TerraMosaic Daily Digest: July 28, 2026
Daily Summary
Failure thresholds in slopes and debris systems are being resolved at finer mechanical and spatial scales. UAV repeat surveys resolve the stop-and-go kinematics of the Pietrapertosa earthflow at metre to sub-metre scale, while shaking-table, base-friction, and centrifuge experiments sharpen how cohesion, bedding geometry, and partially mobilized slip surfaces govern coseismic landslide initiation and mode switching. Debris-flow studies also move beyond generic loading assumptions by quantifying how barrier flexibility redistributes impact pressure and how residual post-event gullies can be coupled to check-dam response across runout, CFD, and flume scales.
Remote-sensing advances center on observability, reliability, and geometry rather than nominal image accuracy. A new InSAR suitability index for northwest Yunnan treats layover, foreshortening, and shadow as measurable constraints on landslide monitoring rather than background nuisances, while a global aerosol study shows that underrepresentation of drylands in reference networks can systematically distort retrieval validation. Flood mapping, multimodal SAR-optical fusion, phase-map segmentation, multi-date stereo reconstruction, and robust RPC adjustment likewise address where and when Earth observation products can be trusted under clouds, noise, seasonality, bright surfaces, or phase discontinuities. The same operational turn appears in phenology-guided crop-damage early warning, evidence-grounded disaster change narration, and coupled flood-health risk mapping.
Across earthquakes, floods, rockburst hazards, and cryosphere change, predictability and uncertainty are becoming explicit analytical quantities. An information-theoretic synthesis reframes earthquake forecasting limits in terms of entropy reduction, Groningen analyses expose delayed and spatially heterogeneous production-seismicity coupling, and regional probabilistic models translate site amplification into required slope resistance for co-seismic stability. Similar thinking appears in flood-frequency remapping below reservoirs, resilience assessment for urban building portfolios, offshore scour reliability analysis, and interpretable or physics-aware learning frameworks that privilege robustness, calibration, and mechanism over benchmark gains alone.
Key Trends
The day's papers repeatedly convert hidden constraints on observability, trigger thresholds, and post-event propagation into explicit state variables that can be mapped, ranked, or stress-tested.
- Observability Is Becoming a Hazard Variable: Several of the strongest papers treat measurement blind spots as part of the hazard problem itself: UAV repeat mapping resolves earthflow kinematics that satellites miss, OGSI maps where InSAR is geometrically trustworthy, global aerosol analysis exposes dryland bias in reference networks, and phase-map segmentation or multi-date geometry methods are redesigned around noisy and seasonally mismatched observations.
- Failure Thresholds Are Being Measured, Not Assumed: Coseismic slope studies quantify how cohesion, bedding dip, partial slip mobilization, and site amplification change failure mode and required resistance, while debris-flow and rockfall papers turn barrier flexibility, residual-source loading, and trajectory uncertainty into explicit design or fatality metrics.
- Residual and Cascading Impacts Are Moving Upstream in Analysis: The literature is shifting from primary event footprints to what follows them: residual debris in post-event gullies, microbial health pathways in urban floods, expansion of potentially dangerous glacial lakes, sediment redistribution after embankment breach, and bank erosion under reservoir regulation are all modeled as first-order hazards rather than afterthoughts.
- Interpretable and Physics-Aware Models Are Replacing Opaque Performance Claims: Entropy-based earthquake predictability, conceptually interpretable hydrologic networks, SHAP-enabled geotechnical tabular foundation models, reliability-aware multimodal fusion, and even negative benchmarks for ConvLSTM all point in the same direction: useful hazard AI is being judged more by mechanism, calibration, and failure-mode transparency than by isolated headline accuracy.
Selected Papers
Across landslides, earthquakes, floods, and cryosphere hazards, the strongest studies turn blind spots and uncertain thresholds into measurable quantities, from UAV-resolved earthflow kinematics and InSAR observability maps to probabilistic coseismic slope resistance and flood-linked public-health exposure. Several papers then push beyond event detection to residual debris-flow risk, glacial-lake expansion, and infrastructure recovery, showing a field moving from snapshot mapping toward mechanism-aware decision support.
