TerraMosaic Daily Digest: August 27, 2026
Daily Summary
The most substantive direct geohazard papers treat slope failure as a cumulative state variable. In paraglacial Alaska, repeated earthquakes continue to damage fjord walls through deglaciation despite glacier buttressing, with convex topography and inherited fractures localizing instability. Landslide mechanics studies similarly resolve failure through internal structure: weak-plane thickness governs displacement, stress, and acoustic-emission evolution in locked-segment slopes, landslide-dam formation depends on grain size, fines, density, and water content that preconfigure later breach behavior, and clustered post-seismic rockfall in Jiuzhaigou is mitigated more effectively by rigid walls with EPS buffers. Hazard observation around slopes is becoming more exacting rather than more permissive: chronic landslide DEMs require site-specific bias correction, Sentinel-1 detection in Patagonia remains useful but loses much of its apparent skill under spatially independent validation, and negative-sample design, multi-temporal land-cover factors, Ligurian soil-moisture evaluation, and SfM-DIC complementarity under InSAR phase aliasing all tighten the evidential basis for landslide mapping and monitoring. A wider geomechanics cohort, from rockburst prediction to deep granite failure, joint roughness, capillary geochemistry, tunnel deformation, and cavern-roof stability, reinforces the same emphasis on latent material state.
Seismic, coastal, and hydroclimatic papers extend this move from local triggers to connected regional fields. Anti-similar earthquakes are documented across additional intermediate-depth settings, implying that flipped-waveform pairs may be more characteristic of such seismicity than previously assumed, while probabilistic and deterministic assessment of the Gulf of Cadiz-Alboran corridor identifies submarine areas where expected peak ground acceleration exceeds 1 g. Along coasts, China's low-elevation zone shows declining aggregate multi-hazard risk but strongly differentiated regional drivers, and a southern South China Sea GNSS/GNSS-IR observatory demonstrates simultaneous tracking of tectonic motion, sea-level change, and tsunami-scale disturbances. Hydroclimatic studies then resolve cascading structure rather than isolated exceedance: extreme precipitation across China propagates through monsoon hubs, river corridors, and mountain passes; drought forecasting improves markedly when analog retrieval, sequence modeling, and stacking are combined; wildfire mapping benefits from super-resolution-enabled multi-satellite transfer; and reservoir sedimentation, bridge-pier scour, saltwater intrusion, floodplain decoupling, post-fire snow change, and permafrost or frost-heave mechanics all show how evolving environmental boundary conditions reshape hazard exposure and response.
A final group links hazard forcing to system operability while a broader methods stream strengthens transferable capability. Water-supply, rail, pole-line, power-grid, LNG-shipping, shelter-planning, hurricane-distribution, and dam studies increasingly evaluate recovery order, functional thresholds, unmet demand, and reliability-based design rather than collapse alone, showing that resilience depends on how restoration resources are staged as much as on structural hardening. In parallel, remote-sensing and scientific-ML papers emphasize physical constraints, geometry, calibration, and uncertainty across operator learning, physics-informed generation, sparse-sensing field reconstruction, inductive kriging, uncertainty-aware subsurface screening, multimodal geolocalization, hyperspectral inpainting, 3D reconstruction, calibration, aerosol retrieval, and foundation-model adaptation. These papers materially expand the analytical toolkit for geoscience and hazard monitoring, but the evidence also shows that transferable methods should not be described as geohazard-validated unless domain testing is explicit.
Key Trends
Five scientific and methodological trajectories define the August 27 literature: cumulative slope conditioning, benchmarked landslide observation, coupled coastal-hydroclimatic regionalization, function-first resilience analysis, and physically structured geospatial AI.
- Hazard conditioning is being read as cumulative internal-state evolution: Across paraglacial rock slopes, locked-segment landslides, landslide dams, clustered rockfall, seepage pathways, frost-heave systems, post-fire snowpacks, and underground rock failure, the decisive controls are inherited structure, material state, moisture, or cumulative damage. The common move is away from trigger-only narratives and toward progressive conditioning that can be monitored or modeled before failure manifests.
- Landslide observation is becoming benchmarked and explicitly transfer-aware: The landslide cohort does not treat remote sensing as plug-and-play. DEM products are bias-tested against UAV or LiDAR reference data, Sentinel-1 detection is stress-tested with spatially independent validation, negative-sample design is formalized, land-cover change is encoded as a preparatory factor, and SfM-DIC is fused with InSAR where phase aliasing breaks standard monitoring assumptions.
