thickness and radiative forcing choices alter spatial snowmelt simulations?
Key Innovation: Demonstrates strong interactions among layer thickness, longwave radiation, albedo, and snow surface temperature in a distributed model.
77. Low-Latency Activation-Regularized Sparse Neural Operators with Distillation Assistance Towards Real-Time Edge-Deployable Virtual Sensing
Core Problem: Neural operators for virtual sensing remain too costly for low-latency edge deployment.
Key Innovation: Sparse-activation SAR layers and distillation improve latency-error-energy tradeoffs.
78. PhysicsBench: A Unified Leaderboard for Generative and Predictive Models in Engineering Design and Simulation
Core Problem: Engineering generative and predictive models lack standardized small-data evaluation and ranking.
Key Innovation: Unified leaderboard across CAD, CFD, and FEA tasks with debiased multi-metric ranking.
79. PlaceSeek: Human-Centered Geospatial Retrieval of Urban Outdoor Places via Semantic Grounding and Affective Alignment
Core Problem: POI-centric retrieval cannot satisfy affective or activity-based outdoor place queries.
Key Innovation: Combines semantic grounding of street-view evidence with affective reranking.
80. KLTNet: Learning Sparse Feature Tracking for Robust and Accurate Monocular Visual-Inertial Odometry
Core Problem: Classical KLT tracking degrades under rapid motion and low texture, hurting monocular VIO.
Key Innovation: Coarse-to-fine dense-to-sparse learned tracking with fixed-reference patches and anisotropic confidence weights.
81. Replicable Conformal Prediction
Core Problem: Independent conformal calibrations produce unstable prediction sets that hinder auditing
Key Innovation: Uses shared-seed grid-rounded thresholds to make conformal outputs replicable with validity guarantees
82. Sequential operator learning under dependent data
Core Problem: Learning linear and nonlinear operators from sequential, adaptive, statistically dependent observations.
Key Innovation: Provides time-uniform self-normalized concentration bounds in Hilbert spaces for sequential operator learning under dependence.
83. Predictability of El Ni\~no from Delayed Observations
Core Problem: Determining how much forecast skill is contained in delayed observations of the Nino-3.4 index.
Key Innovation: Shows delayed-observation models add skill while more complex nonlinear models offer little systematic improvement over simple recurrences.
84. SatDL: Jointly Optimizing Data Redistribution and Training for Satellite-Based Distributed Learning
Core Problem: Reducing end-to-end training time and energy under severe non-IID label imbalance across satellites.
Key Innovation: Jointly optimizes partial data redistribution and model training instead of choosing between full redistribution and no redistribution.
85. MoRF-AST: Calibrated Probabilistic Virtual Sensing for Structural Monitoring under Changing Operating Conditions
Core Problem: Keeping full-field response reconstruction and uncertainty estimates reliable when deployment conditions differ from training conditions.
Key Innovation: Proposes residual flow matching plus affine spread transport to recalibrate posterior uncertainty from sparse sensors at deployment.
86. Freeze-thaw infiltration governed by soil freezing characteristic curves and hydraulic impedance in unsaturated sands
Core Problem: Freeze-thaw infiltration in partially saturated sands is poorly constrained because frozen hydraulic parameters are uncertain.
Key Innovation: Experimentally constrains freezing curves and impedance factors within a validated coupled thermo-hydraulic model.
87. A 225-year (1799-2024) homogenized daily water level series of the Vistula River in Warsaw
Core Problem: Long river-stage records are disrupted by gauge shifts, datum changes, calendar differences, and missing data.
Key Innovation: Builds a 225-year homogenized daily water-level series with LSTM-based reconstruction of missing values.
88. Generation of angular-normalized, cloud-filled, 0.01°-downscaled land surface temperature from 2018 to 2023 based on official FY-4A dataset
Core Problem: Existing land-surface-temperature products suffer from angular effects, cloud gaps, and coarse resolution.
Key Innovation: Generates hourly all-weather 0.01-degree angular-normalized LST from FY-4A data.
89. Lightning2EarthCARE: Collocated geostationary lightning and EarthCARE observations of electrically active storms
Core Problem: Lightning observations and detailed storm vertical structure are rarely linked in a common open dataset.
Key Innovation: Creates a collocated lightning-EarthCARE storm dataset for case studies and broader storm analyses.
90. Hemispheric asymmetry in evapotranspiration reveals terrestrial water-flux redistribution
Core Problem: Global evapotranspiration trends are ambiguous and may conceal regionally important water-flux redistribution.
