Initiated by Dr. Xin Wei, University of Michigan
Ongoing development by the community

TerraMosaic Daily Digest: August 11, 2026

August 11, 2026 TerraMosaic Daily Digest

Daily Summary

Flood and hydrometeorological papers frame August 11 around sequence dependence, urban process realism, and impact accounting. The reconstruction of the 1342-1343 European megafloods argues that flood risk management remains poorly configured for continent-scale cascades, while Chennai simulations show that explicit aerosol microphysics can materially sharpen extreme-rainfall and flood-extent prediction. Complementary studies translate temperature-rainfall scaling into street-scale flood depth, timing, and velocity, learn real-time meteorological bias corrections for runoff across 451 European basins, compile a 1716-record Chinese flood-impact archive, and extend flood analysis into residential fragility, metro susceptibility, levee-breach closure capacity, opportunistic urban weather sensing, and flash-flood-constrained planning.

Slope, sediment, wildfire, and seismic contributions resolve instability through material state and structural context rather than coarse hazard labels. A micro-continuum model reproduces transitions among turbidity currents, mudflows, and mudslides from independently measured rheology; PSI, geotechnics, and finite-element reconstruction show that the Mawlai landslide evolved through progressive rainfall-driven hydro-mechanical weakening before failure; and fracture-resolved laser-scan, DFN, and DEM analysis indicates that a wedge rock mass stable under ambient conditions can become unstable under seismic loading. Related papers improve debris-flow barrier-dam assessment under data scarcity, examine inter-slice forces in probabilistic slope stability, track infiltration-driven failure on photovoltaic slopes, and show that frozen or near-fault infrastructure response depends strongly on coupled stiffness, seepage, and loading state.

Wildfire and cryosphere studies link event dynamics to longer-term system change, while a large methodological cohort strengthens observation and forecasting without claiming prior geohazard validation. Faster wildfire spread is associated with higher burn severity and larger regeneration deficits, supported by a global VIIRS individual-fire inventory and a lightweight multi-platform fire detector. Glacial-lake growth and clustered overflow channels raise corridor-scale GLOF exposure in southeastern Tibet, annual active-layer maps and in situ 4D CT imaging sharpen thaw-related ground assessment, and Arctic sea-ice retreat plus current Petermann Glacier rifting underscore persistent cryosphere instability. Transferable papers then emphasize regime-aware verification of AI weather extremes, simple but competitive cyclone-intensity correction, stable climate emulation, uncertainty-calibrated spatiotemporal forecasting, structurally guided remote-sensing segmentation, geometry-consistent 3D vision adaptation, physics-aware inverse mapping, and domain-tuned subsurface data digitization.

Key Trends

August 11's selections converge on hazard inference that is more sequence-aware, materially constrained, and uncertainty-explicit, while the transferable cohort focuses on structural guidance and physical consistency rather than benchmark gains alone.

  • Flood studies are expanding from event mapping to cascade-, climate-, and impact-aware risk systems: The flood papers move across temporal scales from medieval continental cascades to street-scale inundation and emergency closure engineering, but they converge on the need to model sequencing, forcing realism, and exposed assets together. Aerosol-aware rainfall simulation, temperature-conditioned future storm generation, real-time bias correction, impact databases, building fragility, metro susceptibility, and planning overlays all push flood analysis beyond static hazard footprints.
  • Slope and sediment instability analysis is becoming materially explicit and multi-source: Several direct geohazard papers couple deformation monitoring, geotechnical testing, rheology, and numerical mechanics rather than relying on terrain proxies alone. The selected studies repeatedly show that yield stress, fracture geometry, rainfall infiltration, freeze-thaw evolution, and spatially variable soil strength materially alter runout, failure timing, and inferred safety margins.
  • Wildfire and cryosphere papers tie event kinematics to residual system vulnerability and corridor exposure: The wildfire papers do not stop at detection or burned area; they connect rapid spread to higher severity and weaker postfire recovery potential, while also improving event-scale inventories and edge deployment. Cryosphere papers likewise link glacial-lake growth, active-layer change, sea-ice breakup, and glacier rifting to transport, infrastructure, or longer-term stability consequences.
  • Hazard forecasting methods are being judged by tail skill, real-time bias behavior, and long-horizon stability: The forecasting cohort shifts attention from average skill alone to extremes, calibration, and operational persistence. AI weather models are not found to fail uniformly in the tails, tropical-cyclone intensity improves through targeted correction, reinforcement-style model selection exploits complementary forecasters, and stochastic wrappers stabilize frozen neural weather operators for climate-scale integration while uncertainty-aware interval methods refine reliability.
  • Transferable geospatial AI is prioritizing structural guidance, physical operators, and interpretable uncertainty: The non-hazard-specific papers emphasize how information is represented and trusted rather than claiming immediate geohazard validation. Structural guidance for open-vocabulary segmentation, topographic-shadow normalization, lightweight orientation-aware extraction, pseudo-label filtering for geo-localization, geometry-consistent test-time adaptation, sparse-track inverse deconvolution, spectral gravity inversion, borehole-log digitization, and entropy- or baseline-referenced explanations collectively define a more physically disciplined methodological substrate.

