TerraMosaic Daily Digest: Jan 28, 2026
Daily Summary
This digest synthesizes 298 selected papers and focuses on high-resolution remote-sensing monitoring workflows, landslide process mechanics and slope evolution, flood generation, routing, and hydroclimatic forcing. Top-ranked studies examine earthquake-triggered slope response and liquefaction, operational early-warning thresholding, and flood generation and hydroclimatic forcing.
Across the full set, evidence converges on mechanism-constrained analysis with operational relevance, especially for coastal and submarine hydro-geomechanics and risk, fragility, and resilience quantification. The strongest contributions pair interpretable process evidence with monitoring or forecasting workflows that support warning design and risk prioritization.
Key Trends
- Monitoring workflows rely on integrated remote-sensing products: Multi-source satellite and airborne observations are used for deformation retrieval, change detection, and rapid post-event mapping.
- Landslide studies increasingly resolve process chains: Contributions connect triggering conditions, slope deformation, and mobility outcomes, improving the basis for warning thresholds and scenario testing.
- Flood analyses are becoming event-specific and process-based: Papers emphasize precipitation structure, antecedent wetness, and catchment controls rather than static hazard descriptors.
- Coastal and submarine hazards are treated as coupled systems: Wave, mass-transport, and shoreline processes are analyzed together with engineering implications.
- Risk studies move beyond hazard mapping to consequence pathways: Vulnerability, fragility, exposure, and recovery metrics are integrated to compare interventions under compound hazards.
Selected Papers
This digest features 298 selected papers from 2615 RSS items analyzed across multiple journals. Each paper has been evaluated for its relevance to landslide and broader geohazard research and includes links to the original publications.
1. Physics-informed deep learning links geodetic data and fault friction
Core Problem: Quantitatively assessing frictional heterogeneity from geodetic observations while fully incorporating fault mechanics is challenging due to high-dimensional optimization difficulties, leading to a lack of mechanically consistent fault slip models.
Key Innovation: Uses Physics-Informed Neural Networks (PINNs) to link frictional heterogeneity with geodetic observations, representing spatially variable frictional properties, and successfully estimates heterogeneous friction coinciding with localized SSE nucleation and propagation for the 2010 Bungo SSE, reproducing observed surface displacements and predicting future fault slip evolution.
2. Discovering multiple regimes of landslide motion with critical state soil mechanics
Core Problem: The effects of cyclic loading from seasonal groundwater table fluctuations on slow-moving landslide kinematics and potential failure are under-explored.
Key Innovation: Integration of a bounding surface modified Cam-Clay model with a sliding–consolidation framework to simulate landslide kinematics, identifying four regimes of motion and demonstrating that cyclic failure depends on stress path relative to the critical state, providing a theoretical basis for kinematic-based precursors.
3. Advancing landslide early warning in the Indian Himalayas: SIGMA-based rainfall thresholds for Chamoli district
Core Problem: Landslide early warning in the Indian Himalayas requires robust, spatially calibrated rainfall thresholds that account for distinct triggering mechanisms and geological diversity to reduce false alarms and improve predictive performance.
Key Innovation: A regional landslide early warning system (LEWS) is developed for Chamoli district using a SIGMA-based decision algorithm, defining critical rainfall thresholds as multiples of standard deviation from historical data, employing a dual time-scale framework with seasonally adjustable accumulation windows, and spatially calibrating thresholds for geological diversity, achieving strong predictive performance.
4. Spatial variation of rainfall for landslide early warning: an examination from Majiajing watershed in Yunnan, China
Core Problem: Empirical rainfall thresholds in landslide early warning systems (LEWS) suffer from high false alarm rates due to substantial rainfall uncertainty and spatial variation in mountainous regions.
Key Innovation: A two-year monitoring study in the Majiajing watershed elucidates spatial rainfall distribution patterns, revealing high correlation but significant disparity in rainfall at different elevations (upstream rainfall markedly surpasses downstream), advocating for increased rain gauge density in potential landslide detachment zones to enhance LEWS accuracy.
5. Pore-pressure generation initiating a rainfall-induced landslide: experimental insights and numerical modelling
Core Problem: Landslides with deep-seated clay-rich shear zones are highly susceptible to intense rainfall, but the precise mechanism of pore-pressure generation leading to unexpected sliding needs further investigation.
Key Innovation: This study investigates a rainfall-induced catastrophic landslide through direct shear tests under constant shear stress paths and numerical simulations with a hypoplastic model, revealing that pore-pressure buildup, accelerated by crack infiltration, is the primary driver of landslide initiation, promoting continuous basal sliding.
6. Unraveling shallow landslides in the Qinling Mountains: Novel insights into vegetation-hydrology driven mechanisms
Core Problem: Existing research on the distribution characteristics and driving mechanisms of shallow landslides in the Qinling Mountains, particularly the role of vegetation-hydrology interactions, is limited.
Key Innovation: This study integrates data analysis, field investigations, and remote sensing to analyze spatial patterns and environmental driving factors of shallow landslides in the Qinling Mountains, revealing their concentration in specific altitude/slope ranges, a positive correlation with soil moisture and groundwater enrichment, and a mediating role of vegetation (peak frequency at 50%–60% coverage), enhancing the information quantity model (AUC 0.83).
7. Investigation of Dynamic Failure Mechanism on Three-Section Type of Locked-Segment Slope Using Hilbert-Huang Transform
Core Problem: The dynamic response and failure mechanisms of three-section locked-segment slopes under earthquake loading, particularly from an energy perspective, are not well understood, limiting insights into landslide occurrence.
