TerraMosaic Daily Digest: September 1, 2026
Daily Summary
September 1, 2026 is dominated by geohazards whose consequences depend on delayed triggering, inherited damage, and spatially variable mechanical state. The Hunga study attributes the largest 2022 tsunami to a sudden caldera-subsidence episode roughly an hour after the globally detected explosion, while paired 2013 and 2022 Lushan earthquakes show that earlier shaking can either reduce later landsliding by sediment removal or amplify it through crack damage and incomplete recovery. Related earthquake papers resolve fine-scale Tohoku plate coupling from joint geodetic and intraplate seismic evidence, explain shallow creep and confined rupture in mica schist, and translate near-fault pulse structure, fault-rupture offsets, and combined earthquake-tsunami loading into more explicit hazard metrics for foundations, embankments, tanks, heritage structures, and reinforced-concrete frames.
Across slopes, soils, coasts, and flood systems, the corpus moves from generic susceptibility toward calibrated operational controls. Regional postfire debris-flow models built from 3,788 observations extend rainfall-threshold support across the western United States; differentiable inversion, horn-shaped 3D reliability analysis, and crack-kinetic experiments sharpen diagnosis of heterogeneous slope failure; and centrifuge tests plus surrogate PM4Sand calibration show how liquefaction depends on interlayers, scale, and boundary conditions. Flood, storm-surge, and coastal studies likewise focus on process coupling and geometry, combining exit-level station disruption metrics, Bayesian flood susceptibility, precipitation nowcasting and artifact detection, tide-gauge surge reconstruction, reef and shoreline mapping, estuarine sediment-supply change, and semiarid river adjustment to show that exposure is filtered through infrastructure bottlenecks, atmospheric structure, and evolving landforms.
A secondary methods cohort expands transferable observation and modeling capacity without claiming hazard-specific validation. Open-world landslide diffusion models, open-vocabulary and fog-robust change detection, change-point-aware deformation monitoring, multimodal wildfire segmentation, SAR-to-EO translation, online and semantic 3D reconstruction, cross-modal descriptors, human-in-the-loop segmentation, hyperspectral unmixing, sparse wavefield reconstruction, frequency-aware seismic distillation, kernel-corrected neural operators, optimizer-aware PINNs, and agent-assisted constitutive design all emphasize semantic openness, missing-modality robustness, and physics or geometry constraints. Together they mark a stronger technical substrate for future geohazard analysis, but the direct hazard findings in this digest remain concentrated in the earthquake, landslide, flood, coastal, aeolian, wildfire, and geotechnical studies themselves.
Key Trends
Five trajectories organize the September 1, 2026 corpus: sequenced hazard attribution, inherited-state failure mechanics, regional operational thresholds, geometry-conditioned hydroclimatic risk, and semantically flexible Earth-observation methods that remain largely transferable rather than hazard-validated.
- Multi-trigger hazard attribution: Several studies resolve hazard causation by sequencing triggers rather than treating events as singular shocks: delayed caldera subsidence, sequential earthquake hillslope memory, combined earthquake-tsunami loading, and joint geodetic-seismic coupling all depend on when and where distinct processes interact.
- Failure mechanics as heterogeneous state: Differentiable slope inversion, horn-shaped 3D reliability, mudstone crack kinetics, liquefaction centrifuge tests, PM4Sand surrogate calibration, freeze-thaw contaminated-soil experiments, and rock-bolt pull-out tests all treat instability as a spatially variable, path-dependent material state rather than a single factor of safety.
- Regional and infrastructure-specific thresholds: Postfire debris-flow thresholds across the western United States, West Kowloon exit-based flood disruption metrics, Philippine tide-gauge surge reconstruction, Gobi railway sand-barrier monitoring, and season-limited MICP stabilization translate process studies into place-specific warning, maintenance, or design rules.
- Coupled hydroclimatic and geomorphic controls: Tropical-cyclone upper-level warming, high-resolution precipitation nowcasting, precipitation-artifact screening, reef-opening wave experiments, reef-island shoreline mapping, estuarine sediment-supply analysis, and vegetation-influenced overland-flow modeling all show that hazard intensity depends on coupled atmospheric, hydrodynamic, and morphologic structure.
- Semantic flexibility under incomplete Earth-observation data: EarthLD, zero-shot aerial segmentation, open-vocabulary change detection, fog-robust satellite-UAV benchmarking, SAR-to-EO translation, missing-modality change networks, open-vocabulary 3D reconstruction, sparse-point supervision, and cross-modal descriptors push Earth observation toward adaptable, low-label workflows, but most papers remain transferable methods rather than geohazard validations.
