TerraMosaic Daily Digest: October 19, 2026
Daily Summary
Video precursors, omission costs and inventory age address different decisions in landslide assessment. In two open-pit mine demonstrations, changes in rockfall activity precede rockslides by 21 and 91 seconds. In Deccan Basalt terrain, susceptibility models evaluated with spatial blocks and terrain-representative comparison samples achieve AUC values of about 0.73–0.75, substantially below the random pixel-split scores. A Wanzhou inventory of 1,518 landslides over 70 years shows how the period used to train a map changes its performance in later decades. These studies connect imminent-failure recognition, map classification and the duration of map usefulness to distinct forms of evidence.
Regional observations add a physical and spatial context to those decisions. Gimigliano InSAR time series resolve four kinematic domains with different rainfall-associated responses. Southern California diagnostics find only marginal gains from Random Forest over established manual rainfall thresholds under limited event data. Post-fire mapping across Oregon watersheds benefits from multi-epoch terrain information but underclassifies transport zones. At Mt. Umyeon, physics-derived daily labels supplement an undated inventory; the reported annual-peak AUC measures spatial discrimination, leaving event-time warning performance untested in the available evidence.
Infrastructure studies show why failure mechanisms and observation windows matter when interpreting response indicators. Liquefaction tests record reduced horizontal dam-crest acceleration alongside damaging deformation and vertical response. Tailings-dam experiments trace combined shear and interface sliding, while tunnel-face tests isolate excessive-pressure blowout under shallow dry-sand conditions. Excavation reliability assessments also change when adjacent-structure performance is included. The arch-dam monitoring archive presents a different limit: missing flood-peak observations prevent diagnosis of the peak response.
Key Trends
This issue concentrates on how observations, validation design and failure mechanisms change the interpretation of warning and risk indicators.
- Warning claims need event timing and error accounting: Video rockfall monitoring provides two field demonstrations, while undated-inventory soft labels support a different kind of model assessment. Susceptibility, precursor detection and verified temporal warning skill should remain separate quantities.
- Sampling and inventory age affect susceptibility decisions: The Deccan Basalt study exposes body-level leakage and background-sampling effects; Wanzhou shows loss of temporal transfer. DEM-resolution experiments add a separate warning that predictive gains and explanation reliability do not improve in lockstep.
- Terrain change and rainfall response retain spatial structure: Post-fire transport corridors remain harder to map than initiation and deposition zones. Gimigliano domains respond differently to antecedent rainfall, and Mumbai flood mapping benefits from hydrological predictors under a satellite-derived reference inventory.
- Ground and structure need to be evaluated together: Dam-liquefaction tests separate acceleration filtering from deformation. Tailings interfaces and tunnel-face pressure control specific failure mechanisms, while the excavation study shows how neighbouring structural performance changes an assessed reliability.
- Signals need observations and intervention before failure: The arch-dam record misses the flood peak; the landslide-dammed-lake case identifies a delay between recognized signals and stronger intervention. These are distinct technical and institutional gaps, linked by the need to act on evidence available before failure.
Selected Papers
Landslide warning, susceptibility validation and changing inventories lead this selection. The papers then follow rainfall and terrain response into flood mapping, wildfire, liquefaction, dams and underground stability. Supporting methods address radar geometry, precipitation evaluation, material properties and physical-model diagnostics.
1. From rockfall to landslide: early warning based on video monitoring in open-pit mines
Core Problem: Rockfall activity may provide observable precursors to larger rock-slope failure, but warning performance needs to be assessed at the event and deployment levels.
Key Innovation: A compilation of 36 documented rockslides motivates optical-flow and background-subtraction detection with STA/LTA frequency-change analysis. Two field demonstrations give lead times of 21 and 91 seconds. These cases support practical potential, without establishing a population-wide detection rate, false-alarm rate or sufficient evacuation time for every site.
2. Omission error in landslide susceptibility mapping: a cost-based decision framework for Deccan Basalt Terrain, India
Core Problem: High susceptibility scores can conceal missed landslide terrain when validation samples and classification thresholds differ from the intended decision setting.
Key Innovation: Using 98 landslide bodies, grouped and spatial validation expose leakage and sampling effects obscured by random pixel splits. With terrain-representative comparison samples, spatial AUC falls to roughly 0.73–0.75. Thresholds and assumed omission costs influence decisions more than small differences among the tested algorithms.
3. Shallow landslide diagnostics for Southern California
Core Problem: Regional warning needs credible rainfall and event records before more complex classifiers can be judged against established thresholds.
