TerraMosaic Daily Digest: September 1, 2026

September 1, 2026 TerraMosaic Daily Digest

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

Source: Science Advances Type: volcanic tsunami trigger analysis Geohazard Type: tsunami and submarine volcanic collapse Relevance: 9/10

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

Source: Science Advances Type: sequential-earthquake landslide study Geohazard Type: landslide and earthquake preconditioning Relevance: 9/10

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

Source: Earth Surface Processes and Landforms Type: regional postfire debris-flow hazard modeling Geohazard Type: postfire debris flow and landslide Relevance: 8/10

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

Source: Journal of Geophysical Research: Solid Earth Type: megathrust coupling and seismic hazard monitoring method Geohazard Type: subduction earthquake hazard Relevance: 8/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: landslide Relevance: 8/10

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

Source: Science Advances Type: fault friction and rupture-segmentation study Geohazard Type: earthquake rupture and fault creep Relevance: 8/10

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

Source: Bulletin of Earthquake Engineering Type: near-fault pulse ground-motion model Geohazard Type: earthquake Relevance: 8/10

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

Source: Bulletin of Earthquake Engineering Type: fault-rupture setback physical modeling Geohazard Type: fault rupture / permanent ground deformation Relevance: 8/10

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

Source: Computers and Geotechnics Type: Slope stability inversion Geohazard Type: landslide/slope failure Relevance: 8/10

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

Source: Soil Dynamics and Earthquake Engineering Type: Seismic liquefaction centrifuge experiment Geohazard Type: liquefaction Relevance: 8/10

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

Source: Soils and Foundations Type: Post-event geotechnical damage assessment Geohazard Type: earthquake and tsunami Relevance: 8/10

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

Source: Geophysical Research Letters Type: dust-storm mitigation and air-quality hazard study Geohazard Type: dust storm and aeolian hazard Relevance: 7/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: coastal erosion and shoreline change Relevance: 7/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: wildfire Relevance: 7/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: wildfire Relevance: 7/10

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

Source: Ocean Engineering Type: seismic sloshing engineering study Geohazard Type: earthquake-induced industrial fluid overflow Relevance: 7/10

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

Source: Natural Hazards Type: storm surge inventory and hazard analysis Geohazard Type: storm surge / coastal flooding Relevance: 7/10

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

Source: Natural Hazards Type: forest fire vulnerability mapping Geohazard Type: wildfire Relevance: 7/10

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

Source: Journal of Mountain Science Type: MICP aeolian sand stabilization Geohazard Type: aeolian sand hazard Relevance: 7/10

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

Source: Journal of Mountain Science Type: UAV sand-barrier performance monitoring Geohazard Type: aeolian sand hazard Relevance: 7/10

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

Source: Journal of Mountain Science Type: 3D slope reliability framework Geohazard Type: landslide / slope stability Relevance: 7/10

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

Source: Bulletin of Earthquake Engineering Type: masonry pagoda seismic damage index Geohazard Type: earthquake Relevance: 7/10

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

Source: Bulletin of Earthquake Engineering Type: timber pavilion shaking-table study Geohazard Type: earthquake Relevance: 7/10

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

Source: International Journal of Disaster Risk Reduction Type: rail-hub flood disruption assessment Geohazard Type: urban flooding Relevance: 7/10

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

Source: Journal of Hydrology Type: Flood susceptibility modeling Geohazard Type: flood Relevance: 7/10

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

Source: Soil Dynamics and Earthquake Engineering Type: Seismic embankment fragility analysis Geohazard Type: earthquake-induced embankment failure Relevance: 7/10

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

Source: Soil Dynamics and Earthquake Engineering Type: Liquefaction constitutive-model calibration Geohazard Type: liquefaction Relevance: 7/10

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

Source: Earth Surface Processes and Landforms Type: estuarine erosion-control analysis with interpretable ML Geohazard Type: coastal and estuarine erosion Relevance: 6/10

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

Source: Geophysical Research Letters Type: tropical cyclone intensity dynamics study Geohazard Type: tropical cyclone hazard Relevance: 6/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: extreme precipitation / flood and landslide triggering Relevance: 6/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: wildfire Relevance: 6/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: landslide and disaster change mapping Relevance: 6/10

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

Source: Remote Sensing (MDPI) Type: InSAR deformation change-point detection Geohazard Type: ground deformation / landslide monitoring Relevance: 6/10

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

Source: Engineering Geology Type: mudstone weathering crack-evolution study Geohazard Type: landslide / slope weathering Relevance: 6/10

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

Source: International Journal of Disaster Risk Reduction Type: post-landslide/flood mitigation planning study Geohazard Type: landslide and flood Relevance: 6/10

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

Source: Earth Surface Processes and Landforms Type: semiarid river morphodynamics and bank erosion study Geohazard Type: bank erosion and channel instability Relevance: 5/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: multi-hazard aerial mapping / flood-damage transfer Relevance: 5/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: extreme rainfall and flood Relevance: 5/10

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

Source: Ocean Engineering Type: probabilistic wave forecasting Geohazard Type: coastal wave hazard transfer Relevance: 5/10

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

Source: Ocean Engineering Type: coastal wave attenuation experiment Geohazard Type: reef-mediated wave hazard transfer Relevance: 5/10

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

Source: Geoscientific Model Development Type: precipitation-model improvement Geohazard Type: rainfall-triggered landslide and flood transfer Relevance: 5/10

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

Source: Remote Sensing (MDPI) Type: SAR-coherence surface-water mapping method Geohazard Type: flood and surface-water mapping Relevance: 5/10

