TerraMosaic Daily Digest: July 10, 2026
Daily Summary
The July 10 literature resolves slope failure across scales that are usually treated separately. Propagating geological alternatives changes numerical stability margins; dense near-field seismology reconstructs progressive sinkhole collapse; and UAV mapping records 478 rainstorm-triggered soil slips on agricultural terraces. Rock-ice shear tests and a 290-site thaw-hazard inventory further show how warming, roughness, water and infrastructure heat reorganize failure susceptibility in cold terrain, while operational TBM data convert excavation-response anomalies into collapse warnings.
Liquefaction and earthquake studies converge on measurement fidelity. Controlled sand experiments separate density and depositional-fabric effects on SPT and shear-wave velocity, membrane-corrected torsional tests remove a systematic bias in gravelly-soil resistance, and energy-based tests resolve cyclic degradation of reclaimed calcareous sand. At larger scales, ambient-noise Green's functions and HVSR mapping recover basin amplification where earthquake records are sparse; injection tests, near-fault record selection and foundation shaking experiments connect those ground-motion and pore-pressure constraints to induced slip and infrastructure response.
Hydroclimatic hazards are increasingly reconstructed as evolving fields rather than isolated events. River-network sediment routing, 35 years of shoreline observations and regional total-water-level extremes resolve storage, migration and coastal-flood frequency, while non-stationary runoff indices and connected flash-drought tracking retain shifts that stationary summaries suppress. The transferable methods address the same observational bottlenecks upstream: precipitation and soil-moisture correction, GNSS water-vapour monitoring, optical-SAR fusion, point-cloud registration, satellite orthorectification and physics-constrained reconstruction improve the evidence supplied to hazard models before prediction begins.
Key Trends
Five methodological shifts connect the physical, observational, and computational studies.
- Failure models are carrying geological alternatives forward: Slope stability, loess collapse, rock-ice shearing and cyclic sand response are evaluated across plausible structures, state variables and measurement biases rather than one nominal ground model.
- Dense observations are exposing event sequence: Near-field seismic arrays, UAV inventories, fibre sensing, long satellite records and excavation telemetry distinguish nucleation, propagation and recovery that snapshots cannot resolve.
- Infrastructure is both a load and an observing system: Pipelines alter permafrost heat, injection changes fracture slip, foundations reshape liquefaction, and TBM or structural response supplies diagnostic time series for warning and design.
- Non-stationarity is entering hazard metrics: Runoff indices, coastal water levels, flash droughts and sediment records retain temporal shifts, compound forcing and spatial connectivity instead of assuming an invariant historical distribution.
- Geometric and physical consistency is moving upstream: Orthorectification, point-cloud registration, LiDAR odometry, multimodal fusion and physics-guided reconstruction improve the observations before classification, inversion or forecasting.
Selected Papers
The selected papers connect slope, sinkhole, thaw and tunnel-collapse mechanics with liquefaction, seismic response, river and coastal dynamics, hydroclimatic extremes, cryosphere monitoring, and geometry-aware remote sensing. This issue contains 92 selected papers from 2156 papers analyzed.
1. Dynamic failure mechanisms of an urban sinkhole collapse revealed by dense near-field seismic observations
Core Problem: Urban sinkhole failure evolves too rapidly for sparse regional networks to resolve the collapse sequence and its diagnostic seismic signatures.
Key Innovation: A dense near-field array reconstructs the nucleation and progressive roof failure of an urban collapse, linking distinct seismic phases to subsurface damage and providing an observational basis for real-time sinkhole recognition.
2. On the consideration of geological and geotechnical uncertainty in numerical slope stability analysis
Core Problem: Deterministic slope models suppress uncertainty in stratigraphy, material properties, and failure geometry, producing stability estimates that can appear more precise than the evidence permits.
Key Innovation: The study propagates geological and geotechnical uncertainty through numerical slope-stability analysis and shows how alternative ground models alter failure mechanisms and safety margins.
3. Soil slips of terrace slope triggered by an extreme rainstorm on the Loess Plateau of China
Core Problem: Extreme rainstorms can generate dense clusters of shallow failures on agricultural terraces, yet their abundance, erosion yield, and management controls are poorly quantified.
Key Innovation: Field and UAV mapping identifies 478 soil slips and erosion intensities of 144-4,704 t ha-1; vegetation, soil bulk density, cohesion, and terrace management explain the strongest contrasts.
4. Thaw hazards along major transportation corridors in the northern Da Xing'anling Mountains in Northeast China
Core Problem: Climate warming and infrastructure heat disturb degrading permafrost beneath roads, railways, and pipelines, but corridor-scale hazard inventories remain sparse.
Key Innovation: Field mapping, ground temperatures, resistivity, and terrain data document 290 thaw-hazard sites and show that pipeline thermal effects extend about 16 m laterally, compared with about 10 m for roads and railways.
5. Mechanisms and modeling of rock-ice interface shearing at low temperature: Experimental insights and strength prediction based on surface morphology
Core Problem: Retreating glaciers expose rock-ice interfaces whose low-temperature shear strength controls newly destabilized slopes.