Selected Papers

The selected papers divide into direct hazard studies and transferable sensing, mechanics, and AI methods. Direct contributions address floods, slope failures, wildfire, seismic and underground infrastructure, and cryosphere change, whereas the transferable cohort contributes observation, inversion, forecasting, and uncertainty tools whose geohazard relevance remains prospective in this set.

1. Cascading continental-scale floods across Europe in 1342-1343

Source: Nature Type: continental-scale flood history study Geohazard Type: flood Relevance: 9/10

Core Problem: Modern flood management is not prepared for cascading sequences of extreme floods spanning large regions.

Key Innovation: Reconstructs the 1342-1343 European megaflood cascade as an analog for compound continental flood risk.

2. A micro-continuum physics-based model for cohesive sediment gravity flows across mudslide, mudflow, and turbidity current regimes

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: mudslide / mudflow / turbidity current Relevance: 8/10

Core Problem: Unifying prediction of cohesive sediment gravity flows across low-density currents, high-density currents, mudflows, and mudslides.

Key Innovation: Micro-continuum CFD model driven by independently measured yield-stress relations reproduces flow regime, morphology, speed, and runout across cohesive sediment types.

3. AIFS-TC: A simple correction competitive with the operational frontier for tropical cyclone intensity forecasting

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

Core Problem: AI weather models track tropical cyclones well but strongly underestimate intensity.

Key Innovation: Simple post-processing correction to AIFS-Single that reaches operationally competitive intensity forecasts, including rapid intensification.

4. Do AI weather models miss extremes?

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

Core Problem: Claims that AI weather models systematically miss extremes have not been rigorously audited across observations.

Key Innovation: Station-based, regime-aware comparison of 11 AI and physical forecast systems focused on tail skill.

5. Can aerosols improve urban flood forecasting? A case study of 2015 Chennai extreme event

Source: NHESS Type: urban flood forecasting study Geohazard Type: flood Relevance: 8/10

Core Problem: Urban-scale rainfall simulations remain inaccurate for extreme flood forecasting.

Key Innovation: Explicit aerosol microphysics improved Chennai flood extent mapping by up to 50 percent during the 2015 extreme event.

6. CFDID: a China Flood Disaster Impact Database from 1994 to 2023

Source: ESSD Type: flood disaster impact database Geohazard Type: flood Relevance: 8/10

Core Problem: China lacks systematic long-term records of flood disaster impacts.

Key Innovation: An LLM-assisted and manually verified database of 1716 flood records compiled from 66 official documents spanning 1994-2023.

7. A process-informed framework linking temperature-rainfall projections and urban flood modeling

Source: HESS Type: climate-to-urban-flood modeling framework Geohazard Type: flood Relevance: 8/10

Core Problem: Translate temperature-rainfall projections into realistic future urban flood scenarios.

Key Innovation: A process-informed framework combining rainfall scaling, storm-frequency estimation, and urban flood modeling in a transferable workflow.

8. Faster, bigger, more severe: Extreme wildfire spread sets the stage for forest ecosystem change in western and boreal North America

Source: Science Advances Type: wildfire hazard science Geohazard Type: wildfire Relevance: 8/10

Core Problem: Test whether rapidly spreading fires also create distinct postfire ecological risk.

Key Innovation: Linked daily spread rate to higher burn severity, larger seed-source gaps, and reduced forest recovery potential.

9. GeoHazards, Vol. 7, Pages 97: Evolution of Glacial Lakes and GLOF Hazards to Transportation Routes in the Southeastern Tibetan Engineering Corridor

Source: GeoHazards (MDPI) Type: GLOF hazard assessment Geohazard Type: glacial lake outburst flood Relevance: 8/10

Core Problem: Track glacial-lake growth and identify GLOF threats to Tibetan transport corridors.

Key Innovation: Mapped 1990-2020 lake expansion, ranked susceptibility, and highlighted clustered overflow channels.

10. Understanding slope instability near Mawlai bypass, Meghalaya: a FEM-PSI based landslide investigation

Source: Natural Hazards Type: landslide investigation Geohazard Type: landslide Relevance: 8/10

Core Problem: Reconstruct the mechanism of a 2024 rainfall-induced landslide in Meghalaya.