Key Innovation: Quality-controlled Southern California inventories are combined with the regionally calibrated MHEBC precipitation dataset. Random Forest variants improve only marginally on manually derived operational threshold procedures under the available limited data. The contribution is regional diagnostic evidence, not proof that machine learning generally replaces forecaster-built thresholds.
4. Same area, different timeframes: unveiling the evolution of landslide patterns and their impact on susceptibility maps
Core Problem: A susceptibility map trained on one historical inventory may lose relevance as landslide patterns and environmental conditions change.
Key Innovation: An inventory of 1,518 Wanzhou landslides spanning 70 years supports maps based on five inventory periods and validation in subsequent decades. Temporal baseline changes alter mapped susceptibility and model-derived factor importance. Shorter-baseline maps can perform better initially but lose predictive value rapidly; the result supports temporal re-evaluation rather than a universal inventory-window length.
5. Post-wildfire debris-flow hazard mapping using machine learning in mountainous watersheds
Core Problem: Debris-flow mapping after fire must resolve initiation, transport and deposition while testing transfer across watersheds.
Key Innovation: Seven Columbia River Gorge watersheds are examined with multi-epoch airborne lidar and vegetation features under spatially structured validation. Temporal differences improve classification, but transport zones remain underclassified. This is evidence for mapping debris-flow terrain, not independent validation of future trigger-time forecasts.
6. Activity and Rainfall Response of the Gimigliano Landslide (Southern Italy): Insights from SBAS-InSAR Data Analysis
Core Problem: Landslide-wide displacement summaries can obscure different rainfall responses among kinematic sectors of one unstable slope.
Key Innovation: Gimigliano SBAS-InSAR time series identify four domains that correspond to mapped geomorphology. Rainfall associations are strongest over antecedent accumulation windows with domain- and component-dependent delays. These correlations support interpretation of spatially heterogeneous activity; they are not universal rainfall-trigger thresholds or causal estimates.
7. DEM Resolution Drives AI Variability: Tradeoffs Between Performance and Explainability in Landslide Susceptibility Prediction
Core Problem: Finer terrain grids may improve classification without making model explanations equally reliable.
Key Innovation: A fixed slope-unit and label design compares DEM resolutions from 1 to 90 m across four classifiers. Predictive gains at finer resolution do not consistently improve SHAP faithfulness, stability or geomorphic consistency. This available abstract updates an earlier title-level entry; resolution is treated as an objective-dependent choice.
8. A rapid kinematic analysis method considering the spatial proximity and rock joint connectivity using non-contact measurement
Core Problem: Conventional kinematic tests can overcount possible joint intersections when discontinuity connectivity and spatial proximity are ignored.
Key Innovation: A point-cloud workflow reconstructs fragmented joint patches and screens planar, wedge and toppling configurations across three field datasets. Orientations agree with reference measurements within 5°. Spatial filtering reduces candidate intersections, but complete unstable-block delineation still requires hidden-joint geometry and site-specific mechanical evidence.
9. An extensive shaking table study of the progression of damage caused by liquefaction in a concrete-faced rockfill dam on deep overburden
Core Problem: Reduced horizontal acceleration above a liquefying foundation can coexist with substantial deformation and other damaging responses.
Key Innovation: Bidirectional shaking tests of a concrete-faced rockfill-dam model show loss of horizontal amplification and a shift to lower dominant frequencies as overburden liquefies. Vertical energy remains amplified and damage includes spreading and slab–rockfill separation. Lower crest acceleration is therefore not interpreted as evidence of improved overall seismic safety.
10. Decadal Creep Acceleration on the Northern Japan Trench Driven by Erosion of Heterogeneously Locked Asperities
Core Problem: Decadal changes in aseismic slip may reflect evolving locked asperities rather than a fixed interseismic locking pattern.
Key Innovation: A mechanically motivated inversion of 1998–2009 northern Japan Trench GNSS velocities fits shrinking effective asperities and slip accelerations consistent with repeating earthquakes. The shrinking areas are model-inferred, not direct observations of asperity boundaries, and the retrospective interpretation is not a validated earthquake-warning rule.
11. Antarctic marine-based ice destabilized when warming outpaces solid-Earth feedbacks
Core Problem: The pace of warming may matter when stabilizing bedrock uplift lags ice retreat.
Key Innovation: Coupled ice–Earth simulations identify rate-induced retreat in vulnerable East Antarctic basins, with thresholds sensitive to mantle viscosity. This is model evidence about long-term instability, not a dated prediction of collapse or coastal inundation.