Key Innovation: Direct-shear and acoustic-emission tests isolate roughness, normal stress, and rate effects; increasing joint roughness raises peak strength by 56% and shifts damage from tensile to shear-dominated failure.
6. Experimental investigation of the effects of density and soil fabric to SPT and Vs measurements in sandy soils for liquefaction assessment
Core Problem: SPT resistance and shear-wave velocity are widely used for liquefaction assessment, but their dependence on density and depositional fabric is not interchangeable.
Key Innovation: Controlled sand specimens separate density and fabric effects on SPT and Vs, clarifying when the two field proxies provide complementary rather than equivalent evidence of liquefaction resistance.
7. Effects and Benefits of Overburden Isolated Grout Injection: Recent Field Experience from the Mitigation of Mining-Induced Seismicity in the Ultra-thick Cretaceous Strata
Core Problem: Mining-induced seismicity in ultra-thick strata is difficult to mitigate because stress redistribution above extracted panels is spatially distributed and poorly observable.
Key Innovation: Recent field experience shows how overburden-isolated grout injection modifies load transfer and reduces induced seismic response, providing an operational mitigation route tied to measured strata behaviour.
8. Engineering geological characteristics of weakly cemented strata and multifield evolution of mud-sand inrush disasters induced by coal mining
Core Problem: Mining in weakly cemented, water-sensitive strata can connect roof fractures to aquifers and trigger sudden mud-sand inrush.
Key Innovation: Laboratory tests, distributed fibre sensing, and coupled simulation resolve a stress-change-to-inrush sequence involving dilation, bound-water release, gravitational drainage, and progressive conduit formation.
9. Combining horizontal-to-vertical spectral ratios with empirical amplification functions to develop high resolution site amplification maps: Application to Sion, Switzerland
Core Problem: Earthquake recordings are too sparse to map basin-scale amplification at the resolution required for seismic microzonation.
Key Innovation: Ambient-noise HVSR clustering, canonical correlation, and kriging reproduce frequency-dependent amplification across Sion; canonical correlation best matches independent estimates but underestimates the strongest response by as much as a factor of 2.5.
10. Numerical analysis and test of seismic response of bucket foundation-offshore wind turbine system in liquifiable site
Core Problem: Offshore wind-turbine bucket foundations can settle, tilt, and pull out when surrounding sand liquefies under combined environmental and seismic loading.
Key Innovation: Shaking-table tests and numerical simulations distinguish confined from unconfined sand response and show that extending the bucket into non-liquefiable strata materially improves seismic performance.
11. Membrane penetration reduction technique for undrained cyclic torsional shear tests on gravelly soils
Core Problem: Membrane penetration biases cyclic torsional tests on gravelly soils and can substantially overestimate liquefaction resistance.
Key Innovation: A 4 mm sand coating on both specimen surfaces suppresses penetration bias, enabling more reliable pore-pressure and strain measurements across gravel contents of 0-70%.
12. Pore water pressure and dissipated energy of saturated calcareous sand in island-road subgrades under wave-induced rotational stress paths
Core Problem: Wave-induced rotation of principal stresses drives pore-pressure accumulation in reclaimed calcareous-sand subgrades, but conventional quartz-sand relations do not capture the response.
Key Innovation: Undrained hollow-cylinder tests establish a density-adjusted relation between normalized pore pressure and dissipated energy, providing an energy basis for liquefaction assessment under rotational loading.
13. Shear Slip Induced by CO2 and Water Injection into Volcanic Rocks Under Geothermal Environment
Core Problem: Geothermal CO2 and water injection can reactivate pre-existing volcanic-rock fractures, requiring direct comparison of induced-seismicity potential.
Key Innovation: High-temperature injection tests show slower slip under CO2 than water in most configurations and no greater induced-seismicity hazard for CO2 when slip velocity is used as the proxy.
14. An ambient-noise-based physics-data integrated framework for far-field basin ground motion simulation
Core Problem: Far-field basin ground motions are difficult to simulate where geological heterogeneity is strong and local earthquake records are limited.
Key Innovation: Ambient-noise Green's functions, finite-fault physics, and Bayesian updating reproduce Kanto Basin long-period response at roughly three orders of magnitude lower cost than the tested 3D finite-difference workflow.
15. Effects of confluence discharge ratio and sediment concentration on riverbed scour and deposition intensity
Core Problem: Hyperconcentrated tributary inflows alter confluence scour and aggradation, affecting flood conveyance in the Inner Mongolia reach of the Yellow River.
Key Innovation: Flume experiments quantify coupled discharge-ratio and sediment-concentration controls; at the highest discharge ratio, increasing concentration to 400 kg m-3 reduces maximum scour depth by about 81% while intensifying deposition.
16. Early warning of TBM tunnel collapses based on SBOA-optimized long short-term memory network and Bayesian theorem
Core Problem: TBM collapses emerge from changes in excavation response that are difficult to distinguish from normal operational variability.
Key Innovation: An optimized LSTM predicts the torque-penetration index and a Bayesian sliding-window layer converts prediction degradation into warning probabilities, detecting eight of nine held-out collapse sections with few false alarms.