- Regional hazard fields are being organized as coupled coastal and hydroclimatic systems: Marine seismic hazard, China's low-elevation coastal multi-hazard patterns, tsunami-capable GNSS-IR monitoring, shoreline SAR, extreme-precipitation propagation networks, reservoir-capacity tracking, scour under climate change, and saltwater-intrusion impacts all frame risk as spatially connected system behavior. The emphasis is on relay corridors, exposure patterns, adaptive capacity, and cross-basin or cross-jurisdictional coordination rather than isolated sites alone.
- Infrastructure resilience research is shifting from damage states to functional recovery: Seismic and cyclone studies increasingly ask how lifeline systems continue operating and how they should be restored, not only how badly components break. Water networks, rail track-bridge systems, pole-line networks, LNG shipping, power grids exposed to landslides, hurricane-prone distribution systems, shelters, and large dams are assessed through recovery order, unmet demand, operational thresholds, or reliability-based design choices.
- Transferable geospatial AI is privileging physics, geometry, and calibrated uncertainty: The methods cohort repeatedly embeds known structure into the model or evaluation: boundary conditions in neural operators, physical-law guidance in diffusion, Bayesian fusion for sparse sensing, leakage-free kriging splits, conformal or probabilistic subsurface screening, geometry-aware registration and calibration, and uncertainty-aware 3D reconstruction. These papers broaden the technical basis for future hazard work without implying that generic geospatial AI has already been validated on geohazard tasks.
Selected Papers
The selected papers combine direct studies of landslides, rockfall, seismic hazard, tsunami monitoring, drought, wildfire, coastal risk, and infrastructure resilience with a substantial secondary stream in remote sensing, inversion, and scientific machine learning. Read the first group as hazard-tested findings and the second as transferable analytical capability whose geohazard value depends on explicit application, calibration, and validation.
1. Numerical Modelling of Late Pleistocene to Holocene Earthquake-Induced Progressive Rock Slope Damage in the Paraglacial Serpentine Valley, Prince William Sound, Alaska
Core Problem: How repeated earthquakes and deglaciation jointly accumulate damage that predisposes modern rock-slope failure.
Key Innovation: Long-timescale distinct-element simulations linking glacier buttressing, inherited structure, and progressive coseismic damage to present instability.
2. UAV-Based Evaluation of National and Global Digital Elevation Models for Chronic Landslides in the Indian Himalayas
Core Problem: Which national and global DEMs represent chronic landslide topography accurately enough for analysis.
Key Innovation: UAV and LiDAR referenced, bias-corrected multi-DEM evaluation across Indian and global landslides.
3. Probabilistic and Deterministic Seismic Hazard Assessments of the area comprised between west Gulf of Cádiz and east Alboran Sea
Core Problem: Estimate seismic hazard across the Gulf of Cadiz-Alboran Sea corridor using probabilistic and deterministic methods.
Key Innovation: Jointly maps hazard and reports peak ground accelerations above 1 g in some submarine areas.
4. Dynamic analysis and protective structure optimization for clustered rockfall hazards triggered by the 2017 Jiuzhaigou earthquake: insights from 3D numerical modeling
Core Problem: How clustered post-earthquake rockfalls behave dynamically and how protective structures can be optimized.
Key Innovation: Couples UAV-derived 3D slope reconstruction with impact DEM modeling to show EPS-buffered retaining walls improve impact resistance.
5. Multi-hazard risk assessment and governance implications in China’s low-elevation coastal zone
Core Problem: How multi-hazard risk evolved across China's low-elevation coastal zone and which factors drive regional differences.
Key Innovation: Integrates hazard, exposure, vulnerability, and adaptive capacity in a spatiotemporal framework that isolates region-specific risk drivers.
6. Deep learning-based landslide detection using Sentinel-1 SAR imagery in the Chilean Patagonia
Core Problem: How to build all-weather landslide detection from Sentinel-1 SAR in cloudy mountainous terrain without overstating model transferability.
Key Innovation: Uses a modified U-Net plus block cross-validation to quantify cross-region performance collapse and propose stricter evaluation practice.
7. Numerical Analysis and Instability Prediction of Slopes with Locked Segments Containing Counter-Tilted Weak Planes
Core Problem: How counter-tilted weak-plane thickness influences failure evolution and how impending instability can be predicted.
Key Innovation: Extends FLAC3D with acoustic-emission analysis and applies a critical-displacement criterion that performs well in retrospective prediction.
8. Assessing present-day tectonic stability, sea-level change, and tsunami hazards from a coastal multi-hazard GNSS station in the southern South China Sea
Core Problem: Coastal hazard assessment needs present-day constraints on tectonic stability and vertical land motion.
Key Innovation: Uses a coastal multi-hazard GNSS station to jointly assess tectonic stability, sea-level change, and tsunami hazard.
9. Global Prevalence of Anti-similar Earthquakes at Intermediate Depth
Core Problem: How widespread anti-similar intermediate-depth earthquakes are and what mechanisms produce them.
