TerraMosaic Daily Digest: September 20, 2026
Daily Summary
Landslide mapping is moving from static image cues toward process-aware inference. FuSTNet combines InSAR time-series evolution, deformation rate, terrain geometry and spatial continuity to distinguish slow slope motion from atmospheric and vegetation noise, reporting 93.70% mIoU and flagging potential low-amplitude candidates that manual mapping may overlook. LSMamba adds lightweight local calibration to a global state-space backbone and leads three optical landslide benchmarks, while a Litang Fault study couples deterministic ground-motion simulation with random forests to map prospective coseismic-landslide susceptibility along the G318 corridor. Title-level records extend this physics-data convergence to shallow-landslide-debris-flow cascades, landslide-dammed-lake recovery and mechanics-embedded coseismic susceptibility, but their validation remains unavailable.
Flood and drought studies connect dynamic hazards to exposure and recovery. HydroKAN-Net resolves inundation boundaries across four UAV flood benchmarks, whereas monthly XGBoost-SHAP mapping in Hunan achieves a temporally independent AUC of 0.82 and shows that population exposure substantially reorders susceptibility hotspots. A nationwide agricultural-drought framework unifies propagation and logistic recovery thresholds, with regional recovery classifiers yielding mean AUC values of 0.79-0.88; a South Korean dam-storage study tests a six-month forecast horizon and finds TCN to be the strongest of four compared architectures, while rapid storage fluctuations remain difficult. Probabilistic levee-breach modeling based on 487 USACE overtopping events further ties hydraulic and erosion controls to breach initiation.
Observation systems increasingly target the physical quantity required by hazard models. Earthquake-coda HVSR stabilizes offshore estimates below 0.1 Hz and recovers shear-wave structure to roughly 10 km depth near Fukushima. LiDAR-constrained geolocation reduces InSAR height errors on a Dutch dike by as much as 97%, and the open SPAMS10 dataset converts peat InSAR time series into motion parameters with explicit covariance information. Across Greenland, bare-ice duration covaries with summer darkening and melt, while multi-model cyclone downscaling projects a redistribution of North Atlantic landfall risk whose regional pattern depends on the spatial structure of warming rather than its basin mean.
Method development emphasizes sparse supervision, sensor transfer and explicit uncertainty. DenseRS-CLIP, hyperspectral change detectors, cloud-tolerant fusion, cross-sensor adaptation and compact reconstruction models broaden the reusable remote-sensing toolkit, but most remain untested on geohazards. Geotechnical studies add random-field material-point simulation of tunnel collapse and title-level formulations for rock-ice avalanche run-up, three-dimensional slope stability, liquefaction-aware bridge fragility and probabilistic site response. Their shared direction is to couple mechanics, uncertainty and scalable computation; however, cross-site performance is still unevenly documented.
Key Trends
Across the collection, the strongest advances arise when temporal evolution, physical constraints and decision-relevant exposure are modeled together.
- Landslide AI is becoming dynamics-aware: InSAR evolution, deformation rates, terrain constraints and spatial continuity now complement optical texture, reducing reliance on visually plausible but physically inconsistent segmentation.
- Physics is entering both susceptibility and cascade models: A deterministic ground-motion case study shows direct physics-data coupling, while title-level records point toward slope-failure mechanics and shallow-landslide-debris-flow cascades; their validation remains unresolved.
- Risk maps are resolving time and exposure: Monthly flash-flood risk, cyclone redistribution and propagation-recovery drought metrics treat hazard state, affected population and recovery capacity as separate, evolving quantities.
- Validation is moving toward the target physical state: Coda-wave HVSR, LiDAR-corrected InSAR, multimission altimetry and parameterized peat motion reduce the gap between sensor observations and the subsurface or surface state required for decisions.
- Transferable remote-sensing models remain evidence-limited: Foundation embeddings, dense vision-language distillation, open-set classifiers and efficient 3D mapping aim to reduce annotation and compute burdens, but most hazard benefits remain prospective rather than demonstrated.
Selected Papers
The 20 September collection is led by dynamics-aware InSAR landslide detection, state-space optical segmentation, physics-informed coseismic susceptibility and a title-level probabilistic framework for rainfall-induced shallow-landslide-debris-flow cascades. Companion studies address flash-flood exposure, UAV inundation mapping, agricultural-drought propagation and recovery, cyclone-risk redistribution, offshore seismic structure, peat subsidence data, dike-deformation geolocation, levee breach, rock-ice avalanche mitigation and transferable remote-sensing methods.
1. A Spatiotemporal Fusion Network Based on InSAR Time-Series Evolution for Landslide Detection in Noise-Dominated Mountainous Areas
Core Problem: Image-like segmentation confuses noise with deformation when amplitude and morphology overlap.
Key Innovation: FuSTNet combines deformation evolution, rate maps, DEM slope constraints, and geospatial continuity to make detection physically plausible.
2. Bridging Global Context and Local Detail in State Space Models for Fine-Grained Landslide Segmentation in Remote Sensing Imagery
Core Problem: Global SSM interactions miss fine local boundaries needed for landslide delineation.
Key Innovation: LSMamba adds shallow structural calibration and local-detail enhancement to a VSSD backbone with an efficient decoder.
3. Potential Earthquake-Triggered Landslide Susceptibility Mapping Integrating Deterministic Ground Motion Simulation and Machine Learning: A Case Study of the Litang Fault Zone
Core Problem: Near-fault shaking heterogeneity must be coupled with geology and topography before an event.
Key Innovation: Couples curvilinear-grid finite-difference PGA simulation with fault-geometry-aware Random Forest transfer from Wenchuan inventories.
4. HydroKAN-Net: A Spectral-Gated Kolmogorov-Arnold Network With Boundary-Aware Cross-Scale Aggregation for Flood Segmentation in UAV Images
Core Problem: Reflections, wet terrain, debris, and diffuse boundaries cause flood-mask errors.
Key Innovation: Spectral-gated KAN blocks, boundary-aware cross-scale aggregation, and a water-aware frequency loss.
5. Dynamic Monthly Population-Exposure-Based Flash-Flood Risk Mapping Using an Explainable XGBoost Framework
Core Problem: Static risk maps miss strong seasonal changes in both flash-flood drivers and exposed population.
Key Innovation: Explainable XGBoost integrates dynamic monthly exposure with flood-conditioning variables.
6. Assessment of agricultural drought resilience in China based on coupled drought propagation and recovery processes
Core Problem: Conventional drought resilience metrics do not jointly represent meteorological-to-agricultural propagation and recovery under spatially heterogeneous conditions.
Key Innovation: Combines meteorological-to-agricultural propagation thresholds with soil-moisture-state-dependent recovery thresholds in a national resilience index.
7. Future Redistribution of North Atlantic Tropical Cyclone Risk in Two Contrasting CMIP6 Scenarios
Core Problem: Basin-average warming does not determine where tropical-cyclone landfall risk will rise or fall, creating uncertainty for regional planning.