12. Specific energy earthquake hazard of India for energy-based design of structures
Core Problem: Force-based hazard indicators do not fully describe cumulative energy demands relevant to structural design.
Key Innovation: An India-wide probabilistic assessment combines regional energy-based ground-motion models, site amplification and a logic tree to produce specific-energy hazard maps and design spectra. The outputs are design-oriented hazard estimates, not a validation that energy-based zoning alone reduces losses or predicts particular earthquakes.
13. Integration of time-series SBAS-InSAR and GACOS for monitoring and analysis of land surface deformation in Qingdao City from 2017 to 2024
Core Problem: Long-term deformation monitoring needs atmospheric correction and comparison with independent ground observations.
Key Innovation: A Qingdao Sentinel-1 time series with GACOS correction is compared with leveling, yielding reported RMSEs of 6–24 mm. The analysis identifies spatially concentrated deformation and associations with groundwater and development. These associations do not uniquely establish causation or remove the assumptions in interpreting motion from an ascending radar geometry.
14. Human accessibility and land-use interfaces shape wildfire ignition hazard across Mediterranean National Parks and surrounding landscapes in Spain
Core Problem: Ignition patterns around protected areas depend on accessibility, land-use interfaces and weather, which require spatially independent evaluation.
Key Innovation: A Spain park-and-buffer analysis uses nested spatial blocks and temporally stratified comparison locations, reporting ROC AUC of 0.88. Road and interface variables dominate fitted explanations. SHAP associations and conditional warming scenarios are not treated as causal effects or certain future ignition probabilities.
15. Fracture network evolution in room-and-pillar karst mining: quantifying geological controls and physics-based rapid assessment
Core Problem: Mining-induced fractures may connect to aquifers through geology-dependent paths that a single transferable threshold cannot represent.
Key Innovation: Laboratory inputs, staged simulations and a physics-guided surrogate examine six excavation configurations. Microseismic spatial agreement supports the fracture pattern, while permeability remains a proxy and constitutive choices affect response magnitude. The result supports comparative site screening rather than a universally validated coupled hydraulic-failure predictor.
16. Rethink of reliability assessment of deep excavation with the input of the performance of adjacent engineered structures
Core Problem: An excavation may satisfy its own performance criteria while placing nearby structures in unacceptable states.
Key Innovation: A framework links excavation and adjacent-structure performance through graph-based probabilistic models and Bayesian inference, demonstrated at a Xuzhou Metro project. The assessment can change substantially when structural performance is included; this is a model-conditional reliability comparison, not a directly measured reduction in failure probability.
17. Stability analysis of geotextile tube tailings dams under inclined stacking: composite failure mechanism and oblique horizontal slice method
Core Problem: Inclined tube interfaces can redirect failure surfaces in stacked geotextile tailings dams.
Key Innovation: Reduced-scale tests identify combined tailings shear and interface sliding. An interface-aligned slice formulation reproduces tested slip surfaces and safety factors, with outward inclination reducing calculated stability. The findings are tied to the tested configurations and do not establish field-wide safe inclination limits.
18. Experimental Investigation of Tunnel Face Blowout under Different Cover Ratios and Sand Densities
Core Problem: Excessive face support can cause passive blowout, requiring a different assessment from insufficient-support collapse.
Key Innovation: Three-dimensional half-tunnel tests and particle-image measurements track failure propagation toward the surface as sand density and cover vary. A wedge-and-block limit-equilibrium representation agrees reasonably with the shallow-cover dry-sand tests. Application to saturated or deeper ground remains unestablished by this experiment.
19. Fine-Scale River-Ice-Type Mapping During the Freeze-Up Period Using UAV Multispectral Imagery
Core Problem: Transparent ice and open water can look spectrally similar, limiting reliable mapping of heterogeneous freezing reaches.
Key Innovation: Multispectral UAV imagery and texture features map four surface classes across Yellow River reaches. Spatially independent tests include a separate complex-ice reach. These are ice-type mapping results that can support ice-jam assessment, not demonstrated forecasts of jam formation or flood timing.
20. Spatiotemporal probabilistic analysis for landslide early warning using physics-informed recurrent neural network model: case study Mt. Umyeon, Korea
Core Problem: An undated landslide inventory does not supply observed failure times for temporal warning validation.
Key Innovation: Physics-derived daily soft labels supplement a recurrent model, while the reported annual-peak validation AUC is 0.812 versus 0.794 for a conventional GRU. These are primarily spatial comparisons. Simulated daily labels do not replace dated failures, and independent event-year warning validation remains necessary.