Key Innovation: Extended numerical simulations of shaking table tests using Hilbert-Huang transform and marginal spectrum theory to elucidate the dynamic response and progressive failure mechanism of three-section locked-segment slopes under seismic loading, revealing the influence of input waves and weak interlayers on energy distribution and identifying a progressive failure mode.
8. Optimizing check dams construction and afforestation for debris flows mitigation in slag and landslides- prone gullies
Core Problem: Mitigating debris flows in gullies prone to slag and landslides.
Key Innovation: Optimizing the construction of check dams and afforestation strategies for debris flow mitigation in specific vulnerable gullies.
9. Coupled seepage–deformation analysis of the dynamics of embankment with elastoplasticity based on the full formulation
Core Problem: Analyzing the coupled seepage-deformation dynamics of embankments, considering elastoplasticity and using a full formulation.
Key Innovation: A full formulation-based coupled seepage-deformation analysis for elastoplastic embankment dynamics.
10. From typhoon rainfall to slope failure: optimizing susceptibility models and dynamic thresholds for landslide warnings in Zixing City, China
Core Problem: Existing landslide warning systems inadequately capture the distinct rainfall dynamics of typhoon-specific rainfall-induced landslides, leading to limitations in predicting and warning for these critical hazards.
Key Innovation: An integrated framework combining optimized susceptibility predictions (using machine learning with buffer-based negative sampling and variable weighting) and dynamic rainfall thresholds tailored to typhoon patterns, demonstrating high spatial efficiency for landslide warnings in cyclone-prone regions.
11. Investigation of hydromechanical behaviour of unsaturated hydrophobic soil from micro to macro
Core Problem: The implications of convex menisci in unsaturated hydrophobic soils (often found after wildfires) for their hydromechanical behavior, particularly shear strength, are not fully understood, leading to potential overestimation of strength.
Key Innovation: An investigation of unsaturated hydrophobic soil behavior using a custom direct shear apparatus and µCT, revealing that positive, stress-dependent pore water pressure reduces dilatancy (60%) and peak shear strength (24%) compared to dry grains, highlighting the risk of overestimating strength if wettability is ignored.
12. A framework for evaluating thaw settlement in permafrost regions
Core Problem: Thawing permafrost, driven by climate change, poses escalating challenges for northern infrastructure, and traditional design strategies are unsustainable, requiring improved methods for assessing thaw settlement.
Key Innovation: A Thaw Settlement Evaluation Framework for permafrost regions, comprising regional thaw settlement hazard assessments and site-specific thaw strain estimations, consolidating established methodologies into adaptable workflows to provide actionable insights for infrastructure planning and design.
13. The analysis of failure mechanism and crack propagation patterns of borehole grout stopper ring structure induced by non-uniform geostress in nearshore areas
Core Problem: Understanding the failure mechanisms and crack propagation in borehole grout stopper rings caused by non-uniform geostress in nearshore environments.
Key Innovation: Analyzing the failure mechanism and crack propagation patterns induced by non-uniform geostress in borehole grout stopper rings.
14. Spatiotemporal Evolution of Surface Subsidence in Large-Scale Mining Areas Under Rainfall Influence and Optimization Model Development
Core Problem: Traditional monitoring methods struggle with large-scale surface subsidence monitoring in mining areas, and settlement prediction models face difficulties in development and hyperparameter acquisition, especially under rainfall influence.
Key Innovation: Integration of rainfall data and SBAS-InSAR technology to analyze spatiotemporal evolution of surface subsidence in mining areas, and development of an APO-BiLSTM settlement prediction model (Arctic Puffin Optimization for BiLSTM hyperparameters), reducing single-step prediction RMSE by 79.8% and MAE by 79.1% compared to LSTM and BiLSTM, with 78.3% of points having <4mm error.
15. Analysis of the Elastoplastic Fracture Initiation Characteristics at the Tip of Non-saturated Slanted Rock Open Cracks Under Ice–Gas Pressure
Core Problem: Complex and unobservable internal crack propagation within surface rock fissures sealed by freezing in alpine regions, driven by ice-gas pressure, makes it challenging to understand and predict fracture initiation characteristics.
Key Innovation: Derived elastoplastic fracture mechanics formulas to analyze crack initiation characteristics (GPIF, plastic zone, SIF, initiation angle, stress) at the tip of non-saturated slanted rock cracks under ice-gas pressure, revealing the significant influence of freezing temperature, gas volume ratio, and crack geometry on fracture behavior in alpine regions.
16. Robustness of critical soil moisture to curve-fitting methods and its variability with soil depth, soil texture, and climatic conditions: insights from lysimeter data in Germany
Core Problem: Understanding the robustness and variability of critical soil moisture, a key factor in slope stability and landslide initiation.
Key Innovation: Insights from lysimeter data in Germany on how curve-fitting methods, soil depth, soil texture, and climatic conditions influence critical soil moisture.
17. Influence of oak trees on soil hydraulic characteristics: A spatial analysis in a Mediterranean silvo-pastoral ecosystem
Core Problem: Analyzing the spatial influence of oak trees on soil hydraulic characteristics within a Mediterranean silvo-pastoral ecosystem.
Key Innovation: Spatial analysis revealing the influence of oak trees on soil hydraulic characteristics.
18. On the integration of an ACST-based bounding surface model
Core Problem: Integrating an ACST-based bounding surface model for geotechnical applications.
Key Innovation: Integration of an ACST-based bounding surface model.
19. Extension of explicit Runge-Kutta substepping stress integration for viscoplastic model of saturated soils
Core Problem: Extending explicit Runge-Kutta substepping stress integration for viscoplastic models of saturated soils.
