TerraMosaic Daily Digest: October 21, 2026
Daily Summary
Today's landslide papers examine how regional models respond to changing ground conditions and forcing. Long-term railway-slope simulations in the Brazilian Amazon link instability to cumulative infiltration and local groundwater rise, while an InSAR workflow explores short-term updates to a regional baseline. Earthquake-induced susceptibility mapping adds fault-direction information. Newly available modelling studies also show why geological representation, mesh choice and terrain coordinates can shape the results of slope and runout simulations.
Field evidence sharpens the distinction between related but different hazard processes. Surveys of the Hantai event in western Beijing identify a water-driven debris flood with a coarse mobile bedload, despite basin morphometry overlapping with debris-flow ranges. In avalanche simulations, visible run-up and the height of substantial impact pressure follow different mechanisms. Notched-cliff simulations offer a coastal counterpart: the position of a notch can change whether it amplifies wave run-up.
Hydrological and volcanic studies place greater emphasis on event timing, dependence and evaluation scale. Upper Danube simulations show that changing dependence among catchments contributes to concurrent flood and drought risk, while US observations reveal that upper-decile rainfall often does not produce upper-decile discharge. Separate Etna eruption catalogs support different temporal models for flank activity and clustered summit events. A China-wide drought dataset further illustrates the gap between strong average reconstruction scores and the detection of rare extremes.
Key Trends
The strongest contributions connect event observations and process mechanics with the assumptions used to construct regional hazard estimates.
- Slope assessment spans several timescales: Amazon railway simulations address decades of rainfall and seepage, while InSAR-based updating targets shorter deformation changes. Brienz modelling highlights how sensitive predicted motion can remain to geological structure, numerical resolution and parameter choices.
- Similar flow signatures can conceal different mechanisms: The Beijing field investigation separates water-driven debris floods from more strongly mixed debris flows. Avalanche modelling distinguishes run-up from pressure height, and cliff simulations identify a notch-position control on wave run-up.
- Dependence changes compound risk: Upper Danube joint-tail modelling shows why marginal discharge trends alone cannot characterize concurrent extremes. African hydroclimate analysis extends the question to event memory and persistent wet or dry states.
- Validation must match the intended decision: The SPEI reconstruction includes spatial-buffer and forward-period tests but retains low extreme-drought recall. Rainfall-fusion gains depend on the reference period, and UAV exercise results differ sharply between building and road assessment.
- Observations expand across sensors and environments: Field surveys, LiDAR, radar, video velocimetry and acoustic–seismic measurements provide complementary views of terrain, flow and shallow structure. Their measurement accuracy and sampling limitations remain part of any subsequent hazard interpretation.
Selected Papers
The selected studies cover landslide dynamics and monitoring, sediment-laden floods, avalanche loading, volcanic timing and hydroclimatic extremes. Infrastructure applications and methods for improving observations or physical models follow within the same score-ordered list. The emphasis is on what each study establishes about the process being measured or modelled.
1. Physics-inspired machine learning for earthquake-induced landslide susceptibility mapping: integrating fault-controlled seismic energy attenuation and directional effects
Core Problem: Fault distance alone can miss directional differences in seismic forcing across slopes.
Key Innovation: A fault-azimuth attenuation factor adds direction–distance information to susceptibility models, with reported XGBoost AUC of 0.926. SHAP associations support interpretation within the study; the abstract does not establish causal controls or independent transfer to another earthquake.
2. Long-Term Seepage and Stability of Tropical Railway Cut Slopes Using Satellite Hydroclimatic Data
Core Problem: Sparse field monitoring complicates long-term assessment of tropical residual-soil cut slopes.
Key Innovation: More than 40 years of daily seepage–stability simulations combine satellite hydroclimate inputs with laboratory relations and groundwater observations. Only the steepest tested slope becomes unstable in the simulations; rainfall-accumulation criteria are preliminary model-derived references, not operational warning thresholds.
3. Field characteristics and initiation mechanisms of debris floods: Insights from western Beijing, China
Core Problem: Basin morphometry alone may not distinguish water-driven debris floods from debris flows.