Selected Papers
The selected papers combine direct studies of volcanic tsunamis, earthquake-ground interactions, landslides, floods, wildfire, coastal and aeolian change, cryosphere dynamics, and geotechnical degradation with engineering analyses of structures and infrastructure under coupled loading. A secondary cohort contributes transferable remote-sensing and scientific-machine-learning methods for mapping, change detection, fusion, reconstruction, and operator learning that may support geohazard workflows but are not presented here as hazard-specific validation studies.
1. Delayed submarine caldera subsidence creates extreme tsunami hazard
Core Problem: Determine which eruption processes generated the 2022 Hunga tsunami and when they occurred.
Key Innovation: Combines hydro-acoustic, seismic, oceanographic, atmospheric, and eyewitness observations to identify delayed sudden caldera subsidence as the dominant tsunami trigger.
2. Hillslope memory of sequential earthquakes
Core Problem: Determine whether one major earthquake stabilizes or destabilizes hillslopes for the next nearby event.
Key Innovation: Uses the paired 2013 and 2022 Lushan earthquakes to reveal competing removal and damage effects and three path-dependent hillslope recovery pathways.
3. Regional Models for Postfire Debris-Flow Likelihood and Rainfall Thresholds Across the Western United States
Core Problem: How can postfire debris-flow likelihood and rainfall thresholds be recalibrated so rapid assessments work reliably outside the original southern California training region?
Key Innovation: Expanded the calibration inventory to 3788 observations from 67 burned areas and produced regionalized models that outperform the legacy USGS M1 framework.
4. Detection of Interannual and Fine-Scale Plate Coupling Variations Using Intraplate Earthquakes and Geodetic Data: Application to the Tohoku-Oki Plate Boundary
Core Problem: How can geodetic slip-deficit models and intraplate earthquake mechanisms be combined to resolve fine-scale, interannual coupling changes before large earthquakes?
Key Innovation: Introduced a joint geodetic-seismic coupling diagnostic that detects temporal and fine-scale heterogeneity not recoverable from geodesy alone.
5. EarthLD: Towards Unified Open-World Landslide Understanding via Vision-Language Guided Diffusion Models
Core Problem: Automated landslide detection and mapping suffer from irregular morphology, ambiguous spectra, and cross-sensor domain shift.
Key Innovation: Vision-language-guided diffusion framework plus a harmonized global open-world landslide benchmark for joint recognition and mapping.
6. Shallow creep and confined ruptures in frictionally unstable mica schist
Core Problem: Explain how phyllosilicate-rich mica schist controls aseismic creep and depth-limited seismic rupture under mid-crustal conditions.
Key Innovation: Integrates triaxial experiments and thermal-structure modeling to show instability at 300 to 400 C consistent with recent East Anatolian rupture behavior.
7. Empirical pulse model for near-fault ground motion considering magnitude, distance, and site characteristics
Core Problem: Pulse-like ground motions are hard to identify consistently and parameterize for hazard analysis.
Key Innovation: Couples pulse extraction with GCWT and builds empirical pulse predictors from magnitude, distance, and site conditions.
8. Setback distances for rectangular foundations subjected to earthquake-induced normal-fault surface rupture: 1 g physical model tests
Core Problem: Setback distances for foundations under normal-fault surface rupture lack physical-model evidence.
Key Innovation: Converts rupture evolution, soil pressures, tilt, strain, and contact stress into quantified setback envelopes.
9. Differentiable inverse analysis of spatial-heterogeneous slope stability fusing monitoring data
Core Problem: Infer spatially heterogeneous slope-stability conditions by fusing monitoring data
Key Innovation: Uses differentiable inverse analysis to integrate monitoring data into heterogeneous slope stability estimation
10. Centrifuge modeling of seismic liquefaction and weakening in deep sandy soils with fine-grained interlayers
Core Problem: Understand liquefaction and weakening behavior of deep sandy soils with fine-grained interlayers during earthquakes
Key Innovation: Uses centrifuge modeling to isolate the effect of fine-grained interlayers on deep-soil liquefaction
11. Damage to geotechnical structures induced by combined earthquake and tsunami loading during the 2024 Noto Peninsula earthquake
Core Problem: Characterize damage to geotechnical structures under combined earthquake and tsunami loading during the 2024 Noto Peninsula earthquake
Key Innovation: Centers combined earthquake-tsunami loading effects on geotechnical damage in a real disaster case
12. Dust Pollution Mitigation by Solar Farm Deployment in Northern China
Core Problem: Can solar-farm deployment in northern China reduce dust emissions and downstream particulate pollution?