21. Wetness-conditioned rainfall-trigger priors for probabilistic landslide forecasting
Core Problem: Antecedent wetness may alter how rainfall-trigger information should enter probabilistic forecasts.
Key Innovation: Title-level scope: wetness-conditioned rainfall-trigger priors. The conditioning design, event data, calibration and forecast improvement cannot be assessed from the available bibliographic record.
22. Seismic response, instability, and kinematics of catastrophic loess-sandstone landslide: a case study of the 1927 M8 Gulang earthquake on the NE Tibetan Plateau, China
Core Problem: Reconstructing a catastrophic landslide requires distinguishing seismic response, instability and post-failure movement.
Key Innovation: Title-level scope: the 1927 Gulang earthquake landslide case on the northeastern Tibetan Plateau. Material mechanisms, reconstructed motions and validation are not inferred from the title.
23. From near-miss to collapse: diagnosing temporal governance gaps in landslide-dammed lake risk management
Core Problem: Recognizing hazard signals does not ensure that structural intervention is activated before a dammed lake collapses.
Key Innovation: A Matai’an Creek study traces 97 risk-reduction actions across four phases using an escalation index. Monitoring and coordination dominate before collapse, while stronger intervention appears later. Documentary coding supports a temporal governance diagnosis, not a controlled causal estimate of which intervention would have prevented the disaster.
24. Hydrologically enhanced machine learning framework for urban flood inundation mapping using multi-sensor remote sensing data: a case study of Mumbai, India
Core Problem: Urban flood mapping needs hydrological context, while the construction of reference labels limits what accuracy metrics can establish.
Key Innovation: Mumbai models combine SAR, optical, terrain and rainfall inputs with relative-elevation and wetness predictors. The best reported AUC is 0.803 against a satellite-derived inventory. Predictor ablations support the hydrological features, while the available abstract does not document independent field verification or prospective operational forecasting.
25. Multisensor Change Detection in Flood Monitoring: A Reliability-Aware Dual-Regularized Fusion Network
Core Problem: Optical and SAR observations may have unequal reliability during floods, complicating multisensor change detection.
Key Innovation: A SAR-guided, dual-regularized fusion network addresses sensor imbalance and local feature inconsistency. A public dataset and four regional flood scenarios support evaluation. This abstract updates a previous title-only citation; benchmark evidence is not generalized to all sensors, flood types or operational deployments.
26. End-to-End Evolving Cross-Modal Graph Network for Unsupervised Multimodal Remote Sensing Change Detection
Core Problem: Fixed cross-modal graph structures may inadequately represent differences between heterogeneous remote-sensing observations.
Key Innovation: An evolving graph network learns relationships with reconstruction and cross-modal consistency losses, tested on four benchmarks and a Wangjiaba flood dataset. The current abstract establishes a specific flood application beyond the earlier title-only citation, without supplying a general operational flood-mapping guarantee.
27. Comparative Evaluation of SAW and AHP Approaches for GIS-Based Liquefaction Susceptibility Mapping in Chiang Rai, Thailand
Core Problem: Weighting choices should be compared against a transparent baseline and evaluated beyond a single inventory when possible.
Key Innovation: SAW and AHP maps use the same four factors and 51 documented occurrences from one Chiang Rai earthquake. Their primary AUC difference is not statistically significant. Holdouts strengthen within-event assessment, but do not establish transfer to independent earthquakes or prove expert weighting generally superior.
28. Advantages of converting old homogeneous earth dams into upstream-faced dam typologies
Core Problem: Retrofitting an older homogeneous dam requires assessing how seepage controls interact with earthquake loading.
Key Innovation: A case-based analysis tests upstream waterproofing and a foundation diaphragm as a rehabilitation option. Calculated performance depends strongly on foundation–embankment permeability contrast. These scenario results do not prove that an upstream liner alone is sufficient for all existing dams.
29. Seismic evaluation of transmission tower structures under coupled ice-earthquake action
Core Problem: Ice accretion changes structural mass and dynamic response before earthquake excitation is applied.
Key Innovation: A numerical assessment uses 216 Türkiye ground-motion records to examine ice density and motion type. Larger modelled responses and damage depend on loading intensity and pulse content. The results concern the tested structural scenarios, not observed failure frequencies for an entire transmission network.
30. Integrating morphometric analysis and AHP–Monte Carlo framework for multi-hazard risk reduction
Core Problem: Expert-weighted watershed rankings need sensitivity analysis before they guide multi-hazard resource allocation.