Key Innovation: Integrated PSI, field geotechnics, and FEM to show progressive hydro-mechanical weakening before failure.

11. 3D-DEM assessment of seismic instability in a naturally fractured wedge rock mass based on terrestrial laser scanning and DFN modeling

Source: Bull. Eng. Geol. & Env. Type: seismic rockslide modeling Geohazard Type: earthquake-triggered rockslide Relevance: 8/10

Core Problem: The study addresses how to model the dynamic failure risk of a naturally fractured wedge rock mass when topography alone is insufficient and fracture networks control instability.

Key Innovation: It integrates terrestrial laser scanning, quantitative discontinuity extraction, DFN generation, and 3D DEM simulation to show stable natural conditions but predicted seismic-triggered landslide failure.

12. Ensemble and symbolic regression-based stability assessment of debris-flow barrier dams considering data scarcity

Source: Geoenvironmental Disasters Type: debris-flow dam stability modeling Geohazard Type: debris-flow barrier dam breach Relevance: 8/10

Core Problem: The study tackles how to reliably classify debris-flow barrier dam stability when observations are sparse and existing models adapted from landslide dams miss key controlling behavior.

Key Innovation: It combines physical-model ensembles with symbolic regression, uses inter-model inconsistency to build interpretable features, and derives dominant instability thresholds for dam and lake volume.

13. Investigation of limit equilibrium methods in probabilistic slope stability analyses with spatially variable soils incorporating various inter-slice forces

Source: RESS Type: probabilistic slope stability methodology Geohazard Type: slope failure susceptibility Relevance: 8/10

Core Problem: The study asks how different inter-slice force assumptions within limit equilibrium methods affect convergence, factor-of-safety statistics, and failure identification in random-field slope analysis.

Key Innovation: It systematically compares LEM variants under spatial variability, shows where convergence breaks down, and highlights a finite-element-stress-based approach that avoids factor-of-safety non-convergence.

14. Impact of Wind Forcing Resolution on Modeled Wave Extremes at the Coastal Boundary

Source: GRL Type: coastal wave hazard modeling study Geohazard Type: coastal extreme waves / cyclone impacts Relevance: 7/10

Core Problem: Assessing how wind-forcing resolution alters modeled maximum wave heights under tropical and extratropical cyclones at the coast.

Key Innovation: Event-based SWAN experiments across BARRA and ERA5 resolutions show high-resolution winds are essential near cyclone cores and complex coasts.

15. GeoSeg-OV: Bridging Geospatial Gaps with Structural Guidance for Open-Vocabulary Remote Sensing Segmentation

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: remote sensing hazard mapping Relevance: 7/10

Core Problem: Geospatial domain shifts weaken open-vocabulary segmentation across regions, resolutions, and sensors.

Key Innovation: Uses auxiliary foundation-model features as structural guidance for cost aggregation instead of direct text matching.

16. An adaptive and evolvable deep reinforcement learning framework for weather prediction

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: multi-hazard weather forecasting Relevance: 7/10

Core Problem: No single AI weather model is consistently best across variables, levels, and lead times.

Key Innovation: Deep-RL framework that learns variable-specific ensemble weights and evolves the model pool over time.

17. Projected climate memory and inherited warm-tail risk in accelerated European summer warming

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: heat hazard and climate extremes Relevance: 7/10

Core Problem: The role of inherited slow-state climate memory in European summer warm-tail risk is uncertain.

Key Innovation: Reduced-dynamics decomposition with causal filters and logistic warm-tail risk models across European regions.

18. Rescene: band-limited stochastic forcing turns a frozen neural weather operator into a climate emulator

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: multi-hazard climate forecasting Relevance: 7/10

Core Problem: Frozen neural weather operators become unstable or unrealistic when freely integrated beyond training horizons.

Key Innovation: Climatology-blending wrapper with band-limited stochastic forcing that stabilizes decades-long climate emulation.

19. A Global Dataset of Individual Fire Events (2012-2025) derived from VIIRS Active Fire Product

Source: ESSD Type: global wildfire event dataset Geohazard Type: wildfire Relevance: 7/10

Core Problem: Existing satellite fire inventories break large fires into discontinuous fragments.

Key Innovation: A sliding-window method that derives global individual fire events from VIIRS active-fire detections for 2012-2025.

20. BiasCast: learning and adjusting real time biases from meteorological forecasts to enhance runoff predictions

Source: HESS Type: forecast-bias correction for runoff prediction Geohazard Type: flood Relevance: 7/10

Core Problem: Biases in meteorological forecasts degrade operational runoff predictions when models switch from observations to forecasts.