Key Innovation: Downscales 12 CMIP6 models with the Columbia HAZard model and separates the effects of genesis and steering-flow changes on U.S. regional landfall patterns.
8. SPAMS10: an InSAR-derived soil motion parameter dataset to model relative peat surface elevation changes
Core Problem: Peat surface motion is difficult to represent compactly across groundwater and weather variability.
Key Innovation: Publishes four displacement-model parameters and six variance-covariance terms at parcel scale, with contextual soil codes enabling reconstruction and uncertainty-aware peat-motion analysis.
9. Multiclass Semantic Segmentation of Wildland Fire Images Using Context-Aware Centralized Copy-Paste Data Augmentation
Core Problem: Random copy-paste augmentation can create physically implausible fire scenes and degrade models trained from scarce annotations.
Key Innovation: Restricts pasted fire to semantically valid regions and matches source-target ash/vegetation context before multi-objective evaluation.
10. Knowledge-Graph-Augmented Chronos-2 for HEC-RAS Surrogate Forecasting
Core Problem: Physics-based hydraulic simulations are expensive, while generic time-series models underuse river-network and project-state knowledge.
Key Innovation: Couples frozen Chronos-2 with graph-conditioned retrieval, exact-state residual decoding and input-aligned correction over 4,675 cross sections in 71 reaches.
11. Assessing and Improving Geolocation of InSAR Scatterers with LiDAR Data
Core Problem: 3D geolocation errors prevent reliable assignment of scatterers to dike components.
Key Innovation: Projects LiDAR candidates into scatterer-specific 3D error ellipsoids, cutting height RMSE by up to 97%.
12. Midterm drought forecasting based on dam storage prediction using deep learning algorithms
Core Problem: Existing operational forecasts provide insufficient lead time for persistent drought.
Key Innovation: Compares four sequence architectures across engineered hydrologic datasets and optimizes the best TCN.
13. Machine Learning-Based Probabilistic Modeling of Levee Breach Initiation Under Overtopping
Core Problem: Breach onset is difficult to predict because of interacting hydraulic, geometric and geotechnical controls and limited high-quality field observations.
Key Innovation: Develops logistic-regression and random-forest models from 487 documented USACE overtopping events, with physically guided features and strong unseen-data discrimination; depth and erosion resistance emerge as dominant controls.
14. Geomorphic transition driven by river self-adjustment: A process-based framework for long-term recovery of river-connected landslide-dammed lakes
Core Problem: Title-level focus: identifies how river self-adjustment drives geomorphic transition and recovery after landslide damming; empirical scope is unavailable.
Key Innovation: Title-signalled approach or contribution: A process-based framework for long-term landslide-dammed-lake recovery is claimed, but mechanisms, cases and validation cannot be confirmed without an abstract. Methods, data and results could not be assessed because no reliable abstract was available.
15. A high-efficiency, physically based probabilistic framework (P-CASCADE) for predicting rainfall-induced shallow landslide-debris flow cascades
Core Problem: Title-level focus: identifies efficient probabilistic prediction of rainfall-induced shallow-landslide to debris-flow cascades as the central challenge.
Key Innovation: Title-signalled approach or contribution: The named P-CASCADE framework combines physical basis, probability and computational efficiency, but equations, test sites and performance cannot be verified without an abstract. Methods, data and results could not be assessed because no reliable abstract was available.
16. Embedding slope-failure mechanics in E-PINN for co-seismic landslide susceptibility assessment
Core Problem: Title-level focus: identifies the need to make co-seismic susceptibility models respect slope-failure mechanics.
Key Innovation: Title-signalled approach or contribution: An E-PINN embedding slope-failure mechanics is claimed, but physics constraints, inventory, spatial validation and gains are unavailable. Methods, data and results could not be assessed because no reliable abstract was available.
17. Numerical study of dry rock-ice avalanche run-up against slit dams by the discrete element method
Core Problem: Title-level focus: identifies prediction of dry rock–ice-avalanche run-up against slit dams.
Key Innovation: Title-signalled approach or contribution: DEM is applied to rock–ice avalanche–barrier interaction, but mixture representation, experiments and mitigation findings are unavailable. Methods, data and results could not be assessed because no reliable abstract was available.
18. Automatic Differentiation-Aided Computational Framework for Large-Scale Three-Dimensional Limit Equilibrium Slope Stability Analysis
Core Problem: Title-level focus: identifies computational scaling of large three-dimensional slope-stability analysis.
Key Innovation: Title-signalled approach or contribution: Automatic differentiation is claimed to accelerate or enable the framework, but solver behavior, benchmarks and case results are unavailable. Methods, data and results could not be assessed because no reliable abstract was available.
19. Shear Wave Velocity Assessment Offshore Fukushima by Analysis of HVSR From Coda Waves
Core Problem: Ocean-bottom ambient noise is unstable below 0.1 Hz, limiting conventional HVSR recovery of deep subsurface structure.
Key Innovation: Derives stable 0.05–10 Hz HVSR from long earthquake codas and inverts it with a seawater-aware diffuse-field algorithm for approximately 10 km depth profiles.
20. High-Resolution Prediction of Spatial Distribution of Soil Thickness in Areas With Heterogeneous Soil Parent Materials: Machine Learning Methods and Variable Optimisation
Core Problem: Predicting soil thickness at high resolution is difficult where heterogeneous parent materials weaken the transferability of terrain-only models.
Key Innovation: Adds mapped parent material to SVM, ANN, RF and XGB comparisons; XGB with parent material explains 81% of observed spatial variation from 97 soil profiles.
21. Generative inversion for early ranking of competing geologic interpretations
Core Problem: Early-stage projects often lack enough observations to distinguish competing geological interpretations before costly drilling.
Key Innovation: Turns text-described interpretations into 1,600-image priors, learns interpretation-specific latent inversions, and ranks alternatives by hydraulic-head compatibility.
22. Adaptive World Memory 3D Foundation Model for Scalable 3D Mapping, Localization, and Rendering
Core Problem: Current 3D foundation models lack durable world memory, long-sequence scalability and a unified renderable representation.
Key Innovation: Combines gated adaptive world memory, spatiotemporal test-time regulation, local submaps, global SL(4) refinement and a Gaussian reconstruction head in one model.
23. DenseRS-CLIP: Enhancing Dense Feature Representation of Remote Sensing CLIP Via Attention-Decoupled Dual-Branch Distillation
Core Problem: Global CLIP alignment leaves patch features weak for dense prediction.
Key Innovation: Region-structured semantic distillation and DINOv3 topology distillation in decoupled branches.
24. AlphaEarth Foundations Provide Superior Accuracy in Vegetation Height Mapping Across Severely Disturbed Forest Areas
Core Problem: Vegetation-height models saturate and require costly multisensor preprocessing.
Key Innovation: Benchmarks AlphaEarth embeddings against Sentinel features with RF and U-Net across disturbed forests.