Key Innovation: Two surveys of the 2023 Hantai event combine UAV/LiDAR evidence, mapped particles and flood reconstruction. Stratified flow traces and coarse bedload support a debris-flood interpretation. Agreement with two high-water marks does not fully constrain the reconstruction, and initiation remains an event-based interpretation of rainfall, sediment supply and connectivity.
4. Challenges in modelling rock slope failures using a distinct element method: A critical analysis based on the 2023 Brienz/Brinzauls ’Insel’ landslide
Core Problem: A model that reproduces broad landslide motion can still be highly sensitive to numerical and geological choices.
Key Innovation: The newly available Brienz/Brinzauls abstract reports 2D parameter tests and a 3D 3DEC model. Small parameter, mesh and preparation changes affect fracture, velocity and deposition; overall kinematic agreement supports scenario analysis but leaves predictive robustness limited. This updates the earlier title-level entry.
5. A terrain-following framework for gravity-driven flow modeling on erodible beds
Core Problem: Coordinate choices can distort depth-integrated flow directions over rugged terrain.
Key Innovation: A smoothed reference surface defines terrain-following coordinates while sub-topography retains landform detail. The framework is incorporated into an erodible-bed two-phase model and demonstrated numerically, including a large event. The matching abstract adds mechanism and scope to the earlier title-level entry without establishing universal runout accuracy.
6. Distinguishing run-up height from pressure distribution during avalanche impact on narrow obstacles: mechanisms and semi-empirical prediction from three-dimensional simulations
Core Problem: Visible avalanche run-up need not define the vertical range of substantial pressure on a narrow obstacle.
Key Innovation: Three-dimensional MPM simulations separate run-up and pressure heights. Fast-flow run-up gains an inertia-driven contribution, while simulated pressure height remains velocity-independent through basal flow deflection. Separate semi-empirical relations are proposed within the tested rheologies and obstacle setting.
7. Temporal models for the occurrence of Etna eruptions and implications for hazard assessment
Core Problem: Flank eruptions and clustered summit paroxysms require different temporal descriptions.
Key Innovation: Separate Etna catalogs support renewal modelling of flank activity and inter- and intra-cluster summit events. Brownian Passage Time fits best among the tested alternatives, informing time-window tephra probabilities. Catalog reproduction and model fit do not constitute a dated eruption prediction.
8. Cessation of Nearby Low‐Frequency Earthquake Activity Following the December 2024 Mw 7.0 Mendocino Earthquake: The Likely Role of Pore Pressure Reductions
Core Problem: Post-earthquake suppression of nearby fault activity can challenge simple effective-friction interpretations.
Key Innovation: Twenty-seven LFE families stopped for roughly 7–25 days after the 2024 Mendocino earthquake. Modelling favors clamping plus pore-pressure reduction over shear-stress suppression. The inferred mechanism does not explain all cessation-time differences and is not a directly measured regional pressure field.
9. Exploring Climate Change Effects on Concurrent Floods and Concurrent Droughts via Statistical Deep Learning
Core Problem: Changing dependence among catchments can alter joint extremes even when marginal changes are known.
Key Innovation: Statistical deep learning models joint tails in four Upper Danube catchments using large-ensemble, hydrologically simulated discharge. Under the tested high-emissions scenario, changing dependence contributes to increased concurrent extremes. Results are conditional climate-model projections rather than observed future event probabilities.
10. The 2026 seismic hazard model for Finland
Core Problem: Low-seismicity regions still need explicit source, magnitude and ground-motion assumptions.
Key Innovation: The Finland model compares three ground-motion relations and alternative minimum magnitudes, with reference hard-rock maps, spectra and disaggregation. Including smaller earthquakes changes mapped local high-frequency hazard. Site amplification and vulnerable structures require additional treatment beyond the reference-rock maps.
11. A 1 km Monthly Multi-Timescale SPEI Dataset for Mainland China Reconstructed from Station Observations and Multi-Source Environmental Predictors
Core Problem: High-resolution drought products need evaluation across locations, periods and event severity.