Key Innovation: Coupled dust emission, transport, and air-quality modeling to quantify present and future PM reductions from solar-farm expansion.
13. Learning the Shoreline: A Very High-Resolution Approach to Reef Island Dynamics
Core Problem: Conventional island stability proxies miss fine-scale, short-term shoreline change that matters for low-lying reef-island risk.
Key Innovation: Develops a transferable VHR Pleiades plus XGBoost shoreline-mapping workflow that resolves subtle island-scale extent, shape, and position changes across multiple atolls.
14. Multimodal RGB-Infrared Combination for UAV-Based Wildfire Segmentation: A Comparative Study on FLAME3
Core Problem: Determining how RGB, infrared, and fusion strategies affect UAV-based wildfire segmentation performance.
Key Innovation: Systematically compares fusion timing and architectures and shows thermal information dominates while feature-level fusion is strongest.
15. Scale-based Approach for Active Wildfire Segmentation on Satellite Imagery
Core Problem: Robustly segmenting sparse active-fire pixels across wildfire size regimes in Landsat-8 imagery.
Key Innovation: Introduces a scale-based robustness protocol and identifies SWIR2 and U-Net as consistently strong choices for active-fire mapping.
16. Experimental and numerical study on overflow of rectangular tank under large-amplitude seismic excitation
Core Problem: Quantify overflow behavior of a rectangular tank under large seismic excitation.
Key Innovation: Combined experimental and numerical analysis of earthquake-driven tank overflow.
17. Investigating historical storm surge occurrences in the Philippines from tide gauge observations
Core Problem: The Philippines lacked a nationwide instrument-based reconstruction of historical storm surge events.
Key Innovation: Builds a 1947-2024 tide-gauge storm surge database and quantifies coastal controls on surge magnitude.
18. An integrated remote sensing and AHP framework for forest fire vulnerability assessment in Gironde, France
Core Problem: Gironde lacked a high-resolution wildfire vulnerability assessment integrating environmental and human drivers.
Key Innovation: Combines remote sensing factors with AHP weighting and validates the resulting map using 699 FIRMS ignitions.
19. Low-temperature limitations and seasonal applicability of MICP for aeolian sand stabilization on the Qinghai-Tibet Plateau
Core Problem: Low temperatures may undermine microbial sand stabilization on the Qinghai-Tibet Plateau.
Key Innovation: Identifies the mismatch between carbonate precipitation and actual sand strengthening, defining a viable summer treatment window.
20. UAV photogrammetry for evaluating sand-blocking performance and accumulation dynamics of railway sand barriers in the Gobi High-Wind Corridor
Core Problem: Sand-barrier designs lack high-resolution evidence on accumulation dynamics and maintenance timing.
Key Innovation: Uses UAV photogrammetry to compare barrier designs and derive performance and maintenance recommendations.
21. A 3D finite element limit equilibrium framework for reliability analysis of complex slopes: Demonstrating the superiority of horn-shaped failure mechanisms
Core Problem: Complex 3D slopes need stability and reliability analysis that handles spatial variability and realistic failure shapes.
Key Innovation: Extends FELEM with horn-shaped failure mechanisms and MLE-based random-field reliability analysis.
22. A novel dual-parameter method for evaluating seismic collapse severity in masonry pagodas
Core Problem: Historic masonry pagodas lack a quantitative collapse-severity metric for earthquake damage assessment.
Key Innovation: Combines area loss and residual height into a validated two-parameter collapse index.
23. Seismic performance of multi-story traditional hybrid pavilion-style timber structures: shaking table tests and numerical analysis
Core Problem: Seismic behavior of multi-story hybrid timber pavilions is poorly constrained.
Key Innovation: Validates a shaking-table-calibrated model and shows how column-foot flexibility governs drift and acceleration tradeoffs.
24. Assessing rail station flood disruption using an exit-based framework: Stress testing Hong Kong's West Kowloon Station
Core Problem: Station-scale flood assessments miss how a small set of exits controls overall hub accessibility.
Key Innovation: Couples 2D flooding, exit operability functions, and throughput-based resilience metrics in an exit-centered framework.