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

Source: Journal of Mountain Science Type: rock glacier inventory reproducibility study Geohazard Type: permafrost / rock glacier hazard Relevance: 5/10

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

Source: ISPRS Journal of Photogrammetry and Remote Sensing Type: fog-robust satellite-UAV change detection benchmark Geohazard Type: cross-cutting disaster mapping Relevance: 5/10

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

Source: Journal of Geophysical Research: Solid Earth Type: active-margin crustal imaging study Geohazard Type: earthquake and tectonic deformation Relevance: 4/10

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

Source: Water Resources Research Type: rainforest moisture-recycling hydroclimate study Geohazard Type: drought and ecohydrologic stress Relevance: 4/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: remote-sensing image registration and change analysis Relevance: 4/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: methodological support for geophysical emulation Relevance: 4/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: geospatial remote sensing Relevance: 4/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: seismic / geophysical wavefield analysis Relevance: 4/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: None Relevance: 4/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: geospatial remote sensing / multi-hazard transfer Relevance: 4/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: subsurface geophysical characterization Relevance: 4/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: terrain reconstruction for landslides and disaster mapping Relevance: 4/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: multi-sensor disaster imagery and cloud-obscured hazard mapping Relevance: 4/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: None Relevance: 4/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: surface-material mapping Relevance: 4/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: None Relevance: 4/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: physical hazard modeling transfer Relevance: 4/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: multi-hazard remote sensing Relevance: 4/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: cross-cutting terrain and site reconstruction Relevance: 4/10

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

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: cross-cutting geomechanics and slope-stability modeling Relevance: 4/10

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

Source: Ocean Engineering Type: multi-AUV sonar survey planning Geohazard Type: None Relevance: 4/10

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

Source: Ocean Engineering Type: coastal hydrodynamics method Geohazard Type: wave interaction with porous coastal structures Relevance: 4/10

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

Source: Hydrology and Earth System Sciences Type: overland-flow hydrological modelling Geohazard Type: runoff and erosion process transfer Relevance: 4/10

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

Source: Remote Sensing (MDPI) Type: terrain-corrected LAI retrieval method Geohazard Type: mountain landslide and vegetation monitoring Relevance: 4/10

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

Source: Remote Sensing (MDPI) Type: mangrove compositing evaluation study Geohazard Type: coastal hazard-buffer monitoring Relevance: 4/10

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

Source: Remote Sensing (MDPI) Type: human-in-the-loop remote-sensing segmentation Geohazard Type: multihazard remote-sensing segmentation Relevance: 4/10

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

Source: Remote Sensing (MDPI) Type: urban building-footprint extraction method Geohazard Type: urban exposure mapping Relevance: 4/10

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

Source: Remote Sensing (MDPI) Type: water-segmentation review Geohazard Type: flood and surface-water mapping Relevance: 4/10

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

Source: Remote Sensing (MDPI) Type: land-cover mapping robustness study Geohazard Type: general geohazard susceptibility mapping Relevance: 4/10

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

Source: Remote Sensing (MDPI) Type: remote-sensing small-object detector Geohazard Type: aerial damage and target detection Relevance: 4/10

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

Source: Remote Sensing (MDPI) Type: SAR-optical cloud removal network Geohazard Type: cross-cutting remote sensing Relevance: 4/10

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

Source: Remote Sensing (MDPI) Type: UAV crop-lodging segmentation Geohazard Type: disaster impact mapping Relevance: 4/10

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

Source: Remote Sensing (MDPI) Type: missing-modality optical-SAR change detection Geohazard Type: cross-cutting change detection Relevance: 4/10

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

Source: Remote Sensing (MDPI) Type: semantic change detection method Geohazard Type: cross-cutting remote sensing Relevance: 4/10

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

Source: Remote Sensing (MDPI) Type: PolSAR land-cover segmentation Geohazard Type: cross-cutting remote sensing Relevance: 4/10

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

Source: Journal of Mountain Science Type: heterogeneous-rock damage constitutive modeling Geohazard Type: rock failure mechanics Relevance: 4/10

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

Source: Journal of Mountain Science Type: High Mountain Asia climate projection Geohazard Type: cross-cutting climate forcing Relevance: 4/10

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

Source: Journal of Mountain Science Type: rock bolt pull-out mechanics Geohazard Type: slope stabilization engineering Relevance: 4/10

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

Source: Geotechnical and Geological Engineering Type: review of lime stabilization in marly clays Geohazard Type: expansive clay / shrink-swell Relevance: 4/10

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

Source: Bulletin of Earthquake Engineering Type: Vs-SPT seismic site correlation Geohazard Type: earthquake site effects Relevance: 4/10

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

Source: Bulletin of Earthquake Engineering Type: ML-assisted seismic design classifier Geohazard Type: earthquake engineering Relevance: 4/10

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

Source: Bulletin of Earthquake Engineering Type: reinforced-concrete frame seismic testing Geohazard Type: earthquake Relevance: 4/10

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

Source: Reliability Engineering & System Safety Type: post-disaster drone routing model Geohazard Type: disaster response Relevance: 4/10

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

Source: Earth-Science Reviews Type: riparian root ecomorphodynamics review Geohazard Type: fluvial erosion / bank stability Relevance: 4/10

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

Source: Remote Sensing of Environment Type: biomass-burning pollutant observation study Geohazard Type: wildfire / smoke Relevance: 4/10

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

Source: Soils and Foundations Type: combined salt and freeze-thaw soil degradation study Geohazard Type: freeze-thaw soil degradation Relevance: 4/10

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.