Key Innovation: A 1 km, five-timescale SPEI reconstruction covers mainland China during 1979–2018. Spatial-buffer and forward-period tests are weaker than random validation; extreme-drought recall is only 0.240–0.387. The technical evaluation supports monitoring use while defining a significant rare-event detection limitation.
12. Application of Multi-Temporal LiDAR for Gully Erosion Monitoring in Mediterranean Intensive Olive Groves
Core Problem: Agricultural tillage and machinery can obscure erosion-related topographic change.
Key Innovation: Multi-temporal LiDAR and orthophoto inventories isolate gully-internal change in Mediterranean olive groves. Inventory-constrained elevation differences and a terrain index perform more consistently for permanent gullies. The resulting estimates concern the delineated gullies, rather than all apparent agricultural surface change.
13. Multi-temporal gully expansion and volumetric soil loss estimation in Kakia and Esamburmbur sub-catchments, Enkare-Narok Basin, Kenya
Core Problem: Historical gully-area change must be connected to volumetric erosion with explicit terrain uncertainty.
Key Innovation: The matching Kenyan abstract adds multi-temporal imagery, UAV topography and DEM differencing. Area–volume extrapolation has only moderate explanatory power, and DEM choice changes estimated volume. The newly available details update the earlier title-level citation; modeled or scaling-derived volumes are distinguished from directly mapped change.
14. Solitary wave runup on coastal cliffs with basal notches: High-fidelity simulations and new predictive equations
Core Problem: Cliff-base notches may change run-up in ways absent from idealized-wall formulas.
Key Innovation: A 160-case 2D CFD experiment identifies notch-roof position as a key modifier: emergent notches can amplify run-up, whereas submerged cases can be inactive. A hierarchical predictor improves held-out wave-condition estimates within the numerical setup; natural profiles and three-dimensional effects remain to be tested.
15. Past-to-future drought across hydrographic microregions in Northeast Brazil: SPI-based diagnostics and bivariate (duration–severity) risk under CMIP6 scenarios
Core Problem: Drought planning needs joint duration–severity information across heterogeneous regions.
Key Innovation: Observed rainfall and bias-corrected CMIP6 scenarios are compared across 82 Northeast Brazilian microregions using a fixed SPI reference. Projected dry sequences become longer and more severe despite fewer new sequences. Local return-period changes depend on scenario, region and event definition.
16. Relating precipitation extremes to streamflow peaks using percentile-based correspondence and interpretable machine learning
Core Problem: A high-percentile rainfall event need not produce a corresponding high-percentile streamflow peak.
Key Innovation: US catchment analysis links upper-decile rainfall to event-specific discharge maxima. Only about 57% of paired peaks exceed the discharge 90th percentile at the median catchment. Snow fraction and elevation modify this correspondence; the diagnostic is not an evaluated operational flood forecast.
17. Dynamic landslide hazard assessment by linking time-series InSAR with short-term deformation prediction
Core Problem: Long-term susceptibility and delayed SAR products provide limited short-term responsiveness.
Key Innovation: A dual-scale workflow updates a susceptibility–InSAR baseline with predicted deformation increments from past motion and dynamic triggers. The supplied abstract describes a two-level update matrix and ablations, but gives no forecast horizon, event-based warning skill or independent failure-time validation.
18. Probabilistic analysis of run-out distance of slope failure using RMPM and conditional random fields
Core Problem: Runout estimates should carry uncertainty in subsurface spatial variability.
Key Innovation: Title-level scope: random material point modelling with conditional random fields for runout-distance analysis. Conditioning data, probability calibration and runout validation remain unassessed without a matching abstract.
19. Perturbation-induced time-delayed instability of rock joints
Core Problem: A transient disturbance may precede delayed instability of a discontinuity.
Key Innovation: Title-level scope: perturbation-induced time-delayed instability of rock joints. The delay mechanism, test conditions and predictive evidence cannot be assessed from the verified title-level record.
20. Dynamic responses and damage evolution of a twin–line biased tunnel slope using experimental and numerical methods
Core Problem: Biased twin-tunnel geometry can couple underground response and slope damage.
Key Innovation: Title-level scope: experimental and numerical analysis of a twin-line biased tunnel slope. Damage progression, model similitude and validation remain unassessed without a matching abstract.