Key Innovation: AI methods that learn and correct real-time forecast biases across 451 European river basins.

21. Near-fault seismic hazard effects on embankment dam response: sensitivity analysis of the Geyve Doğantepe Dam

Source: Geomatics, Nat. Haz. & Risk Type: seismic dam safety Geohazard Type: earthquake Relevance: 7/10

Core Problem: Quantify how near-fault shaking and material parameters control embankment dam response.

Key Innovation: Combined nonlinear time-history and sensitivity analysis to show short-period demand and stiffness dominance.

22. Integrating flash-flood susceptibility into geopark suitability planning: a spatial decision framework for the Himā UNESCO world heritage area, Saudi Arabia

Source: Frontiers in Earth Science Type: flash-flood susceptibility planning Geohazard Type: flash flood Relevance: 7/10

Core Problem: Combine flash-flood hazard with heritage value and access in geopark planning.

Key Innovation: Built a fuzzy AHP plus GIS decision framework that treats flood susceptibility as a planning constraint.

23. Remote Sensing, Vol. 18, Pages 2713: Physics-Aware Deep Learning Reconstructs Ground Contamination from Sparse UAV Radiation Measurements over the Fukushima Ukedo Basin Without Field Training

Source: Remote Sensing (MDPI) Type: radiological contamination mapping Geohazard Type: radiological contamination Relevance: 7/10

Core Problem: Recover ground contamination maps from sparse UAV radiation tracks without field-trained labels.

Key Innovation: Cast reconstruction as inverse deconvolution and used physics-aware pretraining with forward-consistency loss.

24. Remote Sensing, Vol. 18, Pages 2706: EdgeNeXt-Attn: A Lightweight Attention-Enhanced Deep Learning Framework for Fire Detection in Remote Sensing Imagery

Source: Remote Sensing (MDPI) Type: wildfire detection AI Geohazard Type: wildfire Relevance: 7/10

Core Problem: Detect wildfire rapidly and accurately across satellite, UAV, and CCTV imagery.

Key Innovation: Enhanced EdgeNeXt with channel and spatial attention for small and ambiguous fire regions on edge hardware.

25. Rainfall infiltration-runoff boundary condition and progressive failure modes of unsaturated photovoltaic slopes under panel interception effects

Source: Engineering Geology Type: photovoltaic slope failure analysis Geohazard Type: rainfall-induced slope failure Relevance: 7/10

Core Problem: Based on the title, the study investigates how photovoltaic panels modify infiltration-runoff boundary conditions and thus change progressive failure modes in unsaturated slopes.

Key Innovation: From the supplied title, the apparent innovation is explicit treatment of panel-interception effects when analyzing hydrologic forcing and staged failure behavior of photovoltaic slopes.

26. Investigating the seismic behavior and deterioration evolution of slopes in seasonally frozen areas

Source: Engineering Geology Type: frozen-region slope seismic study Geohazard Type: seismic slope instability in frozen ground Relevance: 7/10

Core Problem: The paper appears to address how seasonal freezing conditions influence slope seismic behavior and progressive deterioration in cold-region settings.

Key Innovation: From the title alone, the likely contribution is coupling frozen-ground degradation with seismic slope response to clarify instability evolution in seasonally frozen areas.

27. Flood susceptibility assessment of urban metros using multi-metric Bayesian optimization integrating tree-based machine learning: A case study in Beijing

Source: TUST Type: urban metro flood susceptibility Geohazard Type: urban flooding Relevance: 7/10

Core Problem: The paper appears to ask how to assess flood susceptibility of urban metro systems using tree-based machine learning and Bayesian optimization.

Key Innovation: From the title, the likely contribution is a multi-metric optimization framework that tunes interpretable ML models for metro flood susceptibility mapping.

28. Ground-to-Cable Strain Transfer in Unburied DAS on Earth and the Moon

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: seismic or slope-deformation monitoring Relevance: 6/10

Core Problem: Explaining why unburied distributed acoustic sensing cables often show poor ground-to-cable strain transfer and how deployment controls coupling.

Key Innovation: First analytical and numerical coupling model identifies sag-to-radius parameter Theta and the quarter-radius sag threshold governing strain-transfer efficiency.

29. Evaluating Semantic and Spatial Guidance for Foundation Model Segmentation of Small-Scale PV in Remote Sensing Imagery

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: hazard mapping from remote sensing Relevance: 6/10

Core Problem: Small sparse PV targets are hard to segment in remote-sensing imagery.

Key Innovation: Systematic comparison of textual, spatial, and hybrid prompting for SAM-style RS segmentation.