25. Flood susceptibility assessment using a multi-sampling discriminative Bayesian network
Core Problem: Map flood-prone areas using a probabilistic susceptibility model
Key Innovation: Introduces a multi-sampling discriminative Bayesian network for flood susceptibility assessment
26. Quantifying the seismic failure probability of railway embankments considering uncertainty in near-fault pulse parameters
Core Problem: Quantify railway embankment failure probability under uncertain near-fault pulse motions
Key Innovation: Propagates near-fault pulse uncertainty into seismic failure probability estimates for railway embankments
27. Calibration of the PM4Sand model using surrogate modeling: Assessing soil liquefaction predictions across scales and boundary conditions
Core Problem: Calibrate PM4Sand so liquefaction predictions remain reliable from element tests to field-scale slopes
Key Innovation: Uses surrogate modeling for automated PM4Sand calibration and cross-scale evaluation of liquefaction predictions
28. Stage-Dependent Shifts in Controls on Annual Erosion-Deposition Change Under Long-Term Sediment-Supply Decline in the Minjiang Estuary, Southeastern China
Core Problem: Which factors controlled stage-dependent erosion and deposition changes in the Minjiang Estuary under long-term sediment-supply decline?
Key Innovation: Built a stage-specific interpretable ML framework that combines DEM differencing with nonlinear driver attribution across multiple post-dam periods.
29. Impact of Upper-Level Warming on Tropical Cyclone Intensity
Core Problem: What causes upper-level warming in intensifying tropical cyclones and how does it affect potential intensity?
Key Innovation: Linked overshooting-convection-driven stratospheric subsidence to upper-level warming and quantified its effect on cyclone potential intensity.
30. GenONet: A Generative operator Network for High-Resolution Precipitation Nowcasting
Core Problem: Long-horizon high-resolution precipitation nowcasts become blurry and lose physical consistency.
Key Innovation: DeepONet-based generative nowcaster with adversarial training and moisture-conservation regularization.
31. FractalNet-Based Heterogeneous Federated Learning for Orbital Edge Intelligence in Satellite Mega-Constellations: A Wildfire Case Study
Core Problem: How to train edge intelligence across heterogeneous satellite constellations with limited contacts and unequal SWAP-C budgets.
Key Innovation: A FractalNet-based depth-heterogeneous federated framework with contact-aware scheduling and multi-tier orbital control.
32. ReA-OVCD: Training-Free Open-Vocabulary Change Detection via Semantic-Spatial Reliability Assessment
Core Problem: Improve reliability of pixel-level open-vocabulary change detection from remote-sensing imagery.
Key Innovation: Training-free two-stage semantic-spatial reliability assessment over candidate change regions.
33. A Change-Point-Based Deformation Grouping Strategy in Long-Term Near-Real-Time Deformation Monitoring
Core Problem: Sequential InSAR estimation can smooth or miss abrupt deformation changes in long monitoring series.
Key Innovation: Uses BiLSTM plus BEAST to detect deformation change points and define adaptive processing groups.
34. Evolution of cracks in red-bed mudstone under the coupling of drying-wetting cycle and temperature: From microscopic mechanism to a kinetic model
Core Problem: Drying-wetting and temperature coupling drives crack evolution in red-bed mudstone.
Key Innovation: Appears to link microscopic damage mechanisms with a kinetic crack-evolution model.
35. Multi-hazard mitigation measures plan for Ischia Island after the November 26th, 2022, landslide and flood events
Core Problem: Ischia needs an integrated mitigation plan after the 26 November 2022 landslide and flood disaster.
Key Innovation: Appears to frame coordinated island-scale post-event multi-hazard mitigation measures.
36. Morphological Changes and Alterations in Erosion-Accretion Patterns in a Semiarid Brazilian River
Core Problem: How did dam regulation and local geomorphic controls reshape erosion, deposition, and bank retreat in the Jaguaribe River over 1958-2025?
Key Innovation: Integrated long-horizon imagery, geomorphic mapping, field surveys, and hydrosediment data to separate basin-scale dam effects from reach-scale controls.
37. Restrict, Don't Retrain: Inference-Time VLM Guidance for Zero-Shot Aerial Segmentation
Core Problem: Zero-shot aerial segmentation misses relevant classes and small objects in specialized scenes.
Key Innovation: Inference-time VLM guidance that selects classes and localizes missed small objects without retraining.
38. A Sensor-Adaptive Incremental Learning Framework for Artifact Detection in Satellite Precipitation Data
Core Problem: How to detect sensor-specific artifacts in satellite rainfall products when labeled anomaly data are scarce.
