TerraMosaic Daily Digest: September 15, 2026
Daily Summary
An inventory from coastal Alaska links more than 700 satellite-mapped shallow landslides to autumn atmospheric rivers and shows overrepresentation on windward midslopes steeper than 35 degrees, enabling relative-susceptibility mapping. In Ruijin, adding a Landslide and Terrace Distinction Index increased U-Net F1 to 0.91; PSInSAR analyses in the Himalayas and Western Ghats detected deformation transitions that preceded landslides and correlated with rainfall.
Process-based models clarify landslide mobility and cascading hazards. In flume-validated MPM simulations, weak erodible beds modify slope-break geometry and basal resistance, promoting longer runout; the study identifies scour entrainment and impact bulldozing as distinct mechanisms. A gradation-aware model reproduced the two-stage Baige dam breach with errors below 15% for primary variables and identified the coefficient of uniformity as the most influential input for simulation accuracy.
Joint GNSS, InSAR and microgravity inversion located Askja's modeled source volume at approximately 0.4-4.6 km depth; mush compressibility was inferred only under assumed basaltic intrusion densities because density and compressibility remain in trade-off. At Etna, field mapping, historical imagery, laser scanning and drone photogrammetry reconstructed pulsed dike emplacement during the 1971 eruption. In Hoh Xil, GeoDetector analyses associated thermokarst-lake and thaw-slump abundance or expansion with different combinations of precipitation, ground temperature, soil moisture, ice content and thaw-index change; no single factor dominated the regional pattern.
Uncertainty and evaluation design are explicit in several studies. IRENE achieved lower CRPS than STEPS and DGMR at every lead time, although its adversarial variant retained excess fine-scale power at long leads. The Sentinel-2 benchmark contains 2,148 image-mask pairs from 25 wildfires and uses incident-disjoint splits for rare-class segmentation. For the 2025 Myanmar earthquake, field reconnaissance combined with simulated ground motions distinguished near-fault, liquefaction and bridge-typology damage modes; fragility studies likewise foreground sample representativeness, ground-motion uncertainty and regional construction practice.
Key Trends
Across the day's studies, hazard estimates depend increasingly on directional forcing, evolving material states and uncertainty-aware evaluation.
- Hazard patterns vary with terrain and forcing pathway: Atmospheric-river direction and slope position are associated with shallow-landslide occurrence; thermokarst patterns show factor-specific environmental associations; and solar-radiation-modification scenarios produce different extremes despite similar mean temperatures.
- Landslide workflows span mapping, deformation monitoring and cascading-process simulation: Deep-learning recognition maps historical landslides and PSInSAR detects deformation transitions, while entrainment-aware MPM and gradation-aware dam-breach models extend analysis to mobility and downstream flooding.
- Complementary observations refine latent-state inference: Joint microgravity-geodesy constrains Askja's source properties, radar-optical combinations track inundation and land cover, and TS-InSAR time series reveal persistent subsidence consistent with delayed compaction.
- Forecast and fragility studies increasingly represent uncertainty and site dependence: Probabilistic nowcasting evaluates ensemble calibration, Bayesian fragility modeling propagates missing-attribute and ground-motion uncertainty, and offshore ground-motion models account for site class and installation effects.
- Evaluation is becoming event-disjoint and sensor-transfer aware: The wildfire benchmark uses incident-disjoint splits, while GeoCrossBench tests no-overlap and superset spectral-band transfer, exposing performance losses on unseen bands.
Selected Papers
The 15 September selection is anchored by work on climate-topography associations in shallow landslides, with complementary studies on landslide mapping, PSInSAR-based deformation transitions, entrainment-aware runout and dam-breach simulation. Beyond landslides, the collection spans Earth observation of wildfire and floods, thermokarst change, volcanic plumbing, probabilistic rainfall nowcasting, earthquake reconnaissance, structural fragility and cross-sensor geospatial AI.
1. Shallow Landslides Align With Atmospheric Rivers in Coastal Steeplands
Core Problem: Quantifying how atmospheric-river direction and topographic position organize shallow-landslide occurrence in coastal mountains.
Key Innovation: Links a satellite-mapped inventory of more than 700 landslides to a multi-decadal atmospheric-river database and derives aspect- and slope-conditioned susceptibility patterns.
2. Influence of erodible substrate topography and strength on landslide mobility: insights from MPM modeling
Core Problem: Explaining how substrate strength and slope geometry control entrainment and runout.
Key Innovation: Develops and validates an enhanced multi-material MPM framework and identifies scour-entrainment and impact-bulldozing regimes.
3. An efficient approach for rapid simulation of landslide dam breach flood processes considering wide gradation characteristics
Core Problem: Rapidly simulating breach floods while representing wide grain-size distributions.
Key Innovation: Adds gradation-aware initiation and shear-stress physics to DB-IWHR and validates it against the two-stage Baige failure.
4. Geoinformation-based landslide recognition using deep learning in subtropical regions: a case study of Ruijin City
Core Problem: Regional recognition of historical landslides while reducing confusion with agricultural terraces.
Key Innovation: Integrates a Landslide and Terrace Distinction Index with U-Net, field checks, and optical imagery, raising F1 to 0.91.
5. Characterization of the Plumbing System Beneath Askja Caldera, Iceland, Revealed by Microgravity and Deformation Data During Uplift Between 2022 and 2023
Core Problem: Infer magma and mush properties beneath Askja during renewed uplift.
Key Innovation: Joint GNSS, InSAR, and microgravity inversion with compressible-mush interpretation.
6. Quantifying the changes and driving factors of thermokarst hazards in the Hoh Xil region of the Qinghai-Tibet Plateau
Core Problem: Quantifying changes in thermokarst lakes and retrogressive thaw slumps and their statistical associations with environmental factors.
Key Innovation: Combines thermokarst inventories with GeoDetector analysis to compare environmental associations with hazard distribution and expansion.
7. A Sentinel-2 benchmark dataset for deep-learning active-fire segmentation across 25 California wildfires
Core Problem: Provide rare-class satellite labels and incident-disjoint evaluation for active fire.
Key Innovation: Provides 2,148 Sentinel-2 image-mask pairs across 25 fires, with a mask-blind analyst review of 233 test chips.
8. Multisource Remote Sensing and Geospatial Analysis of Vineyard Wildfire Impacts and Resilience: The 2019 Kincade Fire
Core Problem: Assess vineyard wildfire resilience without conflating descriptive and causal contrasts.
Key Innovation: Event-anchored multisource EO framework with spatial models, smoke, roads, and recovery.
9. Seismic Performance and Damage Assessment of Transportation Systems During the 2025 Myanmar Earthquake (M7.9)
Core Problem: Characterizing transportation-system damage in the 2025 M7.9 Myanmar earthquake.
Key Innovation: A 14-day field database, simulated near-fault motions, mechanism analysis, and bridge fragility assessment.
10. Seismic Performance Assessment of a School Building Strengthened With External Shear Walls During an M L 6.4 Earthquake in Taiwan
Core Problem: Explaining damage and evaluating a retrofitted school during a real ML 6.4 earthquake.
Key Innovation: Field reconnaissance combined with parametric modeling and nonlinear response-history analysis.
11. Multi-Temporal Assessment of Bimodal Monsoon Flood Dynamics and Agricultural Exposure Using Integrated Sentinel-1 SAR and Sentinel-2 Optical Data in Punjab, Pakistan
Core Problem: Monitoring multi-phase monsoon inundation and agricultural exposure under persistent cloud cover.