Key Innovation: Global documentation of multiple new anti-similar earthquake clusters, expanding the phenomenon beyond a few classic locales.
10. A novel approach incorporating credibility and spatial uniformity for landslide negative sampling
Core Problem: How to choose negative samples for landslide susceptibility mapping while preserving both credibility and spatial uniformity.
Key Innovation: Introduces the CSUN strategy combining feature-space credibility with spatially uniform sampling, improving AUC and robustness.
11. Machine learning integration for robust real-time drought monitoring and early warning systems
Core Problem: How to improve long-term meteorological drought forecasting from SPI time series.
Key Innovation: Combines KNN analog retrieval, LSTM sequence modeling, and gradient-boosting stacking for robust SPI forecasts.
12. Generative artificial intelligence and deep ensemble learning for short-term rockburst prediction in underground engineering using microseismic data
Core Problem: How to predict rockburst intensity when strong events are rare and class imbalance degrades model performance.
Key Innovation: Integrates CTGAN data augmentation, deep random forest classification, automated tuning, and SHAP interpretation.
13. A review on the material control of landslide dam formation
Core Problem: How source-debris material properties control landslide-dam formation and inherited internal structure.
Key Innovation: Builds a material-centered framework linking debris descriptors to dam geometry, internal organization, and later breach behavior.
14. Land Use/Land Cover Change as a Preparatory Factor for Shallow Landslide Susceptibility: A Multi-Temporal Approach in the Messina Area (Italy)
Core Problem: How to incorporate multi-temporal land-use and land-cover transitions into shallow landslide susceptibility mapping.
Key Innovation: Provides an open-data GIS workflow to encode LULC trajectories and compare static versus dynamic susceptibility factors.
15. Post-processing UAV-based SfM data using digital image correlation: a complementary approach to InSAR for 3D slope monitoring in mining areas prone to phase aliasing
Core Problem: InSAR phase aliasing can leave unstable mine slopes insufficiently monitored in 3D.
Key Innovation: Combines UAV-SfM post-processing with digital image correlation as a complementary 3D slope-monitoring workflow.
16. Identification of concentrated seepage pathways integrating hydrogeological mechanisms and data-driven modeling: a case study of the Baihetan hydropower station
Core Problem: Concentrated seepage pathways at a major hydropower site are hard to detect from observations alone.
Key Innovation: Integrates hydrogeological process understanding with data-driven modeling to identify seepage pathways.
17. A four-stage heuristic recovery model for seismic resilience evaluation of water supply networks
Core Problem: Water supply resilience after earthquakes depends on recovery sequencing that is hard to evaluate.
Key Innovation: Introduces a four-stage heuristic recovery model for seismic resilience evaluation of water networks.
18. Resilience assessment and enhancement of critical infrastructure under natural hazards with case studies in Nordic power grid
Core Problem: Power distribution systems in landslide-prone regions need quantified resilience and cost-effective enhancement options.
Key Innovation: Introduces the RECINAT framework and optimizes virtual-power-plant deployment for hazard-triggered resilience gains.
19. Integrated Investment and Recovery Optimization for Distribution System Resilience under Hurricane Hazards
Core Problem: Distribution-system resilience under hurricane hazards requires coordinated investment and recovery decisions.
Key Innovation: Frames resilience as an integrated optimization of pre-event investment and post-event recovery.
20. Functional Fragility Analysis and Probabilistic Post-Earthquake Operational Assessment of High-Speed Railway Track-Bridge Systems
Core Problem: Rail track-bridge systems need probabilistic post-earthquake operability assessment beyond component-level fragility.
Key Innovation: Combines functional fragility analysis with probabilistic operational assessment for high-speed rail systems.
21. Mechanical-Electrical Coupled Failure Mechanism of Distribution Pole-Line Systems via Seismic Vulnerability Analysis
Core Problem: Distribution pole-line failures under earthquakes involve coupled mechanical and electrical effects.
Key Innovation: Models the coupled failure mechanism of pole-line systems through seismic vulnerability analysis.
22. Deep learning-based burned area mapping of California wildfires using Sentinel-2 and Landsat-8 imagery enhanced with super-resolution techniques
Core Problem: Burned-area delineation for major wildfires can be limited by spatial resolution and cross-sensor consistency.
Key Innovation: Combines Sentinel-2 and Landsat-8 with super-resolution-enhanced deep learning for burned-area mapping.
23. A network-based framework for deciphering extreme precipitation propagation across China
Core Problem: How extreme precipitation signals propagate across China as a connected network.
Key Innovation: Applies a network-based framework to track spatial propagation of extreme rainfall.
24. Appropriate ground-motion intensity measures for assessing seismic damage of arch dams based on interpretable machine learning technique
Core Problem: How to identify the most appropriate ground-motion intensity measures for predicting seismic damage in arch dams.