1. Analysis of slow-moving landslide evolution using high-resolution UAV monitoring: The Pietrapertosa earthflow (Southern Apennines, Italy)
Core Problem: Slow-moving earthflows are hard to monitor with satellite products and sparse field instruments when deformation is spatially heterogeneous and changes rapidly after reactivation.
Key Innovation: Nine UAV surveys from 2022 to 2025 quantify scarp-terrace evolution, displacement rates, erosion-deposition patterns, and mobilized volume in the Pietrapertosa earthflow at high spatial resolution.
2. How to quantify earthquake predictability? Advances in earthquake forecasting and predictability limits
Core Problem: Earthquake forecasting still lacks a common measure that separates intrinsic predictability from apparent model skill gains.
Key Innovation: The paper casts predictability as an entropy gap within a point-process framework, derives information-theoretic benchmarks for Poisson and ETAS models, and shows how mutual information from pre-event observations can raise forecast skill.
3. Dynamics of Debris Flow Impact on Flexible Barriers with Varying Flexibility
Core Problem: Design formulas for debris-flow flexible barriers do not adequately capture how barrier flexibility changes impact pressure and peak force.
Key Innovation: Flume experiments and numerical back-analysis introduce a dimensionless flexibility number that links barrier deformation to pressure redistribution and to an updated impact-force design equation.
4. Terrain impact analysis and suitability assessment of InSAR geometric distortions in the mountainous gorge region of Northwest Yunnan
Core Problem: Steep canyon terrain creates systematic InSAR layover, foreshortening, and shadow that obscure where landslide deformation can actually be observed.
Key Innovation: The study combines LSM and R-Index analyses into an Observation Geometric Suitability Index that maps ascending-descending complementarities and quantifies InSAR reliability across northwest Yunnan.
5. Propagation characteristic of debris avalanches and dynamic effects of curvature terrain under complex topography in China
Core Problem: Debris-avalanche mobility in high mountains remains difficult to predict because runout depends strongly on geomorphic setting and path curvature.
Key Innovation: By combining regional geomorphic statistics with physical experiments, the paper shows how curved gullies shift momentum laterally, reduce velocity, and alter deposit geometry across major Chinese mountain belts.
6. Investigating instability evolution of anti-dipping rock slopes under seismic dynamics: novel base-friction testing and numerical simulation
Core Problem: Seismic instability of anti-dip rock slopes is controlled by bedding geometry, but failure-mode transitions under shaking remain poorly constrained.
Key Innovation: A new base-friction physical test system and supporting simulations show how increasing bedding dip moves failure from progressive sliding to sliding-toppling and then severe block toppling.
7. Dynamic characteristics of residual debris flows and response mechanisms of check dam in Baima Gully along the Jiumian Expressway
Core Problem: Residual loose material left after a debris-flow event is hard to translate into structure-scale design loads for check dams.
Key Innovation: The paper couples UAV and field source quantification, calibrated MassFlow runout modeling, CFD, and flume tests to predict residual surges and evaluate how a gravity dam reduces velocity, inundation, discharge, and impact pressure.
8. Glacial lake expansion and outburst flood threat along the China–Pakistan Economic Corridor: Hunza Basin, Pakistan (1990–2021)
Core Problem: Rapid glacial-lake growth in the Hunza Basin threatens downstream communities and infrastructure, but its spatiotemporal evolution and climatic drivers need quantification.
Key Innovation: Multi-temporal Landsat mapping and climate-trend analysis identify a 74% increase in lake numbers, expanded lake area, and 20 potentially dangerous glacial lakes along the China-Pakistan Economic Corridor.
9. High sensitivity of seismically induced shallow landslides in pyroclastic fall deposits to soil cohesion: insights from shaking table tests
Core Problem: Why the Hokkaido Iburi earthquake triggered widespread shallow slides in pumice-covered slopes is still uncertain, especially the role of cohesion under seismic loading.
Key Innovation: Shaking-table tests and strain-softening stability analysis show that apparent cohesion scaling strongly controls failure timing and that sliding develops in stepwise bursts late in the mainshock record.
10. Lithofacies-controlled 3D geotechnical modeling for liquefaction potential assessment in a Holocene coastal system (Porto San Giorgio, Italy)
Core Problem: Pointwise CPT-based liquefaction estimates cannot represent lithofacies-controlled spatial variability in heterogeneous Holocene coastal deposits.