25. PRECISi: A High-Resolution Daily Gridded Precipitation Dataset for Sicily (1951-2022) Derived from a Doubly Conditional Geostatistical Framework
Core Problem: Sparse, heterogeneous gauges limit daily precipitation reconstruction in Sicily.
Key Innovation: Doubly conditional geostatistical gridding stratified by hydroclimatic regimes.
26. Surface Soil Moisture from Sentinel-2 Imagery: A Systematic Review Complemented by a Case Study in Sardinia, Italy
Core Problem: Sentinel-2 soil-moisture results are inconsistent because validation is sparse and vegetation-dependent.
Key Innovation: Screens 1,158 studies to 66 eligible papers and tests conclusions with a three-year Sardinia case study.
27. Experimental Assessment of Microwave Penetration Depth in Layered Soil by Wideband (0.3-6 GHz) Transmissometry and Reflectometry
Core Problem: Penetration depth is usually inferred from imperfect dielectric models rather than measured in stratified soil.
Key Innovation: Wideband layered-soil VNA experiment separating configuration effects, model uncertainty, and model-free attenuation estimates.
28. Associations Between the Seasonal Darkening of the Greenland Ice Sheet and the Duration of Bare Ice
Core Problem: Ice-sheet darkening amplifies melt, but the relation between summer albedo and the duration of exposed bare ice needs quantitative constraint.
Key Innovation: Combines automatic-weather-station and satellite albedo records to quantify inverse bare-ice-duration relationships and associated ablation, while documenting rising scatter at long exposure durations.
29. Probabilistic large-deformation modelling of tunnel face collapse using random material point method (RMPM) incorporating cross-correlated random fields
Core Problem: Homogeneous deterministic analyses cannot represent how spatially variable soil strength controls tunnel-collapse runout and surface settlement.
Key Innovation: Couples the material point method with cross-correlated lognormal strength fields and Monte Carlo simulation, validating collapse geometry against physical tests and resolving correlation effects on collapse extent and variability.
30. Nested generative downscaling of passive microwave satellite data for navigation-scale Arctic sea ice forecasting
Core Problem: Title-level focus: indicates that coarse passive-microwave observations must be downscaled to navigation-scale sea-ice forecasts.
Key Innovation: Title-signalled approach or contribution: A nested generative downscaling design is named, but forecast skill, resolution and validation cannot be assessed without an abstract. Methods, data and results could not be assessed because no reliable abstract was available.
31. Relaxing the clear-sky assumption: Cloud-tolerant spatiotemporal fusion via mask-guided feature modulation and temporal-memory collaboration
Core Problem: Title-level focus: identifies cloud contamination and the clear-sky assumption as limitations of spatiotemporal image fusion.
Key Innovation: Title-signalled approach or contribution: The title proposes mask-guided feature modulation with temporal-memory collaboration, but datasets, benchmarks and robustness are unavailable. Methods, data and results could not be assessed because no reliable abstract was available.
32. Kinematic pattern interpretation of railway-corridor deformation from InSAR time series with transfer learning-based TCN
Core Problem: Title-level focus: indicates a need to interpret kinematic patterns in railway-corridor deformation from InSAR time series.
Key Innovation: Title-signalled approach or contribution: A transfer-learning temporal convolutional network is named as the interpretation tool, but study area, labels and accuracy are unavailable. Methods, data and results could not be assessed because no reliable abstract was available.
33. Estimating supraglacial lake bottom ablation rates by integrating ICESat-2 repeat track data and multi-temporal sentinel-2 imagery
Core Problem: Title-level focus: identifies estimation of supraglacial-lake bottom ablation rates as the target; site and validation details are unavailable.
Key Innovation: Title-signalled approach or contribution: The title indicates integration of ICESat-2 repeat tracks with multi-temporal Sentinel-2 imagery, but retrieval performance cannot be verified. Methods, data and results could not be assessed because no reliable abstract was available.
34. Soil thawed depth regulates snowmelt erosion by controlling surface runoff and infiltration in alpine meadow soil
Core Problem: Title-level focus: identifies the role of thawed-soil depth in regulating snowmelt erosion; experimental scope is unavailable.
Key Innovation: Title-signalled approach or contribution: The title advances a runoff–infiltration mechanism linking thaw depth to erosion, but effect sizes and validation cannot be verified. Methods, data and results could not be assessed because no reliable abstract was available.
35. Supply-limited post-fire sediment transport and geochemical fingerprinting during rill erosion
Core Problem: Title-level focus: identifies sediment-supply limitation during post-fire rill erosion; experimental and field context are unavailable.
Key Innovation: Title-signalled approach or contribution: Geochemical fingerprinting is paired with sediment-transport analysis, but source discrimination and quantitative results cannot be verified. Methods, data and results could not be assessed because no reliable abstract was available.
36. Seismic response of underground structures in layered unsaturated-saturated poroelastic sites under varying groundwater levels
Core Problem: Title-level focus: identifies seismic behavior of underground structures in layered poroelastic sites with varying groundwater.
Key Innovation: Title-signalled approach or contribution: Coupled consideration of unsaturated–saturated layering and groundwater variation is implied, but formulation and findings are unavailable. Methods, data and results could not be assessed because no reliable abstract was available.
37. Seismic fragility analysis of continuous rigid-frame bridges considering site liquefaction and hydrodynamic effects
Core Problem: Title-level focus: identifies bridge fragility under combined seismic, liquefaction and hydrodynamic effects.
Key Innovation: Title-signalled approach or contribution: Joint treatment of liquefaction and hydrodynamic loading is indicated, but hazard models, case conditions and fragility shifts are unavailable. Methods, data and results could not be assessed because no reliable abstract was available.
38. TAPe+ML: A Compact Structured Representation for Multi-Task Computer Vision
Core Problem: Pixel-tensor models can require large parameter counts and substantial data and compute across vision tasks.
Key Innovation: Encodes relations among perceptual elements before recognition and reuses the compact TAPe representation across classification, detection and instance segmentation.
39. A Scene Language Model for Open-Vocabulary Scene Mapping
Core Problem: Persistent open-vocabulary mapping typically requires engineered association pipelines and feature-heavy object memories.
Key Innovation: Uses one vision-language model to maintain the entire scene as a structured textual object list, achieving competitive mapping with 6–12 times smaller memory.
40. Predicting the Elastic Properties of a Cemented Granular Material during Chemical Damage (Debonding)
Core Problem: Experiments poorly constrain how reaction-driven debonding changes elastic properties of cemented granular rock.
Key Innovation: Compares discrete-element and phase-field/FFT homogenization frameworks to derive softening laws during chemical weathering.
41. Extending Decoupled Attention to Dense Prediction and Masked Training for Multi-Channel Images
Core Problem: Joint attention over all channel-patch tokens dilutes semantically distinct channels and standard decoupling fails under independent channel masking.
Key Innovation: Uses linear assignment to recover cross-channel patch correspondence, extending decoupled attention to masked pretraining and dense prediction.