Key Innovation: A sensor-adaptive incremental anomaly detector using pretrained vision models with explainability and iterative human refinement.
39. PBG-WaveNet: Probabilistic buoy-to-grid reconstruction and forecasting of significant wave height fields from sparse buoy observations
Core Problem: Reconstructs and forecasts gridded significant wave height fields from sparse buoy observations.
Key Innovation: Proposes a probabilistic buoy-to-grid model that jointly infers wave fields and forecast uncertainty from sparse sensors.
40. Influences of stoss-face opening configurations on irregular wave energy partition and frequency dependence across reef structures: Laboratory investigation
Core Problem: Tests how stoss-face opening configurations alter irregular-wave energy partition across reef structures.
Key Innovation: Resolves frequency-dependent wave-energy redistribution caused by different reef-opening geometries.
41. Implementation of the Generalized Double-Moment scaling Normalization method for raindrop size distribution in a WRF 4.3.1 bulk-type cloud microphysics scheme: a case study over the Korean Peninsula
Core Problem: Improves WRF rainfall simulation by implementing generalized double-moment normalization for raindrop size distributions.
Key Innovation: Injects observation-informed raindrop-size scaling into WDM6 and shows better precipitation, reflectivity, and storm-track performance.
42. Improving Backscatter-Based Surface Water Classification in Arid Environments Through Interferometric Coherence
Core Problem: Improve SAR-based surface-water classification where backscatter alone confuses water with look-alike dry surfaces.
Key Innovation: Shows interferometric coherence complements Sentinel-1 backscatter and that simple logical-AND fusion sharply cuts commission errors.
43. An RGIK rock glacier inventory compiled independently by two novel (trained) operators in Val Grisenche, Italy: Degree of consistency and training needs
Core Problem: Novice operators may map rock glaciers inconsistently under RGIK guidelines.
Key Innovation: Quantifies detection and misclassification errors against an expert benchmark to target training weaknesses.
44. Foggy-UAVCD and DSRF-Net: Benchmarking robust satellite-UAV change detection under fog degradation
Core Problem: Satellite-UAV change detection lacks benchmarks and models robust to fog degradation.
Key Innovation: Appears to introduce a fog-degraded benchmark dataset and a dedicated robust fusion network.
45. Segmented Crust of Southeastern Iran Reveals Complex Dynamics of the Zagros-Makran Transition: Insights From Ambient Noise Full-Waveform Inversion
Core Problem: What crustal and uppermost mantle structure characterizes the Zagros-Makran transition in southeastern Iran?
Key Innovation: Used ambient-noise full-waveform inversion to derive a high-resolution 3D shear-wave velocity model that resolves segmentation, slab geometry, and major sedimentary packages.
46. Enhanced Dry Season Moisture Recycling in the Congo and Amazon Rainforests
Core Problem: How does dry-season moisture recycling change in the Amazon and Congo rainforests during mean and dry years?
Key Innovation: Provided a long-record cross-continental comparison showing greater dry-season reliance on rainforest-sourced rainfall and stronger downwind dependence during dry years.
47. CrossFeat: Bridging Imaging Modalities in Feature Descriptor Space
Core Problem: How can a standard monomodal descriptor be made effective across fundamentally different imaging modalities without pair-specific retraining?
Key Innovation: Learned a crossing function in descriptor space that preserves geometry while translating appearance across modalities, including satellite imagery.
48. Neural means and kernel corrections for operator learning
Core Problem: Can neural operator models be improved by exact kernel correction on residuals and learned features?
Key Innovation: Paired neural means with exact Matern-kernel residual correction and provided theory for when feature-space correction outperforms raw-state correction.
49. Do Satellites See Commuters? A Critical Benchmark of Vision Foundation Models
Core Problem: Which satellite foundation-model features transfer best in origin-destination generation across geographies.
Key Innovation: Controlled multi-encoder benchmark revealing tradeoffs between language-supervised and geographically grounded pretraining.
50. HarmoCore: Functional Latent Diffusion for Sparse Reconstruction of Oscillatory Wave Fields
Core Problem: How to reconstruct highly oscillatory complex wave fields from extremely sparse sensors.
Key Innovation: Functional Tucker latent diffusion with core-space posterior sampling and optional PDE residual guidance.
51. Design and Implementation of a Kalman Filter-Infused Algorithm for Tilt Estimation
Core Problem: Accelerometer noise and gyroscope drift reduce the stability of real-time tilt estimates from low-cost sensors.