Key Innovation: Integrates Sentinel-1 time series and Sentinel-2 classification in GEE, with independent spatial checks and explicit crop impacts.
12. The Evolution of Land Subsidence Under the New Water Regime in the North China Plain: A TS-InSAR and Spatiotemporal Pattern Analysis
Core Problem: Mapping how subsidence patterns evolved during groundwater recovery and management across the North China Plain.
Key Innovation: Combines 2016-2023 TS-InSAR, centroid tracking and emerging-hot-spot analysis to identify persistent localized compaction consistent with a delayed response.
13. Empirical fragility assessment of low-rise hybrid buildings following the 2023 Kahramanmaraş earthquake sequence
Core Problem: Estimating damage fragility for low-rise hybrid buildings not represented by standard typologies.
Key Innovation: Uses building-level ShakeMap demand and empirical logistic fragility functions to evaluate hybrid buildings as a candidate separate vulnerability group.
14. Bayesian updating of seismic fragility curves of the Irpinia 1980 building stock with machine-learning-supported data analysis
Core Problem: Deriving robust building fragility curves despite missing attributes and uncertain ground motion.
Key Innovation: Combines ML-supported data completion with Bayesian MCMC and a spatially correlated latent intensity field.
15. Seismic uplift fragility of scoured bridge pile-group foundations in cohesionless soils
Core Problem: Quantifying pile uplift risk for scoured bridge foundations during earthquakes.
Key Innovation: Builds probabilistic uplift demand, capacity, and fragility models across 49 bridge archetypes.
16. A new ground motion prediction model for Arias intensity and cumulative absolute velocity in the Japan Trench region based on HVSR site classification
Core Problem: Predicting offshore Arias intensity and cumulative absolute velocity in the Japan Trench.
Key Innovation: Uses a large S-net record set, HVSR site classes, installation effects, and spatial-correlation models.
17. Along-Strike Coupling Heterogeneity in Cascadia's Slow-Slip Zone Constrained by GNSS and Reduced-Order Rate-And-State Friction Modeling
Core Problem: Explaining along-strike differences in Cascadia slow-slip recurrence, moment, and unrecovered slip deficit.
Key Innovation: Integrates GNSS coupling inversion, Bayesian inference, physics-based rate-and-state modeling, and reduced-order acceleration to separate controls on recurrence and moment.
18. Early detection of slow-moving landslides in the Himalayas and Western Ghats using PSInSAR
Core Problem: Detecting pre-failure transitions in slow-moving and reactivated mountain slopes.
Key Innovation: Uses PSInSAR time series and a statistical PS-group transition framework, with rainfall-linked precursor periods.
19. From Manual Construction to AI-Driven Scenario Emergence: Rethinking Catastrophe Risk Modeling
Core Problem: Testing whether AI weather forecasts can generate coherent extreme-weather scenarios at lower computational cost than conventional construction.
Key Innovation: Self-iterated AI weather forecasts that generate continuous global fields from which extremes emerge.
20. Climatology and trends of extreme precipitation in France: evaluation of an explicit-convection regional climate model
Core Problem: Evaluating long-term daily and hourly precipitation extremes and trends over France.
Key Innovation: Evaluates a 2.5-km convection-permitting regional climate model for daily extremes over 1959-2022 and hourly extremes over 1990-2022.
21. Pulsed dike emplacement and lava tube formation during the 1971 Etna eruption
Core Problem: Reconstructing feeder-dike emplacement and lava-tube formation during an Etna eruption.
Key Innovation: Combines field evidence, historical imagery, laser scanning and drone photogrammetry to reconstruct pulsed emplacement and quantify cavity geometry and volume.
22. Regional landslide early warning through machine-learning-based rainfall threshold and hydromorphological slope units
Core Problem: Title-level focus: developing regional early warning that combines rainfall thresholds with slope-unit representation.
Key Innovation: Title-signalled approach or contribution: The title proposes machine-learned rainfall thresholds coupled to hydromorphological slope units. Methods, data and results could not be assessed because no reliable abstract was available.
23. Influence of static block remobilization on DFN-based rockfall propagation modelling: the 2015 Manikaran case study, India
Core Problem: Title-level focus: determining how remobilization of static blocks affects discrete-fracture-network-based rockfall propagation.
Key Innovation: Title-signalled approach or contribution: The title indicates explicit incorporation of static-block remobilization into DFN propagation modeling in a documented case study. Methods, data and results could not be assessed because no reliable abstract was available.
24. Seismogenic sensitivity of the buried basement faults in the Ganga foreland basin and their Himalayan tectonic connection: A morpho-tectonic constraint
Core Problem: Title-level focus: constraining the seismic potential and Himalayan linkage of buried foreland-basin faults.
Key Innovation: Title-signalled approach or contribution: The title indicates a morphotectonic analysis of seismogenic basement structures. Methods, data and results could not be assessed because no reliable abstract was available.
25. P-FLOOD: a probabilistic framework for flood hazard mapping in levee-protected floodplains
Core Problem: Title-level focus: mapping probabilistic flood hazard in levee-protected floodplains.
Key Innovation: Title-signalled approach or contribution: The title presents a dedicated probabilistic framework for residual flood hazard behind levees. Methods, data and results could not be assessed because no reliable abstract was available.
26. Response-pattern-aware graph neural network for rapid compound rainfall-tide urban drainage prediction
Core Problem: Title-level focus: rapidly predicting urban drainage responses under compound rainfall and tide forcing.
Key Innovation: Title-signalled approach or contribution: The title proposes a response-pattern-aware graph neural network for operationally fast prediction. Methods, data and results could not be assessed because no reliable abstract was available.
27. The progressive failure behavior of tunnel lining structure under strike-slip faulting: Insights from mechanical analysis and model test
Core Problem: Title-level focus: understanding progressive tunnel-lining failure under strike-slip faulting.
Key Innovation: Title-signalled approach or contribution: The title indicates combined mechanical analysis and physical model testing of fault-induced damage. Methods, data and results could not be assessed because no reliable abstract was available.
28. From simulation to urban action: evaluating a decision-support platform for tsunami-resilient coastal cities
Core Problem: Title-level focus: evaluating decision support for tsunami-resilient coastal-city planning.
Key Innovation: Title-signalled approach or contribution: The title indicates an evaluated platform linking simulation outputs to urban action. Methods, data and results could not be assessed because no reliable abstract was available.
29. Future changes and drivers of socioeconomic exposure to extreme precipitation in the Huai River Basin
Core Problem: Title-level focus: assessing future changes and drivers of socioeconomic exposure to extreme precipitation.
Key Innovation: Title-signalled approach or contribution: The title indicates a forward-looking driver analysis for basin-scale exposure. Methods, data and results could not be assessed because no reliable abstract was available.
30. Water Transport in MICP and Vegetation Synergistic Slope Protection Under Varying Cementation Solution Concentrations
Core Problem: Title-level focus: understanding water transport in combined MICP–vegetation slope protection under varying treatment concentrations.
Key Innovation: Title-signalled approach or contribution: The title identifies a coupled biological slope-protection approach and a controlled cementation-concentration comparison. Methods, data and results could not be assessed because no reliable abstract was available.