17. Regional-scale coastal hazard modelling using a univariate framework for extreme total water levels including sea-level rise: a case study in Portugal
Core Problem: Regional coastal-flood estimates must combine extreme total water levels with sea-level rise without making the analysis operationally intractable.
Key Innovation: Peak-over-threshold and block-maxima models are evaluated across 400 profiles and 120 km of Portuguese coast, showing a strong increase in extreme-flood frequency even under the most conservative sea-level-rise scenario.
18. Effects of pressurization rate on injection-induced fracture slip under different experimental control modes
Core Problem: Natural faults respond to fluid injection under boundary conditions that laboratory constant-stress tests do not fully represent.
Key Innovation: Triaxial direct-shear tests contrast displacement- and stress-controlled boundaries, showing progressive creep-slip in the former and episodic rapid slip in the latter; faster pressurization concentrates asperity failure and peak slip rate.
19. Precursor identification and early warning of coal-rock mass instability using FBG-based strain tensor reconstruction
Core Problem: Deep coal-rock failure is preceded by spatial strain localization and directional reorganization that conventional point sensors cannot reconstruct.
Key Innovation: Multi-point fibre-Bragg-grating tensors resolve principal-strain rotation and localization through the full failure process, supporting three jump-ratio indicators for warning under different confinement levels.
20. Mechanical and Energy Absorption Behavior of Foam‒Aluminum Composite Rock Bolts: Insight from Mechanical Testing
Core Problem: Conventional rockburst supports cannot combine high working resistance with the deformation capacity needed to absorb dynamic energy.
Key Innovation: An aluminium-foam-filled thin-walled cartridge forces efficient concertina deformation, raising specific energy absorption by 48.74% and sustaining 270-280 kN over 200-210 mm of deformation.
21. Mechanical response and progressive instability evolution study for a central pillar subjected to mining-induced stresses
Core Problem: Mining sequence redistributes confinement and damage through central pillars, but local safety factors alone do not capture migration toward global instability.
Key Innovation: A calibrated FLAC3D analysis combines displacement-vector angle, integral safety factor, and stress-information entropy to trace weak-zone migration and progressive pillar destabilization during excavation.
22. The platform EUCLIDE - EUCentre for loss impact and damage evaluation - main integrated IT tools and functionalities
Core Problem: National seismic-risk assessment requires exposure, vulnerability, hazard, damage, and visualization components to operate on a common data foundation.
Key Innovation: EUCLIDE integrates Italian exposure databases and custom vulnerability models to run probabilistic and deterministic seismic-damage analyses for structures and infrastructure.
23. Updated fiber models for post-earthquake residual displacement estimation of reinforced concrete piers
Core Problem: Conventional fibre models can misrepresent permanent pier deformation after strong shaking, biasing post-earthquake serviceability decisions.
Key Innovation: Updated constitutive and numerical fibre formulations improve residual-displacement estimates for reinforced-concrete piers across nonlinear cyclic response regimes.
24. Performance evaluation of simplified SSI models during strong earthquake events and nonlinear soil response
Core Problem: Simplified soil-structure-interaction models are widely used in design, but their validity under strong motion and nonlinear soil response remains uncertain.
Key Innovation: Comparisons against nonlinear reference simulations identify where common reduced-order SSI representations preserve structural demand and where soil yielding invalidates their assumptions.
25. Predictive tools for seismic fragility and damage probability maps of bridges with reinforced concrete circular piers
Core Problem: Bridge fragility maps require rapid estimates of residual capacity across many pier geometries and shaking levels, but repeated nonlinear analysis is computationally expensive.
Key Innovation: A numerical database of circular reinforced-concrete piers supports predictive fragility functions and spatial damage-probability mapping across intensity levels.
26. Nonlinear seismic response of atmospheric storage tanks: Effects of geometry and roof configuration within a coupled FEM-SPH framework
Core Problem: Fluid-structure interaction and roof geometry can shift the onset and location of seismic damage in liquid-storage tanks.
Key Innovation: Coupled FEM-SPH simulations resolve how aspect ratio, fill level, and roof configuration redistribute shell plastic strain, with demand rising sharply beyond the identified strong-motion transition.
27. Nonlinear load-settlement behavior and efficiency of pile groups in permafrost regions
Core Problem: Linear superposition of single-pile capacity ignores interaction and temperature-dependent adfreeze behaviour in permafrost pile groups.
Key Innovation: A nonlinear load-transfer model couples pile interaction with temperature- and ice-content-dependent shaft resistance to predict group settlement and efficiency.
28. A prompt-conditioned multimodal framework for streamflow prediction with lightweight task adaptation
Core Problem: Streamflow models usually require basin-specific retraining and struggle to reuse heterogeneous meteorological and hydrological inputs across forecasting tasks.
Key Innovation: A prompt-conditioned multimodal architecture adapts a shared predictor with lightweight task parameters, enabling transfer across basins and forecast configurations without full model retraining.
29. Multiscale pore structure of aeolian loess and its role in metastable structure and collapse behavior
Core Problem: Loess collapsibility is governed by multiscale pore connectivity that bulk void ratio cannot resolve.