Key Innovation: A Pampa Basin assessment ranks 31 sub-watersheds and perturbs AHP weights in 10,000 Monte Carlo iterations, with comparison to reported household damage. Rank stability addresses weight uncertainty; it is not a calibrated probability of future hazard occurrence or proof of transferable rankings elsewhere.
31. Geophysical site characterization and scenario-based seismic hazard mapping with slope-based topographic amplification factors: case study of the hilltop Trakošćan castle, Croatia
Core Problem: Hilltop seismic assessment needs subsurface characterization and explicit treatment of terrain-related amplification.
Key Innovation: Title-level scope: geophysical characterization and scenario mapping at Trakošćan Castle using slope-based amplification factors. Factor validation, site response and hazard changes remain unassessed without a matching abstract.
32. Integrated seismic and electrical imaging with clustering for the hydro-mechanical screening of a high-altitude earthfill dam
Core Problem: Geophysical contrasts need a defensible connection to hydro-mechanical conditions before they guide dam screening.
Key Innovation: Title-level scope: seismic and electrical imaging combined with clustering. The diagnostic accuracy, hydraulic interpretation and dam-safety conclusions remain unassessed.
33. Probabilistic stability and progressive failure of pile-supported embankments under borehole-constrained stratigraphic uncertainty
Core Problem: Sparse boreholes leave uncertainty about stratigraphic continuity relevant to embankment stability.
Key Innovation: Title-level scope: probabilistic stability and progressive failure under borehole-constrained stratigraphic uncertainty. Failure probabilities, conditioning choices and field validation remain unassessed.
34. An engineering-scale numerical framework for predicting shield tunnelling response in gassy ground: Integrating advanced constitutive modelling and 3D simulation
Core Problem: Gas-bearing ground can complicate predictions of excavation response and the associated construction risks.
Key Innovation: Title-level scope: an engineering-scale constitutive and three-dimensional simulation framework. Gas coupling, measured response and validation accuracy are not inferred from the bibliographic record.
35. Single-sided slotted precast concrete joints with friction dissipation devices: Seismic performance and parametric analysis
Core Problem: Precast connections need a defined mechanism for dissipating earthquake energy while limiting damage.
Key Innovation: Title-level scope: single-sided slotted joints with friction-dissipation devices. Test results, dissipation gains and parameter effects remain unassessed.
36. A novel calculating method for the post-earthquake traffic capacity of high-speed railway simply supported bridges
Core Problem: Physical bridge damage must be connected to a usable estimate of post-earthquake railway traffic capacity.
Key Innovation: Title-level scope: a capacity-calculation method for simply supported high-speed railway bridges. Validation data, operating restrictions and recovery benefits are not available for assessment.
37. Mission-dependent functional fragility of forest and wildland–urban interface transportation networks: from physical damage to accessibility and recovery
Core Problem: Road damage does not uniquely determine whether a link can serve a particular evacuation, response or recovery mission.
Key Innovation: A probabilistic service interface is tested synthetically and on event-era Mati wildfire road topology. Mission definitions alter estimated accessibility and recovery. The real-network example checks computational applicability, while service probabilities remain scenario assumptions rather than calibrated historical closure records.
38. More Frequent Atmospheric Rivers and Associated Precipitation Extremes Induced by the Baroclinic Annular Mode
Core Problem: Subseasonal atmospheric variability can change the background frequency of landfalling moisture corridors and heavy rain.
Key Innovation: The Baroclinic Annular Mode is associated with substantial changes in atmospheric-river frequency and related precipitation extremes, linked to storm-track energetics. The analysis motivates extended-range prediction research; it does not demonstrate a calibrated local landslide or flood-warning product.
39. Probable Maximum Precipitation: The Smethport Challenge
Core Problem: A small number of exceptional historical storms strongly influence probable-maximum-precipitation estimates.
Key Innovation: The Smethport study combines the record storm context with radar analyses of related northwest-flow events to identify environments favouring intense moisture convergence. Its proposed convection-permitting modelling challenge is not a completed demonstration that a revised PMP bound or future storm climatology has been established.
40. Climate change impacts on groundwater simulated using the AquiFR modelling platform
Core Problem: Aquifers can respond differently to climate forcing across regions, including changes in both high and low groundwater levels.
Key Innovation: AquiFR simulations driven by 36 regional climate projections indicate a north–south contrast across represented French aquifers. High- and low-level extremes are scenario- and model-dependent, with current return periods referenced to 1975–2004. They are conditional projections rather than dated forecasts or observations of future flooding.