Key Innovation: The study builds a lithofacies-constrained 3D geotechnical model that interpolates liquefaction susceptibility across a coastal system rather than leaving the assessment at isolated sounding locations.
11. Regional probabilistic modeling of required slope resistance to co-seismic failure including site amplification effects
Core Problem: Regional coseismic landslide screening rarely incorporates site amplification when estimating the slope resistance needed to avoid failure.
Key Innovation: This paper formulates a probabilistic framework that converts amplified seismic demand into required slope resistance at regional scale, yielding spatially explicit co-seismic failure thresholds.
12. Seismic behavior of rock slopes containing a partially mobilized sliding surface via centrifuge shaking table testing: Unraveling interaction between seismic load transfer and structural deterioration
Core Problem: Rock slopes with partially mobilized sliding surfaces fail under earthquakes through coupled load transfer and structural deterioration that conventional models underresolve.
Key Innovation: Centrifuge shaking-table tests isolate this interaction and show how a pre-existing partly activated sliding surface amplifies dynamic instability in rock slopes.
13. Coupled hydrological and public health risks from urban flooding: integrated remote sensing, machine learning, and hydrodynamic–ecological modelling
Core Problem: Urban flood assessments often ignore the way contaminated floodwater couples hydraulic exposure with microbial public-health risk.
Key Innovation: The study integrates remote sensing, machine learning, and hydrodynamic analysis to map where flood inundation and public-health exposure co-intensify in urban environments.
14. Flood extent mapping using multi-temporal sentinel-2 remote sensing and deep learning
Core Problem: Flood extent mapping from optical satellite data remains unreliable when inundation timing and surface conditions change through an event sequence.
Key Innovation: Using multi-temporal Sentinel-2 inputs and deep learning, the framework explicitly exploits pre-, syn-, and post-flood image context to improve flood delineation in the Karkheh River Basin.
15. Intensity measures considering directivity of near-fault ground motion and optimal selection for bridges -- case study in the Wushi earthquake
Core Problem: Near-fault bridge seismic assessment still lacks intensity measures that capture directivity in two horizontal directions instead of along a single axis.
Key Innovation: The paper defines omnidirectional ground-motion intensity measures and uses the Wushi earthquake to identify directivity-aware predictors better suited to bridge response screening.
16. A dual-framework approach to regional flood susceptibility: coupling frequency ratio with logistic regression in Tokat, Türkiye
Core Problem: Regional flood susceptibility maps often depend on a single statistical model even when flood controls are nonlinear and spatially heterogeneous.
Key Innovation: The study couples frequency ratio and logistic regression within a dual-framework susceptibility analysis to cross-check and strengthen regional flood zoning.
17. Quantitative assessment of geological risks in the coastal waters from Huizhou to Shanwei, north of the South China Sea
Core Problem: Coastal waters on the northern South China Sea host multiple marine geological hazards, but their relative distribution and severity are not well quantified for planning.
Key Innovation: By integrating single-channel seismic profiles, boreholes, and regional geology, the paper produces a quantitative inventory of four major marine geohazard types between Huizhou and Shanwei.
18. Analysis of the Influence of Weak Stratum State on the Stability of Dip Slope and Anti-Dip Sides of Rock Foundation Pits
Core Problem: Foundation pits in limestone terrain can be destabilized by weak interlayers, yet the separate roles of weak-layer thickness and burial depth are not well constrained.
Key Innovation: 3DEC simulations disentangle how weak-stratum state changes deformation and stability on both dip and anti-dip sides of rock foundation pits in subway engineering.
19. Phenology-Guided Early Prediction of Crop Damage Under Long-Duration Inundation Using Multi-Source SAR–Optical Imagery
Core Problem: Crop losses from long-duration flooding are difficult to anticipate early because inundation persistence and vegetation response evolve differently across sensors.
Key Innovation: The study uses phenology-guided fusion of SAR and optical imagery to predict crop damage before the full post-flood signal is visible in conventional assessments.
20. Production-seismicity relationship in the Groningen field, NL: a multi-scale statistical and hazard assessment
Core Problem: The Groningen production-seismicity link is known in aggregate, but its timing and spatial variability across fault-bounded domains remain unclear.