42. XCalib Depth-Guided Geometric Optimization for Dense Thermal-Visible Video Registration
Core Problem: Global homographies fail under depth-dependent parallax, while unconstrained dense flow can be physically implausible and temporally unstable.
Key Innovation: Uses monocular metric depth and optimized virtual camera geometry as an implicit constraint, plus normalized-edge correlation for cross-spectral alignment.
43. Determination of Physical Height Differences from Time Transfer via the ACES Mission - A Simulation Study
Core Problem: Clock-based geopotential-height recovery needs realistic performance assessment for intermittent space-to-ground microwave and optical links.
Key Innovation: Simulates ACES clock/link noise and evaluates common-view and split non-common-view estimators, finding decimeter-to-centimeter potential for optical links.
44. Object Detection Benchmarks are Incomplete: The Role of Label Errors and Annotation Uncertainty
Core Problem: Standard object-detection benchmarks hide substantial missing-label error and deterministic annotation ambiguity.
Key Innovation: Re-annotates four major datasets with high-recall multi-annotator soft labels and releases uncertainty-aware detection and real-label-error benchmarks.
45. Chronosphere: Space-Time Tessellation of Local Climate Experts
Core Problem: Fixed-resolution location encoders allocate capacity poorly where environmental complexity varies across space and season.
Key Innovation: Learns an adaptive tessellation on the space-time torus with shared local basis functions, improving transfer across climate tasks.
46. The Role of Radiometric Features in Cross-Site Leaf-Wood Segmentation of LiDAR Point Clouds
Core Problem: Geometry-only leaf-wood classifiers trained on dense terrestrial scans fail when transferred to sparse top-down airborne LiDAR.
Key Innovation: Shows that radiometric attributes encode sensor-crossing material cues, increasing wood recall by 119% and improving component connectivity across terrestrial and RPA-LiDAR sites.
47. Riemannian Simultaneous Inference for Tangent Vector Field Regression
Core Problem: Vector responses on curved manifolds occupy different tangent spaces, complicating nonparametric regression and simultaneous inference.
Key Innovation: Parallel-transports responses, corrects manifold volume effects and derives a feasible simultaneous confidence tube, illustrated through global wind reconstruction.
48. Subpixel-Guided Network for Hyperspectral Image Change Detection Based on Multitemporal Unmixing
Core Problem: Mixed pixels and weak temporal exploitation limit hyperspectral change detection.
Key Innovation: Shared-endmember multitemporal unmixing plus aligned pixel–subpixel feature fusion.
49. An Explainable Network for Adaptive Change Detection in Unannotated Hyperspectral Images
Core Problem: HSI change detection suffers from mixed pixels, scale variation, redundancy, and absent labels.
Key Innovation: Low-rank explainable representation, subpixel features, self-generated weak labels, and adaptive fusion.
50. Frequency-Decoupled Enhancement and Large Kernel Cross-Scale Fusion Network for Remote Sensing Change Detection
Core Problem: Pseudochanges, blurred boundaries, and scale variation degrade bitemporal detection.
Key Innovation: Frequency decomposition, cross-attention refinement, and large-kernel cross-scale decoding.
51. Unified Model Compression Framework for Hyperspectral Image Super-Resolution
Core Problem: HSI super-resolution is too costly for constrained platforms.
Key Innovation: Block base-vector and channel-aware scale-adaptive quantizers with FPGA/GPU kernels.
52. Rapid Assessment of Blast-Induced Structural Damage Using a Physics-Based Multimodal Network
Core Problem: Optical imagery alone misses internally damaged structures and needs extensive labels.
Key Innovation: Fuses simulated blast loading with pre/post-event imagery through transfer learning.
53. Interpretable Spectral-State Learning via Unmixing-Guided Mamba Attention Networks for Hyperspectral Image Classification
Core Problem: HSI classifiers need long-range spectral modeling and scientific interpretability.
Key Innovation: Bidirectional Mamba with endmember-guided pretraining, spatial attention, and multilevel explanations.
54. Frequency-Aware and Conditional Entropy-Guided Network for Multispectral Object Detection
Core Problem: Fusion methods underuse modality-specific contours and textures.
Key Innovation: Wavelet enhancement plus conditional-entropy-guided cross-attention.
55. SSH-Net: Spectral-Spatial Enhanced Hierarchical Network for Robust Polymorphic River Extraction From Sentinel-2 Imagery
Core Problem: Variable river widths and complex backgrounds hinder water extraction.
Key Innovation: Spectral enhancement, spatial-continuity context, multiscale integration, and Tversky-BCE loss.
56. Self-Contrastive Graph Diffusion Clustering for Hyperspectral Images
Core Problem: Augmentation noise and false negatives weaken HSI contrastive clustering.
Key Innovation: Superpixel graph diffusion with structure-and-feature-aware pair selection.
57. Band-Selected Spectral-Spatial Transformer With Temporal Adaptive Modulation for Hyperspectral Image Change Detection
Core Problem: All-band encoding is redundant and fixed temporal fusion is inflexible.
Key Innovation: Discriminative band selection plus sample-adaptive temporal modulation.
58. Cross-Sensor Hyperspectral Image Classification via Multiscale Spectral-Spatial Representation Learning and HyperLoRA-Based Adaptation
Core Problem: Limited labels and cross-sensor variability impair HSI generalization.
Key Innovation: Multiscale fusion with HyperLoRA and adversarial robustness enhancement.
59. Highly-Constrained Unsupervised Hyperspectral Band Selection via Quality-aware Latent Feature Clustering
Core Problem: Selecting fewer than ten bands often destroys downstream performance.
Key Innovation: Quality-aware graph distance, latent clustering, and adaptive priority refinement.
60. Methane Inversion inter-Comparison (MICA): A Multi-model Estimation of Regional and National Emissions in Asia
Core Problem: Asian methane budgets remain uncertain because model structure and observational constraints yield divergent regional and national estimates.
Key Innovation: Harmonizes seven atmospheric inversions over 2010–2021 and releases a transparent comparison dataset spanning regional and national emissions.
61. Seasonal evolution of a warm alpine snowpack: a multi-instrument dataset of snow microstructure at Col de Porte, French Alps
Core Problem: Warm alpine snow evolution lacks dense multi-instrument observations.
Key Innovation: Weekly tomography, SnowMicroPen, and optical measurements over two winters.
62. Parameter estimation for land-surface models using Neural Physics
Core Problem: Land-surface parameters are difficult to identify from incomplete observations.
Key Innovation: Neural Physics inversion with observability tests across soil depths and fluxes.
63. A deep learning framework for gridding daily climate variables from a sparse station network
Core Problem: Sparse stations leave spatial gaps in daily climate variables.
Key Innovation: Deep-learning gridding framework tailored to sparse observational networks.
64. Stratigraphy-guided joint inversion of acoustic impedance and resistivity for carbonate reservoir characterization
Core Problem: Seismic elastic properties alone poorly discriminate fluids and resistivity is sparse.