31. A Deep Reinforcement Learning Framework for Coastal Multi-Hazard Emergency Response
Core Problem: Title-level focus: supporting emergency-response decisions under interacting coastal hazards.
Key Innovation: Title-signalled approach or contribution: The title proposes a deep reinforcement learning framework for coastal multi-hazard response. Methods, data and results could not be assessed because no reliable abstract was available.
32. Supporting Natural Hazard Comprehension in Africa through Geovisualization: A Capacity-Building Case Study of a Summer School Training in Uganda
Core Problem: Title-level focus: improving understanding of natural hazards through geovisualization training.
Key Innovation: Title-signalled approach or contribution: The title presents a Uganda-based capacity-building case study focused on geovisual hazard communication. Methods, data and results could not be assessed because no reliable abstract was available.
33. A multi-level probabilistic propagation framework for functionality loss risk assessment of railway bridge networks under earthquakes
Core Problem: Title-level focus: assessing how earthquake impacts propagate into functionality loss across railway bridge networks.
Key Innovation: Title-signalled approach or contribution: The title proposes a multi-level probabilistic propagation framework at network scale. Methods, data and results could not be assessed because no reliable abstract was available.
34. A novel change homogenizing flood index and an expedited locally tuned Markov random field model for unsupervised rural floodwater detection in sentinel-1 data: An application to flood impact assessment
Core Problem: Title-level focus: detecting rural floodwater from Sentinel-1 without supervision while supporting impact assessment.
Key Innovation: Title-signalled approach or contribution: The title proposes a change-homogenizing flood index and an expedited locally tuned Markov random field model. Methods, data and results could not be assessed because no reliable abstract was available.
35. Functional data-based estimation of coastal erosion and accretion at Playa Salguero using J-Net Dynamic
Core Problem: Title-level focus: estimating coastal erosion and accretion at Playa Salguero.
Key Innovation: Title-signalled approach or contribution: The title indicates a functional-data approach using J-Net Dynamic. Methods, data and results could not be assessed because no reliable abstract was available.
36. Bluff erosion in an Arctic delta: A case study of the Coppermine delta, Nunavut
Core Problem: Title-level focus: characterizing bluff erosion in the Coppermine delta.
Key Innovation: Title-signalled approach or contribution: The title indicates a focused Arctic-delta erosion case study. Methods, data and results could not be assessed because no reliable abstract was available.
37. Improving flood forecasts: the combined impact of data assimilation and machine learning post-processing
Core Problem: Title-level focus: improving flood forecasts by combining state or observation updating with learned post-processing.
Key Innovation: Title-signalled approach or contribution: The title indicates joint use of data assimilation and machine-learning post-processing. Methods, data and results could not be assessed because no reliable abstract was available.
38. The impact of soil-structure interaction on the hybrid force-displacement seismic design method for steel moment-resisting frames
Core Problem: Title-level focus: determining how soil–structure interaction affects a hybrid seismic design method for steel frames.
Key Innovation: Title-signalled approach or contribution: The title evaluates soil–structure interaction within a combined force–displacement design framework. Methods, data and results could not be assessed because no reliable abstract was available.
39. Evaluation of high vibration mode and hybrid foundation reinforcement effects for monopile-supported offshore wind turbines under earthquake loading
Core Problem: Title-level focus: evaluating higher-mode response and reinforced-foundation effects for monopile wind turbines during earthquakes.
Key Innovation: Title-signalled approach or contribution: The title jointly considers high vibration modes and hybrid foundation reinforcement. Methods, data and results could not be assessed because no reliable abstract was available.
40. Seismic response of virgin and pre-shaken saturated sandy soil at identical relative density using shaking table tests
Core Problem: Title-level focus: comparing seismic response of previously shaken and virgin saturated sand at matched density.
Key Innovation: Title-signalled approach or contribution: The title indicates controlled shaking-table comparison that isolates loading history from relative density. Methods, data and results could not be assessed because no reliable abstract was available.
41. Hydrodynamic performance of TPMS structures as bioinspired submerged breakwaters for incident wave energy attenuation
Core Problem: Title-level focus: evaluating submerged breakwaters designed to attenuate incident wave energy.
Key Innovation: Title-signalled approach or contribution: The title applies bioinspired triply periodic minimal-surface structures to breakwater design. Methods, data and results could not be assessed because no reliable abstract was available.
42. Dynamic responses of rubble mound breakwaters reinforced with geogrid-wrapped slopes and sheet piles under sequential earthquake loading
Core Problem: Title-level focus: assessing reinforced rubble-mound breakwater response under sequential earthquakes.
Key Innovation: Title-signalled approach or contribution: The title combines geogrid-wrapped slopes, sheet piles, and repeated seismic loading. Methods, data and results could not be assessed because no reliable abstract was available.
43. Mechanisms of flood-induced sediment resuspension in a restored coastal lagoon: Implications for flood resilience
Core Problem: Title-level focus: explaining sediment resuspension during floods in a restored coastal lagoon.
Key Innovation: Title-signalled approach or contribution: The title links process mechanisms directly to flood-resilience implications in a restoration setting. Methods, data and results could not be assessed because no reliable abstract was available.
44. Probabilistic mapping of volcanic ash dispersion using MODIS and HYSPLIT to support aeronautical safety operations: Popocatépetl volcano airspace case study
Core Problem: Title-level focus: mapping volcanic-ash dispersion probabilistically to support airspace decisions around Popocatépetl.
Key Innovation: Title-signalled approach or contribution: The title integrates MODIS observations with HYSPLIT dispersion modeling in an operational case study. Methods, data and results could not be assessed because no reliable abstract was available.
45. Assessment of tropical cyclone vulnerability in East Africa coastal areas using a GIS-based Analytic Hierarchy Process: the case of Dar es Salaam Metropolitan City, Tanzania
Core Problem: Mapping urban coastal vulnerability from exposure, sensitivity, and adaptive capacity.
Key Innovation: Combines 17 indicators in a 30 m GIS-AHP framework to identify actionable vulnerability hotspots.
46. Pathway-dependent responses of temperature extremes under solar radiation modification over Northern Africa
Core Problem: Determine whether restoring regional mean temperature through solar-radiation modification also restores the structure of hot and cold extremes.
Key Innovation: Compares eight extreme-temperature indices across CMIP6 and GeoMIP6 pathways and resolves pathway-dependent changes in distribution tails, frequencies and day-night asymmetry.
47. Mechanical behavior analysis of defective pipelines in oblique slip fault zones based on experiments and simulation
Core Problem: Assessing buried pipeline response to coupled corrosion defects and oblique-slip fault displacement.
Key Innovation: Links two-box experiments with ABAQUS simulations to isolate defect-location, diameter, and burial-depth effects.
48. Seismic performance of corrosion-resistant concrete beam-column joints reinforced with hybrid FRP and stainless steel bars
Core Problem: Improving ductility and energy dissipation of corrosion-resistant beam-column joints.
Key Innovation: Tests hybrid GFRP and stainless-steel reinforcement and identifies replacement-rate control of failure mode.
49. HUMAID-NER: A Disaster Tweet Dataset for Joint Named Entity Recognition and Event Classification via Uncertainty-Weighted Multitask Learning
Core Problem: Extracting operational entities and event categories from disaster social-media streams.
Key Innovation: A 60,000-tweet NER dataset plus uncertainty-weighted joint entity and event classification.