Key Innovation: A multi-scale statistical analysis from field-wide to well-centered buffers shows that seismic response strengthens at lag times of about three to four years and varies sharply with local fault architecture.
21. Landslide image recognition based on residual networks with transfer learning and spatial-channel dual attention mechanism
Core Problem: Landslide image recognition models often miss subtle morphological cues when training data are limited or visually heterogeneous.
Key Innovation: The paper combines transfer learning with a residual network and spatial-channel dual attention to improve image-based landslide identification.
22. Emergency monitoring index evaluation for collapse scenarios: A TriFAHP-DBN model addressing data-scarce conditions
Core Problem: Geological-disaster emergency monitoring begins under severe data scarcity, making collapse-response index evaluation unstable and subjective.
Key Innovation: A TriFAHP-DBN framework fuses sparse evidence and uncertainty reasoning to rank monitoring priorities during early collapse emergencies.
23. A New Evaluation Method of Active Energy-Releasing Capacity of Boreholes for Roadway Rockburst Mitigation
Core Problem: Borehole destressing is widely used against roadway rockburst, but its active energy-release capacity has lacked a quantitative evaluation standard.
Key Innovation: The study derives borehole and roadway burst criteria, defines new dynamic risk and critical energy density indices, and introduces an R-B energy-efficiency ratio to quantify how much roadway energy storage can be actively released.
24. Integrating 3D rockfall modelling and probabilistic analysis for vehicle fatality risk quantification
Core Problem: Rockfall risk analyses rarely translate 3D trajectory uncertainty into the probability of vehicle occupant fatality on exposed roads.
Key Innovation: The paper couples three-dimensional rockfall simulation with probabilistic fatality analysis to quantify vehicle risk rather than only block trajectories or impact zones.
25. A Comprehensive Seismic Resilience Assessment Framework for Urban Building Portfolios Considering Building Functional Importance
Core Problem: Urban seismic resilience metrics usually underweight the functional importance of different building types during post-earthquake recovery.
Key Innovation: The framework introduces building functional importance and functional sensitivity factors, then uses them to simulate prioritized recovery and derive a normalized portfolio-scale resilience score.
26. A centrifuge study of seismic interactions among three structures on liquefiable ground
Core Problem: The seismic response of closely spaced structures on liquefiable ground is altered by structure-soil-structure interaction, but this effect is poorly resolved experimentally.
Key Innovation: Centrifuge tests on three neighboring structures show how liquefiable soil mediates coupled seismic behavior in dense urban settings.
27. Hydraulic response and failure mechanism of red-bed mudstone slope under cycling freeze–thaw conditions: a comparison study
Core Problem: Red-bed mudstone slopes degrade under repeated freeze-thaw and hydraulic cycling, but the coupled water-response and failure mechanism remains insufficiently resolved.
Key Innovation: The study identifies how cycling freeze-thaw and wetting-drying processes drive hydraulic change and eventual failure in red-bed mudstone slopes.
28. Impacts of geomorphology and land use/cover change on sediment connectivity in China's Loess Plateau hilly–gully region
Core Problem: Changes in geomorphology and land use can reorganize sediment pathways across the Loess Plateau, but their combined effect on sediment connectivity is hard to isolate.
Key Innovation: The paper uses sediment connectivity as the analytical lens to quantify how topography and land-use/cover change jointly reshape source-to-sink coupling across the plateau.
29. Rapid Morphological Changes in the Intertidal Zone of the Ganges‐Brahmaputra Delta Revealed by SWOT Altimetry Mission
Core Problem: Intertidal morphology in the Ganges-Brahmaputra Delta changes too rapidly for conventional monitoring to capture regional patterns with enough temporal frequency.
Key Innovation: SWOT water-surface and elevation observations are used to reveal rapid intertidal morphological change across the delta at broad spatial scale.
30. Mapping Flood Peak Distributions From Natural Conditions to Post‐Reservoir States and Vice Versa: A Simplified Analytical Framework for Uncontrolled Reservoirs With Spillways Outlet
Core Problem: Reservoir regulation changes downstream flood peaks, but hydrologists lack a clean way to map probability distributions between natural and post-reservoir states.
Key Innovation: The paper derives a transformation framework that carries flood-peak distributions back and forth between regulated and unregulated flow regimes.