Key Innovation: Implements Kalman fusion of MPU6050 accelerometer and gyroscope signals on an RP2040 platform and validates noise and drift suppression in simulation and hardware tests.
52. RingMoClaw: An Experience-Inspired Multi-Agent Framework for Self-Evolving Research in Remote Sensing
Core Problem: Remote-sensing model improvement still depends heavily on manual trial-and-error rather than autonomous research iteration.
Key Innovation: Closed-loop multi-agent framework with research, quality-control, and dual-stream experience memory for self-evolving remote-sensing experimentation.
53. A Frequency-Aware Dynamic Knowledge Distillation Framework: An Effective Tool for Bridging Low- and High-Frequency Seismic Information
Core Problem: Full-band distillation of seismic data misses the different roles of low-frequency structural information and high-frequency detail.
Key Innovation: Proposes a teacher-student framework that decomposes seismic features by frequency and distills low- and high-frequency knowledge separately with cross-domain alignment.
54. On-the-Fly3R: Towards Robust Online 3D Reconstruction with Feed-Forward 3R Models for Large-Scale UAV Scenarios
Core Problem: Feed-forward 3D reconstruction methods do not scale well to large, weakly ordered UAV image streams typical of cross-strip mapping.
Key Innovation: Enables training-free progressive reconstruction from unordered UAV imagery using retrieval-guided subset selection, validation-rejection-retry consistency control, and pose-graph optimization.
55. ReFlowSET: Representation-Aligned Latent Flow Matching for SAR-to-EO Image Translation
Core Problem: Latent SAR-to-EO translation quality is constrained by mismatched autoencoders inherited from natural-image pretraining.
Key Innovation: Selects the latent codec through a SAR-EO reconstruction audit and trains a compact conditional flow-matching transformer with foundation-model representation alignment.
56. Lightweight Interpretable RGB-Guided Hyperspectral Super-Resolution under Real Cross-resolution Misalignment
Core Problem: Residual RGB-hyperspectral misregistration can inject false high-frequency detail into guided spectral super-resolution.
Key Innovation: Combines cross-modal flow alignment, confidence-weighted spectral regression, and gated fusion in a lightweight framework that transfers across spectral supports without retraining.
57. Agentic Multimodal Models for Environmental Hyperspectral Unmixing
Core Problem: Modular hyperspectral unmixing pipelines often retain redundant or ambiguous endmembers and propagate upstream errors.
Key Innovation: Adds an LVLM agent that iteratively inspects spectral and spatial evidence, edits the endmember set, and re-estimates abundances.
58. Predicting Subsurface Abnormalities Growth using Physics-Informed Neural Networks
Core Problem: Forecasting the temporal evolution of subsurface abnormalities from GPR data requires both spatial-temporal learning and electromagnetic consistency.
Key Innovation: Embeds electromagnetic-wave constraints in a CNN-attention-ConvLSTM architecture for physics-informed prediction of GPR abnormality growth.
59. Gradient-Update Mismatch: Rethinking Conflict-Free Training of Physics-Informed Neural Networks
Core Problem: Gradient-surgery conflict removal in PINNs breaks after optimizer transformations.
Key Innovation: Aligns the actual optimizer update, not just the raw gradient direction, with the conflict-free cone and adjusts optimizer state accordingly.
60. C-DiffSET: Leveraging Latent Diffusion for SAR-to-EO Image Translation with Confidence-Guided Reliable Object Generation
Core Problem: How to translate noisy SAR imagery into reliable EO-like imagery without overfitting limited paired datasets.
Key Innovation: A pretrained latent-diffusion framework with confidence-guided loss to suppress temporal mismatch artifacts and unreliable object generation.
61. Ov3R: Open-Vocabulary Semantic 3D Reconstruction from RGB Videos
Core Problem: Build globally consistent open-vocabulary semantic 3D reconstructions from RGB video.
Key Innovation: CLIP-informed dense point reconstruction plus 2D-to-3D semantic feature lifting.
62. LLM-driven design of physics-constrained constitutive models: two agents are better than one
Core Problem: Automate constitutive model design while enforcing fundamental physical constraints.
Key Innovation: Creator-Inspector multi-agent workflow that audits nine physics constraints during model generation.
63. Quality-aware cooperative sonar coverage planning for multiple AUVs under ocean-current and dynamic constraints
Core Problem: Cooperative AUV surveys must balance sonar coverage quality, communication, collision avoidance, ocean currents, and executable motion constraints.