Key Innovation: Uses interpretable machine learning to relate candidate intensity measures to dam seismic damage.
25. UniGeo: A Multi-modal Large Language Model for Text-Guided Cross-View Geo-Localization
Core Problem: Open-ended text queries and highly similar image candidates make direct text-image matching unreliable for fine-grained drone geo-localization.
Key Innovation: Builds a unified multimodal model that combines geo-semantic understanding, cross-view semantic generation, and candidate-level verification.
26. Self-Augmented Diffusion Guidance for Physics-Informed Generation
Core Problem: Diffusion models can generate physically implausible spatiotemporal fields because governing-law constraints are not built into the sampling process.
Key Innovation: Conditions generation on the degree of physical-law violation and uses self-augmented guidance to enforce dynamics without solving equations at every denoising step.
27. Hyperspectral Diffusion Equivariant Imaging (HyDiff-EI): A Self-supervised Framework for Hyperspectral Image Inpainting
Core Problem: Hyperspectral inpainting is ill-posed, often depends on large pretraining, and must generalize across sensor configurations with little annotated data.
Key Innovation: Performs test-time diffusion optimization directly on a single corrupted hyperspectral scene while enforcing equivariant consistency priors.
28. Temporal Sensitivity Analysis of Tessera Embeddings
Core Problem: Determine how much temporal context Tessera embeddings need for accurate land-cover mapping.
Key Innovation: Controlled sensitivity study shows task-dependent degradation as observation windows shrink from one year to one day.
29. TRACE: Retrospective Streaming Generation of Physical Fields under Sparse Structured Sensing
Core Problem: Continuous physical fields are hard to reconstruct when sensing arrives as sparse structured streams rather than full batches.
Key Innovation: Combines latent-space Bayesian evidence fusion, Kalman-style filtering, and retrospective smoothing for streaming field reconstruction.
30. Bridging short- and medium-range weather forecasting with machine learning
Core Problem: Separate short- and medium-range forecast systems miss the benefits of a unified high-resolution ML weather model.
Key Innovation: Builds a nested global-plus-CONUS ensemble model that improves near-surface fields and longer-lead storm localization.
31. Enforcing Dirichlet Boundary Conditions in Operator Learning
Core Problem: Neural operators usually rely on data alone to satisfy known Dirichlet boundary conditions.
Key Innovation: Constrains outputs to Dirichlet Laplacian eigenfunction spans on general geometries while preserving operator expressivity.
32. A REVIEW ON THE USE OF SYNTHETIC APERTURE RADAR (SAR) FOR MONITORING SHORELINES AND INTERTIDAL AREAS
Core Problem: Synthesize how SAR can monitor dynamic shorelines and intertidal areas.
Key Innovation: Consolidates SAR capabilities, workflows, and use cases for challenging coastal environments.
33. Comparing Modelled and Remotely Sensed Soil Moisture Products Using In Situ Observations in Liguria, Italy: Evaluation via SWI Filtering and Rescaling Techniques
Core Problem: Which satellite and model soil-moisture products best reproduce in situ observations in Liguria.
Key Innovation: Shows SWI filtering plus rescaling, especially CDF matching and linear regression, improves agreement with ground measurements.
34. Continuous Satellite Monitoring of Reservoir Capacity Loss Using Deep Learning and Stochastic Mapping: The Poechos Reservoir and Regional Transferability in Northern Peru
Core Problem: How to update reservoir elevation-area-volume curves continuously without frequent bathymetric surveys.
Key Innovation: Fuses Sentinel-1 segmentation, PlanetScope calibration, and SWOT altimetry with stochastic mapping for near-continuous EAV reconstruction.
35. Beyond hazard thresholds: Impact-based and cross-jurisdictional shelter planning for tropical cyclone resilience
Core Problem: Threshold-based shelter planning misses impact differences across jurisdictions during tropical cyclones.
Key Innovation: Frames shelter planning as impact-based and cross-jurisdictional rather than threshold-only.
36. A micro-scale analytical framework for identifying potential stranding signals during extreme rainfall using LBS data
Core Problem: Emergency managers lack micro-scale detection of where people become stranded during extreme rainfall.
Key Innovation: Uses location-based-service dwell-time anomalies to identify and classify potential stranding hotspots.
37. Probabilistic analysis of the impact of climate change on local scour under bridge piers: A case study in southern Sweden
Core Problem: Climate change alters hydraulic forcing and uncertainty in local scour beneath bridge piers.
Key Innovation: Applies a probabilistic climate-impact framework to bridge-pier scour in southern Sweden.