21. Hydrodynamic loads on tandem twin pipelines induced by submarine landslides: Insights from physical flume tests and SPH simulations
Core Problem: Adjacent offshore pipelines can interact with a moving submarine landslide.
Key Innovation: Title-level scope: flume testing and SPH simulation of loads on tandem twin pipelines. Load magnitudes, geometry dependence and predictive accuracy cannot be inferred from the available record.
22. Progressive soil collapse and tunnel-system response in an underground complex under varying excavation configurations
Core Problem: Progressive soil collapse may interact with several components of an underground complex.
Key Innovation: Title-level scope: excavation-configuration effects on soil collapse and tunnel-system response. Failure thresholds and the evidence supporting model predictions remain unassessed.
23. Domain-decomposition-free CF-PINN for tunnel seepage and induced ground response in layered anisotropic strata
Core Problem: Layering and anisotropy complicate coupled tunnel seepage and ground response.
Key Innovation: Title-level scope: a domain-decomposition-free CF-PINN approach for layered anisotropic strata. Boundary-condition handling, physical validation and field generalization are not established by the title alone.
24. Study on the effects of metro-induced vibrations on adjacent utility tunnels in fissure-affected sites
Core Problem: Metro-induced vibration may affect neighboring underground structures at fissured sites.
Key Innovation: Title-level scope: vibration effects on adjacent utility tunnels in fissure-affected ground. Site conditions, response severity and applicable safety limits remain unassessed without an abstract.
25. Comprehensive risk assessment of snowmelt-runoff and storm-runoff flood disasters in the China-Pakistan Economic Corridor
Core Problem: Mountain transport corridors face floods with different seasonal generation mechanisms.
Key Innovation: Title-level scope: combined risk assessment of snowmelt-runoff and storm-runoff floods in the China–Pakistan Economic Corridor. Risk weights, inventories and validation cannot be assessed from the available record.
26. Spatial heterogeneity and attribution analysis of urban flood risk throughout the full lifecycle: insights from a safety-economy-environment framework and an interpretable MCDM-XGBoost-GWRF model
Core Problem: Flood risk evolves across exposure, response and recovery stages.
Key Innovation: Title-level scope: spatial heterogeneity and attribution across the urban-flood lifecycle using a multi-criteria and interpretable-learning framework. Component definitions, attribution assumptions and independent validation remain unassessed.
27. How urban flooding impacts healthcare accessibility and spatial inequalities?
Core Problem: Flooded transport networks can alter healthcare access unevenly across communities.
Key Innovation: Title-level scope: urban flooding, healthcare accessibility and spatial inequalities. The study area, accessibility metric, scenario realism and effect sizes are unavailable in the matching title-level record.
28. Climate Change Restructures Extreme-event Memory and Hydroclimatic Regime Persistence across Africa
Core Problem: The duration of extreme-event influence can matter as much as isolated event intensity.
Key Innovation: An event-conditioned probability framework compares African observations with climate projections, identifying different memory changes in semi-arid and humid regions and more persistent dry states. These are region- and scenario-dependent statistical results, not proof that all local droughts will intensify uniformly.
29. DisasterRoute: Evidence-Guided Dual-System Routing for Visual Disaster Question Answering
Core Problem: Limited edge computing makes uniform use of expensive reasoning inefficient.
Key Innovation: DisasterRoute routes UAV-image questions between direct classification and selective reasoning on DisasterBench. Reported accuracy–latency improvements support benchmark evaluation; they do not establish real-world rescue-decision reliability or a calibrated operational triage policy.
30. Key Findings for Small UAS for Post-Disaster Damage Assessment from the CURSE'26 Exercise
Core Problem: UAV building and road classifiers may fail differently under field acquisition constraints.
Key Innovation: CURSE'26 tests imagery from 23 teams. Building classification transfers across two states, while road assessment suffers from map misalignment, lighting and environmental variation. Exercise results should not be read as equal performance across damage types or independent real-disaster validation.
31. GuidedSAM-Plume: An Expert-Guided Segmentation Framework for Volcanic Plumes
Core Problem: Diffuse plume boundaries make fully automatic segmentation difficult.