30. A Spectral-Domain Pseudo-Inverse Method for True 3D Gravity Inversion

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

Core Problem: Surface gravity observations are underdetermined for true 3D inversion and weak in depth resolution.

Key Innovation: Upward-continued 3D data volume plus stable spectral-domain pseudo-inverse filtering.

31. Derivative Computation in PINNs: Automatic Differentiation, Finite Differences and Beyond

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: hazard process modeling Relevance: 6/10

Core Problem: Automatic differentiation in PINNs can be costly and even wrong for architectures with inter-sample dependencies.

Key Innovation: Calibrated finite-difference and stochastic finite-difference derivatives as faster, lower-memory alternatives.

32. Entropy-Centric Explainable AI for Remote Sensing Image Segmentation

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: general remote sensing Relevance: 6/10

Core Problem: Semantic-segmentation XAI remains weak and poorly evaluated in remote sensing.

Key Innovation: Entropy-centric segmentation explanations plus a dedicated relevance-evaluation methodology.

33. Every Packet Counts: Dispersing Information for Loss-Resilient Learned Image Compression

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: satellite and emergency communications Relevance: 6/10

Core Problem: Learned image compression degrades badly under packet loss because information is unevenly distributed.

Key Innovation: Inter-channel redistribution and grouping plus shorter autoregressive dependencies for robust decoding.

34. SAR2Agri: Learning SAR Intensity Representations for Agricultural Monitoring

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: SAR remote sensing Relevance: 6/10

Core Problem: SAR-specific self-supervised pretraining for temporal agricultural monitoring is underdeveloped.

Key Innovation: SAR-only temporal pretext pipeline with masking and curriculum learning for phenology-aware features.

35. TSCoNet: A Two-Stage Copula CNN-LSTM for Uncertainty-Aware Spatio-Temporal Forecasting

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: hydrometeorological hazard drivers Relevance: 6/10

Core Problem: Deep spatio-temporal forecasters rarely provide reliable uncertainty without sacrificing accuracy.

Key Innovation: A two-stage CNN-LSTM plus Gaussian copula adds multivariate predictive intervals while preserving point-forecast skill.

36. OpenMesh: wireless signal dataset for opportunistic urban weather sensing in New York City

Source: ESSD Type: urban opportunistic weather-sensing dataset Geohazard Type: flood Relevance: 6/10

Core Problem: Cities lack affordable high-frequency precipitation observations for real-time monitoring.

Key Innovation: A wireless-link signal dataset paired with weather stations showing community networks can sense rain and snow at one-minute resolution.

37. A 1 km resolution dataset of Northern Hemisphere permafrost active layer thickness (2000-2024)

Source: ESSD Type: permafrost active-layer dataset Geohazard Type: permafrost Relevance: 6/10

Core Problem: Large-scale yearly active-layer thickness has been difficult to map continuously across the Northern Hemisphere.

Key Innovation: Combines long-term observations with AI to produce 1 km annual active-layer thickness maps from 2000 to 2024.

38. Remote Sensing, Vol. 18, Pages 2719: TSEC+TC: A Partitioned TSEC-Assisted Topographic Normalization Framework for Rugged Mountainous Terrain

Source: Remote Sensing (MDPI) Type: mountain remote-sensing correction Geohazard Type: mountain hazards Relevance: 6/10

Core Problem: Correct topographic and shadow distortions in rugged mountain imagery.

Key Innovation: Partitioned shadowed and sunlit pixels so TSEC and conventional correction target the same normalized reflectance.

39. Locally tuned Large Language Model (LLM) to empower digitalization of borehole logs for 3D stratigraphic modelling

Source: TUST Type: LLM borehole log digitization Geohazard Type: subsurface characterization for hazard analysis Relevance: 6/10

Core Problem: The paper tackles the manual, subjective bottleneck of converting free-text borehole descriptions into standardized stratigraphic classes for 3D subsurface modeling.

Key Innovation: It develops a locally deployed, domain-tuned LLM called StratumGPT that digitizes borehole logs with near-expert accuracy while preserving data confidentiality.

40. Interpretable prediction of tunnel seismic damage using multi-parameter ground-motion descriptors

Source: TUST Type: tunnel seismic damage prediction Geohazard Type: earthquake damage to underground structures Relevance: 6/10

Core Problem: The paper appears to address how multiple ground-motion descriptors can be used to interpretably predict tunnel seismic damage.

Key Innovation: From the title, the main innovation is an interpretable prediction framework that moves beyond single seismic intensity measures for tunnel damage estimation.