Key Innovation: Extension of explicit Runge-Kutta substepping stress integration for viscoplastic saturated soil models.
20. Shear dilatancy and friction characteristics of calcareous sand under biaxial rotation of principal stress
Core Problem: Investigating the shear dilatancy and friction characteristics of calcareous sand under biaxial rotation of principal stress.
Key Innovation: Characterization of shear dilatancy and friction of calcareous sand under biaxial principal stress rotation.
21. Bedrock ledges, colluvial wedges, and ridgetop wetlands: characterizing geomorphic and atmospheric controls on the 2023 Wrangell landslide to inform landslide assessment in Southeast Alaska, USA
Core Problem: Lack of understanding of regional geomorphic and atmospheric controls on landslide triggering events and runout behaviour, specifically highlighted by the 2023 Wrangell landslide, hindering regional landslide hazard assessment.
Key Innovation: Characterization of geomorphic, hydrologic, and atmospheric conditions (bedrock ledges, colluvial wedges, ridgetop wetlands, rapid snowmelt, drainage) contributing to the Wrangell landslide using field observations, lidar, geotechnical analyses, and climate data, suggesting these factors for future hazard assessment.
22. Landslide recognition with sample augmentation based on a joint DCGAN and Pix2Pix
Core Problem: Improving the accuracy and robustness of landslide recognition, likely addressing challenges related to limited training data.
Key Innovation: A novel sample augmentation method for landslide recognition based on a joint Deep Convolutional Generative Adversarial Network (DCGAN) and Pix2Pix model.
23. Dual-Modality IoT Framework for Integrated Access Control and Environmental Safety Monitoring with Real-Time Cloud Analytics
Core Problem: Traditional physical security and environmental safety monitoring systems are independent silos, leading to operational inefficiencies, delayed emergency responses, and increased management complexity.
Key Innovation: A comprehensive dual-modality IoT framework integrating RFID-based access control with multi-sensor environmental safety monitoring via a unified cloud architecture, demonstrating high accuracy, reliability, and cost-effectiveness for synergistic security-safety integration.
24. An Empirical Investigation of Neural ODEs and Symbolic Regression for Dynamical Systems
Core Problem: Accurately modeling the dynamics of complex systems and discovering their governing differential equations from noisy, limited data are critical tasks for scientific discovery, with existing methods like Neural Ordinary Differential Equations (NODEs) and Symbolic Regression (SR) having limitations in extrapolation and equation recovery.
Key Innovation: An empirical investigation demonstrating that NODEs can extrapolate effectively to new boundary conditions under dynamic similarity, and SR successfully recovers equations from noisy ground-truth data. Crucially, SR recovers two out of three governing equations (and approximates the third) using data generated by a NODE trained on only 10% of the full simulation, suggesting a promising new approach for scientific discovery.
25. Physics-Guided Multimodal Transformers are the Necessary Foundation for the Next Generation of Meteorological Science
Core Problem: Current AI models in meteorological and climate sciences are fragmented, often lack scientific consistency, violate fundamental physical laws, and struggle to scale across heterogeneous data, while existing 'hybrid' attempts are ad-hoc.
Key Innovation: Advocates for a unified paradigm of physics-guided multimodal transformers as the necessary foundation for next-generation AI in meteorological and climate sciences, arguing that this architecture can systematically integrate domain knowledge via physical constraint embedding and physics-informed loss functions, leading to scientifically grounded and robust AI systems.
26. Developing time-dependent fragility curves for cube-armored breakwaters: stochastic modeling and experimental quantification
Core Problem: Assessing the time-dependent vulnerability of cube-armored breakwaters to coastal hazards.
Key Innovation: Developing time-dependent fragility curves for breakwaters using stochastic modeling and experimental quantification.
27. Strength and deformation characteristics of coral sand accounting for particle breakage
Core Problem: Understanding the geotechnical strength and deformation characteristics of coral sand, particularly when particle breakage occurs.
Key Innovation: Investigating the strength and deformation characteristics of coral sand, explicitly accounting for particle breakage.
28. Climate change-driven shoreline change along the Catalan coast (NW Mediterranean): A probabilistic approach for risk-informed coastal management.
Core Problem: Assessing climate change-driven shoreline change and its implications for coastal management along the Catalan coast.
Key Innovation: A probabilistic approach for risk-informed coastal management concerning climate change-driven shoreline change along the Catalan coast.
29. Improving regional tsunami early warning with seismic array techniques
Core Problem: Sparsely instrumented regions like the Southwest Pacific struggle to consistently meet operational tsunami forecasting targets after large earthquakes.
Key Innovation: A method is presented to improve regional tsunami early warning by utilizing novel seismic array processing and a hybrid approach combining array methods, W-phase inversion, and tsunami wave analysis to directly delineate earthquake rupture extent in near-real time, enhancing forecast accuracy and timeliness.
30. The influence of geological factors on fire vulnerability in Northwestern Russia’s boreal region
Core Problem: The specific factors controlling wildfire occurrence and spread in northern latitudes, particularly the role of geological settings, remain insufficiently understood.
Key Innovation: This study provides a comprehensive assessment of wildfire susceptibility in the Republic of Karelia, focusing on geological settings (Quaternary sediments) alongside climatic, topographic, and anthropogenic factors, revealing that large wildfires are predominantly associated with glacio-lacustrine sandy and moraine deposits and that geological factors demonstrate an independent contribution to fire vulnerability.
31. A statistical approach to foreshocks discrimination in Italy
Core Problem: Real-time discrimination between seismic clusters that lead to large mainshocks (foreshocks) and those that wane (swarms) is an outstanding challenge in short-term seismic forecasting.