38. Physics-driven multi-task deep learning framework with a dual time-frequency feature awareness mechanism for structural seismic response prediction
Core Problem: Accurate structural seismic response prediction remains difficult under complex dynamic behavior.
Key Innovation: Combines physics-driven multi-task learning with dual time-frequency awareness for seismic response prediction.
39. Resilience assessment of the global liquefied natural gas shipping network under tropical cyclone disruptions: A data-driven structural and functional analysis
Core Problem: Tropical cyclones can disrupt global LNG shipping structure and function in poorly quantified ways.
Key Innovation: Uses data-driven structural and functional metrics to assess shipping-network resilience under cyclone disruptions.
40. Reliability-oriented stochastic analysis and parameter design of nonlinear rate-independent negative stiffness dampers for seismic isolation system
Core Problem: Negative-stiffness dampers for seismic isolation require stochastic reliability-based parameter design.
Key Innovation: Performs reliability-oriented stochastic analysis and design for nonlinear rate-independent negative stiffness dampers.
41. Parameter-efficient multimodal adaptation of vision foundation models for remote sensing semantic segmentation
Core Problem: Adapting vision foundation models to remote-sensing semantic segmentation efficiently remains difficult.
Key Innovation: Develops parameter-efficient multimodal adaptation for remote-sensing segmentation with foundation models.
42. Decoupling ice lens initiation from growth via interface asymmetry
Core Problem: The physical controls on ice lens initiation versus growth are not cleanly separated.
Key Innovation: Uses interface asymmetry to decouple ice-lens initiation from subsequent growth dynamics.
43. Generating a consistent long-term L-band soil moisture record through Bayesian merging of SMOS and SMAP data
Core Problem: How to build a consistent long-term L-band soil-moisture record by merging SMOS and SMAP.
Key Innovation: Uses Bayesian merging to create a harmonized long-duration satellite soil-moisture product.
44. Economic impacts of saltwater intrusion on the Guangdong-Hong Kong-Macao Greater Bay Area under climate change scenarios
Core Problem: How saltwater intrusion under climate-change scenarios affects the economy of the Greater Bay Area.
Key Innovation: Translates future saltwater-intrusion scenarios into regional economic impact estimates.
45. PerLA: A thermo-hydro-mechanical framework coupling land-atmosphere interactions for evaluating highway and railway subgrade performance in permafrost regions
Core Problem: How coupled land-atmosphere, thermal, hydraulic, and mechanical processes control highway and railway subgrade performance in permafrost.
Key Innovation: Introduces the PerLA framework to couple land-atmosphere interactions with thermo-hydro-mechanical subgrade response.
46. A scale-undistorted test prediction method for ultra-high core earth-rockfill dam using multi-scale dynamic centrifuge model tests
Core Problem: How to predict the response of an ultra-high core earth-rockfill dam using multi-scale dynamic centrifuge tests.
Key Innovation: Proposes a scale-undistorted prediction method that links multi-scale centrifuge modeling to dam seismic response.
47. Mapping the Crustal Thickness of the Andes Using pmP Phases Identified From Adaptive Teleseismic Array Data
Core Problem: How to densify Moho-depth constraints in a large subduction zone without local arrays.
Key Innovation: Automatic pmP phase picking from teleseismic arrays to build a continental-scale Andean crustal thickness map.
48. NeuDonatello: Uncertainty-Aware Framework for Accurate Neural SDF Learning
Core Problem: How to reconstruct accurate surfaces from RGB alone when geometry is uncertain.
Key Innovation: Monte Carlo estimated signed-distance uncertainty that adaptively regularizes neural surface learning and density conversion.
49. CGS-SLAM: Collaborative Gaussian Splatting based SLAM for Multi-Agent Reconstruction
Core Problem: Most 3DGS SLAM systems rely on RGB-D sensors and lack practical collaborative mapping with only consumer-grade RGB and inertial inputs.
Key Innovation: Builds a decentralized-centralized multi-agent SLAM pipeline using RGB-inertial tracking, monocular metric depth, shared keyframe encodings, and learned submap alignment.
50. DINOcular: Self-Supervised Visuospatial Representations
Core Problem: Learn self-supervised representations that combine RGB semantics with explicit depth geometry.
Key Innovation: Inter-patch and intra-patch fusion of depth-derived priors into a visual backbone for better 3D awareness.
51. A Dynamic Likelihood Approach to Filtering for Advection-Diffusion Dynamics
Core Problem: Advection-dominated diffusion systems are hard to estimate accurately when observations are sparse.
Key Innovation: Extends dynamic-likelihood filtering to advection-diffusion with split-step propagation of observation-informed likelihoods.
52. Physics-Informed Stochastic Configuration Machine: A Backpropagation-Free Neural Network with Fast Training for Nonlinear Differential Equations
Core Problem: Standard PINNs are slow and optimization-heavy for nonlinear differential equations.