41. Is every flood rough on built environment? Flood fragility and vulnerability modeling for residential buildings

Source: Journal of Hydrology Type: flood building fragility modeling Geohazard Type: flood impact vulnerability Relevance: 6/10

Core Problem: The paper appears to examine how flood intensity translates into variable damage and vulnerability across residential building stocks rather than assuming uniform flood effects.

Key Innovation: From the title, the likely contribution is a fragility-based modeling framework that differentiates building vulnerability under flooding conditions.

42. Assessment of structural sediment connectivity using different weighting factors and field data in a mountain-piedmont catchment of the Pampean Ranges of Argentina

Source: Earth Surf. Proc. & Landforms Type: sediment connectivity method study Geohazard Type: sediment connectivity / debris-flow source-transfer Relevance: 5/10

Core Problem: Determining which index-of-connectivity weighting formulation best represents structural sediment connectivity in a heterogeneous mountain-piedmont catchment.

Key Innovation: Field-validated comparison of Manning-based weighting schemes showing ICz best captures spatial connectivity variability with medium-resolution DEMs.

43. Geospatial Foundation Embeddings as Transferable Catchment Descriptors

Source: GRL Type: geospatial foundation-model hydrology study Geohazard Type: catchment characterization for flood or debris-flow analysis Relevance: 5/10

Core Problem: Replacing hand-crafted static basin attributes with transferable geospatial embeddings for rainfall-runoff prediction across gauged and ungauged catchments.

Key Innovation: Shows foundation-model embeddings can match traditional descriptors alone and improve PUB performance when fused with standard catchment attributes.

44. Anomalous Sea Ice Retreat in the Refreezing Western Arctic by a 30-Year Record Cyclone in October: Oceanic, Atmospheric, and Wave Influences

Source: GRL Type: Arctic cyclone sea-ice event study Geohazard Type: cryosphere storm impacts Relevance: 5/10

Core Problem: Explaining anomalous October 2022 sea-ice retreat during an extreme western Arctic cyclone.

Key Innovation: Combines surface-energy and ocean-heat-content estimates to show oceanic forcing exceeded atmospheric forcing, with mechanical breakup likely closing the gap to observations.

45. Retrieval-Corrected Conformal Prediction for Time Series

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: forecast uncertainty and early warning Relevance: 5/10

Core Problem: Standard conformal prediction calibrates inefficiently when time-series errors are dependent and regime-varying.

Key Innovation: Retrieves analogous residuals for local evidence, then applies scalar conformal correction to restore coverage.

46. Cross-View Feature Matching: Survey, Benchmarking, and Foundation-Model Perspectives

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: photogrammetry and change detection Relevance: 5/10

Core Problem: Cross-view feature matching methods lack a unified taxonomy and fair benchmarking framework.

Key Innovation: Integrated survey, benchmarking protocol, and foundation-model perspective on cross-view correspondence.

47. Deep Learning-Based Statistical Downscaling of Sea Surface Temperature Using a Residual Corrective Neural Network

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: coastal and ocean hazards Relevance: 5/10

Core Problem: Coarse seasonal climate forecasts miss fine-scale coastal SST structure and extremes.

Key Innovation: Two-stage U-Net plus residual corrective refinement with an extreme-aware loss for SST downscaling.

48. The Kuramoto Neural Operator: Learning to Solve PDEs via Coupled Oscillator Dynamics

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: general geohazard process modeling Relevance: 5/10

Core Problem: Fixed-basis neural operators struggle with locally interacting PDE dynamics.

Key Innovation: Operator learning through a latent field of coupled oscillators.

49. Stochastic Emulation of a Fully Coupled Preindustrial E3SMv3 Simulation

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: climate-driven hydro and coastal hazards Relevance: 5/10

Core Problem: Coupled climate emulators lose internal variability and rare-event realism.

Key Innovation: Stochastic atmosphere-ocean emulator preserving ENSO, eddy SST, sea ice, and heavy-precipitation behavior.

50. Efficient Weak-Entropy PINN for Solving Hyperbolic Conservation Laws

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: debris flow, flood, and avalanche process modeling Relevance: 5/10

Core Problem: Standard PINNs struggle with shocks and discontinuities in hyperbolic conservation laws.

Key Innovation: Weak-form entropy-constrained PINN with FFT-based efficient integration.

51. Rapid assessment method for lateral bearing capacity of row piles in levee-breach closure based on an improved p-y curve

Source: Ocean Engineering Type: levee-breach pile capacity method Geohazard Type: flood Relevance: 5/10

Core Problem: Rapidly estimate lateral bearing capacity of row piles used in levee-breach closure works.

Key Innovation: An improved p-y curve based assessment method tailored to row piles in breach-closure conditions.