50. IRENE: A Convolutional GRU Ensemble Model for Radar Precipitation Nowcasting over Italy
Core Problem: Improve calibrated high-resolution radar precipitation ensembles.
Key Innovation: ConvGRU ensemble with importance sampling, adversarial sharpness, and spectral constraints.
51. GeoCrossBench: Cross-Band Generalization for Remote Sensing
Core Problem: Evaluating whether Earth-observation models generalize to unseen spectral bands and satellites.
Key Innovation: GeoCrossBench plus a self-supervised band-agnostic baseline evaluated in a large controlled campaign.
52. Thermo-Hydro-Mechanical-Chemical Coupling in Reactive Fractured Rocks With Double Porosity
Core Problem: Represent coupled reaction, flow, heat, and deformation in dual-porosity rock.
Key Innovation: Thermodynamically derived THMC constitutive model with fracture-matrix exchange.
53. Persistent-Homology-Based Descriptors of Pore Heterogeneity for Generalized Permeability Prediction
Core Problem: Describe heterogeneous pore networks well enough to generalize permeability prediction.
Key Innovation: Persistent-homology descriptors weighted by functional connectivity.
54. Multi-Label Proportion Learning for Sea-Ice Type Prediction
Core Problem: Predict sea-ice composition from polygon-level ice charts.
Key Innovation: Multi-label proportion learning fuses SAR, microwave, and reanalysis data.
55. Adaptive and accuracy-aware multiple data assimilation in a three step framework
Core Problem: Balance accuracy and cost in ensemble smoother multiple assimilation.
Key Innovation: Three-step framework and accuracy-triggered catch-up mechanism.
56. VPRef: A Cross-Domain Benchmark for Referring Remote Sensing Image Segmentation
Core Problem: Benchmark referring segmentation under coupled visual and textual domain shift.
Key Innovation: VPRef dataset plus parameter-efficient SAM3 adaptation with pseudo-label and prompt mixing.
57. AsyncCouple-Flow: Asynchronous Cross-Modal Coupling and Flow Matching for Spatio-Temporal Forecasting
Core Problem: Forecast spatiotemporal systems without forced grid interpolation or complete sensors.
Key Innovation: Offset-aware coupling graph, token sparsification, and flow-matching forecast head.
58. G3AR: Graph-Guided Neural Visual Geometry for Scalable Multi-Sequence Aerial Registration
Core Problem: Scale dense neural geometry to irregularly overlapping aerial collections.
Key Innovation: Verified proximity graph guides chunks and Sim(3) registration topology.
59. HLC-GS: Risk-Map-Guided Height-Layer Consistency Gaussian Splatting for DSM Reconstruction from Optical Satellite Imagery
Core Problem: Prevent nonphysical height-layer mixing in satellite 3D Gaussian splatting.
Key Innovation: Risk maps regularize dominant layers and suppress unsupported secondary layers.
60. SPEAR NeXT Causal Latent Forecasting Across Multiple Horizons for Spectral Temporal Earth Representation Learning
Core Problem: Predict future EO latent states rather than reconstructing past observations.
Key Innovation: Compact optical-radar-environmental embeddings with causal multi-horizon Transformer forecasting.
61. From Foundation Embeddings to Cropland Maps: Label Efficiency, Temporal Transferability and Independent Human Validation
Core Problem: Map cropland accurately with frozen AlphaEarth representations and limited labels.
Key Innovation: Spatially separated evaluation, multi-year transfer, and blind human consensus.
62. A Self-Diagnosing Structural Error-Aware Parameter Estimation Method for Earth System Models
Core Problem: Estimating climate-model parameters despite structural error, sparse ensembles, and uncertain observations.
Key Innovation: Automated interpretable history matching that diagnoses structural inconsistency and excludes unsuitable variables.
63. Nationally Consistent, Locally Incomplete: A Bayesian Remote-Sensing Audit of Rooftop Photovoltaic Registries
Core Problem: Auditing incomplete rooftop photovoltaic registries from imperfect remote-sensing detections.
Key Innovation: A Bayesian correction framework that converts detector output into uncertainty-aware capacity estimates.
64. A global gridded dataset of significant wave height via fusion of multi-mission altimetry and numerical hindcast
Core Problem: Producing spatially complete, accurate global significant-wave-height fields.
Key Innovation: Fusion of multi-mission altimetry with numerical hindcasts, including a version aimed at AI training.
65. Multi-Scale Validation of Satellite-Based Precipitation Products and Their Impacts on Hydrological Simulation in a Humid Mountainous Basin
Core Problem: Quantifying precipitation-product uncertainty and its propagation into mountainous-basin runoff simulations.
Key Innovation: Multi-scale validation links four satellite products to gauges and hydrologic-model performance, including extremes.
66. Segmented Moho Uplift and Implications for Magmatic Underplating Beneath the Datong Basin
Core Problem: Resolving how upper-crustal faulting and lower-crustal magmatism shape Moho structure beneath an intracontinental rift.
Key Innovation: Uses dense-array receiver-function reverse-time migration and statistical spatial analysis to relate Moho gradients to seismicity and volcanic-area amplitudes.
67. Spatiotemporal Patterns of the Vegetation-Atmosphere Decoupling Coefficient Across the Contiguous United States
Core Problem: Mapping how vegetation–atmosphere coupling varies across ecosystems, climate gradients, and wet–dry conditions.
Key Innovation: Integrates 52 AmeriFlux sites, XGBoost, satellite data, and reanalysis to produce eight-day continental maps and characterize asymmetric drought responses.
68. Structure-preserving feature disentanglement with an adaptive transformer for unsupervised cross-domain object detection in heterogeneous SAR imagery
Core Problem: Title-level focus: enabling unsupervised cross-domain object detection across heterogeneous SAR imagery.
Key Innovation: Title-signalled approach or contribution: The title proposes structure-preserving feature disentanglement with an adaptive transformer. Methods, data and results could not be assessed because no reliable abstract was available.
69. Spot the differences: a registration-robust object-level change detection framework for remote sensing images
Core Problem: Title-level focus: performing object-level remote-sensing change detection robustly under registration error.
Key Innovation: Title-signalled approach or contribution: The title proposes a registration-robust object-level change-detection framework. Methods, data and results could not be assessed because no reliable abstract was available.
70. An RTE-guided mixture-of-experts framework for AMSR2 passive microwave soil moisture retrieval
Core Problem: Title-level focus: retrieving soil moisture from AMSR2 passive microwave observations.
Key Innovation: Title-signalled approach or contribution: The title proposes a radiative-transfer-equation-guided mixture-of-experts framework. Methods, data and results could not be assessed because no reliable abstract was available.
71. Automating glacier facies classification: Benchmark dataset and deep learning baseline from pan-European sample
Core Problem: Title-level focus: automating classification of glacier facies across a pan-European sample.
Key Innovation: Title-signalled approach or contribution: The title indicates both a benchmark dataset and a deep-learning baseline. Methods, data and results could not be assessed because no reliable abstract was available.
72. Inverse modeling for post-wildfire hydrologic process understanding over snow-dominated watersheds
Core Problem: Title-level focus: understanding post-wildfire hydrologic processes in snow-dominated watersheds.
Key Innovation: Title-signalled approach or contribution: The title indicates inverse modeling targeted at post-fire process attribution. Methods, data and results could not be assessed because no reliable abstract was available.