31. SeasonStereo: Robust Dense Stereo Matching for Multi-Date Satellite Imagery via Generative AI
Core Problem: Dense stereo matching for multi-date satellite imagery breaks down when seasonal and illumination differences destroy traditional correspondence assumptions.
Key Innovation: SeasonStereo uses generative-AI-assisted matching to recover robust diachronic correspondences and improve 3D reconstruction from multi-date satellite pairs.
32. EvoPINN: Agentic Discovery of Executable Algorithms for Physics-Informed Neural Networks
Core Problem: Physics-informed neural networks still depend heavily on hand-designed solver components and algorithm choices.
Key Innovation: EvoPINN uses agentic search to discover executable algorithmic variants for PINNs instead of treating architecture and training design as fixed manual inputs.
33. PIKS: Universal Physics-Informed Kernel Methods
Core Problem: Physics-informed learning methods often trade scalability for analytical structure when solving operator-constrained problems.
Key Innovation: PIKS proposes a kernel-based physics-informed framework that retains operator awareness while broadening the class of learnable structured dynamics.
34. Equivariant Eikonal Neural Networks: Grid-Free, Scalable Travel-Time Prediction on Homogeneous Spaces
Core Problem: Travel-time prediction for seismic and wavefront problems is usually tied to grids that limit geometric flexibility and scalability.
Key Innovation: Equivariant Eikonal Neural Networks solve grid-free travel-time prediction by embedding Eikonal structure within symmetry-aware neural fields.
35. Global monitoring of methane point sources using deep learning on hyperspectral radiance measurements from EMIT
Core Problem: Global methane super-emitters are numerous, transient, and hard to monitor consistently from space at point-source scale.
Key Innovation: The study applies deep learning to hyperspectral radiance data for global monitoring of methane point sources, enabling large-scale detection of concentrated emissions.
36. TabPFN Extensions for Interpretable Geotechnical Modelling
Core Problem: Geotechnical site characterization must work with sparse heterogeneous borehole data while still exposing prediction uncertainty and variable importance.
Key Innovation: The paper evaluates TabPFN and its extensions on classification and iterative imputation tasks, then couples embeddings, SHAP, uncertainty decomposition, and consolidation reliability metrics into one interpretable workflow.
37. Stress-path dependent lateral response of monopiles under time-evolving scour morphologies: A field-data-driven numerical investigation
Core Problem: Offshore monopile lateral capacity is known to decline under scour, but the influence of evolving scour-hole morphology on stress paths and response is poorly constrained.
Key Innovation: Field-informed analysis shows how different time-evolving scour geometries change monopile lateral response through stress-path-dependent mechanisms rather than through depth loss alone.
38. Post-scour reliability assessment of offshore monopiles incorporating field monitoring data and multivariate Gaussian modelling
Core Problem: Reliability assessment of offshore monopiles after scour is difficult because monitoring data and multimodal uncertainties are rarely integrated.
Key Innovation: The study combines field monitoring with probabilistic modeling to update post-scour reliability instead of relying on deterministic foundation checks.
39. Dynamic response analysis of tripod offshore wind turbines considering source-to-structure seismic input and near-fault effects
Core Problem: Seismic design of near-fault offshore wind turbines is challenging because source, propagation, seabed, and soil-structure interaction effects act together.
Key Innovation: This work analyzes tripod offshore turbine response with source-to-structure seismic inputs so that near-fault rupture and subsoil effects are carried through the same dynamic model.
40. Remote sensing of aerosols over drylands: Challenges, uncertainties, and paths forward
Core Problem: Global aerosol retrievals are validated against AERONET sites concentrated in urban and vegetated regions, leaving bright, coarse-aerosol drylands systematically underrepresented.
Key Innovation: The study quantifies that sampling imbalance against MODIS-POLDER disagreement and radiative effects, then argues for validation stratified by surface type and aerosol regime rather than a nominally global aggregate.
41. An interpretable predictive model for shale gas hazard risk in tunnel excavation based on an improved broad learning system
Core Problem: Tunnel excavation in shale formations can trigger gas hazards, but risk prediction is often opaque and weakly linked to geological controls.