52. Petermann Glacier on the brink: Progress, challenges and insights

Source: Science Advances Type: glacier stability review Geohazard Type: glacier instability Relevance: 5/10

Core Problem: Synthesize what controls Petermann Glacier stability and future retreat risk.

Key Innovation: Integrated surface, ocean, and bed evidence to frame floating-tongue vulnerability and buttressing loss.

53. Remote Sensing, Vol. 18, Pages 2715: CGWT-DETR: Context-Guided Wavelet Transform DETR for Small Object Detection in Aerial RGB and Thermal Infrared Imagery

Source: Remote Sensing (MDPI) Type: small-object aerial detection Geohazard Type: multi-hazard transfer Relevance: 5/10

Core Problem: Improve small-object detection in aerial RGB and thermal imagery under real-time constraints.

Key Innovation: Added wavelet fusion and context-guided downsampling to RT-DETR for better detail preservation at lower compute.

54. Remote Sensing, Vol. 18, Pages 2716: LOA-Net: Lightweight Orientation-Aware Network for Road Extraction from Remote Sensing Imagery

Source: Remote Sensing (MDPI) Type: road extraction network Geohazard Type: multi-hazard exposure mapping Relevance: 5/10

Core Problem: Extract road networks accurately with a very lightweight model.

Key Innovation: Introduced road-aligned deformable convolution with explicit orientation supervision.

55. Remote Sensing, Vol. 18, Pages 2712: Characterizing the Mismatch Between ECOSTRESS-Derived Land Surface Temperature and ENVI-Met-Simulated UTCI Across Local Climate Zones

Source: Remote Sensing (MDPI) Type: urban heat proxy evaluation Geohazard Type: urban heat Relevance: 5/10

Core Problem: Test whether satellite land-surface temperature is a reliable proxy for pedestrian heat exposure.

Key Innovation: Used scene-adjusted comparison against ENVI-met UTCI to show weak and unstable correspondence.

56. Remote Sensing, Vol. 18, Pages 2711: Surface Subsidence Monitoring and Interpretable Factor Analysis in Coal Mining Areas of Henan Province Based on SBAS-InSAR

Source: Remote Sensing (MDPI) Type: mining subsidence InSAR Geohazard Type: land subsidence Relevance: 5/10

Core Problem: Monitor mining-induced subsidence and explain its spatial heterogeneity.

Key Innovation: Combined SBAS-InSAR with XGBoost-SHAP and wavelet coherence to attribute deformation patterns.

57. In-situ 4D CT tracking of microstructural evolution in granular materials subjected to freeze-thaw-seepage-mechanical coupling

Source: Acta Geotechnica Type: freeze-thaw soil microstructure imaging Geohazard Type: permafrost ground weakening Relevance: 5/10

Core Problem: The study asks how freeze-thaw-seepage-mechanical cycling changes particle, pore, and seepage structure in granular frozen materials and how those changes affect strength.

Key Innovation: It introduces an in situ sub-micron 4D CT triaxial system plus pore-network analysis to directly track evolving shear bands, particle metrics, permeability, and pore topology through coupled cycling.

58. FICA: Fine-Grained Invariant Causal Adaptation for few-shot remote sensing scene classification

Source: ISPRS J. Photogrammetry Type: few-shot remote sensing ML Geohazard Type: general remote-sensing transfer Relevance: 5/10

Core Problem: Based on the title, the study tackles few-shot remote-sensing scene classification under distribution shift by learning fine-grained invariant causal structure.

Key Innovation: The likely innovation is an invariant causal adaptation framework designed to improve generalization when only a few labeled remote-sensing samples are available.

59. Resistance capacity estimation for monopile-supported wind turbines in liquefiable soil under seismic and environmental loads

Source: Soil Dyn. & Earthquake Eng. Type: seismic liquefaction fragility study Geohazard Type: earthquake and liquefaction vulnerability Relevance: 5/10

Core Problem: The study asks how earthquake type, soil deformability, and model fidelity influence seismic fragility of monopile-supported offshore wind turbines in liquefiable ground.

Key Innovation: It performs cloud-based fragility analysis across multiple earthquake classes and shows that crustal and interface events plus soft soils strongly increase vulnerability.

60. Unsupervised Detection of Groundwater Storage Anomalies in Ghana Using GRACE Satellite Data

Source: ArXiv (Geo/RS/AI) Type: Preprint Geohazard Type: drought and subsidence proxy Relevance: 4/10

Core Problem: Characterize national groundwater anomalies in a data-scarce setting.

Key Innovation: GRACE-based anomaly detection using ensemble Isolation Forest plus spatiotemporal analysis.