73. A dual post-processing framework for probabilistic streamflow forecasting using deep learning ensembles: Integrating residual correction and Vine Copula-based BMA
Core Problem: Title-level focus: post-processing ensemble streamflow forecasts to improve calibration and probabilistic reliability.
Key Innovation: Title-signalled approach or contribution: The title combines deep-learning residual correction with Vine-Copula-based Bayesian model averaging. Methods, data and results could not be assessed because no reliable abstract was available.
74. Improving and evaluating the performance of change detection methods for soil moisture retrieval using Sentinel-1A over a forest-dominated mountainous study area in Beijing, China
Core Problem: Title-level focus: improving and evaluating Sentinel-1A change-detection methods for soil-moisture retrieval in forested mountains.
Key Innovation: Title-signalled approach or contribution: The title indicates targeted method improvement and evaluation in a difficult forest-dominated mountain setting. Methods, data and results could not be assessed because no reliable abstract was available.
75. Shoreline reflection of free infragravity waves over smooth and rough beds: A SWASH-based numerical study
Core Problem: Title-level focus: determining how bed roughness affects shoreline reflection of infragravity waves.
Key Innovation: Title-signalled approach or contribution: The title indicates a SWASH-based comparison of smooth and rough beds. Methods, data and results could not be assessed because no reliable abstract was available.
76. Hillshade-Based Gated Matching Network for Robust Optical-DEM Cross-Modal Registration
Core Problem: Title-level focus: robustly registering optical images and DEMs despite cross-modal appearance differences.
Key Innovation: Title-signalled approach or contribution: The title uses hillshade and gated matching to bridge optical and elevation modalities. Methods, data and results could not be assessed because no reliable abstract was available.
77. Predicting Minnesota Lake Ice Phenology With Deep Learning, Explainable Methods, and a Physically Based Benchmark, 1980-2018
Core Problem: Reconstruct lake-specific ice phenology where observations are sparse.
Key Innovation: LSTM benchmarked against a physical model over 625 lakes and unseen lakes and years.
78. A panoramic aerodynamic performance prediction method for turbomachinery cascades using transformer-enhanced neural operator
Core Problem: Predict full turbomachinery flow fields and downstream performance.
Key Innovation: Transformer-enhanced neural operator approximates CFD at four-orders lower cost.
79. Drift Field Net: Learning Ocean Lagrangian advection fields from in-situ and satellite observations
Core Problem: Predict ocean surface advection and particle trajectories.
Key Innovation: Simulation pretraining plus advection-consistent Lagrangian fine-tuning.
80. Robust Fault Detection in Mechanical Multimodal Time Series via Self-Supervised Cross-Modal Reconstruction
Core Problem: Detect faults under heterogeneous sampling and deployment shift.
Key Innovation: Self-supervised cross-modal reconstruction without strict temporal alignment.
81. Generative models for simulation based filtering: Formulations and Empirical Comparisons
Core Problem: Compare generative transports for nonlinear Bayesian filtering.
Key Innovation: Unified derivation of interpolant, flow-matching, and Schrödinger-bridge filters.
82. Recovering Physical Parameters from Fragmented Observations via Exact Distributed Spline Merging
Core Problem: Recover governing parameters from distributed spatiotemporal measurements.
Key Innovation: Exact fixed-basis spline-statistic merging with derivative regression.
83. Stable by Construction: Variational Latent Markov Operators for Long-Horizon PDE Prediction
Core Problem: Prevent autoregressive error accumulation in physical-field forecasts.
Key Innovation: Variational latent Markov operator with structured probabilistic transitions.
84. Can Deep Learning Achieve Cross-Physics Mapping?
Core Problem: Learn operators between systems governed by different PDE classes.
Key Innovation: Dimensionless latent alignment tested bidirectionally across diffusion and wave fields.
85. GloSVeT: a global 0.05° monthly mean surface soil and vegetation component temperature dataset (2003-2023)
Core Problem: Separating global surface soil and vegetation temperatures over two decades.
Key Innovation: A global 0.05-degree monthly component-temperature dataset for 2003–2023.
86. Failure mechanism and support strategy of weak carbonaceous slate in a large-span underground powerhouse cavern: A case study
Core Problem: Title-level focus: diagnosing failure and support needs in weak slate around a large cavern.
Key Innovation: Title-signalled approach or contribution: The title indicates a mechanism-based case study linking failure diagnosis to support design. Methods, data and results could not be assessed because no reliable abstract was available.
87. Multiscale pore-network modeling of MICP grouting in heterogeneous soils
Core Problem: Predicting MICP grouting in heterogeneous soil from pore to specimen scale.
Key Innovation: A coupled pore-network model resolves microbial kinetics, reactive transport, precipitation, and evolving flow paths.
88. Intercomparison and evaluation of global land surface phenology (LSP) products
Core Problem: Title-level focus: comparing and evaluating global land-surface phenology products.
Key Innovation: Title-signalled approach or contribution: The title indicates a global product intercomparison and evaluation effort. Methods, data and results could not be assessed because no reliable abstract was available.
89. A Darcy-Fw hydro-thermal-mechanical model for freezing-suction-driven frost-heave cracking of U-shaped canals
Core Problem: Title-level focus: modeling frost-heave cracking of canals driven by freezing suction.
Key Innovation: Title-signalled approach or contribution: The title proposes a coupled hydro-thermal-mechanical Darcy-Fw model. Methods, data and results could not be assessed because no reliable abstract was available.
90. Effects of rock-bridge inclination angle on dynamic fracture behavior of frozen fissured sandstone
Core Problem: Title-level focus: determining how rock-bridge inclination affects dynamic fracture in frozen fissured sandstone.
Key Innovation: Title-signalled approach or contribution: The title isolates inclination angle as a control on dynamic fracture behavior under frozen conditions. Methods, data and results could not be assessed because no reliable abstract was available.
91. Stress wave propagation characteristics in layered rock masses with filled joints considering in-situ stress gradients and multiple reflections
Core Problem: Title-level focus: modeling stress-wave propagation through layered, filled-joint rock under stress gradients and repeated reflections.
Key Innovation: Title-signalled approach or contribution: The title integrates in-situ stress gradients and multiple reflections in a jointed-rock propagation analysis. Methods, data and results could not be assessed because no reliable abstract was available.
92. Improved ResNet-based identification of ballastless track subgrade settlement using vehicle vibration responses
Core Problem: Title-level focus: identifying track-subgrade settlement from vehicle vibration responses.
Key Innovation: Title-signalled approach or contribution: The title proposes an improved ResNet-based identification method using operational vibration data. Methods, data and results could not be assessed because no reliable abstract was available.
93. Orientation-dependent frost heave and thaw settlement of anisotropic silty clay under freeze-thaw: Experimental investigation and deformation modeling
Core Problem: Title-level focus: quantifying orientation-dependent frost heave and thaw settlement in anisotropic silty clay.
Key Innovation: Title-signalled approach or contribution: The title combines controlled experiments with deformation modeling and explicitly treats anisotropy. Methods, data and results could not be assessed because no reliable abstract was available.
94. Underwater crack detection method for bridges based on multi-beam sonar and PIST
Core Problem: Title-level focus: detecting underwater cracks in bridge components.
Key Innovation: Title-signalled approach or contribution: The title combines multibeam sonar with PIST for underwater defect detection. Methods, data and results could not be assessed because no reliable abstract was available.