41. Assessment of climate-induced coastal risks and adaptation strategies in the Caribbean Basin
Core Problem: Coastal protection decisions must account for ecological condition, maintenance and institutional constraints as well as immediate engineering benefits.
Key Innovation: A Caribbean review screens 665 records and examines 43 studies across engineered, ecosystem-based and other adaptation approaches. It identifies conditions favouring integrated responses; the synthesis does not establish one universally superior measure or quantify a new prospective regional loss reduction.
42. Post-fire burned forest area mapping with Sentinel-2 imagery: evaluating object-based machine learning models and transferability
Core Problem: Burned-area classifiers need transfer tests rather than accuracy estimates confined to their development sites.
Key Innovation: Object-based Random Forest and gradient boosting are evaluated in Türkiye and transferred to comparable Mediterranean sites in Greece and Spain, where overall accuracy remains above 83%. The demonstrated scope is burned-forest mapping under similar environments, not all ecosystem types or subsequent debris-flow prediction.
43. Scale, threshold and the recoverable social gradient of projected heat exposure: a paired census-tract analysis of European outermost and metropolitan territories
Core Problem: Spatial scale and exposure thresholds can change the social gradient recovered from climate-risk maps.
Key Innovation: Title-level scope: paired census-tract comparisons between European outermost and metropolitan territories. Gradient direction, effect magnitude and causal interpretation are not available for assessment.
44. The 2015/16 El Nin˜o Drought impacts on Household Food Security, Adaptation Expenditure and Coping Strategies: A Gender-Based Perspective
Core Problem: Drought impacts and coping resources may differ by household gender context and the timing of the drought episode.
Key Innovation: A South African household panel for 2010–2017 is combined with gridded drought data and fixed-effects analysis. Food-security impacts vary across household groups and years, while adaptation expenditure shows no significant drought response. The findings do not establish gender as an isolated causal mechanism or demonstrate that increased adaptation spending occurred.
45. Lightweight Machine Learning for Classifying Disaster-Related Social Media Posts
Core Problem: Short, imbalanced social-media messages can obscure information categories needed during disasters.
Key Innovation: A controlled CrisisBench comparison evaluates lightweight TF-IDF classifiers, label-level errors and computational requirements. The selected model reports Macro-F1 of 0.658 on the held-out split. Lower margins relate to errors, but the work does not validate real-time emergency decisions or universal transfer across languages and disaster types.
46. Field-monitored deformation, seepage, and stress response of a concrete arch dam during an extreme inflow event
Core Problem: A monitoring archive cannot diagnose peak dam behaviour if critical channels or observation times are missing.
Key Innovation: Reassessment of an arch-dam inflow event finds gaps around the flood peak, unreconciled datum information and surveys without exact dates. Available measurements document parts of the event but do not establish piping, elastic recovery or structural safety. The contribution is a specific diagnostic limit of the monitoring record.
47. Full-scale experimental investigation of progressive failure characteristics and failure mechanisms of hybrid fiber-reinforced concrete (HFRC) lining with reduced reinforcement ratio
Core Problem: Progressive failure needs to be assessed when changes in fibre reinforcement and steel ratio alter lining behaviour.
Key Innovation: Title-level scope: full-scale testing of hybrid fibre-reinforced concrete linings with reduced reinforcement. Capacity, failure sequences and safe design limits remain unassessed without the abstract.
48. Scattering‐to‐Intrinsic Transition Governing Coda Attenuation in the US Continental Crust
Core Problem: Coda-wave attenuation combines scattering and intrinsic energy loss, with their relative importance varying by frequency.
Key Innovation: USArray analysis finds scattering-dominated behaviour at lower tested frequencies and greater intrinsic absorption at higher frequencies, associated with heat flow. The work informs crustal interpretation; it does not directly validate site-specific ground-motion hazard or earthquake prediction.
49. How Microporosity Shapes Carbonate Hydraulic Behavior: A Digital Rock Physics Perspective
Core Problem: Unresolved microporous regions may connect larger pores and alter bulk hydraulic behaviour.
Key Innovation: A GPU lattice-Boltzmann formulation combines resolved and continuum flow in three carbonate samples. Spatial permutation controls help isolate the role of pore organization. The demonstrated effects concern sample-scale permeability, not direct prediction of karst collapse, slope failure or aquifer-scale flow.
50. SPIRIT: Sparse Physics-Informed Inverse Rendering for SAR Imaging
Core Problem: Sparse radar observations can leave geometry ambiguous even when image-domain reconstruction looks plausible.