Key Innovation: Combines MAPPO cooperative planning with command-level feasibility projection and cell-level sonar-quality accounting under a fixed mission budget.
64. A smoothed particle hydrodynamics porous media model to simulate wave-permeable structure interactions with an adaptive particle splitting and merging scheme
Core Problem: Simulates wave interactions with permeable structures using smoothed particle hydrodynamics in porous media.
Key Innovation: Adds adaptive particle splitting and merging to a porous-media SPH framework for wave-structure interaction.
65. Evaluating different roughness approaches and infiltration parameters for vegetation-influenced overland flow in hydrological model
Core Problem: Tests roughness, infiltration, and initial-moisture treatments for vegetation-influenced overland flow modelling.
Key Innovation: Separates how vegetation-induced roughness and infiltration can be represented while exposing initial soil moisture as the main unresolved control.
66. A Terrain-Corrected Vegetation Index Strategy for Improving Leaf Area Index Estimation in Mountainous Areas
Core Problem: Improve leaf area index estimation in mountainous areas where terrain distorts reflectance and vegetation indices.
Key Innovation: Combines topographic correction models with terrain-corrected vegetation indices and random-forest retrieval to improve LAI accuracy across slope and aspect conditions.
67. Controlled Evaluation of Sentinel-2 Annual Compositing Strategies for Deep Learning-Based Mangrove Mapping in China
Core Problem: Determine which annual Sentinel-2 compositing rule best supports national-scale deep-learning mangrove mapping.
Key Innovation: Holds model and validation settings fixed to isolate compositing-rule effects, showing median compositing yields the most balanced national-scale output.
68. iSAGE: A Human-in-the-Loop Framework for Remote Sensing Semantic Segmentation via Sparse Point Supervision
Core Problem: Cut annotation cost for remote-sensing semantic segmentation without relying on self-generated pseudo-labels that can reinforce confident mistakes.
Key Innovation: Introduces an iterative expert-click loop that targets confident errors and recovers near-dense performance from extremely sparse point labels.
69. Building Footprint Extraction in High-Density Urban Areas Based on Multi-Source Remote Sensing Data Fusion and ACM-PSPNet
Core Problem: Extract accurate building footprints in dense urban areas with shadows, adjacency, and complex boundaries.
Key Innovation: Fuses orthophotos and nDSM data in an attention-augmented PSPNet variant that improves boundary recovery and adjacent-building separation.
70. Deep Learning for Water Body Segmentation in Remote Sensing Imagery: A Review
Core Problem: Assess the current state, datasets, metrics, and weaknesses of deep-learning water body segmentation in remote-sensing imagery.
Key Innovation: Provides a structured synthesis of U-Net, DeepLab, Transformer, Mamba, and SAM-based water segmentation approaches and their failure modes.
71. Land-Cover and Land-Use Mapping Under Limited Data Highlights Hyperparameter Stability and Predictor Design
Core Problem: Build stable long-term land-cover and land-use maps when observations and reference data are limited.
Key Innovation: Shows that predictor design and hyperparameter-stability analysis matter more than single tuned settings for sparse-sample Landsat classification.
72. MFRA-YOLOv11: Remote Sensing Small Object Detection Algorithm Based on Multiscale Feature Extraction and Region Awareness
Core Problem: Increase small-object detection accuracy in remote-sensing images without losing practical efficiency.
Key Innovation: Adds multiscale feature extraction, triple-attention region awareness, and adaptive receptive fields to YOLOv11 for modest but consistent gains.
73. TFCRNet: Dual-Discriminator SAR-to-Optical Translation and Region-Gated Cross-Attention Fusion for Thick-Cloud Removal
Core Problem: Thick clouds remove optical surface information, while naive SAR-optical fusion introduces spectral and structural artifacts.
Key Innovation: Two-stage SAR-to-optical translation with dual discriminators and cloud-mask-guided cross-attention fusion.
74. Geometry-Guided Semi-Supervised Multimodal Segmentation for UAV-Based Rice-Lodging Mapping
Core Problem: Rice lodging is difficult to segment with few labels and ambiguous RGB appearance.
Key Innovation: Reliability-aware DSM prompting, pseudo-label calibration, and boundary regularization for semi-supervised UAV mapping.
75. MARC-Net: A Modality-Availability-Aware Robust Change Network for Missing-Optical Bi-Temporal Optical-SAR Change Detection of Reclaimed Cropland
Core Problem: Bi-temporal change detection degrades when one optical image is unavailable or poor quality.