95. A data-driven framework for stochastic wind field simulation incorporating higher-order statistics and correlation features
Core Problem: Title-level focus: simulating stochastic wind fields while retaining non-Gaussian statistics and spatial or temporal correlation.
Key Innovation: Title-signalled approach or contribution: The title explicitly incorporates higher-order statistics and correlation features in a data-driven framework. Methods, data and results could not be assessed because no reliable abstract was available.
96. A Geospatial-Spectral Image Fusion Framework Guided by Deep Reinforcement Learning-Selected Key Information for Forest Aboveground Carbon Stock Estimation
Core Problem: Title-level focus: estimating forest aboveground carbon stock through geospatial–spectral image fusion.
Key Innovation: Title-signalled approach or contribution: The title uses deep reinforcement learning to select key information guiding fusion. Methods, data and results could not be assessed because no reliable abstract was available.
97. DBNINet: A Decoder-Guided Boundary Enhancement and Node Interaction Network for Road Extraction from Remote Sensing Images
Core Problem: Title-level focus: extracting roads accurately from remotely sensed imagery, including boundary structure.
Key Innovation: Title-signalled approach or contribution: The title proposes decoder-guided boundary enhancement and node interaction. Methods, data and results could not be assessed because no reliable abstract was available.
98. Semantic-Structure Guided Mamba for Hyperspectral and Multispectral Image Fusion
Core Problem: Title-level focus: fusing hyperspectral and multispectral imagery while preserving semantic structure.
Key Innovation: Title-signalled approach or contribution: The title proposes semantic-structure guidance within a Mamba-based fusion model. Methods, data and results could not be assessed because no reliable abstract was available.
99. Evaporation Dynamics From Flowing Water Surfaces
Core Problem: Quantify evaporation from turbulent flowing water surfaces.
Key Innovation: Controlled flume isolates nonlinear effects of turbulence, radiation, and wind.
100. The record is part of the task: matched-record evaluation of text classifiers across maintenance, safety and recall reporting
Core Problem: Measure how record choice changes classifier conclusions for the same case.
Key Innovation: Matched-record evaluation across maintenance, aviation safety, and recalls.
101. Scaling Laws for Physics-Aware ACOPF Surrogate Learning
Core Problem: Understand accuracy-feasibility tradeoffs as surrogate size scales.
Key Innovation: Scaling laws compare MSE and augmented-Lagrangian training.
102. Racing in Volume with Flow Ensembles
Core Problem: Stream reconstructions of fast outdoor subjects from sparse fixed cameras.
Key Innovation: Uncertainty-weighted fusion of matches, tracks, and optical flow in 3DGS.
103. Bounded Adjustment with Reliability-Guided Embedding for Imbalanced Learning with Noisy Labels
Core Problem: Learn balanced classifiers under mislabeled minority examples.
Key Innovation: Bounded prior-adjusted score with reliability-guided angular geometry.
104. GraLoD: Graphics-Inspired Continuous Level-of-Detail Learning for Image Restoration
Core Problem: Avoid fixed restoration scales across content, stage, and degradation.
Key Innovation: Continuous level-of-detail fields query neighboring encoder scales.
105. Can Knowledge Transfer Parameters Be Learned? LePoKet for Efficient Robotic Vision
Core Problem: Transfer pretrained geometry to compact motion estimation.
Key Innovation: Learnable structural parent-child transfer improves RAFT optical flow.
106. Bridging the Perceptual Gap: Residual-Enhanced Downscaling and Manifold-Aware Perception Alignment Adaptation for NR-IQA
Core Problem: Expose perceptual degradation hidden by semantic CLIP features.
Key Innovation: Manifold adapter plus residual-enhanced downscaling.
107. Efficient Quantization-Aware Distillation with Cross-Modal Alignment for Edge Vision-Language Models
Core Problem: Compress open-vocabulary multimodal models without modality imbalance.
Key Innovation: Unified teacher-anchored quantization and RGB-guided non-RGB adapter.
108. Continuous-Time Machine Learning: A Unified Mathematical Perspective
Core Problem: Unify mathematical families of continuous-time machine learning.
Key Innovation: Taxonomy links vector fields, stochasticity, memory, and discretization with controlled benchmarks.
109. FSANet: Frequency-Spatial Aware Network for Image Segmentation
Core Problem: Segment difficult scenes under noise, poor light, and occlusion.
Key Innovation: Dual-domain awareness, structural priors, edge estimation, and a new stress-test dataset.
110. PSMP-CLIP: Patch-Prompt SAM and Multi-Semantic Prompting for CLIP-Based Zero-Shot Anomaly Detection
Core Problem: Detect anomalies without target-domain training samples.
Key Innovation: Patch-prompt SAM2 and multi-semantic CLIP prompt regularization.
111. Noise2Noise Revisited: Training Pair Distributions Dominate Loss Choice in Self-Supervised Denoising
Core Problem: Determine what drives Noise2Noise generalization.
Key Innovation: Controlled analysis separates loss effects from training-pair distribution.
112. Hyper-RED: Scalable Event Pre-training via Semantic Hypergraph Distillation
Core Problem: Scale event representation learning despite scarce labels and modality mismatch.
Key Innovation: Hypergraph relational distillation from pretrained image features.
113. Accelerated Decoding of Centroid Positional Encoding for Instance Segmentation
Core Problem: Remove post-inference bottlenecks in centroid-encoded instance segmentation.
Key Innovation: Parallel CUDA decoder tailored to positional embeddings.
114. PiPS: Post-Hoc Prototypical Explanations for Interpretable Semantic Segmentation
Core Problem: Explain dense segmentation decisions while preserving model performance.
Key Innovation: Model-agnostic spatial prototypes extracted post hoc.
115. Evaluating Mesh Reconstruction Methods for Crop Phenotyping
Core Problem: Evaluate mesh reconstruction fidelity for remotely observed crops.
Key Innovation: Controlled seven-pipeline comparison across geometric and perceptual metrics.
116. sensVLA: Spatially-Grounded Vision-Language-Action Model for Autonomous Wheel Loader
Core Problem: Control heavy equipment from language, RGB, proprioception, and 3D geometry.
Key Innovation: Separate BEV cross-attention path into a flow-matching action expert.
117. Bi-FlowGS: Bridging Generative View Completion and Gaussian Geometry through Bidirectional Flow Co-Refinement
Core Problem: Prevent photorealistic rendering from hiding erroneous 3D Gaussian geometry.
Key Innovation: Bidirectional flow exchange between video completion and Gaussian geometry.
118. High-Fidelity Digital Twin Data Models by Randomized Dynamic Mode Decomposition and Deep Learning with Applications in Fluid Dynamics
Core Problem: Build nonintrusive high-fidelity reduced models from simulation outputs.
Key Innovation: Randomized dynamic mode decomposition combined with deep learning.
119. Event-based Selective Attention for Multi-resolution Fast Region of Interest (ROI) Detection
Core Problem: Detect salient regions directly from aggressively downscaled event streams.
Key Innovation: Training-free multiscale bottom-up attention.
120. ResLRP: The Role of Residual Cancellation in Attribution Instability in Vision Transformers
Core Problem: Prevent residual-path cancellation from destabilizing ViT explanations.
Key Innovation: Conservative residual-aware LRP with causal validation.