Key Innovation: Micro-CT, electron microscopy, pore-network flow, and discrete-element deposition show that vertically connected macropores dominate infiltration and explain the greater wetting-collapse susceptibility of shallow aeolian loess.
30. Scenario-based ground motion selection to match conditional spectral shape parameters considering pulse effects
Core Problem: Near-fault velocity pulses can bias structural demand when ground-motion suites reproduce spectral shape but not directivity-sensitive intensity measures.
Key Innovation: The selection framework conditions spectral-shape distributions jointly on PGA and PGV/PGA, producing hazard-consistent ensembles for the tested pulse-affected scenarios without imposing an arbitrary pulse-record quota.
31. Relationship between mean particle size and fines contribution factor in steady state of sand
Core Problem: Equivalent granular void-ratio methods become inaccurate and burdensome for sands containing large fractions of fines.
Key Innovation: Inverse steady-state analysis links the fines contribution factor to mean particle size and yields a simpler relation that agrees with published high-fines datasets.
32. Micromechanical stress probing analysis of memory surface evolution in undrained cyclic loading
Core Problem: Memory-surface constitutive models underpredict strain accumulation after cyclic mobility because their evolution laws lack direct micromechanical evidence.
Key Innovation: DEM stress probing shows initial expansion followed by sharp contraction of the memory surface as coordination number falls, identifying fabric or strain-liquefaction variables needed in improved models.
33. Gradation-state-dependent triaxial modelling and experimental investigation of calcareous sands under monotonic and undrained cyclic loading
Core Problem: Particle breakage changes calcareous-sand gradation during cyclic loading, invalidating constitutive models calibrated only for monotonic response.
Key Innovation: A bounding-surface model couples breakage, density, fabric, and post-liquefaction degradation and reproduces monotonic and undrained cyclic tests across compression and extension paths.
34. Brief communication: Drought economic assessments must include human health impacts
Core Problem: Economic drought assessments commonly omit health losses and therefore undervalue resilient water infrastructure.
Key Innovation: Valuation of groundwater access in Northeast Brazil estimates avoidable diarrhoea-related losses equal to 9.92% of local GDP and as much as US$1.15 billion at state scale.
35. Enhanced identification of watershed hydrological extremes using an improved drought index incorporating hydrological non-stationarity
Core Problem: Stationary runoff indices misclassify floods and droughts when seasonal controls and runoff distributions shift through time.
Key Innovation: A GAMLSS-based non-stationary runoff index tests nearly 300 configurations and improves detection and discrimination while revealing a transition from mean- to variability-dominated hydrological extremes.
36. Daily drought prediction in the Huaihe River Basin using VMD-informer-LSTM
Core Problem: Daily drought sequences combine non-stationary long-term evolution with short-lived anomalies that single sequence models struggle to preserve.
Key Innovation: Variational-mode decomposition with Informer and LSTM components improves R2, RMSE, MAE, and MAPE by 28.4-50.8% over the LSTM baseline, although long-lead intensity remains underestimated.
37. Decadal-scale variability and predictors of large wood storage in meandering rivers across Europe
Core Problem: Single inventories cannot represent the strongly time-varying large-wood loads that influence channel obstruction and flood risk in meandering rivers.
Key Innovation: Decadal observations across 65 bends in four European rivers show that river-specific context dominates wood storage, while local bend geometry and individual flow metrics have weak, inconsistent predictive power.
38. Modelling bedload transport at the network scale in a glacier-fed Alpine river system
Core Problem: Section-scale transport-capacity formulas assume unlimited sediment supply and overpredict bedload in steep Alpine river networks.
Key Innovation: An adapted D-CASCADE network model constrained by six years of outlet observations reproduces realistic longitudinal connectivity, storage, and annual transport patterns across a glacier-fed catchment.
39. Holistic analysis of shoreline change and mudbank dynamics across the Guiana coastline
Core Problem: Localized or short shoreline records cannot resolve complete mudbank-driven erosion-accretion cycles along the Guiana coast.
Key Innovation: Annual Landsat mapping over 1,500 km from 1988 to 2023 revises mudbank migration rates and shows that major coastal sectors do not follow the accepted 30-year oscillatory model.
40. Spatial patterns of sediment load in the Yarlung Tsangpo River: Transport-deposition in the middle reach and a ‘sediment factory’ in the lower
Core Problem: Sparse gauges obscure where sediment is produced, stored, and transferred through the high-relief Yarlung Tsangpo network.
Key Innovation: A calibrated global sediment simulator identifies alternating transport and deposition zones and attributes 92.9% of outlet load to the lower-reach sediment-production system.
41. Multi-Criteria evaluation and ranking of 28 precipitation products over China
Core Problem: Hydrological and hazard models often select precipitation products from a few aggregate metrics, masking regional and event-scale trade-offs.
Key Innovation: A multi-criteria comparison ranks 28 products over China across complementary accuracy dimensions, providing spatially explicit guidance for forcing selection rather than a single universal winner.