Key Innovation: Availability-aware network with temporal proxy stabilization and condition-balanced optimization across missing-data cases.
76. FD-ProtoSCD: Semantic Change Detection in High-Resolution Remote Sensing Images via Frequency-Domain Disentanglement and Dynamic Prototype Learning
Core Problem: Appearance-induced pseudo-changes and imbalanced class transitions degrade semantic change detection.
Key Innovation: Frequency-domain disentanglement plus dynamic class-prototype learning for rare transitions.
77. Complex-Valued HRU-Net with Cross-Gated Attention for PolSAR Semantic Segmentation
Core Problem: Standard U-Net segmentation loses fine spatial detail and fuses encoder-decoder features too crudely.
Key Innovation: Complex-valued HRNet-U-Net with cross-gated attention for sharper and more selective PolSAR segmentation.
78. A scalar elastic damage model for heterogeneous rocks
Core Problem: Existing damage models struggle to represent heterogeneous rock under combined tensile and shear failure modes.
Key Innovation: Introduces a Weibull-based unified scalar damage variable driven by weighted tensile-shear equivalent strain and validates it against sandstone tests.
79. Projection of surface air temperature and precipitation in High Mountain Asia and its major upstream river basins through multi-model ensemble from CMIP6
Core Problem: Future temperature and precipitation trajectories in High Mountain Asia and major headwater basins remain uncertain.
Key Innovation: Uses a CMIP6 multi-model ensemble to resolve basin-specific warming and precipitation trends through 2100.
80. Effects of encapsulation length and grout stiffness on the load-bearing mechanics of rock bolts
Core Problem: Effects of encapsulation length and grout properties on bolt performance lack systematic experimental evidence.
Key Innovation: Large-scale full-factorial pull-out tests with internal strain measurements separate length and grout effects.
81. Lime Stabilisation of Calcareous and Marly Clays: A Critical Review
Core Problem: Marl stabilization practice is often extrapolated from non-calcareous clay frameworks that may not apply.
Key Innovation: Synthesizes carbonate-controlled lime demand, dual carbonation behavior, and sulphate-heave risks into a marl-specific framework.
82. Empirical correlation between shear wave velocity (VS) and SPT results using 1D MASW in fine-grained soils of Cochabamba, Bolivia
Core Problem: Direct shear-wave velocity data are often unavailable in urban site characterization.
Key Innovation: Derives a local MASW-SPT correlation for fine-grained soils in Cochabamba.
83. Performance-based machine learning-aided seismic design methodology for reinforced concrete moment frames
Core Problem: Peak drift limits alone do not capture seismic performance levels across diverse RC frames.
Key Innovation: Learns ASCE41-style performance levels from connection test data and projected drifts.
84. Experimental study on seismic performance of reinforced concrete frame with ceramsite aerated concrete block infill wall
Core Problem: Ceramsite aerated-concrete infill walls change frame stiffness, ductility, damage distribution, and energy dissipation in ways that require experimental constraint.
Key Innovation: Combines quasi-static tests and ABAQUS validation to isolate infill-wall effects and assess EPDM flexible connections for reducing earthquake damage.
85. A unified model for multi-task drone routing in post-disaster road assessment
Core Problem: Drone road-assessment missions after disasters involve multiple tasks that are hard to optimize jointly.
Key Innovation: Appears to unify multi-task drone routing for post-disaster assessment.
86. The role of riparian plant roots in river ecomorphodynamics: A review
Core Problem: Synthesizes the geomorphic role of riparian roots in river ecomorphodynamics.
Key Innovation: Appears to consolidate root-driven controls on fluvial form and process across settings.
87. Geostationary observations of air pollutants from biomass burning: A synergy of GIIRS and GEMS over Southeast Asia
Core Problem: Southeast Asian biomass-burning pollution needs better geostationary observation coverage.
Key Innovation: Appears to fuse GIIRS and GEMS observations for regional smoke and pollutant retrievals.
88. Macroscopic properties and microstructure of stabilized contaminated soil under the combined action of chloride salt erosion and freeze-thaw cycles
Core Problem: Combined chloride exposure and freeze-thaw cycling can degrade strength, hydraulic behavior, pore structure, and contaminant retention in stabilized soil.
Key Innovation: Links strength loss, conductivity change, contaminant release, and pore-crack evolution through coupled macro-scale tests, XRD, SEM, NMR, and chemical-fraction analysis.