Key Innovation: The paper builds an interpretable predictive model that connects shale-gas hazard risk to geological and construction variables instead of offering a black-box warning score.
42. Influence of geological attributes and structural elements on the behavior of a metamorphic colluvium granules from Himachal Pradesh India
Core Problem: The engineering behavior of metamorphic colluvium from landslide-prone Himalayan terrain is not well understood from geological attributes alone.
Key Innovation: The study ties the mechanical response of a metamorphic colluvial material to its geological fabric and structural elements using samples from a documented landslide setting.
43. Climate-driven amplification of monthly sediment transport peaks in a permafrost dominated watershed
Core Problem: Permafrost thaw is expected to amplify sediment pulses, but the monthly timing and magnitude of climate-driven transport peaks remain uncertain.
Key Innovation: The paper demonstrates how warming and thaw reorganize monthly sediment-transport peaks in a permafrost-dominated watershed rather than only changing annual totals.
44. Experimental analysis, numerical simulation and prediction model of tunnel lining internal forces under the coupled effect of water-rich Karst and Goaf
Core Problem: Tunnel linings crossing water-rich karst and mined-out coal seams face coupled geohazards that standard single-defect analyses cannot capture.
Key Innovation: Experiments, simulations, and a prediction model are combined to quantify lining internal forces under simultaneous karst and goaf conditions.
45. Shear-driven dynamic fracture in rock joints with a single locked convex asperity structure under constant normal stiffness
Core Problem: Fault-controlled tunnel rockbursts depend on joint-scale dynamic fracture processes that are difficult to observe under combined static and dynamic loading.
Key Innovation: The study isolates shear-driven fracture in joints with a locked convex asperity to clarify one microscale mechanism of high-intensity fault rockburst initiation.
46. The application of geospatially-integrated machine learning models for flood prediction: A review
Core Problem: Flood-prediction studies using geospatial machine learning are proliferating faster than their strengths, limitations, and data dependencies are being synthesized.
Key Innovation: This review organizes geospatially integrated machine-learning approaches for flood prediction and evaluates where they are useful, fragile, or data-limited.
47. Constraining the Permeability and Outer‐Rise Hydration at the Central America Margin
Core Problem: Water transport into the subduction system controls seismic and magmatic hazards, but outer-rise permeability and hydration are difficult to constrain offshore.
Key Innovation: The study estimates permeability and hydration state at the central America margin to tighten geohazard interpretations of subduction-zone fluid delivery.
48. Seismic Constraints on Igneous Intrusions and Fluid/Gas Distribution in a Continental Back‐Arc System: Evidence From the Yaeyama Rift in the Southern Okinawa Trough
Core Problem: Continental back-arc hazard assessment depends on knowing where intrusive bodies and fluid-rich zones are concentrated in the crust.
Key Innovation: Seismic imaging is used to map igneous intrusions and fluid-gas distributions, providing structural constraints on a back-arc system's tectonic and magmatic hazard setting.
49. On the Importance of the Jacobian of a DNN Observation Operator in Land Data Assimilation
Core Problem: Neural observation operators can degrade land data assimilation if their Jacobian structure distorts the update step, even when forward prediction looks accurate.
Key Innovation: The paper shows that the Jacobian of a DNN observation operator materially affects assimilation behavior and must be evaluated as part of model design.
50. MetaKoopman: Bayesian Meta-Learning of Koopman Operators for Modeling Structured Dynamics under Distribution Shifts
Core Problem: Forecasting nonlinear dynamics under distribution shift remains unstable when learned latent operators are not transferable across regimes.
Key Innovation: MetaKoopman uses Bayesian meta-learning of Koopman operators to preserve structured dynamics while adapting to shifted conditions.
51. From Conceptual Hydrologic Models to Conceptually Interpretable Neural Networks: A Snow-Water Mass-Conserving-Perceptron Framework for Discovering Catchment-Scale Precipitation-Storage-Runoff Representations
Core Problem: Neural hydrologic models often gain accuracy at the cost of conceptual interpretability and mass-conservation transparency.
Key Innovation: The study reformulates conceptual snow-water models as physically constrained neural networks, preserving interpretable hydrologic components within a learning system.