Key Innovation: GuidedSAM-Plume uses expert prompts with SAM2 and evaluates both segmentation and volcanological descriptors. The expert-guided workflow and re-annotated dataset support plume measurement; the method does not independently predict an eruption or ash-dispersion hazard.
32. MNOPNet: A Lightweight Rotated Object Detection Network for Small-Scale Wildfire Detection
Core Problem: Small flames and smoke require efficient detection with accurate localization.
Key Innovation: A lightweight rotated detector reports 91.5% mAP50 on 7,808 UAV-derived images. Hardware-dependent throughput and benchmark detection gains do not establish time-to-warning, false-alarm performance or prospective field reliability.
33. Soil Erosion Evolution and Its Response to Human Activities over the Past ~250 Years in Two Karst Depressions of Chongqing, Southwest China
Core Problem: Present erosion trends can reflect past land use and depletion of erodible soil.
Key Innovation: Dated sediment profiles in two Chongqing depressions reconstruct roughly 250 years of erosion. Historical land use, rainfall and restoration offer interpretations of changing rates; the regional decline may also reflect depleted soil stocks rather than uniform ecosystem recovery.
34. Performance of Strip Footings on Sand Slopes Reinforced with Protrusion-Reinforced Geocells
Core Problem: Soil–geocell interaction affects bearing response and crest deformation.
Key Innovation: Small-scale tests compare protrusion-reinforced and conventional geocells across footing setbacks and reinforcement lengths. Improvements depend on geometry and mobilized settlement; the proposed interaction mechanism was not directly measured and full-scale transfer remains open.
35. Impact of pounding interaction on the seismic fragility of low-rise adjacent unreinforced masonry buildings
Core Problem: Adjacent buildings can redistribute earthquake demand through pounding and contact.
Key Innovation: Equivalent-frame models compare isolated units with seven masonry-cluster arrangements. A weaker unit can benefit while increasing a stronger neighbor's fragility, with progressive-collapse interactions. The findings are configuration-dependent model results rather than a universal protection effect.
36. Influence of pre-freezing water redistribution on the frost heave behavior of unsaturated coarse-grained soils
Core Problem: Nominally non-frost-susceptible coarse soil may heave under particular moisture boundaries.
Key Innovation: Covered, top-down freezing tests link heave to pre-freezing water redistribution, gradation and retained moisture. The experiment challenges classification-based assumptions under these conditions without establishing a universal field-heave rate.
37. Mitigation of liquefaction-induced settlement in reclaimed coral sand foundations by electrolytic desaturation: Field model tests and saturation-dependent mechanisms
Core Problem: Reclaimed coral sand may need treatment targeted at saturation-dependent cyclic response.
Key Innovation: Title-level scope: electrolytic desaturation assessed through field model tests and saturation-dependent mechanisms. Settlement reduction, treatment durability and construction constraints remain unassessed.
38. Dynamic response of jacket-supported offshore wind turbines subjected to obliquely incident SV and Rayleigh waves
Core Problem: Wave incidence and wave type can affect the response of jacket-supported structures.
Key Innovation: Title-level scope: offshore wind-turbine response to obliquely incident SV and Rayleigh waves. Response amplification, modelling assumptions and validation are unavailable in the matching title-level record.
39. Rockburst characteristics of coal-rock combination subjected to true triaxial unloading
Core Problem: True-triaxial unloading can change the instability of composite coal–rock systems.
Key Innovation: Title-level scope: rockburst characteristics under true-triaxial unloading. The unloading path, burst indicators and quantitative results remain unassessed without a matching abstract.
40. Rapid Infiltration Signatures From Global Infiltration Observations
Core Problem: Rapid infiltration signatures can vary with measurement method and local context.
Key Innovation: A dual-mechanism fit identifies rapid signatures in 26.6% of 4,992 curves, often alongside non-rapid curves at the same locations. The classification is an observational/model signature and needs independent physical evidence before identifying specific preferential-flow pathways.