121. Neural Field Ensembles for Aerodynamic Surface Prediction: Winning Solution to the ONERA CRM Wall Distribution 2025 Challenge
Core Problem: Predict surface physical fields without expensive CFD.
Key Innovation: Fourier-encoded conditional neural-field ensemble with extensive ablation.
122. A unified framework for global and local interpretability using adaptive derivative-ordered random explanation
Core Problem: Explain nonlinear feature and sample interactions efficiently.
Key Innovation: First- and second-derivative attribution with randomized SVD and sparsity.
123. DecoGS: Adaptive Static-Dynamic Decoupling of 3D Gaussians for Free-Viewpoint Video Streaming
Core Problem: Efficient online 3D reconstruction from streaming video without destabilizing static regions.
Key Innovation: Adaptive static-dynamic Gaussian updates, gradient gating, and visibility filtering.
124. Exploring 2D backbone effects for indoor semantic occupancy prediction
Core Problem: Selecting image backbones for voxel-level semantic occupancy prediction.
Key Innovation: A controlled backbone comparison showing DINOv2 and BLIP2 dominate common CLIP variants.
125. Quantum-Inspired Trainable and Parameter-Efficient Tensor Networks for Image Inpainting
Core Problem: Efficient reconstruction of images observed through fixed sampling masks.
Key Innovation: Trainable invertible quantum-inspired tensor-network transforms with low parameter count.
126. PanoGS-SLAM: Panoramic 3D Gaussian Splatting SLAM
Core Problem: Robust real-time dense mapping with panoramic cameras under rapid motion.
Key Innovation: Spherical-domain 3D Gaussian splatting SLAM with omnidirectional photometric constraints.
127. SSC-Priors: Exploring Semantic and Visibility Priors to Boost Lidar Semantic Scene Completion
Core Problem: Improving LiDAR semantic scene completion without heavy architectural redesign.
Key Innovation: Semantic pseudo-label and sensor-visibility priors supplied directly to existing completion networks.
128. ORCA: Occlusion-Aware Refinement and Completion for Novel View Synthesis
Core Problem: Completing newly exposed regions during single-image novel-view synthesis.
Key Innovation: Occlusion-aware Gaussian-anchor reconstruction that distinguishes small disocclusions from regions needing generation.
129. Det-LIME: Detector-Aware, Multi-Instance Local Interpretable Model-Agnostic Explanations for Automated Marine Mammal Detection
Core Problem: Explaining multi-instance marine-mammal detections in ecological imagery.
Key Innovation: Detector-aware LIME with instance-specific, box-aligned explanations.
130. ENCP: Episode-Normalized Conformal Prediction for Vision-and-Language Navigation
Core Problem: Obtaining valid uncertainty sets over dependent, variable-length navigation episodes.
Key Innovation: Episode-normalized conformal prediction calibrated on a per-episode maximum score.
131. 3D Field Data Reduction with Adaptive Sample-Based Gaussian-Encoded Reconstruction
Core Problem: Fixed-budget reduction of structured, unstructured, and particle-based 3D fields.
Key Innovation: A unified adaptive sample-based Gaussian encoding across multiple field-data formats.
132. Computer-assisted global regularity across nonlinear families of three-dimensional periodic Navier-Stokes flows
Core Problem: Certifying global regularity across continuous families of 3D periodic Navier-Stokes flows.
Key Innovation: Computer-assisted bounds covering parameter families and infinitely many perturbation modes.
133. GPEvac: GNN-Based PPO for Adaptive Evacuation Routing During Shooting Events
Core Problem: Adaptive evacuation routing under moving threats and crowding.
Key Innovation: A graph-neural PPO policy with global message passing that generalizes across building layouts.
134. Geometry vs Structure: Graph-Based Diagnostics for LiDAR Point-Cloud Simulation Fidelity
Core Problem: Measuring structural fidelity of simulated LiDAR point clouds against real scans.
Key Innovation: Graph communities and topology-aware matching complement conventional geometric distance metrics.
135. Physics Informed Random Feature Neural Networks for Solving PDEs
Core Problem: Reducing spectral bias and collocation cost in physics-informed PDE solution.
Key Innovation: Random-feature neural PDE solvers with approximation analysis and high-probability error bounds.
136. The Neverwhere Visual Parkour Benchmark Suite
Core Problem: Reproducible evaluation of visual locomotion in realistic complex environments.
Key Innovation: A suite of more than sixty 3D Gaussian-splat urban scenes with closed-loop evaluation.
137. Structure-Driven Inversion: A New Paradigm for Solving Inverse Problems
Core Problem: Solving inverse problems without relying solely on iterative objective minimization.
Key Innovation: A structure-driven inversion paradigm using mathematical pseudo-inverses and physics-based projections.
138. Continual Learning for Traversability Prediction with Uncertainty-Aware Adaptation
Core Problem: Continual traversability prediction across novel terrains without catastrophic forgetting.
Key Innovation: Uncertainty-aware generative experience recall that avoids storing past terrain data.
139. HuMemSLAM: Efficient Human-Inspired Semantic Place Recognition for Robust Visual SLAM
Core Problem: Maintaining visual place recognition under perceptual aliasing and appearance change.
Key Innovation: Human-inspired semantic place recognition integrated with ORB-SLAM3 at low latency.
140. Unlocking Zero-shot Potential of Semi-dense Image Matching via Gaussian Splatting
Core Problem: Training robust semi-dense image matchers under extreme viewpoint change.
Key Innovation: Geometrically corrected 3D Gaussian splats generate precise correspondences and align 2D and 3D representations.
141. Multi-View Foundation Models
Core Problem: Making foundation-model features consistent across views of the same 3D scene.
Key Innovation: Intermediate 3D-aware attention converts single-view encoders into multi-view foundation models.
142. CFGPNet: Cross-Attention-Based Fused Gradient Programmed Network Framework for Multispectral Object Detection
Core Problem: Accurate, efficient object detection by fusing visible and thermal imagery.
Key Innovation: Cross-scale attention exchange, selective aggregation, and removable gradient supervision.
143. Nonnegative matrix factorizations and related compositional models: Equivalence, identifiability, and an application on the grain-size analysis of sediments
Core Problem: Clarifying equivalence and identifiability among compositional matrix-factorization models.
Key Innovation: Proofs linking NMF, end-member analysis, latent-class models, and an application to sediment grain sizes.
144. A global urban built-up area dataset for cities with populations exceeding 300,000 (2000-2025)
Core Problem: Creating comparable long-term built-up-area maps for large cities worldwide.
Key Innovation: A unified urban-settlement definition applied to satellite and population data for 1,611 cities.
145. Assessing the stability of LSTM runoff projections in Switzerland under climate scenarios
Core Problem: Testing whether observation-trained LSTMs remain credible under future climate scenarios.
Key Innovation: Cross-catchment evaluation of runoff shifts and limitations for extremes in Switzerland.
146. Setting the bar: benchmarks for model performances in large-sample hydrology
Core Problem: Defining fair performance expectations for river-flow models across diverse catchments.
Key Innovation: Empirical lower and upper benchmarks that account for local limits on achievable skill.
147. A global decline in atmospheric dust during the 21st century
Core Problem: Reconstructing recent changes in the global atmospheric dust cycle.
Key Innovation: A multi-product inverse framework integrates satellite sensors, reanalyses, and AERONET.