52. Robust RPC Bundle Adjustment for Multi-Date Satellite Imagery with Season-Invariant Correspondences
Core Problem: Refining RPC camera models for multi-date satellite scenes is difficult because seasonal change weakens the correspondences needed for bundle adjustment.
Key Innovation: The method performs robust RPC bundle adjustment using season-invariant correspondences so that geolocation can be improved across temporally separated imagery.
53. Skillful forecasting of offshore winds from satellite scatterometer constellations
Core Problem: Intraday offshore-wind forecasts are limited by sparse in situ observations even though satellite scatterometers repeatedly sample the same marine regions.
Key Innovation: The paper shows that constellations of scatterometer observations can support skillful offshore-wind forecasting rather than functioning only as retrospective measurements.
54. SpectraDINO: Modality-Conditioned Adaptation of RGB Vision Foundation Models Across Infrared Bands
Core Problem: RGB-pretrained vision foundation models lose accuracy when applied directly to infrared imagery because spectral statistics and semantics shift across bands.
Key Innovation: SpectraDINO adapts RGB foundation representations to infrared bands through modality-conditioned tuning instead of training a new model from scratch.
55. Mapping small reservoirs across Brazil from 1984 to 2025
Core Problem: Small reservoirs are widespread and hydrologically important in Brazil, yet they are systematically undercounted in national water inventories.
Key Innovation: The paper maps small reservoirs across Brazil from 1984 to 2025, turning diffuse small-scale impoundment into a measurable national dataset.
56. AerialMetric: Benchmarking and Adapting UAV Monocular Metric Depth Estimation in the Real World
Core Problem: Real-world UAV monocular depth estimation lacks benchmarks that expose metric accuracy limits under aerial imaging conditions.
Key Innovation: AerialMetric provides a benchmark and adaptation framework for monocular metric depth estimation tailored to practical UAV scenes.
57. Interferometric Phase Map Segmentation Algorithm Based on an Adaptive Optimization Tolerance Granularity Space Model
Core Problem: Unified phase unwrapping over an entire InSAR phase map propagates noise from low-quality areas into otherwise interpretable regions.
Key Innovation: The paper segments interferometric phase maps with an adaptively optimized tolerance-granularity model so that region-specific unwrapping can reduce global error propagation.
58. Reliability-Aware Cross-Modal Fusion of Sentinel-1 SAR and Sentinel-2 Optical Imagery for Robust Patch-Level Land Cover Scene Classification
Core Problem: SAR-optical fusion models fail unpredictably when one modality is degraded by noise, masking, or channel loss.
Key Innovation: A reliability-aware fusion gate dynamically reweights Sentinel-1 and Sentinel-2 features using confidence and optical quality indicators, improving patch-level classification under degraded conditions.
59. ChangePixel: Pixel-Level Evidence-Grounded Disaster Change Narration via Single-Backbone Transfer
Core Problem: Disaster change detection products usually mark changed pixels without explaining the evidence behind the inferred change.
Key Innovation: ChangePixel frames disaster mapping as pixel-level evidence-grounded narration, aiming to pair change localization with explicit explanatory support from a single backbone.
60. Enhancing unconfined compressive strength of granite residual soil through synergistic application of microorganisms and jute fibers
Core Problem: MICP can strengthen granite residual soil but often leaves the treated material too brittle for slope and earthwork applications.
Key Innovation: The study combines microorganisms and jute fibers to increase unconfined compressive strength while markedly improving ductility through calcite-fiber-soil bonding.
61. Regional cryosphere strategy for the Hindu Kush Himalaya
Core Problem: Cryosphere hazards across the Hindu Kush Himalaya are usually managed separately even though glaciers, snow, permafrost, water, and risk co-evolve regionally.
Key Innovation: The paper proposes a regional cryosphere strategy that treats the Hindu Kush Himalaya as an integrated multi-hazard system rather than a set of isolated sectors.
62. Synergistic mechanism between root mechanical and soil ecological effects on the dynamic evolution of root reinforcement
Core Problem: Vegetation-based slope reinforcement is often reduced to root mechanics alone, ignoring ecological effects that change soil structure over time.
Key Innovation: The study analyzes how root mechanical action and soil ecological processes work together during the dynamic evolution of root reinforcement.