41. Acoustic‐Seismic Coupling Signals From Aircraft Reveal Firn‐Ice Transition and Near‐Surface Elastic Structure of Antarctic Ice
Core Problem: Shallow ice structure is difficult to map using surface observations alone.
Key Innovation: Aircraft-generated acoustic–seismic signals recorded by a 36-node ground array support joint inversion of the upper roughly 100 m. The method still uses deployed ground instruments; it is not stand-alone airborne imaging or a direct ice-instability forecast.
42. A National Scale Hybrid Machine Learning Model for Enhanced Groundwater Depth Estimation
Core Problem: Sparse wells limit spatially detailed estimates of shallow water depth.
Key Innovation: A Denmark hybrid model corrects hydrological simulations and reduces national mean bias from 1.72 to 0.22 m. Better depth estimates can inform hazard inputs, but flood occurrence or groundwater-triggered slope failures were not directly predicted in the supplied evidence.
43. Analysis of Dual‐Porosity Model With Fissure Skin Effect in Confined Karst Aquifer
Core Problem: Fissure-skin storage and exchange can distort pumping-test interpretation.
Key Innovation: An analytical dual-porosity model incorporates a fissure-skin effect and improves fit to a mining-area constant-rate test. Better drawdown fitting supports hydraulic characterization without demonstrating sinkhole or water-inrush prediction.
44. Auto‐STIV: A Deep Learning Framework for Fully Autonomous Two‐Dimensional Streamflow Velocity Vector Field Estimation
Core Problem: Video-based surface velocimetry often needs manual direction and parameter selection.
Key Innovation: A two-stage CNN estimates flow direction and speed, with laboratory training and Colorado River footage evaluation. Surface-velocity performance does not alone establish discharge accuracy, channel cross-section knowledge or warning reliability.
45. Register Anything Model for Generalizable and Robust Point Cloud Registration
Core Problem: Registration models can lose accuracy across sensors, scene scales and point densities.
Key Innovation: RAM combines geometry-aware normalization with coarse-to-fine registration and reports strong results across 12 benchmarks. Cross-domain registration is useful for terrain workflows; landslide-change detection and field-survey accuracy are separate tasks.
46. BlazeVGGT: 13x Faster VGGT at 2,000 Frames via Camera Token without Training
Core Problem: Dense cross-view attention limits reconstruction from long image sequences.
Key Innovation: BlazeVGGT sparsifies key–value views while retaining all frames as queries, reporting faster inference and lower Chamfer distance at 2,000 frames. The acceleration is benchmark- and sequence-dependent, with no direct geohazard validation.
47. OpenEarth-OVS44: A Multimodal Remote Sensing Dataset for Open-Vocabulary Segmentation
Core Problem: Fine-grained open-vocabulary mapping needs consistent categories and cross-region evaluation.
Key Innovation: OpenEarth-OVS44 describes expert-verified RGB segments and registered RGB–SAR pairs across 40 cities, with held-out-region and synthetic-cloud tracks. The abstract promises a future data/code release; it does not establish a currently accessible archive or disaster-specific performance.
48. Tropical Cyclone Attributes in Sparse Autoencoder Features of an AI Weather Model
Core Problem: Skillful weather predictions do not explain how a model represents physical event attributes.
Key Innovation: Sparse-autoencoder features in GraphCast are related to cyclone position, intensity and pressure structure across independent decompositions. These reproducible associations describe representation structure; they do not establish causal control or improved cyclone forecasts.
49. Computational and experimental study of multi-grain-size sediment transport and sorting process
Core Problem: Uniform grain-size approximations can miss sorting and bed change.
Key Innovation: A large-eddy/multi-size sediment model is compared with flume grain distributions and bed elevations before a meandering-river application. Smaller experimental scour/deposition errors support the tested transport model, rather than field flood-risk prediction.
50. Data-driven and physics-guided hindcasting of measured wave spectra
Core Problem: Physical constraints can help different wave-spectrum objectives by different amounts.
Key Innovation: Buoy-spectrum hindcasts use known ERA5 winds and held-out years at three Baltic stations. Relaxation gives modest objective-dependent gains, whereas upper bounds can worsen errors. This is hindcasting, not demonstrated operational forecasting or independent-station transfer.