Key Innovation: This study analyzes small and large mainshock clusters in Italy, suggesting that seismicity spreads over larger areas, has higher magnitude variance and entropy, and grows in number and cumulative seismic moment with impending mainshock magnitude, proposing a statistical approach to assess the probability of a cluster culminating in a large mainshock.
32. Preparation of flood potential maps using machine learning and comparison of their performance
Core Problem: Accurate flood susceptibility maps are crucial for flood control and management, but there is a need to evaluate and compare the performance of various machine learning models for their creation.
Key Innovation: This study creates flood potential maps for the Borujerd-Dorud basin using various machine learning models (Deep Learning, CatBoost, XGB, RF, KNN, SVM), evaluates their performance with spatial group k-fold cross-validation, and identifies Random Forest as the most accurate model (AUC = 0.71), demonstrating its reliability for prioritizing flood mitigation efforts.
33. Pillar Stability Assessment in Deep Mining Using a Stacked Ensemble Model with Interpretable Features
Core Problem: Accurately and interpretably assessing pillar stability in deep hard rock mining is crucial for safety and resource recovery, but existing predictive methods may lack performance and transparency.
Key Innovation: Developed a stacked ensemble machine learning (EnML) framework, optimized by a Modified Equilibrium Optimizer (m-EO), for deep mining pillar stability assessment, achieving high accuracy and providing global and local interpretability through SHAP and ICE, identifying stress-to-strength ratio as the dominant failure predictor.
34. Stress-Induced Failure Mechanisms of a Shallow, Large-Span Cavern: Insights from Field Observations and Numerical Back analysis
Core Problem: Understanding the stress-induced failure mechanisms in shallow, large-span caverns excavated in horizontally bedded rock under elevated horizontal in situ stress is challenging, especially with limited direct observations of laminated rock beam behavior.
Key Innovation: Conducted a detailed investigation combining field observations and 3DEC numerical back analysis to elucidate the tensile-shear interaction governing stress-induced failure in a shallow, large-span cavern, revealing higher-than-predicted horizontal stresses, validating the CWFS model for laminated sandstone, and proposing a bedding shear threshold for re-bolting.
35. Application of Building Information Modelling to Tension-Based Ground Reinforcement Systems
Core Problem: The application of Building Information Modelling (BIM) in geotechnical engineering, particularly for tension-based ground reinforcement systems (e.g., soil nails, rock bolts), is hindered by a lack of standardized modelling methods and functional data structures that cover the full project lifecycle.
Key Innovation: Developed a novel, generalizable BIM data model for tension-supporting ground reinforcement elements, integrating the full project lifecycle (design, installation, inspection, maintenance) and defining LOD requirements for geometry and metadata, validated through real-world Norwegian infrastructure projects including slopes.
36. Strain localization in rock: From multi-scale measurement to AI-driven prediction
Core Problem: Understanding and predicting strain localization in rock, a precursor to failure, across multiple scales is challenging, and current methods may lack predictive capabilities.
Key Innovation: Developed a framework that integrates multi-scale measurement with AI-driven prediction to better understand and forecast strain localization in rock.
37. The Role of Social Infrastructure in Community-Based Disaster Resilience: A Case Study of the 2024 Noto Peninsula Earthquake
Core Problem: The specific role of social infrastructure in fostering community-based disaster resilience, particularly in the aftermath of major events like the 2024 Noto Peninsula Earthquake, needs further investigation.
Key Innovation: Investigated the role of social infrastructure in community-based disaster resilience, using the 2024 Noto Peninsula Earthquake as a case study.
38. Cross-system modeling and analysis of cascading failure propagation in large-scale metro stations under extreme flooding events
Core Problem: Analyzing and modeling cascading failure propagation in large-scale metro stations when subjected to extreme flooding events.
Key Innovation: A cross-system modeling and analysis approach to understand cascading failure propagation in metro stations during extreme flooding events.
39. A comprehensive framework for assessing critical behavior of lead rubber bearings under Multiaxial-Loading: Numerical, theoretical, collapse probability, and seismic analysis
Core Problem: Developing a comprehensive framework to assess the critical behavior of lead rubber bearings under multiaxial and seismic loading, including collapse probability.
Key Innovation: A comprehensive framework for assessing critical behavior and collapse probability of lead rubber bearings under multiaxial and seismic loading.
40. Three-dimensional analytical model for the dynamic interaction between the unsaturated subsoil and surface foundations
Core Problem: Developing a three-dimensional analytical model for the dynamic interaction between unsaturated subsoil and surface foundations.
Key Innovation: A 3D analytical model for dynamic interaction between unsaturated subsoil and surface foundations.
41. Visco-elastic interface and dynamic response of a lined tunnel in anisotropic rock mass under unloading waves
Core Problem: Analyzing the dynamic response of a lined tunnel in anisotropic rock mass under unloading waves, considering a visco-elastic interface.
Key Innovation: Analysis of dynamic response of lined tunnels in anisotropic rock mass under unloading waves with a visco-elastic interface.
42. Experimental study and performance assessment of the high polymerization displacement amplification damping system
Core Problem: Experimentally studying and assessing the performance of a high polymerization displacement amplification damping system.
Key Innovation: Experimental study and performance assessment of a high polymerization displacement amplification damping system.
43. Mesh, Hydrodynamic Boundary, and Uncertainty Analysis of the 2D‐SWEs: Taking Numerical Simulation of River Networks as an Example
Core Problem: Insufficient predictive ability and inefficient computation in two-dimensional shallow water equations (2D-SWEs) for river network simulations, stemming from improper boundary settings and mesh selection.