Key Innovation: Uses observation-constrained basin-scale ET estimates to reveal opposing hemispheric trends and project their persistence to 2050.
91. Radargrammetric 3D Positioning of Pseudo Corner-Reflector Scatterers in KOMPSAT-5 Stacks with Per-Target Conditioning Diagnostics
Core Problem: Naturally occurring pseudo corner reflectors are difficult to detect and position reliably in SAR stacks.
Key Innovation: Combines contrast-ranked pseudo-CR detection, bundle triangulation, and per-target conditioning diagnostics for 3D positioning.
92. Submerged Hazard Identification and Processing Using Augmented Image-Based Detection (SHIP-AID)
Core Problem: Optically visible shallow-water wrecks and debris are difficult to detect consistently in imagery.
Key Innovation: Builds a modular GeoAI benchmark and detection framework with physically informed augmentation for submerged hazards.
93. How Accurately Can Smartphone LiDAR Document the Exposed Coarse Root Architecture of Scots Pine? A Low-Cost Field Workflow
Core Problem: Root architecture is important for anchorage yet difficult to document without costly scanners.
Key Innovation: Shows that consumer smartphone LiDAR can recover several exposed coarse-root metrics with usable field accuracy.
94. Vegetation Mapping Through Multiscale Remote Sensing
Core Problem: Vegetation mapping across scales remains challenging for long-term monitoring of ecosystem change.
Key Innovation: Presents a multiscale remote-sensing perspective on vegetation-type mapping and monitoring.
95. Mapping Vegetation Alliances Using Deep Learning and Multi-Source Remote Sensing Data
Core Problem: Traditional workflows map broad vegetation classes but struggle to classify detailed alliances efficiently at large scale.
Key Innovation: Integrates field data with multi-source and multi-temporal remote sensing in a deep-learning framework for 32 vegetation alliances.
96. Seismic resonance in base-isolated structures under long-period excitation: application of novel earthquake predominant periods
Core Problem: Whether base-isolated structures can undergo clear resonance under near-fault long-period excitation.
Key Innovation: Use of displacement-based earthquake predominant periods to examine resonance in isolated structures.
97. Angular anisotropy of Landsat and MODIS land surface temperature toward multi-source LST consistency
Core Problem: How angular anisotropy differences between Landsat and MODIS distort land-surface-temperature consistency.
Key Innovation: A cross-sensor treatment of LST anisotropy aimed at improving multi-source thermal consistency.
98. Combining knowledge-based training data generation and deep learning algorithms for mapping surface water bodies, wetlands, and paddy rice in Northeast Asia
Core Problem: How to generate training data and map water bodies, wetlands, and paddy rice across Northeast Asia.
Key Innovation: Knowledge-based training-data generation combined with deep learning for large-area water-related mapping.
99. Temporal frequency analysis identifies land degradation hotspots and sustainable use bright spots
Core Problem: How temporal frequency analysis can identify degradation hotspots and sustainable-use bright spots.
Key Innovation: A temporal-frequency framework for distinguishing persistent degradation from positive land-use trajectories.
100. Research on seismic performance and simplified mechanical model of infilled frame considering the influence of prefabricated constructional columns and wall-opening forms
Core Problem: How prefabricated constructional columns and wall-opening forms alter infilled-frame seismic behavior.
Key Innovation: A simplified mechanical model that incorporates column configuration and opening geometry.
101. A practical Equivalent Linear Spring-Damper model for nonlinear soil-structure interaction under seismic loading
Core Problem: How nonlinear soil-structure interaction affects building response and how it is modeled in practice.
Key Innovation: A practical synthesis of SSI modeling approaches, code treatment, and research gaps rather than a new hazard method.
102. Effects of fluctuating groundwater levels and freeze-thaw cycles on water and salt migration in saline soil subgrade of oasis flood-irrigation areas
Core Problem: How groundwater fluctuations and freeze-thaw cycles drive water and salt migration in saline soil subgrade.
Key Innovation: A coupled analysis of groundwater and freeze-thaw effects on moisture-salt redistribution.
103. Differential hydro-mechanical responses of adjacent operational metro twin tunnels to staged dewatering and excavation
Core Problem: How staged dewatering and excavation differentially affect adjacent operational metro twin tunnels.
Key Innovation: A staged coupled hydro-mechanical analysis of asymmetric responses in neighboring active tunnels.