51. Numerical investigation of the correlation between the complete anisotropic stress–strain curves and the CPTu response of marine clays
Core Problem: Cone resistance can reflect anisotropy and post-peak behavior beyond a single strength value.
Key Innovation: NGI-ADPSoft penetration simulations, checked at Tiller-Flotten, explore anisotropy, softening and rate effects. The results inform uncertainty in CPT-derived stress–strain relations; they do not establish a unique inversion or field-validated monopile response.
52. Physics-Guided Generative Pansharpening for Remote Sensing Images Under Unknown Sensor Degradation
Core Problem: Unknown blur and spectral response complicate physically consistent pansharpening.
Key Innovation: Self-calibration and posterior diffusion enforce observation consistency while producing image ensembles and uncertainty maps. Cross-sensor image tests support reconstruction assessment; uncertainty maps still require task-specific calibration before hazard decisions.
53. Rational Utilization of Multitemporal Satellite Remote Sensing Data for Fully Satellite-Derived Bathymetry
Core Problem: Independent per-image depth fitting can lose temporal consistency across image sequences.
Key Innovation: Jointly trained multitemporal Sentinel-2 and ICESat-2 models produce time-defined depth maps evaluated at three island sites. The calibration and reference-depth procedures should be retained when interpreting the fully satellite-derived label; the method is not shown to provide survey-free ground truth.
54. Benchmarking convection-permitting climate simulations for hydrometeorological characterizations: a comparative study of WRF-SAAG and observation-based products in Chile
Core Problem: Kilometer-scale climate models can perform differently by climate zone and hydrological variable.
Key Innovation: WRF-SAAG is compared with station-informed products and hydrological observations across Chile. Arid-region and soil-moisture weaknesses persist; part of CR2MET's advantage reflects overlap with evaluation stations, limiting claims of fully independent product ranking.
55. Evaluation of SWOT HR PIXC version D water level time series of small lakes
Core Problem: Product filters can remove useful water-level observations as well as errors.
Key Innovation: A customized SWOT PIXC workflow retains more timestamps than filtered LakeSP at 37 Danish lakes, with similar median error against gauges. More retained samples improve temporal coverage, while regional roll-error indications constrain interpretation of individual passes.
56. Long-Term Evaluation of AIRS Precipitable Water Vapor over China (2005–2025): Moisture-State-Dependent Bias and Temporal Consistency
Core Problem: High monthly correlation can coexist with biased long-term anomalies and trends.
Key Innovation: AIRS–radiosonde comparisons over China identify moisture-state-dependent anomaly damping and changing bias despite high monthly agreement. One GNSS site supports local divergence analysis, not nationwide causal attribution of all trend differences.
57. FSLA-RTDETR: An Improved Four-Scale RT-DETR Network for Dense Small-Object Seabed Pockmark Detection in Shaded-Relief Bathymetric Imagery
Core Problem: Dense small pockmarks can be difficult to detect in shaded bathymetry.
Key Innovation: A four-scale detector is tested on 880 northern Gulf of Mexico images with 8,013 annotated pockmarks. Improved object localization supports geomorphological mapping; a detected depression is not by itself proof of active gas leakage or slope instability.
58. Dynamically Weighted Fusion of Daily Multi-Source Precipitation Products over the Haihe River Basin Using a Physically Constrained Two-Stage Framework
Core Problem: Fusion skill can depend on the reference period and shared errors among candidate products.
Key Innovation: Two-stage Haihe reconstruction improves several error metrics, but its event-skill advantage largely disappears when the MaxPrec reference is restricted to pre-evaluation years. Softmax weights and relative-uncertainty diagnostics are not calibrated probabilities or proof of true uncertainty reduction.
59. Towards Reliable Strength Prediction of Crystalline Rock Using an Image-Based Framework
Core Problem: Limited specimens constrain estimation of strength variability.
Key Innovation: Granite texture synthesis and CASRock simulations estimate a UCS range and Weibull variability. Synthetic images derived from one source image do not provide independent geological samples, and laboratory agreement needs wider material and site evaluation.