Key Innovation: A deep coupling of mesh and hydrodynamic boundary, proposing a hydrodynamic boundary classification framework, and systematically quantifying uncertainty and computational performance of various meshes for 2D-SWEs, with recommendations for mesh usage in river channels and floodplains.
44. Evaluating the feasibility of scaling the FIER framework for large-scale flood inundation prediction
Core Problem: Traditional flood forecasting methods face computational and data challenges for large geographic areas, particularly in data-scarce regions, hindering accurate and timely flood inundation forecasts.
Key Innovation: A novel approach to scaling the data-driven FIER framework for large-scale flood inundation prediction by creating individual FIER models using watershed boundaries and mosaicking results, demonstrating improved accuracy for flood and low flow cases in the Upper Mississippi Alluvial Plain, offering a promising solution for data-scarce regions.
45. Geospatial assessment and mapping of water-induced soil erosion in a semiarid region of the MENA using GIS-based RUSLE modeling
Core Problem: Quantifying the spatial patterns and annual rates of water-induced soil loss in a semiarid watershed to support effective land-management and erosion-control strategies.
Key Innovation: A GIS-based implementation of the Revised Universal Soil Loss Equation (RUSLE) to produce a basin-wide map of potential soil erosion, revealing spatial heterogeneity and identifying priority areas for conservation.
46. Insights into the Arabia–Anatolia plate collision from integrated SAR analysis and detailed modelling of the 2023 Türkiye–Syria earthquakes
Core Problem: Understanding the complex surface deformation and fault kinematics of the 2023 Türkiye–Syria earthquakes and their implications for future seismic hazard, addressing limitations of previous SAR analyses.
Key Innovation: Integrated SAR analysis, including high-azimuth-resolution SAOCOM-1 data, to derive a detailed displacement field and an elaborated 22-segment fault model, providing new constraints on strain partitioning and future seismic hazard along the East Anatolian Fault Zone.
47. Probabilistic Sensing: Intelligence in Data Sampling
Core Problem: Extending sensor intelligence to the data-acquisition process (deciding whether to sample or not) for transformative energy-efficiency gains, while avoiding information loss from deterministic decisions and overcoming sub-sampling-rate response time limits.
Key Innovation: Presents a probabilistic sensing paradigm employing a probabilistic neuron (p-neuron) driven by an analog feature extraction circuit, enabling real-time intelligent autonomous activation of data-sampling, validated on active seismic survey data to achieve lossless probabilistic data acquisition with significant energy and sample savings.
48. Evolving beyond collapse: An adaptive particle batch smoother for cryospheric data assimilation
Core Problem: Particle-based data assimilation schemes for cryospheric applications, such as the Particle Batch Smoother (PBS), suffer from ensemble collapse, limiting their robustness and adaptability to complex, distributed multiyear simulations.
Key Innovation: Presents the Adaptive Particle Batch Smoother (AdaPBS), a new adaptive particle-based data assimilation scheme that extends PBS with the AMIS algorithm, improving resilience against ensemble collapse, enabling early-stopping strategies, and automatically adapting computational cost, demonstrating superior or matching performance in assimilating snow depth observations across various sites.
49. Cheap2Rich: A Multi-Fidelity Framework for Data Assimilation and System Identification of Multiscale Physics -- Rotating Detonation Engines
Core Problem: Bridging the sim2real gap between computationally inexpensive models and complex physical systems, especially in multi-scale settings where reduced-order models capture only dominant dynamics.
Key Innovation: Cheap2Rich, a multi-scale data assimilation framework that reconstructs high-fidelity state spaces from sparse sensor histories by combining a fast low-fidelity prior with learned, interpretable discrepancy corrections, demonstrated on rotating detonation engines.
50. GVGS: Gaussian Visibility-Aware Multi-View Geometry for Accurate Surface Reconstruction
Core Problem: Accurate surface reconstruction using 3D Gaussian Splatting remains challenging due to unreliable multi-view constraints under large geometric discrepancies and scale ambiguity/local inconsistency of monocular depth priors.
Key Innovation: GVGS, which introduces a Gaussian visibility-aware multi-view geometric consistency constraint and a progressive quadtree-calibrated monocular depth constraint, enabling more accurate and stable geometric supervision and improving geometric accuracy over prior methods.
51. Test-Time Adaptation for Anomaly Segmentation via Topology-Aware Optimal Transport Chaining
Core Problem: Anomaly segmentation methods, especially those based on thresholding, produce brittle masks under distribution shift, and existing methods lack robust adaptation under domain shift.
Key Innovation: TopoOT, a topology-aware optimal transport framework that integrates multi-filtration persistence diagrams with test-time adaptation, using Optimal Transport Chaining to align PDs and yield geodesic stability scores for robust adaptation under domain shift.
52. TINNs: Time-Induced Neural Networks for Solving Time-Dependent PDEs
Core Problem: Standard space-time PINNs for time-dependent PDEs reuse a single network with shared weights across all times, forcing the same features to represent different dynamics, degrading accuracy and destabilizing training.
Key Innovation: Time-Induced Neural Networks (TINNs), a novel architecture that parameterizes network weights as a learned function of time, allowing the effective spatial representation to evolve while maintaining shared structure, improving accuracy and convergence for time-dependent PDEs.
53. Unsupervised Anomaly Detection in Multi-Agent Trajectory Prediction via Transformer-Based Models
Core Problem: Identifying safety-critical scenarios in autonomous driving is challenging due to the rarity of such events, and existing methods lack a systematic way to verify if statistical anomalies reflect physical danger.