Key Innovation: CNN–Transformer joint inversion conditioned on interpreted horizons.
65. Real-Time Target Detection in Compressed Domain for Streak Tube LiDAR by Two-Pass Labeling and Sparse Attention
Core Problem: Full decompression makes high-frame-rate LiDAR processing too slow.
Key Innovation: Selective decoding, linear-time component labeling, and a 6,724-parameter sparse-attention network reaching 1,536 FPS.
66. Bridging Temporal Gaps in Thermal Remote Sensing: A Generative Adversarial Network Approach to Land-Cover-Stratified Diurnal Heat Retention in Peri-Urban Landscapes
Core Problem: UHI indicators are coarse, daytime-biased, and insensitive to land-cover-specific behavior.
Key Innovation: Conditional GAN fusion plus a land-cover-stratified Relative Diurnal Thermal Index.
67. Reliability-Aware Dual-Stream Self-Training for Semi-Supervised Semantic Segmentation of High-Resolution Remote Sensing Imagery
Core Problem: Pseudo-label reliability and unlabeled-data coverage trade off against each other.
Key Innovation: Stable-sample offline supervision, full-data online learning, adaptive thresholds, and prototype contrast.
68. Frequency-Domain Feature Enhancement Combined with Contrastive Learning for Hyperspectral Image Open-Set Classification
Core Problem: Known and unknown spectra are hard to separate under open-set conditions.
Key Innovation: Wavelet feature enhancement, multiscale weighted ResNet, and dynamic hard-sample contrastive learning.
69. Evaluating Remote Sensing Foundation Model Embeddings for Cross-City Thematic Mapping of Eucalyptus Plantations in Guangxi, China
Core Problem: Plantation mapping fails when target-region labels are unavailable.
Key Innovation: Fuses two foundation embeddings with conventional features under repeated cross-city tests.
70. Discrete Boltzmann Modeling of Two-Layer Shallow Water Flows
Core Problem: Conventional depth-averaged and lattice-Boltzmann approaches struggle to represent layered flow structure and remain stable under supercritical conditions.
Key Innovation: Introduces a D2Q16-2 discrete Boltzmann formulation with interlayer momentum coupling and dynamic wetting-drying, tested on eight hydraulic benchmarks.
71. An Interpretable and Lightweight Dynamic Framework for Streamflow-Meteorology Networks to Enhance Daily Streamflow Forecasting
Core Problem: Streamflow models must balance predictive accuracy, computational efficiency, interpretability and transfer across heterogeneous catchments.
Key Innovation: Uses reservoir computing to infer dynamic streamflow–meteorology networks and injects the inferred structure plus catchment similarity into forecasts, reporting strong peak-flow gains with few trainable parameters.
72. Linking Brinell Creep Indentation to Macroscopic Rock Creep Compression Using Viscoelastic Modelling and Simulation-Informed Machine Learning
Core Problem: Localized indentation responses are difficult to translate to macroscopic creep properties, while conventional long-duration compression tests are costly.
Key Innovation: Combines Burgers viscoelastic simulations, sensitivity analysis and a neural network trained on 459 simulations to predict macroscopic parameters from indentation inputs, with experimental comparisons.
73. Bias-variance trade-off in radiative transfer model inversion drives uncertainty in leaf area index estimation
Core Problem: Title-level focus: identifies the bias–variance trade-off in radiative-transfer inversion as a driver of leaf-area-index uncertainty; study design is unavailable.
Key Innovation: Title-signalled approach or contribution: The title suggests an explicit decomposition of retrieval uncertainty, but methods, datasets and results cannot be verified without an abstract. Methods, data and results could not be assessed because no reliable abstract was available.
74. Optically consistent multi-parameter aerosol retrieval from dual-satellite polarimetric observations enabled by physics-guided deep learning
Core Problem: Title-level focus: indicates a need for optically consistent recovery of multiple aerosol parameters from polarimetric satellite observations.
Key Innovation: Title-signalled approach or contribution: The title combines physics-guided deep learning with dual-satellite polarimetry, but architecture, validation and gains are unavailable. Methods, data and results could not be assessed because no reliable abstract was available.
75. Global-scale influence of the freeze-thaw process on soil salinity and sodicity
Core Problem: Title-level focus: identifies the global influence of freeze-thaw on soil salinity and sodicity; data and causal design are unavailable.
Key Innovation: Title-signalled approach or contribution: The apparent contribution is global-scale synthesis or mapping, but its observations, model and findings cannot be verified from the title. Methods, data and results could not be assessed because no reliable abstract was available.
76. Contrasting glacier responses to morpho-topographic controls: Evidence from small and large glacier clusters in the central Himalaya
Core Problem: Title-level focus: indicates comparison of small and large glacier-cluster responses to morpho-topographic controls.
Key Innovation: Title-signalled approach or contribution: The stated contribution is a size-stratified Himalayan comparison, but observations, periods and findings are unavailable. Methods, data and results could not be assessed because no reliable abstract was available.
77. Probabilistic assessment of sediment dynamics in natural and human-modified sub-basins of the Godavari Basin, India
Core Problem: Title-level focus: indicates probabilistic comparison of sediment dynamics between natural and modified sub-basins; variables and period are unavailable.
Key Innovation: Title-signalled approach or contribution: A probabilistic basin-scale assessment is claimed, but modeling approach, uncertainty treatment and findings cannot be verified. Methods, data and results could not be assessed because no reliable abstract was available.
78. Vegetation restoration weakens climatic control on sediment connectivity in ecologically fragile regions
Core Problem: Title-level focus: identifies how vegetation restoration modifies climatic control over sediment connectivity; sites and methods are unavailable.
Key Innovation: Title-signalled approach or contribution: The title claims that restoration weakens climate control, but magnitude, causal evidence and generality cannot be assessed. Methods, data and results could not be assessed because no reliable abstract was available.
79. Revealing drying-induced crack networks and hydrological pathways in loess using image recognition and continuous X-ray CT
Core Problem: Title-level focus: identifies the need to resolve drying-induced cracks and their hydrologic pathways in loess.
Key Innovation: Title-signalled approach or contribution: Combines image recognition with continuous X-ray CT, but resolution, experiments and mechanistic findings are unavailable. Methods, data and results could not be assessed because no reliable abstract was available.
80. 3D mesostructural modeling and DEM simulation of soil-rock mixtures using Minkowski sums
Core Problem: Title-level focus: identifies realistic three-dimensional representation and DEM simulation of soil–rock mixtures.
Key Innovation: Title-signalled approach or contribution: Minkowski sums are named as the mesostructural construction tool, but realism, calibration and mechanical findings are unavailable. Methods, data and results could not be assessed because no reliable abstract was available.
81. Gaussian-process trend framework for efficient simulation of 3D multivariate conditional random fields
Core Problem: Title-level focus: identifies efficient simulation of three-dimensional multivariate conditional random fields.
Key Innovation: Title-signalled approach or contribution: A Gaussian-process trend framework is proposed, but computational gains, conditioning behavior and validation are unavailable. Methods, data and results could not be assessed because no reliable abstract was available.