Key Innovation: SPIRIT fits explicit three-dimensional surfaces while modelling visibility and coherent wave propagation. Synthetic and measured data support reconstruction under sparse acquisition. Reported collection savings concern tested configurations; they do not demonstrate landslide detection or monitoring performance.
51. Parameter misspecification in physics-informed neural networks: A closed-form audit and its limits
Core Problem: A low equation-residual loss does not establish that the encoded physical parameter was correctly specified.
Key Innovation: A closed-form audit estimates a residual-linear parameter from a frozen network and tests its limits across several equations. Some misspecifications remain undetected when architecture and equation share the same incorrect rate. This is a diagnostic of specified model structure, not a general certificate of physical truth.
52. Super-resolution as Spatial Downscaling: Coherence, Structure and Task Utility in Earth Observation
Core Problem: Pixel fidelity and observation consistency may fail to reveal invented spatial detail in reconstructed satellite imagery.
Key Innovation: Spatially blocked comparisons assess spectral structure, coherence and downstream tasks across 735 tiles from 190 scenes. Learned super-resolution can amplify spurious change without significant building-change gains. Reconstruction metrics alone therefore do not establish improved hazard mapping.
53. Beyond the record: evaluating probabilistic AI weather forecasts deep in the tail
Core Problem: Skill on record-breaking weather needs common verification that retains false alarms as well as observed exceedances.
Key Innovation: Four probabilistic AI systems are compared with IFS-ENS for 2022 over land at 1.5° resolution. Relative skill depends on variable, threshold depth and lead time. Better tail scores can coexist with probability bias; the work does not establish local rainfall-trigger or disaster-warning accuracy.
54. Dual Reverse Time Migration: Efficient Least-squares Imaging with Multiple Scattering in Dual Space
Core Problem: First-order scattering assumptions and repeated inverse computations constrain seismic imaging of complex structures.
Key Innovation: Dual RTM reformulates the problem in scattering-data space and incorporates higher-order interactions, tested on synthetic benchmarks. Its computational formulation and imaging gains do not demonstrate independent field hazard characterization or remove uncertainty in the background Earth model.
55. Wet-Biased Data Pipelines Mislead the Training and Evaluation of Generative Precipitation Nowcasting
Core Problem: Selecting rainy target cases can bias both what a generative model learns and how its calibration appears.
Key Innovation: Analytical results and a fixed satellite-to-radar diffusion model show that wet-biased sampling can reverse calibration assessments, including an out-of-domain test after a seven-year gap. The result diagnoses an evaluation mechanism rather than demonstrating an improved flood-warning system.
56. Permafrost augmented simulation and forecast using hybrid modeling through spectral neural networks
Core Problem: Long simulations of coupled water and heat transfer can be expensive, while surrogate stability and observation agreement need separate checks.
Key Innovation: A spectral surrogate is combined with experimental-data assimilation and compared with measured soil-temperature behaviour at a forested permafrost site. Climate-scenario forecasts remain conditional; improved thermal simulation does not independently validate thaw-collapse or infrastructure-failure predictions.
57. BridgeCast: Bridging Ocean Wave Forecasts to Reanalysis via Flow Matching with Exogenous Variables
Core Problem: Bias correction of ocean-wave forecasts needs to distinguish agreement with reanalysis from verification against independent observations.
Key Innovation: BridgeCast uses conditional flow matching to translate physical forecasts toward reanalysis-like fields with atmospheric covariates. Regional and lead-time comparisons support the correction task; the available abstract does not establish independent storm-wave or coastal-damage validation.
58. Small-strain stiffness characteristics of granular materials under monotonic and cyclic loading: effects of particle breakage
Core Problem: Particle breakage and cyclic contact rearrangement may affect small-strain stiffness in competing directions.
Key Innovation: Oedometer and bender-element tests on recycled concrete and brick particles isolate breakage-related stiffness loss from contact and stress-ratio effects. A proposed model is tied to one-dimensional monotonic compression; its performance is not generalized to natural slopes or all cyclic loading paths.
59. Georegistration of Terrestrial LiDAR Point Clouds With Satellite Images Without GNSS
Core Problem: GNSS-degraded point-cloud maps need a spatial reference, but road-map alignment is different from survey-grade control.
Key Innovation: A road-skeleton workflow aligns LiDAR with map references, with independent map-provider comparisons. Reported planimetric consistency does not establish survey-grade three-dimensional accuracy, especially where vertical anchoring relies on coarse terrain data or road geometry is sparse.