Key Innovation: An unsupervised anomaly detection framework based on a multi-agent Transformer that models normal driving and measures deviations through prediction residuals, with a dual evaluation scheme for detection stability and physical alignment, identifying subtle multi-agent risks.
54. A Foundation Model for Virtual Sensors
Core Problem: Existing virtual sensor approaches are application-specific, require hand-selected inputs, cannot leverage task synergies, and lack consistent benchmarks. Meanwhile, emerging time series foundation models are computationally expensive and limited to predicting their input signals, making them incompatible with virtual sensors.
Key Innovation: The first foundation model for virtual sensors, a unified model that can simultaneously predict diverse virtual sensors, exploit synergies, learn relevant input signals for each sensor (adding explainability), and maintain computational efficiency. It achieves significant reductions in computation time (415x) and memory (951x) while maintaining or improving predictive quality on large-scale datasets.
55. MuRAL-CPD: Active Learning for Multiresolution Change Point Detection
Core Problem: Traditional Change Point Detection (CPD) methods often rely on unsupervised techniques, lacking adaptability to task-specific definitions of change and unable to benefit from user knowledge, which limits their accuracy and interpretability in real-world applications.
Key Innovation: MuRAL-CPD, a novel semi-supervised method that integrates active learning into a multiresolution CPD algorithm. It leverages a wavelet-based multiresolution decomposition to detect changes across multiple temporal scales and incorporates user feedback to iteratively optimize key hyperparameters, improving both accuracy and interpretability on several real-world datasets with minimal supervision.
56. FC-PINO: High Precision Physics-Informed Neural Operators via Fourier Continuation
Core Problem: Standard Physics-Informed Neural Operators (PINO) using spectral differentiation struggle with non-periodic and non-smooth Partial Differential Equations (PDEs) due to periodicity assumptions, leading to significant errors and Gibbs phenomena.
Key Innovation: FC-PINO (Fourier-Continuation-based Physics-Informed Neural Operator) which integrates Fourier continuation to transform non-periodic signals into periodic functions, enabling fast and accurate derivative computations and providing robust, high-precision solutions for a wider range of PDE problems.
57. Dense-SfM: Structure from Motion with Dense Consistent Matching
Core Problem: Traditional Structure from Motion (SfM) methods rely on sparse keypoint matching, limiting both accuracy and point density, especially in texture-less areas, for 3D reconstruction.
Key Innovation: Dense-SfM integrates dense matching with a Gaussian Splatting (GS) based track extension for more consistent, longer feature tracks, and uses a multi-view kernelized matching module for robust track refinement, significantly improving accuracy and density in 3D reconstruction from multi-view images.
58. Range Image-Based Implicit Neural Compression for LiDAR Point Clouds
Core Problem: Efficiently compressing LiDAR point clouds for high-precision 3D scene archives is challenging, as conventional image compression techniques are limited by differences in bit precision and pixel value distribution in range images (RIs).
Key Innovation: Proposes a novel implicit neural representation (INR)-based RI compression method that divides RIs into depth and mask images, compressing them using patch-wise and pixel-wise INR architectures with model pruning and quantization, outperforming existing methods in 3D reconstruction and detection quality at low bitrates.
59. DeepBooTS: Dual-Stream Residual Boosting for Drift-Resilient Time-Series Forecasting
Core Problem: Time-series forecasting methods exhibit compromised robustness to concept drift due to the pronounced non-stationarity of time-series data, despite the use of instance normalization.
Key Innovation: DeepBooTS, a novel end-to-end dual-stream residual-decreasing boosting method for time-series forecasting that progressively reconstructs the intrinsic signal, where each deep model block acts as an ensemble of learners with an auxiliary output branch correcting residuals of previous blocks, enhancing robustness to concept drift.
60. TBC: A Target-Background Contrast Metric for Low-Altitude Infrared and Visible Image Fusion
Core Problem: Traditional no-reference metrics for low-altitude Infrared and Visible Image Fusion (IVIF) in UAV reconnaissance fail in complex low-light environments (the "Noise Trap"), paradoxically assigning higher scores to degraded images and misguiding algorithm optimization.
Key Innovation: The Target-Background Contrast (TBC) metric, inspired by Weber's Law, which focuses on the relative contrast of salient targets rather than global statistics, penalizes background noise, rewards target visibility, and exhibits high "Semantic Discriminability" and computational efficiency for intelligent UAV systems.
61. The Ensemble Schr{\"o}dinger Bridge filter for Nonlinear Data Assimilation
Core Problem: Nonlinear data assimilation, especially in chaotic and high-dimensional environments, requires robust and efficient filtering methods that can handle nonlinear dynamics without structural model error or high computational cost.
Key Innovation: The Ensemble Schrödinger Bridge nonlinear filter, a novel approach for nonlinear data assimilation that marries a standard prediction procedure with diffusion generative modeling for the analysis procedure, realizing one filtering step that is derivative-free, training-free, highly parallelizable, and performs well in highly nonlinear, chaotic environments.
62. EVEREST: An Evidential, Tail-Aware Transformer for Rare-Event Time-Series Forecasting
Core Problem: Forecasting rare events in multivariate time-series data is challenging due to severe class imbalance, long-range dependencies, and distributional uncertainty.
Key Innovation: EVEREST, a transformer-based architecture, provides probabilistic rare-event forecasting with calibrated predictions and tail-aware risk estimation. It integrates a learnable attention bottleneck, an evidential head for uncertainty, an extreme-value head for tail risk, and a lightweight precursor head for early-event detection, achieving state-of-the-art performance on space-weather data.