61. Remote Sensing, Vol. 18, Pages 2714: Progressive Pseudo-Label Filtering with Reciprocal Neighborhood Retrieval for Cross-View Geo-Localization

Source: Remote Sensing (MDPI) Type: cross-view geo-localization Geohazard Type: multi-hazard transfer Relevance: 4/10

Core Problem: Learn cross-view geo-localization without paired ground-satellite labels.

Key Innovation: Progressively filtered pseudo-labels with reciprocal neighborhoods and consistency cues.

62. Remote Sensing, Vol. 18, Pages 2718: Interpretation of Gravity Changes at the Dongchuan Station

Source: Remote Sensing (MDPI) Type: gravity monitoring Geohazard Type: earthquake monitoring Relevance: 4/10

Core Problem: Interpret gravity changes at a tectonically active station without confusing local disturbances for geophysical signal.

Key Innovation: Combined gravity, GNSS, hydrologic correction, imagery, and forward modeling to isolate anthropogenic mass effects.

63. Remote Sensing, Vol. 18, Pages 2710: A Vision Transformer with Dynamic Masking and Cross-Modal Semantic Learning for Remote Sensing Scene Classification

Source: Remote Sensing (MDPI) Type: self-supervised scene classification Geohazard Type: multi-hazard transfer Relevance: 4/10

Core Problem: Use unlabeled remote-sensing images more effectively for scene classification.

Key Innovation: Combined dynamic masking pretraining with cross-modal semantic learning in a Vision Transformer.

64. Remote Sensing, Vol. 18, Pages 2709: A Coordinate-Based Framework for Sea Surface Wind Speed Reconstruction from Sparse Multi-Source Observations

Source: Remote Sensing (MDPI) Type: wind field reconstruction Geohazard Type: storm-wind transfer Relevance: 4/10

Core Problem: Reconstruct sea-surface wind from sparse multi-source observations at arbitrary locations.

Key Innovation: Used a coordinate-based deep model for non-gridded fusion that beats several baselines under sparse input.

65. Impacts of engineering activities on the permafrost environment in the Hola Basin in the Northern Da Xing'anling Mountains in Northeast China

Source: Engineering Geology Type: permafrost engineering impact study Geohazard Type: permafrost degradation Relevance: 4/10

Core Problem: The paper appears to examine how engineering activities alter the permafrost environment in the Hola Basin and thereby affect cold-region ground conditions.

Key Innovation: From the supplied information, the likely contribution is a case-specific assessment of anthropogenic disturbance to permafrost systems in a geomaterial and engineering geology context.

66. Integrating vulnerability and prioritization policies in post-earthquake recovery analysis of accessibility-based healthcare urban networks

Source: IJDRR Type: earthquake recovery network analysis Geohazard Type: post-earthquake infrastructure recovery Relevance: 4/10

Core Problem: The paper appears to study how vulnerability and policy-based prioritization affect restoration of healthcare access networks after earthquakes.

Key Innovation: From the title, the main contribution is likely a recovery-analysis framework that combines accessibility metrics with vulnerability and prioritization policies.

67. A prefire assessment of postfire threats to support proactive planning for fire recovery

Source: IJDRR Type: postfire threat planning Geohazard Type: postfire cascading hazards Relevance: 4/10

Core Problem: The paper appears to ask how prefire analysis can identify likely postfire threats early enough to guide recovery planning before a fire occurs.

Key Innovation: From the title, the main value is a proactive planning framework that shifts postfire threat assessment earlier in the disaster cycle.

68. Staged stress-release mechanism and optimal configurations of longitudinal joints in arc-toe trapezoidal canals under frost heave

Source: Cold Regions Science and Technology Type: Journal article / preprint record Geohazard Type: Frost heave and cold-region infrastructure deformation Relevance: 4/10

Core Problem: Longitudinal joints mitigate frost-heave cracking in cold-region canals, but designers lack quantitative guidance on where to place them and how joint number and width should vary with soil, groundwater, and temperature.

Key Innovation: A validated coupled hydro-thermo-mechanical model and 273-case design study isolate the roles of joint position, width, and number, showing that stress-based placement and climate-dependent joint counts outperform simple lining thickening.

69. InSAR-driven characterization of “ungauged” aquifer systems: Insights from the Gediz River Basin, Türkiye

Source: Journal of Hydrology Type: InSAR aquifer characterization Geohazard Type: ground deformation proxy / aquifer systems Relevance: 4/10

Core Problem: The paper appears to ask whether InSAR observations can characterize poorly monitored aquifer systems and reveal their behavior without dense ground data.

Key Innovation: From the title, the likely innovation is using deformation signals as a remote-sensing proxy to infer properties of otherwise ungauged aquifer systems.