Key Innovation: Replaces backpropagation with analytically linearized physics losses solved through generalized linear least squares.
53. DP-JMRNet: A Deep Unfolding Network for Differential Phase Preservation in Sparse Bitemporal SAR Reconstruction
Core Problem: Sparse SAR reconstruction often preserves magnitude while degrading the differential phase needed for deformation analysis.
Key Innovation: Jointly reconstructs bitemporal SAR with a differential-phase objective, exchange-equivariant interaction, and coherence-aware cross-epoch sharing.
54. Leakage-Free Evaluation and Distribution-Robust Spatio-Temporal Graph Learning for Inductive Kriging
Core Problem: Standard inductive-kriging evaluations leak spatial information and underestimate out-of-distribution difficulty at unseen locations.
Key Innovation: Introduces a leakage-free spatio-temporal split and a robust kriging model that explicitly handles structural shift and ambiguous propagation.
55. Dams trigger long-term river-floodplain decoupling in dynamic Andean foreland rivers
Core Problem: How dams alter long-term channel-floodplain coupling in dynamic Andean foreland rivers.
Key Innovation: Hydro-morphodynamic modeling predicts persistent post-dam incision, secondary floodplains, and floodplain decoupling.
56. Adaptive particle swarm algorithm for solving three-dimensional coordinates of infrasound sources based on acoustic field velocity potential theory
Core Problem: How to localize infrasound precursor sources in 3D using a compact sensor array.
Key Innovation: Derives velocity-potential-based localization equations and solves them with an adaptive particle swarm optimizer.
57. Frost heave modeling and field verification for arc-bottom trapezoidal lined canals in cold regions
Core Problem: Lined canals in cold regions need reliable prediction of frost-heave deformation under field conditions.
Key Innovation: Combines frost-heave modeling with field verification for arc-bottom trapezoidal lined canals.
58. Differentiated contribution of waviness and unevenness to rock joint shear strength: a DEM investigation
Core Problem: How waviness and unevenness contribute differently to rock-joint shear strength.
Key Innovation: Uses DEM to separate roughness components and quantify their distinct effects on joint shear strength.
59. A calibrated regional screening framework for SPT N -values from public borehole records under strict spatial cross-validation
Core Problem: How to predict regional raw SPT N-values and stiff-layer occurrence from sparse public borehole data without spatial leakage.
Key Innovation: Combines strict spatial cross-validation, conformal uncertainty calibration, and survival modeling for deployable regional subsurface screening.
60. Flow-regime transitions in size-density heterogeneous wet granular materials governed by effective cohesion and inertial driving
Core Problem: How effective cohesion and inertial driving govern flow-regime transitions in heterogeneous wet granular materials.
Key Innovation: Introduces an equivalent Bond number for size-density heterogeneous wet mixtures and organizes continuous, intermittent, and cohesive flow states on a Bond-Froude regime map.
61. A data-driven multiscale and probabilistic stress-force-fabric modeling approach for geomaterials using DEM-FDM simulations
Core Problem: How to represent stress, force, and fabric behavior of geomaterials in a data-driven multiscale probabilistic framework.
Key Innovation: Combines DEM-FDM simulations with probabilistic multiscale modeling of stress-force-fabric relationships.
62. GIA-Induced Mantle Circulation and Surface Horizontal Motions in North America: Effects of 3D Laterally Varying Elastic Thickness and Viscosity
Core Problem: How 3D lithosphere and viscosity structure shape postglacial horizontal crustal motions.
Key Innovation: Global viscoelastic GIA modeling with laterally varying elastic thickness and viscosity showing box-like mantle circulation and phase-lagged horizontal motions.
63. Process-Based Model Parameters for Post-Fire Snow Hydrology: Looking Beyond Snow Albedo and Vegetation
Core Problem: Which process-model parameters control burned-forest snowpack evolution.
Key Innovation: Global sensitivity plus LiDAR-constrained calibration showing postfire shifts in albedo decay, canopy interception, unloading temperature, and snow roughness.
64. Mapping Woody Vegetation from Multi-Source Imagery and Prediction Fusion for Enhanced Data Efficiency and Accuracy
Core Problem: How to map woody vegetation accurately from inconsistent multi-source imagery with limited labels.
Key Innovation: Image composition, prediction fusion, and multi-source label transfer that sharply improve data efficiency and robustness.
65. Learning Woody Clearing With Loss Alignment for Zero-Shot Regrowth and Woody Segmentation
Core Problem: How to align woody-clearing models with end-user metrics and transfer them to scarce-label regrowth and segmentation tasks.
Key Innovation: Loss-alignment coefficient plus imagery generation and augmentation enabling zero-shot regrowth and woody segmentation.