63. Progressive-Resolution Policy Distillation: Leveraging Coarse-Resolution Simulations for Time-Efficient Fine-Resolution Policy Learning
Core Problem: In autonomous excavation using reinforcement learning, fine-resolution simulations are accurate but time-consuming, while coarse-resolution simulations are fast but less realistic, creating a trade-off in policy learning.
Key Innovation: Progressive-Resolution Policy Distillation (PRPD) is a novel framework that progressively transfers policies through middle-resolution simulations with conservative policy transfer, leveraging coarse-resolution simulations for pre-training to reduce sampling time by over 7x while maintaining task success rates comparable to fine-resolution learning.
64. Probabilistic inference of wave parameters and extreme wind-wave analysis for OWTs under typhoon conditions
Core Problem: Accurately inferring wave parameters and analyzing extreme wind-wave conditions for Offshore Wind Turbines (OWTs) during typhoons.
Key Innovation: Developing a probabilistic inference method for wave parameters and an extreme wind-wave analysis specifically for OWTs under typhoon conditions.
65. Spectra-Diffusion: A physics-consistent two-stage framework for directional wave spectrum prediction
Core Problem: Accurately predicting directional wave spectra.
Key Innovation: Introducing Spectra-Diffusion, a physics-consistent two-stage framework for directional wave spectrum prediction.
66. On the buoyancy production term for Reynolds-averaged modelling of breaking waves
Core Problem: Improving the accuracy of Reynolds-averaged modeling of breaking waves, specifically concerning the buoyancy production term.
Key Innovation: Investigation and refinement of the buoyancy production term for Reynolds-averaged modeling of breaking waves.
67. Multifactor Spatial Downscaling of Satellite Precipitation Based on Vegetation Index and Elevation
Core Problem: Many environmental applications require high spatial resolution and accurate precipitation data, but satellite remote sensing products are often too coarse, and ground observations are unavailable in many regions.
Key Innovation: The landcover similarity and distance weighted regression with residual correction model (LDWRR) for spatial downscaling of satellite precipitation (IMERG from 10 to 1 km), incorporating vegetation index and topography, and including a residual correction procedure, achieving improved accuracy (correlation coefficients > 0.7 daily, >4% improvement over IMERG) over the Pearl River Basin.
68. A GNSS PPP-Based Quantification Method for Tropospheric Delay Nonisotropy
Core Problem: Tropospheric delay nonisotropy is a major error source in high-precision GNSS positioning, and existing quantification methods rely on complete SPD grid data, limiting their application in real-time precise point positioning (PPP).
Key Innovation: A GNSS PPP-based quantification method for tropospheric delay nonisotropy that directly solves for the nonisotropic value using GNSS-observation data, overcoming dependence on full azimuthal coverage, improving static PPP positioning accuracy by 12.42% (east), 4.93% (north), and 10.57% (up), and reducing convergence time by 17.50%.
69. Spatiotemporal Heterogeneity in Greenland Firn From the Synthesis of Satellite Radar Altimetry and Passive Microwave Measurements
Core Problem: Understanding the spatiotemporal properties of the Greenland Ice Sheet firn layer is crucial for assessing overall ice sheet mass balance and internal meltwater storage capacity, but recovering vertical firn density heterogeneity over time is challenging.
Key Innovation: Investigation into recovering vertical firn density heterogeneity over a decade from the synthesis of passive microwave (SMOS) and active radar altimetry (SARAL, CryoSat-2) measurements, using the mismatch between observations and a forward model as a proxy, demonstrating clear long-term patterns of firn heterogeneity linked to extreme melt seasons.
70. A Case Study of the Tornadic Supercell in the Province of Pampanga, Philippines (27 May 2024)
Core Problem: Understanding the atmospheric conditions and mechanisms conducive to tornadic activity in complex tropical environments like the Philippines is limited.
Key Innovation: This study provides an integrated damage assessment, visual evaluation, environmental context, and remote sensing analysis of a specific tornadic supercell event, revealing its path, intensity, and atmospheric features, and highlighting similarities to well-documented tornadic events in North America.
71. The Puerto Rico Disaster Research Network (PR-DRN)
Core Problem: Despite increased disaster-related research in Puerto Rico, significant gaps remain in understanding long-term recovery, socio-economic impacts, and multi-hazard interactions, hindering comprehensive disaster risk understanding and effective policy.
Key Innovation: This project systematically catalogs scholarly literature on disasters in Puerto Rico, identifies key contributors and critical research gaps (e.g., long-term recovery, multi-hazard integration), and establishes the Puerto Rico Disaster Research Network (PR-DRN) to foster interdisciplinary collaboration and enhance disaster research and response strategies.
72. Multi-decadal geospatial trend analysis reveals anthropogenic control of groundwater-level decline across the Malwa region of Northwestern India
Core Problem: The long-term sustainability of groundwater in the Malwa Region of Northwestern India is uncertain, and the drivers of its decline need to be systematically identified.
Key Innovation: This study uses 25 years of monitoring-well data and CHIRPS precipitation data to map groundwater-level decline across the Malwa region, revealing dramatic expansion of deep water table zones and identifying tubewell density (r = 0.67) and paddy acreage (r = 0.58) as the strongest anthropogenic controls, explaining 71% of depletion trends.
73. Calculation of fill foundation deformation based on a modified unsaturated soil constitutive model
Core Problem: Existing computational and analytical methods inadequately capture the complex mechanical behavior of high fill foundations comprising soil–stone mixtures, leading to inaccurate deformation predictions.
Key Innovation: This study modifies incremental nonlinear and elastoplastic constitutive models for unsaturated soils and develops a calculation method for vertical and lateral deformations of high fill foundations using a layered summation approach, demonstrating that the modified models accurately capture stress-strain behaviors and are applicable to both elastic and elastoplastic foundation states.