63. STCAFormer: Spatio-temporal cross attention based transformer for fusion of remote sensing images
Core Problem: Remote-sensing monitoring suffers when high spatial and high temporal resolution observations must be fused without preserving change consistency.
Key Innovation: STCAFormer uses spatio-temporal cross attention to improve image fusion across times and resolutions for geospatial monitoring tasks.
64. Sediment deposition induced by embankment-breach flooding under the influence of drainage trench networks
Core Problem: Embankment-breach floods can reorganize farmland sedimentation, but the effects of drainage trench networks on deposit patterns are poorly known.
Key Innovation: The paper quantifies how drainage trenches modulate sediment deposition during breach flooding, linking flood hydraulics to agricultural geomorphic impact.
65. Gravel cover regulates vegetation growth and bank-slope erosion resistance in the Three Gorges Reservoir Area
Core Problem: Reservoir banks in the Three Gorges region undergo repeated hydrodynamic stress, yet the role of gravel cover in vegetation growth and erosion resistance remains unclear.
Key Innovation: The study shows how gravel cover alters plant establishment and strengthens bank-slope resistance to erosion under reservoir water-level fluctuations.
66. Assessment of erosion and accretion rates along the Pacific coast of Panama
Core Problem: Long coastlines can hide fundamentally different erosion controls, making regional rates hard to interpret from shoreline change alone.
Key Innovation: Using satellite-based erosion and accretion estimates along Panama's Pacific coast, the paper shows that sediment supply and mangrove buffering can dominate over simple wave-energy expectations.
67. Deformation and failure characteristics of high-geo-stress layered surrounding rock tunnels: Laboratory testing and the material point method
Core Problem: Layered rock tunnels under high geostress exhibit asymmetric deformation and failure that standard continuum treatments often oversimplify.
Key Innovation: Laboratory characterization and material point method simulations identify how bedding angle and lateral pressure jointly control large-deformation tunnel failure.
68. Benchmarking ConvLSTM for One-Day-Ahead IMDAA Rainfall-Field Prediction across Four Indian Cities
Core Problem: ConvLSTM is widely used for rainfall forecasting, but its value for one-day-ahead daily gridded rainfall fields is rarely benchmarked against simpler baselines.
Key Innovation: A four-city IMDAA comparison shows that ConvLSTM often fails to outperform persistence or FC-LSTM, especially for high-rainfall detection, providing a useful negative benchmark.
69. DACE-Net: Detail-Aware Calibrated Enhancement Network for Building Change Detection
Core Problem: Building change detection in high-resolution imagery is still vulnerable to pseudochanges caused by texture variation and imperfect detail recovery.
Key Innovation: DACE-Net combines long-range feature extraction with explicit and implicit detail calibration to suppress false changes and improve building-change mapping.
70. Comparative Performance Analysis of Mainstream Deep Learning and Vision Foundation Models for Small-Sample Vegetation Segmentation in High-Resolution Remote Sensing Imagery
Core Problem: Small-sample vegetation segmentation in remote sensing has unclear performance boundaries between classic deep networks and newer vision foundation models.
Key Innovation: The paper benchmarks eight models, including SAM variants, and proposes cross-temporal pseudo-label self-training to improve small-sample segmentation under scarce annotation.
71. Impact of biochar particle size and content on small strain stiffness characteristics of amended expansive clay
Core Problem: Biochar amendments are being used to modify expansive clay, but their effects on small-strain stiffness and stiffness degradation are not well quantified.
Key Innovation: The study varies biochar particle size and content, then links stiffness evolution to SEM and porosimetry observations and a modified Hardin-Drnevich model.
72. Borehole geophysical studies in glaciers. Part II: Borehole in situ observatories
Core Problem: Cryosphere monitoring needs in situ observatories that can validate remote sensing and resolve internal glacier processes over long periods.
Key Innovation: This review synthesizes borehole observatory methods for glaciers, comparing sensors, deployment strategies, accuracy, and limitations for long-term ice monitoring.
73. The effect of grain size range on the accuracy of sediment sources determination using the composite fingerprinting approach
Core Problem: Sediment-source fingerprinting can be biased when the chosen grain-size fraction fails to represent all contributing sources.
Key Innovation: Rainfall simulations and artificial mixtures show when bulk or combined grain-size windows outperform dominant fractions for accurate sediment-source apportionment.