42. Geographically-weighted weakly supervised Bayesian High-Resolution Transformer for 200 m resolution pan-Arctic sea ice concentration mapping and uncertainty estimation using Sentinel-1, RCM, and AMSR2 data
Core Problem: Pan-Arctic sea-ice concentration labels are spatially coarse and uncertain, while radar signatures vary geographically and between sensors.
Key Innovation: A geographically weighted Bayesian transformer combines Sentinel-1, RCM, and AMSR2 under weak supervision to produce 200 m concentration maps with pixel-level uncertainty.
43. Large-Scale monitoring of dry snow in High Mountain Asia (2015-2024) from Sentinel-1 SAR observations
Core Problem: Cloud cover and sparse field observations limit consistent mapping of dry-snow conditions across High Mountain Asia.
Key Innovation: A 2015-2024 Sentinel-1 SAR record resolves large-scale dry-snow occurrence and persistence at high spatial and temporal coverage, supplying a transferable cryosphere-monitoring baseline.
44. Spatiotemporal Downscaling of SMAP Soil Moisture With an Integrated Machine Learning and Land Data Assimilation Framework
Core Problem: Coarse SMAP observations miss field-scale wetness patterns required by flood, drought, and irrigation models.
Key Innovation: Machine-learning downscaling is embedded in a land-data-assimilation framework so fine spatial structure remains dynamically consistent with the surface water balance.
45. Progressive scale convolutional network for spatio-temporal downscaling of soil moisture: A case study over the Tibetan Plateau
Core Problem: Soil-moisture downscaling must recover local spatial structure without discarding the temporal behaviour of the parent SMAP product.
Key Innovation: A progressive-scale convolutional network combines ERA5-Land temporal information with staged spatial refinement over the Tibetan Plateau.
46. Deep learning-based downscaling of SMAP data for surface and root-zone soil moisture mapping
Core Problem: Surface and root-zone wetness require different temporal memory, yet most satellite downscaling targets only the surface layer.
Key Innovation: A CNN-LSTM framework generates daily 100 m maps at 5 cm and 20 cm depths by coupling spatial predictors with temporally persistent soil-moisture dynamics.
47. Seasonal prior-guided temporal hypergraph network with predictive fidelity loss for soil moisture prediction
Core Problem: Soil-moisture controls and memory vary by season, weakening models that impose one temporal dependency structure throughout the year.
Key Innovation: Seasonal priors guide a temporal hypergraph network, while a predictive-fidelity loss constrains both magnitude and temporal evolution.
48. Filling data gaps in soil moisture monitoring networks via integrating spatio-temporal contextual information
Core Problem: Sensor outages break the temporal continuity required for watershed modelling and satellite-product validation.
Key Innovation: ST-GapFill selects spatially correlated stations and uses an LSTM to reconstruct missing observations from neighbouring and temporal context.
49. A novel high-resolution downscaling method for daily extreme precipitation using Sentinel-5P cloud parameters
Core Problem: Daily precipitation extremes are difficult to downscale because optical predictors disappear beneath storm cloud and common surface proxies respond too slowly.
Key Innovation: DeepGeo Downscaler uses Sentinel-5P cloud variables as event-time predictors to recover high-resolution daily extremes.
50. Enhancing Radar Observations of Extreme Precipitation Systems Using Nonuniform Pulse Repetition Time Waveforms
Core Problem: Uniform pulse-repetition timing creates a fixed trade-off between unambiguous velocity and range during fast, broad-spectrum convective storms.
Key Innovation: Nonuniform pulse-repetition waveforms extend observable velocity while retaining range information for supercells, hailstorms, tornadoes, and downbursts.
51. Prior-guided multi-domain mixture-of-experts for multimodal Earth observation data gaps
Core Problem: Earth-observation fusion fails when clouds, revisit gaps, or sensor outages remove one or more modalities at inference time.
Key Innovation: DAMoE activates domain-specific experts according to the available sensors and physical priors, reconstructing missing optical, SAR, or spectral information without a single fixed input pattern.
52. Intelligent identification of rock discontinuities via normal geometric feature-based edge modelling of point clouds in tunnel
Core Problem: Automated tunnel point-cloud analysis struggles to separate true rock discontinuity boundaries from sampling noise and surface roughness.
Key Innovation: Normal-vector geometry and edge modelling isolate discontinuity traces directly from tunnel point clouds, supporting reproducible rock-mass characterization without manual tracing.
53. CDHNet: An implicit calibration and difference-guided hierarchical perception change detection network for remote sensing images
Core Problem: Bitemporal misalignment and weak difference features cause remote-sensing change detectors to confuse geometric error with real surface change.
Key Innovation: CDHNet couples implicit spatiotemporal calibration with hierarchical difference perception to strengthen change evidence while suppressing registration artefacts.
54. Structural-prior-constrained adaptive 3D reconstruction and digital leakage inspection for utility tunnels
Core Problem: Tunnel imagery contains repetitive texture, occlusion, and weak illumination that produce voids and blurred geometry in conventional SfM-MVS models.
Key Innovation: Structural priors constrain centreline and cross-section geometry, while adaptive reconstruction supports leakage inspection on a consistent tunnel surface.