60. Dynamic Strain-Softening Modeling of Ultra-deep Sandstone Under Extreme Confinement: Insights from Asynchronous Strength Evolution
Core Problem: High confinement changes both post-peak sandstone behavior and the evolution of strength parameters.
Key Innovation: Tests on deep Tarim sandstone support a strain-softening formulation with asynchronous friction and cohesion evolution. FLAC3D comparisons reproduce laboratory curves within the tested pressure range, without directly validating field wellbore failure.
61. A TBM Tunnel Rock-Mass Classification Method Integrating Excavatability and Surrounding-Rock Stability
Core Problem: TBM classification needs to represent excavatability alongside surrounding-rock stability.
Key Innovation: Nearly 3,000 records from 18 TBMs support a revised classification and comparisons with established systems. Construction efficiency, classification correspondence and a safety-factor relation are distinct from independently validated collapse prediction.
62. Experimental Investigation on Low-Temperature Frost Heave Behavior of Flawed Sandstone under Constant Compression Stress
Core Problem: Flaw geometry and compression modify low-temperature cracking around caverns.
Key Innovation: Sandstone tests at −30°C and fixed compression reveal inclination-dependent heave pressure and crack propagation. The proposed criterion is bounded by the tested flaws, loads and thermal cycle rather than a general cavern-safety limit.
63. Defining the chalk–shale boundary in intact rock samples for engineering geology practice
Core Problem: Descriptive chalk–shale labels may obscure moisture-sensitive mechanical transitions.
Key Innovation: Mineralogical and mechanical data from the Ghareb Formation identify a transitional clay/carbonate-ratio range. This site-based functional classification can improve characterization, but its numerical boundary is not established as universal across formations.
64. Robust reinforcement learning multi-baseline phase unwrapping algorithm with JPEG-LS
Core Problem: Noisy multi-baseline interferograms challenge integer phase-ambiguity estimation.
Key Innovation: Reinforcement-learning cut selection and JPEG-LS-based correction improve tested phase-unwrapping errors and efficiency. Improvements in interferogram reconstruction do not directly quantify displacement accuracy or landslide-warning skill.
65. Elastic continuum and shell theory analytical solutions for tunneling effects on existing tunnels and large-diameter pipes
Core Problem: Beam idealizations can miss cross-sectional deformation of buried cylindrical structures.
Key Innovation: An elastic-continuum/shell solution resolves longitudinal and circumferential modes, including the cavity's missing soil. Parametric comparisons extend analytical representation; the supplied evidence does not establish field damage-prediction accuracy.
66. From distributed microdamage to localized fracture in a full-scale rebar-free SFRC segmental tunnel lining: AE–MS evidence and multi-source interpretation
Core Problem: Internal damage, visible cracks and deformation may evolve asynchronously.
Key Innovation: A full-scale SFRC lining test synchronizes AE, microseismicity, imaging and deformation to identify localization. Its system-normalized AE ratio and convergence interval are experiment-specific observations, not transferable alarm thresholds.
67. Global patterns of reservoir normal water storage capacities estimated by statistical and machine learning models
Core Problem: Global assessments often lack reported normal operating storage capacity.
Key Innovation: Models fitted to 801 reported reservoirs estimate normal storage for 7,250 reservoirs. Normal storage and its ratio are not the same as available flood-control capacity, current water volume or an independently observed operating rule.
68. Limitations of the state-parameter framework for well-graded sandy-gravel: A confinement-dependent assessment of NorSand and SaniSand
Core Problem: State-parameter calibrations on uniform sand may not transfer across confinement in well-graded gravel.
Key Innovation: Drained tests spanning 10–1,000 kPa expose regime-dependent NorSand/SaniSand misfit. Partitioned or stress-dependent parameters improve response representation, with a wall example illustrating settlement differences. Transfer still depends on material and stress range.
69. Generative prediction of rock discontinuities with a tabular foundation model
Core Problem: Sparse tabular observations limit statistical prediction of rock discontinuities.
Key Innovation: Title-level scope: generative discontinuity prediction with a tabular foundation model. Data diversity, uncertainty calibration and geological transfer remain unassessed; the foundation-model label does not establish a disruptive hazard advance.