82. Seismic response of a monopile-supported wind turbine in dry granular soil: A discrete-element study
Core Problem: Title-level focus: identifies wind-turbine monopile response to seismic excitation in dry granular soil.
Key Innovation: Title-signalled approach or contribution: A discrete-element treatment is stated, but excitation scenarios, validation and failure implications are unavailable. Methods, data and results could not be assessed because no reliable abstract was available.
83. Shape effects in granular flow down a rough incline: Understanding the role of particle flatness and elongation
Core Problem: Title-level focus: identifies how flatness and elongation affect granular flow down rough slopes.
Key Innovation: Title-signalled approach or contribution: A shape-resolved mechanics study is indicated, but experimental or numerical approach and mobility results are unavailable. Methods, data and results could not be assessed because no reliable abstract was available.
84. Non-ordinary state-based peridynamics method for predicting excavation damage zone of deep tunnels with initial in-situ stress equilibrium reconstruction
Core Problem: Title-level focus: identifies prediction of deep-tunnel excavation damage while preserving initial in-situ stress equilibrium.
Key Innovation: Title-signalled approach or contribution: A non-ordinary state-based peridynamic formulation with stress-equilibrium reconstruction is named, but verification and field performance are unavailable. Methods, data and results could not be assessed because no reliable abstract was available.
85. 2D time-domain SH-wave full-waveform inversion for geotechnical site characterization: a practical comparison of optimization strategies
Core Problem: Title-level focus: identifies practical optimization choices for two-dimensional time-domain SH-wave inversion.
Key Innovation: Title-signalled approach or contribution: A comparative optimization assessment is indicated, but data, strategies and reconstruction quality are unavailable. Methods, data and results could not be assessed because no reliable abstract was available.
86. A damage model for full-stage mechanical behavior of rocks and application to excavation response analysis of deep-buried tunnels
Core Problem: Title-level focus: identifies constitutive representation of rock behavior across the full damage process.
Key Innovation: Title-signalled approach or contribution: A damage model applied to deep-buried tunnel excavation is claimed, but formulation, calibration and performance are unavailable. Methods, data and results could not be assessed because no reliable abstract was available.
87. Visualization of fracture structures in tunnel surrounding rocks: Deep learning of borehole images and heterogeneous zonal reconstruction
Core Problem: Title-level focus: identifies visualization of heterogeneous fracture structures around tunnels from borehole images.
Key Innovation: Title-signalled approach or contribution: Deep learning and heterogeneous zonal reconstruction are combined, but architecture, spatial accuracy and field validation are unavailable. Methods, data and results could not be assessed because no reliable abstract was available.
88. ScourFusion: A trustworthy large language model-augmented framework for automated extreme-aimed scour depth prediction under heterogeneous information
Core Problem: Title-level focus: identifies automated prediction of extreme scour depth from heterogeneous information.
Key Innovation: Title-signalled approach or contribution: A trustworthy LLM-augmented fusion framework is claimed, but the LLM role, safeguards, data and predictive skill are unavailable. Methods, data and results could not be assessed because no reliable abstract was available.
89. MarsFM: Shading-Regularized Flow Matching for Martian Relief Estimation
Core Problem: We present MarsFM, an image-conditioned latent flow-matching model for local Martian relief estimation from single-band HiRISE RED orthoimagery.
Key Innovation: The method combines a pretrained generative prior with stereo-derived geometric supervision and a differentiable Lunar--Lambert shading objective.
90. 4DGS-Fixer: Generative Sparse-View 4D Gaussian Splatting with Iterative Refinement Guided by Video Diffusion Priors
Core Problem: Sparse-view videos provide poor geometric initialization and missing observations for dynamic 4D Gaussian splatting.
Key Innovation: Fuses multi-view depth for dense initialization and iteratively uses a pretrained video-restoration prior as pseudo-supervision.
91. Multi-viewpoint Geo-localization with Event Cameras
Core Problem: Event-based place recognition lacks robust viewpoint handling and suitably challenging datasets.
Key Innovation: Fine-tunes an event-based vision transformer on five image-to-event converted geotagged datasets and releases a multi-orientation event-VPR benchmark.
92. Combining Object Detection with Geometry-Aware Clustering to Distinguish Overlapping Plants in UAV Imagery
Core Problem: Reliable plant-level information from unmanned aerial vehicle (UAV) imagery is important for automated crop monitoring.
Key Innovation: This study presents a geometry-aware post-detection framework for resolving overlapping plant instances using standard RGB UAV imagery.
93. S3VD: Semantic-Guidance Spatio-Temporal Scanning for Video Deraining
Core Problem: Rain corrupts high-frequency detail and temporal consistency, while one-dimensional Mamba scanning weakens spatial semantics.
Key Innovation: Combines DINOv2 semantic priors with a decoupled-gating spatiotemporal Mamba scan and reports state-of-the-art deraining gains.
94. Cube-Splat: High-Fidelity 360° Gaussian Splatting SLAM via Cubemap Factorization and Adjoint-Consistent Optimization
Core Problem: Existing dense Gaussian-splatting SLAM pipelines are poorly adapted to 360-degree imagery and cross-view geometric consistency.
Key Innovation: Factorizes panoramas into a fixed cubemap with adjoint-consistent multi-face optimization and releases the synthetic SynPano benchmark.
95. Think Locally, Refine Globally for Memory-Efficient 3D Reconstruction
Core Problem: Global attention makes long-sequence 3D reconstruction memory intensive, while local windows accumulate camera-pose drift.
Key Innovation: Adds sparse cross-window attention and register-token global camera refinement to VGGT for bounded-memory streaming reconstruction.
96. VoxelTTO: Voxel-Aligned Feed-Forward 3D Gaussian Splatting with Test-Time Optimization
Core Problem: Pixel-aligned Gaussian regression produces overlap and artifacts, while camera-pose errors misalign novel-view reconstructions.
Key Innovation: Aggregates image features in a global voxel representation, decodes voxel-aligned Gaussians, and adapts lightweight LoRA modules at test time using optional pose supervision.
97. 2D GauSS-MI: Efficient Active Scene Reconstruction with Balanced Visual and Geometric Quality
Core Problem: Active reconstruction must balance visual fidelity, geometric completeness and onboard computational cost.
Key Innovation: Defines a view-dependent reliability model and Shannon-mutual-information criterion over oriented 2D Gaussian splats for active view selection.
98. Refine Then Fusion: Training-Free 3D Point Cloud Adaptation with Priority Refinement and Multi-Modal Knowledge Fusion
Core Problem: Pretrained multimodal 3D features contain redundant channels and sample-dependent modality reliability.
Key Innovation: Performs training-free channel refinement, reliability-aware multimodal fusion and cache-based few-shot inference.
99. SFVO: Decoupled Confidence-Guided Stereo-Flow Visual Odometry with Bidirectional PnP
Core Problem: Monocular VO has scale ambiguity, while learned stereo VO is expensive and underuses stereo-flow geometric complementarity.