55. Analysis of Spatial and Temporal Characterization of GNSS Tropospheric Parameters During the Extreme Rainfall on ‘9.7’ in Hong Kong
Core Problem: Short-range extreme-rainfall warning requires water-vapour indicators that resolve both timing and spatial concentration before peak rainfall.
Key Innovation: GNSS-derived PWV, zenith wet delay, and vapour concentration lead the 2023 Hong Kong rainfall peak by 1-3 hours and spatially coincide with the heaviest-rainfall zone.
56. Flash Drought Dynamics in China’s Major Agricultural Plains: Spatiotemporal Patterns and Crop Photosynthetic Recovery Across Cropping Systems
Core Problem: Regional flash-drought tracking rarely connects event propagation to crop-specific photosynthetic recovery.
Key Innovation: Multi-source observations from 2001-2024 combine soil-moisture event tracking and Random Forest-SHAP attribution; SIF responds 6-9 days before GPP and exposes contrasting risks across rice, rainfed, and rotation systems.
57. Interannual Dynamics of Boreal Forest Wildfires (2020-2024) Derived from MODIS-Constrained Sentinel-2 Burned-Area Mapping
Core Problem: Annual boreal burned-area products can omit fires or assign them to the wrong year when spatial detail and temporal fire constraints are separated.
Key Innovation: BS2BAM constrains Sentinel-2 classification with MODIS fire products, reducing omission error from 30.82% to 15.98% and producing 10 m annual maps for 2020-2024.
58. Global patterns and driving mechanisms of vegetation sensitivity to compound atmospheric drought and heat events over 1982-2020
Core Problem: The global sensitivity of vegetation to compound atmospheric drought and heat is not captured by drought or temperature indicators alone.
Key Innovation: A multi-source sensitivity index shows increasing vulnerability across 63.6% of ecosystems since 1982 and identifies temperature as the dominant control across 34.5% of terrestrial areas.
59. A SIF-Constrained Noah-MP land surface model for simulating water use efficiency and drought Resilience: Application in a Semi-Arid Basin
Core Problem: Noah-MP overestimates evapotranspiration in water-limited systems when physiological drought stress is weakly represented.
Key Innovation: Embedding satellite SIF in the stomatal-resistance scheme lowers simulated evapotranspiration, strengthens carbon-water coupling, and reduces modelled drought-related water-use-efficiency loss from 13% to 5% in the Awash Basin.
60. SyneFloor: Synergizing zero-shot geometric priors and active learning for weakly supervised floorplan reconstruction
Core Problem: Polygon-level floorplan annotation is too costly to support reconstruction across diverse buildings and point-cloud densities.
Key Innovation: SyneFloor combines zero-shot foundation-model geometry with active learning to reconstruct structured floorplans under sparse supervision.
61. MiMNet-STASF: a spatiotemporal decoupled network for sub-pixel canopy structure retrieval of urban forests with multi-source long-term remote sensing time series
Core Problem: Urban forest structure is difficult to retrieve from mixed 10 m pixels because sensor, season, and canopy composition vary simultaneously.
Key Innovation: MiMNet-STASF separates spatial and temporal encoding across 24 months of Sentinel-1/2 data to retrieve sub-pixel canopy height and vertical composition.
62. A boundary-aware heterogeneous graph framework for fine-grained segmentation of tree crown-trunk components using UAV LiDAR and multi-view imagery
Core Problem: Dense plantations cause crown overlap and trunk occlusion, degrading component-level segmentation from either UAV LiDAR or imagery alone.
Key Innovation: MVCTNet builds a heterogeneous graph across LiDAR and four image views to separate crown and trunk geometry and derive under-branch height.
63. High-resolution forest height retrieval from L-band interferometric SAR time series using deep learning over Northern Spain
Core Problem: L-band forest-height retrieval must separate canopy structure from backscatter variability across acquisitions and baselines.
Key Innovation: A physics-informed U-Net family combines ALOS-2 interferometric time-series features with airborne-laser reference heights, outperforming intensity-only retrieval.
64. A theory-consistent data-driven framework for nonlinear mapping of SAR image spectra to normalized two-dimensional wave spectra
Core Problem: Azimuth cutoff removes part of the directional wave spectrum from SAR imagery, forcing conventional retrievals to depend on external priors.
Key Innovation: A theory-consistent learned inversion maps SAR spectra directly to normalized two-dimensional wave spectra using simulated image-spectrum pairs.
65. Clean and consistent MLS point clouds from the start: Moving-object removal coupled with LIDAR-inertial odometry for urban scenes
Core Problem: Moving vehicles and pedestrians leave duplicate and distorted geometry in mobile-laser maps when filtering occurs after odometry.
Key Innovation: Moving-object removal is coupled directly to LiDAR-inertial odometry so dynamic returns are rejected before they contaminate the urban map.
66. Allroad-SLAM: Factor graph-based 3D reconstruction in diverse road scenarios with a consumer-grade GNSS-LiDAR-inertial system
Core Problem: Road reconstruction must remain metrically stable across open roads, urban canyons, and GNSS degradation using low-cost sensors.