66. Bayesian methods and Markov chain Monte Carlo algorithms for curve reconstruction and point cloud data analysis
Core Problem: How to reconstruct curves from noisy, incomplete point clouds while quantifying uncertainty.
Key Innovation: Fully Bayesian point-cloud model with tailored Markov chain Monte Carlo samplers that return posterior uncertainty instead of a single fit.
67. DPA-I2P: Depth-Guided Projective Alignment for Image-to-Point-Cloud Registration in Autonomous Driving
Core Problem: How to align camera images with sparse 3D point clouds more reliably across modalities.
Key Innovation: Depth-guided metric encoding, projection-consistent vision lifting, and query pruning for cross-modal pose estimation.
68. SimCast-S2S: An Efficient Generative Model for Subseasonal Precipitation Forecasting via Transfer Learning from Climate Simulations
Core Problem: How to generate accurate and efficient probabilistic subseasonal precipitation forecasts from limited data.
Key Innovation: Latent diffusion forecasting pretrained on climate simulations and adapted to reanalysis with low-rank adaptation.
69. Text-to-seed generation: Training-free open-vocabulary seeded semantic segmentation via re-purposing diffusion as text-guided seed generator
Core Problem: How to segment open-vocabulary targets more reliably than coarse-mask prompting allows.
Key Innovation: Training-free pipeline that turns diffusion attention into seed points and lets SAM expand them into full masks.
70. Generative Semantic Scene Completion
Core Problem: Outdoor semantic scene completion must infer dense 3D semantics from extremely sparse LiDAR while handling severe long-tail imbalance.
Key Innovation: Unifies paired sparse-dense data synthesis, diffusion-based semantic completion, and one-step refinement in a generative scene-completion framework.
71. KISS-GS: 3D Gaussian Splatting Compression Kept Simple
Core Problem: 3D Gaussian Splatting scenes are too large to deploy easily, and existing compression pipelines make the source of gains hard to reuse or analyze.
Key Innovation: Separates pruning and encoding into a modular pipeline, then uses image-native SOG-XT codebooks and smoothing for major size reductions.
72. Per-View Gaussian Predictions Enable Training-Free Distractor Filtering in Feed-Forward 3DGS
Core Problem: Feed-forward 3DGS captures transient objects seen in only some views, creating blurred, duplicated, or floating artifacts in reconstructed scenes.
Key Innovation: Uses per-view Gaussian exclusion and rendering-based verification to detect and remove inconsistent distractors without retraining or scene-specific optimization.
73. TEMPLAR Wales: A georeferenced environmental and toponymic dataset of Welsh settlements
Core Problem: Place names and environmental measurements are hard to reuse quantitatively without a reproducible structure separating places, lexical annotations, and spatial attributes.
Key Innovation: Releases a relational Wales-wide dataset linking stable settlement identifiers to lexical detections and multi-scale terrain, hydrology, land-cover, and woody-cover attributes.
74. UrbanGround: From Local Perception to Spatial Agency in a Real-Scale City
Core Problem: Test whether local street-view perception composes into reliable city-scale navigation and spatial action.
Key Innovation: Closed-loop benchmark built from territory-wide 3D geospatial data exposes long-horizon grounding failures in MLLM agents.
75. Multi-Dataset Inverse Problem Solving with Distributed Generative AI
Core Problem: Shared unknown parameters are difficult to infer consistently from multiple heterogeneous datasets with different forward operators.
Key Innovation: Extends distributed generative inverse solvers so each dataset has its own operator and discriminator while sharing global parameters.
76. Camera Calibration Using Inaccurate and Asynchronous Discrete GPS Trajectory from Drones
Core Problem: Stationary cameras are hard to calibrate from drone GPS tracks because altitude bias, time offset, and discrete sampling all distort the reference trajectory.
Key Innovation: Jointly estimates camera orientation, GPS altitude bias, and timing offset with a specialized maximum-likelihood scheme.
77. Data-driven Koopman mode approximation: A neural power iteration algorithm
Core Problem: Dominant Koopman modes are hard to learn accurately with expressive neural function classes.
Key Innovation: Applies a neural power-iteration scheme that learns dominant Koopman modes without explicit operator projection.
78. Active Diffusion-Based Inference for Ill-Posed Inverse Problems under Incomplete Priors
Core Problem: Inverse solvers fail when the true parameters lie outside assumed prior bounds.
Key Innovation: Uses posterior-uncertainty-driven active diffusion to expand the parameter domain and recover misspecified solutions.
79. Accurate Measurement of 3D and 2D Circular Centers With Application to LiDAR-Camera Extrinsic Calibration
Core Problem: Common circular-target calibration pipelines bias both 3D LiDAR center estimates and 2D projected circle centers.