60. A Dual-Polarization-Integrated Time-Series InSAR Framework for Arc-Level Parameter Estimation and Network Densification
Core Problem: Single-polarization time-series processing can limit coherent-point density and network connectivity.
Key Innovation: Dual-polarization projection is incorporated into arc-level parameter estimation on 45 Sentinel-1 images over an airport. Network densification and internal continuity improve under tested configurations. The newly available abstract does not establish independent displacement truth or direct landslide-warning performance.
61. Glacier Volume Changes Across the Greater Caucasus Between 1960 and 2025
Core Problem: Long glacier-area records do not directly reveal how regional ice volume and thickness have changed.
Key Innovation: Five inventory periods and the PalaeoIce framework reconstruct Greater Caucasus ice volume for 1960–2025. The reported regional volume decline is a model-supported reconstruction, not a direct glacier-wide thickness survey or a prediction of specific avalanches and lake outbursts.
62. Support-Aware Dual-SWIR Downscaling from Sentinel-2 and Submetre Jilin-1 Imagery on Non-Nested Grids
Core Problem: Combining fine RGBN imagery with coarse SWIR observations requires accounting for the actual spatial support of both grids.
Key Innovation: An area-overlap inverse allocation improves controlled comparisons using retained 20 m SWIR references. The 0.75 m output remains observation-consistent conditional allocation; coarse-reference accuracy does not independently validate each fine-scale spectral value or downstream hazard map.
63. Monitoring Lakes in Mountains with Sentinel-1 SAR Imagery
Core Problem: Snow, ice and terrain can obscure mountain-lake state and area in optical imagery.
Key Innovation: The OASIS paired-SAR index tracks lake-surface conditions, with temperature and optical-area references in the French Alps. Ice-cover and area monitoring can support further hazard studies, but the work does not demonstrate glacial-lake outburst or drainage forecasts.
64. A Downscaling–Calibration Framework for GPM Precipitation Data Based on Geographically Weighted Machine Learning: A Case Study of the Yellow River Basin, China
Core Problem: Coarse precipitation products need spatial calibration before they can support finer regional analyses.
Key Innovation: A geographically weighted downscaling and calibration workflow evaluates annual and monthly GPM estimates in three contrasting Yellow River Basin years. Improved pooled estimates do not establish event-scale rainfall extremes or landslide-trigger performance at every 1 km grid cell.
65. Strain-Normalized Modulus Framework for Settlement Prediction in Fine Sands Using In Situ Tests
Core Problem: Stiffness inferred from different field tests must be compared at a compatible strain level before settlement estimates are combined.
Key Innovation: A three-site Florida programme normalizes test-derived moduli through a shared framework, with a low-strain reference obtained by extrapolation beyond the measured range. Site-specific settlement comparisons support regional calibration; the reference modulus is not a direct low-strain measurement or a universal sand correlation.
66. Aspect-associated contrasts in soil thermal regimes and phase-sensitive dielectric responses during seasonal freeze–thaw periods
Core Problem: Aspect-related thermal differences can complicate the interpretation of dielectric measurements during phase transitions.
Key Innovation: Title-level scope: soil thermal regimes and phase-sensitive dielectric responses during seasonal freezing and thawing. Calibration, water-content interpretation and slope-instability implications remain unassessed.
67. An intelligent tunnel lining crack inspection framework via deep learning segmentation and geometric feature quantification
Core Problem: Image segmentation needs geometric interpretation before detected cracks can inform engineering decisions.
Key Innovation: Title-level scope: deep-learning segmentation and geometric quantification for lining-crack inspection. Detection accuracy, crack-size uncertainty and structural-safety relevance remain unassessed.
68. Coupled hydrochemical and freeze-thaw deterioration of confined red-bed mudstone with respect to disintegration behavior and post-disintegration hydraulic properties
Core Problem: Chemical exposure and freezing–thawing may interact in rock disintegration and the hydraulic properties of its products.
Key Innovation: Title-level scope: confined red-bed mudstone under coupled hydrochemical and freeze–thaw deterioration. Damage trends, conductivity changes and field-instability implications are not inferred without the abstract.
69. A comparative study of the dynamic properties of coral sand and coral gravelly sand using centrifuge shaking table tests
Core Problem: Coral sand and gravelly sand may respond differently under dynamic loading.
Key Innovation: Title-level scope: comparative centrifuge shaking-table tests of dynamic properties. Property differences, liquefaction behaviour and generalization to other deposits remain unassessed.