Key Innovation: Allroad-SLAM couples LiDAR-inertial odometry with a factor-graph backend that integrates consumer-grade GNSS constraints across diverse road environments.
67. A hierarchical framework employing voting mechanism for large-scale ALB and MBES point cloud registration
Core Problem: Airborne LiDAR bathymetry and multibeam sonar differ in density, footprint, and viewpoint, making direct large-area registration unstable.
Key Innovation: A hierarchical voting framework extracts robust correspondences and progressively aligns ALB and MBES point clouds across scales.
68. A review of machine learning applications in inland and nearshore water color remote sensing: from preprocessing to object detection and parameter estimation
Core Problem: Water-colour machine learning remains fragmented across preprocessing, object detection, parameter retrieval, sensors, and validation protocols.
Key Innovation: The review organizes the end-to-end workflow, inventories open resources, and identifies label bias, uncertainty, interpretability, multimodal fusion, and spatial transfer as unresolved constraints.
69. AFDINet: Adaptive frequency disentanglement and interaction in hybrid Mamba-CNN for optical-SAR land cover classification
Core Problem: Optical-SAR land-cover fusion must reconcile nonlinear modality differences without losing high-frequency boundaries or broad spatial context.
Key Innovation: AFDINet combines CNN locality with Mamba sequence modelling and adaptively exchanges frequency components between optical and SAR branches.
70. Self-orthorectification of satellite images using multi-view geometric constraints without external geospatial data
Core Problem: High-resolution satellite images cannot always rely on ground control or an external DEM for geometric correction in remote areas.
Key Innovation: Overlapping views and rational-polynomial camera models jointly recover geometric consistency, enabling self-orthorectification without external control data.
71. R3LIO: Robust reflectivity-assisted rotating LiDAR-inertial odometry for degenerate and unstructured environments
Core Problem: Rotating LiDAR systems gain field of view but suffer motion distortion and geometric degeneracy in sparse or repetitive environments.
Key Innovation: R3LIO uses reflectivity constraints alongside inertial and geometric residuals to stabilize odometry in degenerate, unstructured scenes.
72. GLoC: Global-local graph fusion and clique search for LiDAR place recognition
Core Problem: LiDAR place recognition must distinguish repeated global layouts while remaining tolerant to local occlusion and viewpoint change.
Key Innovation: GLoC combines global graph similarity, local vertex consistency, and maximal-clique search to verify place matches geometrically.
73. MPA: A Mamba Pyramid Aggregator for Stereo Matching of Satellite Remote Sensing Images
Core Problem: Satellite stereo matching must preserve fine disparity boundaries while aggregating context across very large image extents.
Key Innovation: MPA uses a multiscale Mamba pyramid to propagate long-range evidence at linear sequence complexity during disparity estimation.
74. Correlation-Guided Progressive Interaction and Gated Fusion for Multimodal Optical and SAR Remote Sensing Image Object Detection
Core Problem: Optical and SAR detectors lose complementary information when features are fused before correcting semantic and geometric mismatch.
Key Innovation: Correlation-guided interaction aligns modality-specific features progressively, and gated fusion retains only mutually informative responses for detection.
75. DCMArb: Decoupled-Collaborative Mamba for Arbitrary-scale Hyperspectral Super-resolution
Core Problem: Arbitrary-scale hyperspectral super-resolution must model smooth spatial continuity and irregular spectral coupling with different inductive biases.
Key Innovation: DCMArb separates spatial and spectral Mamba pathways and then coordinates them to reconstruct user-specified resolution while preserving spectral signatures.
76. A transfer learning-based contourlet residual-driven terrain-aware method for Arctic seafloor DEM super-resolution
Core Problem: Arctic seafloor DEMs are sparse and heterogeneous, limiting high-resolution terrain analysis where direct training data are scarce.
Key Innovation: Contourlet residuals and terrain-aware transfer learning recover multiscale relief while adapting knowledge from better-sampled elevation domains.
77. Enhancing point clouds for multi-scale object monitoring in large railway infrastructure scenes
Core Problem: Railway point clouds accumulated over hundreds of miles vary in density and contain gaps that obscure assets at different physical scales.
Key Innovation: A multiscale enhancement workflow regularizes large corridor clouds while preserving local object geometry for repeated infrastructure monitoring.
78. Satellite-derived daily shortwave radiation over the Tibetan Plateau considering topographic effects: Method and products
Core Problem: Coarse radiation products smooth topographic shading and slope-aspect effects across the Tibetan Plateau.
Key Innovation: A data-driven terrain correction produces daily shortwave radiation that explicitly represents complex relief within the LessRad-terrain product.
79. Spatiotemporal variability of maximum precipitation altitude and precipitation gradients in global mountain basins
Core Problem: Coarse precipitation grids obscure elevation-dependent maxima and gradients that control runoff and snow accumulation in mountain basins.
Key Innovation: Global basin analysis maps maximum-precipitation altitude and vertical gradients through time, exposing regional topographic controls unavailable from basin means.
80. Hyperbolic Prototype Guidance for Incremental SAR ATR
Core Problem: Incremental SAR recognition must add new classes without erasing prior knowledge and remain robust to clutter and sidelobes.