Key Innovation: Builds geometric pose constraints from pretrained stereo and optical-flow correspondences with separate confidence maps for rotation and translation.
100. Info3R: Information-Adaptive Test-Time Training for 3D Reconstruction
Core Problem: Online transformer reconstruction degrades over long image streams because it treats redundant and informative frames alike and its internal state loses plasticity.
Key Innovation: Introduces information-weighted state updates and confidence-triggered dynamic resets, reducing trajectory error on extended outdoor KITTI sequences.
101. Available Guardrails: Certifying Selective Prediction across ML Systems
Core Problem: Fine-grained selective predictors may be statistically valid in principle yet impossible to certify for small reporting groups with finite calibration data.
Key Innovation: Defines certificate availability via exact-binomial inversion and optimizes the safety-granularity-coverage frontier with dynamic programming and error-budget reallocation.
102. Robust Structureless Monocular Visual Inertial Initialization Exploiting Line Features and Vanishing Points
Core Problem: Visual-inertial odometry initialization becomes ill-conditioned under degenerate motion and often depends on restrictive excitation or costly 3D reconstruction.
Key Innovation: Builds a structureless initializer from tracked 2D lines, vanishing points and line epipolar/projection residuals, with released code and improved degenerate-motion performance.
103. Refining Ground Truth Poses in Autonomous Driving Datasets via Neural Rendering
Core Problem: Residual calibration and ego-pose errors in large multi-sensor datasets silently contaminate model training and evaluation.
Key Innovation: Jointly refines shared rig extrinsics and continuous-time trajectories with a NeRF pipeline and validates changes without pose ground truth using several geometric consistency tests.
104. Sensing the Vertical Dynamics of the Amazon Rainforest Using Multifrequency Microwave Radiometry
Core Problem: Conventional optical indices miss lower-canopy water-stress dynamics.
Key Innovation: Joint L-, C-, and X-band VOD analysis revealing earlier lower-canopy decline.
105. How can remote sensing support urban tree cadaster efforts and diversity analysis of dominant genera?
Core Problem: Municipal tree inventories omit private land and limit diversity analysis.
Key Innovation: Hierarchical genus classification from bitemporal WorldView imagery.
106. Validation and Comparison of ANEM-SBAC Method for Temperature and Emissivity Separation Using MODIS and VIIRS Data in the València Test Site
Core Problem: Temperature–emissivity separation is biased by atmospheric errors and low contrast.
Key Innovation: Pixel-wise ANEM initialization with single-band atmospheric correction and uncertainty analysis.
107. A Multiscale Rotation Ship Detection Network for SAR Images Based on Scale-Aware Gaussian Loss and Direction Decoupled Attention
Core Problem: Ship scale, aspect ratio, and sidelobes degrade rotated detection.
Key Innovation: Frequency–spatial perception, direction-decoupled attention, adaptive fusion, and Gaussian loss.
108. Intelligent Identification of Ocean Surface Rainfall from CFOSAT Scatterometer Observations
Core Problem: Rain contamination degrades scatterometer wind retrieval.
Key Innovation: Optimized XGBoost and LightGBM exploiting multiangle dual-polarization observations.
109. Raw Materials for the World: The Linkage of Ore and Mineral Mining Activities and Fast-Pace Settlement Development in the Democratic Republic of Congo
Core Problem: Quantifies settlement growth around mining regions in the DRC.
Key Innovation: Long-term remote-sensing proxy analysis linking mining and building expansion.
110. SPG-IAD: Scattering-Point-Guided and IoU-Aware Dynamic Pseudolabel Selection for Semi-Supervised SAR Ship Detection
Core Problem: Speckle and target diversity corrupt SAR pseudolabel confidence.
Key Innovation: Scattering-point dual-criterion filtering with IoU-aware teacher–student selection.
111. Few-Shot SAR Target Classification Via Scattering-Aware Interaction and Distribution Alignment
Core Problem: Clutter and limited samples cause class confusion.
Key Innovation: Scattering-prior attention plus optimal-transport distribution alignment.
112. Multisource Satellite Altimetry Reveals Complex Elevation Dynamics at Klutlan Glacier in Alaska Between 2008 and 2024
Core Problem: Rugged terrain and sensor differences complicate long-term glacier altimetry.
Key Innovation: Bias-corrected fusion of Jason-2, CryoSat-2, and ICESat-2 time series.
113. Gaussian-process fusion of ICESat-2 and SMOS/SMAP data for daily estimates of Arctic sea ice thickness and volume
Core Problem: Sea-ice thickness lacks complete daily coverage.
Key Innovation: Gaussian-process fusion of laser altimetry and passive microwave data.
114. SDUST2025_SWOT: Global marine deflection of the vertical and gravity anomaly datasets derived from SWOT/KaRIn wide-swath altimetry using local geoid fitting
Core Problem: Derives global fine-resolution gravity information from SWOT swaths.
Key Innovation: Local-geoid fitting for 1-arc-minute DOV and gravity products.
115. Evaluation of plume rise parameterizations in GEM-MACHv2 with analysis of image data using a deep convolutional neural network
Core Problem: Operational plume-rise schemes can overpredict stack-plume height.
Key Innovation: Two-year imagery analyzed by a deep CNN to evaluate GEM-MACHv2 parameterizations.
116. Modeling of radiative transfer through cryospheric Earth system: software package SCIATRAN
Core Problem: Complex cryospheric surfaces are inadequately represented in radiative transfer.
Key Innovation: SCIATRAN extension validated against measurements.
117. Evaluation of the LandscapeDNDC model for drained peatland forest managements, LDNDC v1.35.2 (revision 11434)
Core Problem: Assesses peatland hydrology and carbon responses to forestry regimes.
Key Innovation: Process-based evaluation across control, rotational, and continuous-cover management.
118. Variational Stokes method applied to free surface boundaries in numerical geodynamical models using the staggered-grid finite-difference discretisation
Core Problem: Staggered-grid geodynamic solvers need accurate and stable representation of deforming free surfaces.
Key Innovation: Applies a variational Stokes formulation to free-surface boundaries within a staggered-grid finite-difference discretization.
119. Lagrangian tracking methods applied to free surface boundaries in numerical geodynamic models
Core Problem: Geodynamic simulations require free-surface tracking that limits mesh distortion and computational cost.
Key Innovation: Evaluates Lagrangian tracking strategies for evolving free-surface boundaries in numerical geodynamic models.
120. Optimization of Farmland Management Zoning in the Black Soil Region: A Climate Adaptability Assessment Considering Crop Growth Response and Topographic Characteristics
Core Problem: Single-year management zones fail under variable hydroclimate.
Key Innovation: Multi-period feature fusion with heterogeneous spatial attention and stability evaluation.
121. SASI-Net: A Sparse Adaptive SAR Imaging Network with Complex-Valued Mask-Aware Completion and Adaptive Truncated Penalty
Core Problem: Dense millimeter-wave SAR sampling is costly.