148. Sea Surface Current Vector Reconstruction from Multitemporal Sentinel-1 Doppler Observations: A Trajectory-Crossing Approach with Spatial Registration
Core Problem: Reconstructing sea-surface current vectors from ascending and descending Sentinel-1 Doppler observations acquired at different times without spatial-mismatch bias.
Key Innovation: Registers multitemporal radial-current gradient fields with maximum cross-correlation before trajectory-crossing reconstruction and evaluates them against CMEMS, SWOT and drifter observations.
149. Unsupervised Scale-Conditioned Hyperspectral and Multispectral Image Fusion via a Frequency-Spatial Dual-Domain Network
Core Problem: Fusing hyperspectral and multispectral imagery at fractional and variable spatial scales.
Key Innovation: An unsupervised scale-conditioned dual-domain network works on native grids with learned physical degradation priors.
150. Retrieval of Optically Active and Inactive Water Quality Parameters Using Remote Sensing and Machine Learning: Evidence from Water Hyacinth-Infested Lake Tana, Ethiopia
Core Problem: Retrieving optically active and inactive water-quality parameters across a data-scarce, water-hyacinth-affected lake.
Key Innovation: Integrates 858 in-situ observations, Sentinel-2 and four machine-learning models to map four water-quality variables and their seasonal and spatial hotspots.
151. Class-Specific Interpretation and Validation of Optical and SAR Sensor Contributions in Multi-Sensor Land-Cover Classification Using SHAP and Leave-One-Sensor-Out Ablation
Core Problem: Interpreting class-specific contributions of optical and SAR sensors in land-cover classification.
Key Innovation: Combines SHAP, permutation importance, feature correlation, and leave-one-sensor-out ablation.
152. Regional Marine Gravity Field Refinement by Integrating Region-Adaptive Fusion and Multi-Relational Graph Residual Learning
Core Problem: Refining regional marine gravity fields with heterogeneous observations and spatial structure.
Key Innovation: Combines region-adaptive background fusion with multi-relational graph residual learning and spatial holdouts.
153. Gas Plume Detection from Infrared Hyperspectral Remote Sensing Data Based on Deep Learning Algorithms
Core Problem: Identifying coexisting gas species pixelwise from noisy 121-band infrared hyperspectral observations.
Key Innovation: Uses a compact spectral 1D-CNN across simulated noise conditions and measured scenes, benchmarked against traditional detectors and a spectral Transformer, as a fixed-platform proof of concept.
154. Dynamic Reconstruction of Vegetation Earth Observation Time Series: Beyond Gap-Filling in Level-3 Products
Core Problem: Reconstructing fast, nonlinear vegetation-observation trajectories from sparse EO sampling without smoothing away short-lived dynamics.
Key Innovation: Synthesizes dynamic reconstruction that conditions Level-3 trajectories on multi-sensor observations, meteorology, spatial context and model priors while exposing uncertainty and scale trade-offs.
155. Stochastic Analysis of Geocell Reinforced Pile Supported Embankments Using Karhunen-Loève Expansion
Core Problem: Quantifying how spatially variable, cross-correlated soil parameters affect settlement and stress reliability in geocell-reinforced pile-supported embankments.
Key Innovation: Couples three-dimensional Karhunen-Loève random fields with Abaqus and Monte Carlo analysis to resolve geocell-induced changes in mean response, variability and governing parameter regime.
156. Advancing ethical and relational disaster risk reduction
Core Problem: Title-level focus: studies advancing ethical and relational disaster risk reduction.
Key Innovation: Title-signalled approach or contribution: Provides an adjacent method or empirical result that may inform selected geohazard workflows. Methods, data and results could not be assessed because no reliable abstract was available.
157. Effects of field-aged biochar on rainfall-induced erosion and soil nutrient spatial redistribution in cold-region black soil
Core Problem: Title-level focus: assessing how field-aged biochar affects rainfall erosion and nutrient redistribution in cold-region black soil.
Key Innovation: Title-signalled approach or contribution: The title indicates an erosion experiment, but the supplied abstract is an unrelated general soil-ecosystem review. Methods, data and results could not be assessed because no reliable abstract was available.
158. Sediment transport capacity equation based on effective suspension number partitioning
Core Problem: Title-level focus: studies sediment transport capacity equation based on effective suspension number partitioning.
Key Innovation: Title-signalled approach or contribution: Provides an adjacent method or empirical result that may inform selected geohazard workflows. Methods, data and results could not be assessed because no reliable abstract was available.
159. Investigating streamflow dynamics in heterogeneous catchments through sensitivity function analysis
Core Problem: Determine how widely available hydroclimatic records and catchment descriptors can group hydrologically similar catchments.
Key Innovation: Combines sensitivity-function-derived response signatures from 280 catchments with six-dimensional characterization and Bayesian clustering into nine hydrologically interpretable classes.
160. Monthly dynamics of global surface water from 2015 to 2023
Core Problem: Title-level focus: mapping monthly global surface-water dynamics over 2015–2023.
Key Innovation: Title-signalled approach or contribution: The title indicates a global, temporally resolved surface-water product or analysis. Methods, data and results could not be assessed because no reliable abstract was available.
161. Characteristics and microscopic mechanisms of mud pumping in ballasted railway subgrades under hydro-thermal-mechanical coupling
Core Problem: Title-level focus: studies characteristics and microscopic mechanisms of mud pumping in ballasted railway subgrades under hydro–thermal–mechanical coupling.
Key Innovation: Title-signalled approach or contribution: Provides an adjacent method or empirical result that may inform selected geohazard workflows. Methods, data and results could not be assessed because no reliable abstract was available.
162. Geometry of Breaking Waves Under Natural Sea Conditions
Core Problem: Characterizing breaking-wave geometry under variable real-world sea states.
Key Innovation: Analyzes 16,369 stereo-imaged breaking waves and shows that geometric scaling varies with sea state.
163. Scouring process in river sinuosities: Insights from a large field dataset from south-east France
Core Problem: Title-level focus: characterizing scour processes associated with river sinuosity.
Key Innovation: Title-signalled approach or contribution: The title indicates evidence from a large field dataset rather than a purely conceptual or small-scale study. Methods, data and results could not be assessed because no reliable abstract was available.
164. A Cummins-equation-based physics-informed neural network for response prediction and system identification of wave energy converters
Core Problem: Title-level focus: predicting response and identifying wave-energy-converter systems from physical and observational constraints.
Key Innovation: Title-signalled approach or contribution: The title embeds the Cummins equation in a physics-informed neural network. Methods, data and results could not be assessed because no reliable abstract was available.
165. A physics-guided graph neural network framework for prediction of offshore wind farm power and fatigue loads
Core Problem: Title-level focus: predicting wind-farm power and fatigue loads across interacting offshore turbines.
Key Innovation: Title-signalled approach or contribution: The title combines physical guidance with a graph neural network for farm-scale prediction. Methods, data and results could not be assessed because no reliable abstract was available.
166. Physics-guided machine learning with sparse data for load-bearing capacity prediction of grouted connections in offshore wind turbine structures
Core Problem: Title-level focus: predicting grouted-connection load capacity when training data are sparse.
Key Innovation: Title-signalled approach or contribution: The title proposes physics-guided machine learning for a data-limited structural mechanics task. Methods, data and results could not be assessed because no reliable abstract was available.