Key Innovation: Improves cross-modal center measurement with conformal-geometric-algebra fitting and a chord-length-variance image estimator.
80. Heterogeneous controls on rock failure under expansive pressure
Core Problem: What physical factors control rock failure efficiency under soundless expansive cracking pressure.
Key Innovation: Identifies texture-heterogeneity evolution as a useful indicator of expansive-pressure rock fracturing performance.
81. True-Triaxial Failure Characteristics of Deep Granite Under Coupled Hydraulic Driving and Principal Stress Confinement
Core Problem: How seepage-pressure difference and principal-stress anisotropy jointly control deep granite failure.
Key Innovation: Introduces a hydraulic deviatoric ratio and identifies three coupled failure regimes from true-triaxial stress-seepage tests.
82. Thermo-Hydro-Mechanical Response of a Transmission Tower Pile Foundation in Seasonally Frozen Ground
Core Problem: How a transmission-tower pile foundation responds to coupled freezing, moisture migration, and stress evolution in seasonally frozen ground.
Key Innovation: Develops a fully coupled THM model that reproduces frost depth and links moisture accumulation to pile uplift by frost heave.
83. A self-adaptive mathematical approach for real-time structural health monitoring and anomaly detection using unsupervised machine learning
Core Problem: Real-time SHM anomaly detection needs adaptable unsupervised methods that avoid heavy case-specific tuning.
Key Innovation: Proposes a self-adaptive mathematical SHM and anomaly-detection approach using unsupervised machine learning.
84. A global system for monitoring radiative transfer and directional temperatures over heterogeneous surfaces
Core Problem: Directional temperature and radiative-transfer effects over heterogeneous surfaces are difficult to monitor globally.
Key Innovation: Builds a global monitoring system for radiative transfer and directional surface temperatures.
85. MMDiff: Multi-modal remote sensing image generation via cross-modality spatial feature transfer
Core Problem: Remote-sensing image generation across modalities often loses spatial consistency and complementary information.
Key Innovation: Uses cross-modality spatial feature transfer for multimodal remote-sensing image generation.
86. Enhanced retrieval of aerosol optical/microphysical parameters for Himawari-8 geostationary satellite measurements with data-driven deep learning method
Core Problem: Retrieving aerosol optical and microphysical parameters from Himawari-8 remains challenging.
Key Innovation: Applies a data-driven deep-learning method to enhance aerosol parameter retrieval from geostationary imagery.
87. A multi-task collaborative method for individual tree segmentation and multi-dimensional attribute extraction in mountainous areas using deep learning technologies
Core Problem: Individual-tree segmentation and attribute extraction in mountainous terrain are difficult to do jointly and accurately.
Key Innovation: Uses multi-task deep learning to segment trees and extract multi-dimensional attributes in mountainous areas.
88. An improved dual-engine predictive approach for strata deformation induced by tunnel excavation considering the complexity of convergence patterns
Core Problem: Idealized tunnel convergence assumptions limit accurate prediction of excavation-induced strata deformation and settlement.
Key Innovation: Combines generalized analytical convergence modeling with deep-learning inversion for refined strata-deformation prediction.
89. Reactive transport processes in unsaturated porous media: inclusion of capillary geochemistry
Core Problem: How to incorporate capillary geochemistry into reactive transport modeling in unsaturated porous media.
Key Innovation: Extends reactive transport formulations to include capillary-driven geochemical effects.
90. A synergetic framework for land-atmosphere coupling-based remote sensing inversion of evapotranspiration, surface resistance, and aerodynamic resistance
Core Problem: How to jointly invert evapotranspiration, surface resistance, and aerodynamic resistance using land-atmosphere coupling.
Key Innovation: Proposes a synergetic inversion framework that retrieves multiple land-surface exchange variables together.
91. Regime shifts in European basins detected from the water budget and the Combined Climatological Deviation Index
Core Problem: How to detect regime shifts in European basin water budgets using a combined climatological deviation index.
Key Innovation: Combines water-budget and Combined Climatological Deviation Index diagnostics with Pettitt and Bai-Perron tests to identify coherent hydroclimatic regime shifts across major European basins.
92. Statistical evaluation of the collapse capacity and the seismic performance of special truss moment frames
Core Problem: How to statistically evaluate collapse capacity and seismic performance of special truss moment frames.
Key Innovation: Applies statistical seismic-performance evaluation to a specific frame system.
93. Three-dimensional upper-bound stability analysis of coral reef limestone cavern roofs under tension-shear failure
Core Problem: How coral reef limestone cavern roofs fail under combined tension-shear conditions in three dimensions.
Key Innovation: Uses a three-dimensional upper-bound approach to quantify cavern-roof stability under tension-shear failure.