Key Innovation: Complex-valued mask-aware echo completion with physics-guided imaging and adaptive penalty.
122. WSF-YOLO: Wavelet-Guided Spatial-Frequency Joint Modeling for Aircraft Detection in SAR Images
Core Problem: Aircraft signatures are weak amid SAR clutter and scale variation.
Key Innovation: Wavelet-guided joint spatial–frequency YOLO design.
123. Assessing the Sensitivity of Sentinel-1 and Sentinel-2 Time Series to Wheat Yellow Rust Using Feature Evaluation, Separability, and Random Forest Classification
Core Problem: Determines when radar and optical signals detect wheat yellow rust.
Key Innovation: Joint separability and Random Forest evaluation across seasonal time series.
124. Disentangling Topographic Constraints and the Climatic and Anthropogenic Pathways Shaping Vegetation Greenness in the Southeastern Qinghai-Xizang Plateau
Core Problem: Separates long-term greening from terrain, climate, and human associations.
Key Innovation: Combines robust trends, breakpoints, geographic detector, and structural-equation modeling.
125. Label-Efficient Semi-Supervised Building Semantic Segmentation for High-Resolution UAV Imagery
Core Problem: Pixel annotation is costly across new urban regions.
Key Innovation: Frozen DINOv2 encoder, EMA teacher–student learning, and boundary-only labeled supervision.
126. Anomaly-Driven Gated Fusion Network for Infrared Small Target Detection
Core Problem: Tiny targets are submerged in clutter and global attention is expensive.
Key Innovation: Statistical local-anomaly blocks and anomaly-gated feature pyramids in a sub-million-parameter network.
127. Fine-Grained Tree Species Classification in Urban Forests via Phenology-Structure Synergistic Fusion of Multi-Source Remote Sensing Data
Core Problem: Spectrally similar species and heterogeneous structure limit fine-grained mapping.
Key Innovation: Reliability-aware fusion of imagery, phenology, LiDAR, and hyperspectral-semantic priors.
128. FuzzyQSM: A Novel Tree Structure Reconstruction Approach Using Uncertainty
Core Problem: Beamwidth and occlusion make tree point clouds uncertain.
Key Innovation: Fits cylinders to fuzzy point clouds using expected Mahalanobis distance.
129. Experimental Study on the Dynamic Response Characteristics of Cross Twin Tunnels Under Subway Train Load
Core Problem: Intersecting tunnel response under train-induced loads is poorly characterized.
Key Innovation: Physical model tests varying force and frequency with a passenger-load response model.
130. Pile Resistance and Ground Vibration During Hammer Driving in Stratified Soils
Core Problem: Pile resistance and vibration during large-diameter hammer driving lack integrated field evidence.
Key Innovation: FBG pile measurements combined with radial surface-vibration monitoring.
131. Improved Simplified Continuum Model for Laterally Loaded Flexible Piles
Core Problem: Simplified pile models neglect vertical soil displacement.
Key Innovation: Variational continuum solution with nonzero horizontal and vertical soil displacement.
132. Isolation performance of non-uniform local resonant block on ground vibrations induced by train operation
Core Problem: Reduces train-induced ground vibration across relevant frequencies.
Key Innovation: Non-uniform locally resonant block for vibration isolation.
133. A coupled hydro-mechanical-chemical-biological model for microbially induced calcite precipitation in soils
Core Problem: Existing MICP models often omit important process couplings, treat precipitation stress phenomenologically, or underrepresent attached biomass.
Key Innovation: Derives a thermodynamically consistent hydro–mechanical–chemical–biological framework and validates it against published simulations and two column experiments, finding limited biomass contribution to porosity reduction.
134. The spatial vibration responses and energy evolution mechanisms of sandstone subgrade under different vibration parameters
Core Problem: Title-level focus: characterizes spatial vibration and energy evolution in sandstone subgrade.
Key Innovation: Title-signalled approach or contribution: Parametric analysis of vibration-response and energy-transfer mechanisms. Methods, data and results could not be assessed because no reliable abstract was available.
135. Expert knowledge-guided decision calibration for accurate fine-grained tree species classification
Core Problem: Title-level focus: identifies fine-grained tree-species classification and decision calibration as the target; data and model details are unavailable.
Key Innovation: Title-signalled approach or contribution: Expert-knowledge guidance is presented as the calibration mechanism, but its implementation and empirical benefit cannot be verified. Methods, data and results could not be assessed because no reliable abstract was available.
136. Exploring the spatial heterogeneity of deep soil drought by using electrical resistivity tomography (ERT) at multiple scales
Core Problem: Title-level focus: indicates investigation of spatially heterogeneous deep-soil drought across scales; design and location are unavailable.
Key Innovation: Title-signalled approach or contribution: Multi-scale ERT is the stated measurement strategy, but resolution, validation and findings cannot be assessed from the title. Methods, data and results could not be assessed because no reliable abstract was available.
137. Soil quality responses to ski piste disturbance in a semi-arid mountainous region of Iran
Core Problem: Title-level focus: indicates assessment of soil-quality responses to ski-piste disturbance in semi-arid mountain terrain.
Key Innovation: Title-signalled approach or contribution: The geographic and disturbance setting are explicit, but indicators, comparisons and findings cannot be verified without an abstract. Methods, data and results could not be assessed because no reliable abstract was available.
138. Interpretable early-stage advance rate prediction for slurry shield tunnelling using TabPFN-SHAP
Core Problem: Title-level focus: identifies data-scarce early-stage prediction of slurry-shield advance rate; dataset and features are unavailable.
Key Innovation: Title-signalled approach or contribution: The title combines TabPFN with SHAP, but model performance and operational usefulness cannot be verified. Methods, data and results could not be assessed because no reliable abstract was available.
139. Theoretical analysis of the time-dependent response of roadway surrounding rock under energy-driven conditions and the synergistic control of support structures
Core Problem: Title-level focus: identifies delayed rock response under energy-driven conditions and interaction with support structures.
Key Innovation: Title-signalled approach or contribution: A synergistic theoretical treatment is implied, but the governing model and validation are unavailable. Methods, data and results could not be assessed because no reliable abstract was available.
140. Corotational FEM-DEM for Multiscale and Multiphase Modelling of Granular Soils
Core Problem: Title-level focus: identifies coupled continuum–discrete representation of granular soils across scales and phases.
Key Innovation: Title-signalled approach or contribution: A corotational FEM–DEM framework is named, but coupling, benchmarks and demonstrated applications are unavailable. Methods, data and results could not be assessed because no reliable abstract was available.
141. A reliable machine learning tool for berm breakwater recession prediction
Core Problem: Title-level focus: identifies reliable prediction of berm-breakwater recession.
Key Innovation: Title-signalled approach or contribution: A machine-learning tool is claimed, but inputs, reliability treatment and validation are unavailable. Methods, data and results could not be assessed because no reliable abstract was available.