Key Innovation: Hyperbolic class prototypes encode hierarchical relations and guide continual updates, reducing catastrophic forgetting under noisy radar observations.
81. WDTS: Water droplet model-driven entropy optimization for individual tree skeletonization from terrestrial laser scanning point clouds
Core Problem: Tree-skeleton extraction from terrestrial laser scans often produces off-centre branches, excessive nodes, or broken topology.
Key Innovation: WDTS treats point-cloud paths as entropy-minimizing water-droplet trajectories to recover compact, centred, and topologically coherent skeletons.
82. Multiscale Sequence Aware Mamba for Hyperspectral Image Classification
Core Problem: Standard Mamba models underrepresent adjacent-band correlation and multiscale spatial structure in hyperspectral scenes.
Key Innovation: MSAM adds scale-aware spectral-spatial sequences and local band interaction while retaining linear-complexity long-range modelling.
83. Hyperspectral object tracking with dimensionality reduction and spatial-spectral-temporal modeling
Core Problem: Hyperspectral tracking must suppress redundant bands while preserving material cues across motion, occlusion, and background change.
Key Innovation: A learned dimensionality-reduction stage feeds spatial-spectral-temporal modelling, concentrating discriminative spectra before sequence tracking.
84. PoseIDON: 6DoF pose estimation with foundation model features for marine sediment burial mapping
Core Problem: Partial burial, poor visibility, and object degradation make seafloor burial geometry difficult to recover from remotely operated vehicle video.
Key Innovation: PoseIDON combines foundation-model features with multiview photogrammetry to estimate object pose and surrounding seabed orientation for sediment-burial mapping.
85. Lightweight Raw Echo Image Preprocessing for Long-Range Airborne Streak Tube Imaging LiDAR Using Adaptive Frequency-Domain Noise Suppression
Core Problem: Long-range airborne streak-tube LiDAR echoes are distorted by atmospheric speckle, detector noise, and weak returns before range estimation begins.
Key Innovation: Geometry-aware region classification and adaptive frequency-domain suppression reduce elevation error in 6 km airborne observations without a heavy learned model.
86. Integrating Deep Generative AI and Hyperspectral–Multispectral Data Fusion for Enhancing Digital Soil Mapping
Core Problem: EnMAP spectra resolve soil properties but its 30 m pixels are too coarse for field-scale digital soil maps under sparse sampling.
Key Innovation: A 1D U-Net fuses EnMAP hyperspectral and SuperDove multispectral imagery, then generative learning maps soil organic matter, salinity, and available phosphorus from 110 samples.
87. CAF-Net: A Unified Framework for Resolving Spatial–Frequency Representation Conflicts in Multimodal Remote Sensing Segmentation
Core Problem: Optical-DSM segmentation loses boundary fidelity when spatial structure and frequency-domain semantics are fused without resolving their mismatch.
Key Innovation: CAF-Net uses DSM-guided structural modulation and adaptive cross-frequency weighting to align boundaries and semantics in one multimodal architecture.
88. DWFSeg: A Dynamic Multiscale Feature Fusion and Dual Attention-Enhanced Network for High-Precision Water Body Segmentation Based on Super-Resolution Remote Sensing Imagery
Core Problem: Medium-resolution imagery misses narrow rivers and ambiguous land-water boundaries, while native high-resolution coverage is temporally sparse.
Key Innovation: Real-ESRGAN first reconstructs imagery to 2.5 m, after which dynamic multiscale fusion and dual attention extract fine water boundaries for hydrological and flood applications.
89. Physically Consistent SAR Image Generation for Unseen Aspect Angles via Attributed Scattering Center Evolution
Core Problem: Sparse aspect-angle coverage leaves SAR recognizers without physically plausible observations at unseen viewpoints.
Key Innovation: Attributed-scattering-centre evolution constrains dominant echoes while a separate background model preserves speckle statistics during cross-angle SAR synthesis.
90. S2M-Net: Dynamic Hyperspectral Unmixing Network Integrating Spectral Sequence Mamba and Local Spatial–Spectral Awareness
Core Problem: Hyperspectral unmixing requires long-range spectral context without the quadratic cost and noise sensitivity of global attention.
Key Innovation: S2M-Net couples spectral-sequence Mamba with a local spatial-spectral branch to estimate endmembers and abundances while filtering redundant bands.
91. A Novel Photogrammetry-Based Data Generation Technique for Post-Disaster Human Detection in UAV Imagery
Core Problem: Post-disaster UAV human detection lacks labelled examples spanning the camera elevations and viewing angles encountered during response.
Key Innovation: Photogrammetric 3D human models generate 3,000 synthetic training patches; accuracy rises from 88.06% to 91.08%, and to 95.21% when combined with conventional augmentation.
92. Advances and perspectives in modern geostatistics for energy, natural resources, and environmental studies
Core Problem: Modern environmental inference must reconcile sparse observations, multiscale spatial dependence, non-Gaussian behaviour, and uncertainty propagation.
Key Innovation: The review organizes recent geostatistical advances across covariance modelling, simulation, machine learning integration, and uncertainty-aware resource and environmental applications.