74. Effect of Shear Rate on Crack Development, Fracture Surface Morphology and Permeability Characteristics in Sandstone
Core Problem: The influence of shear rate on crack development, fracture surface morphology, and permeability characteristics in sandstone during direct shear is not fully understood, which is crucial for predicting rock mass behavior.
Key Innovation: Conducted direct shear tests on sandstone at varying shear rates, using AE monitoring and 3D scanning, to reveal that higher shear rates increase shear crack proportion, promote transgranular cracks, reduce fracture roughness, intensify pore pressure fluctuations, and enhance seepage capacity, with low-roughness surfaces exhibiting higher seepage post-failure.
75. Characterizing Failure Mechanism of Soft and Hard Rocks: Implication from Acoustic Emission and Machine Learning
Core Problem: Differentiating damage evolution and objectively identifying stress states between soft and hard rocks at the laboratory scale is challenging but essential for understanding rock instability failure mechanisms.
Key Innovation: Systematically investigated fracture evolution in soft and hard rocks using acoustic emission (AE) parameters, revealing distinct phased characteristics, microcrack evolution, and failure mechanisms (plastic vs. brittle), and developed a high-accuracy machine learning model to classify rock fracture states based on multi-parameter AE data.
76. Influence of Arch Height on the Stability of Roadway Surrounding Rock
Core Problem: The influence of semicircular arch height on the stability of roadway surrounding rock, particularly its deformation and failure evolution, is not fully understood, leading to challenges in optimal design and safety control.
Key Innovation: Investigated the influence of semicircular arch height on roadway surrounding rock stability using numerical and physical similarity simulations, identifying stress concentration areas, demonstrating increased displacement with arch height, pinpointing primary damage zones, and determining an optimal arch height of 1.5 m for a phosphate mine.
77. A Statistical Moment Based Framework for Probabilistic Assessment of Tunnels with Correlated Input Variables
Core Problem: Accurate probabilistic assessment of tunnel convergence and stability is challenging due to uncertain and potentially correlated geotechnical properties, and existing methods may underestimate failure probability for high variability.
Key Innovation: Introduced a novel statistical moment-based framework for probabilistic assessment of tunnel stability, capable of incorporating correlated input variables, demonstrating improved accuracy compared to point estimation methods for higher coefficients of variation, and offering a generalized approach for both analytical and numerical tunnel models.
78. Effect of soda residue on Skeleton formation and strength development in soil stabilization
Core Problem: Understanding the effect of soda residue on the skeleton formation and strength development during soil stabilization is important for optimizing ground improvement techniques.
Key Innovation: Investigated the effect of soda residue on the skeleton formation and strength development in soil stabilization.
79. Field and numerical investigations of canal damage characteristics and mechanisms under coupled drying-wetting and freezing-thawing cycles
Core Problem: Understanding the damage characteristics and mechanisms of canals under coupled drying-wetting and freezing-thawing cycles is crucial for their long-term stability and performance.
Key Innovation: Conducted field and numerical investigations to analyze the damage characteristics and mechanisms of canals subjected to coupled drying-wetting and freezing-thawing cycles.
80. An innovative transform mapping and visualization of fracture persistence from borehole-group image analysis: MFP<sub>bia</sub>
Core Problem: Accurately mapping and visualizing fracture persistence from borehole-group image analysis is challenging but essential for understanding rock mass behavior.
Key Innovation: Developed an innovative transform mapping and visualization method (MFPbia) for fracture persistence derived from borehole-group image analysis.
81. An updated atmospheric transmittance model including sulfur dioxide absorption developed for geostationary satellite hyperspectral infrared sounders
Core Problem: Developing a more accurate atmospheric transmittance model for hyperspectral infrared sounders by including sulfur dioxide absorption.
Key Innovation: An updated atmospheric transmittance model incorporating sulfur dioxide absorption, specifically developed for geostationary satellite hyperspectral infrared sounders, relevant for volcanic gas monitoring.
82. Annual national tree canopy cover mapping: A novel workflow with temporal transferability and improved uncertainty quantification
Core Problem: Developing a robust and temporally transferable workflow for annual national tree canopy cover mapping with improved uncertainty quantification, relevant for landslide susceptibility.
Key Innovation: A novel workflow for annual national tree canopy cover mapping, featuring temporal transferability and improved uncertainty quantification, providing crucial input for landslide susceptibility models.
83. Characterization of nutrient release from volcanic rocks in soils from the southern Main Ethiopian Rift
Core Problem: Understanding the nutrient release characteristics from volcanic rocks into soils in a specific region.
Key Innovation: Characterization of nutrient release from volcanic rocks in soils located in the southern Main Ethiopian Rift, relevant for understanding volcanic soil properties.
84. Multi-link network modeling of water resource systems: identifying critical linkages driving resilience dynamics
Core Problem: Identifying critical linkages within multi-link network models of water resource systems that drive resilience dynamics.
Key Innovation: A multi-link network modeling approach to identify critical linkages influencing the resilience dynamics of water resource systems, crucial for flood management.
85. Distribution of soil functional genes and enzymes jointly determined by slope positions and latitude in Mollisols belt
Core Problem: Understanding how slope positions and latitude jointly determine the distribution of soil functional genes and enzymes in Mollisols.
Key Innovation: Demonstrates that both slope positions and latitude are key determinants of soil functional gene and enzyme distribution in Mollisols, providing insights into soil properties relevant for slope stability.
86. Topography modifies the effect of mycorrhizal type on soil carbon accumulation in a subtropical mountainous forest
Core Problem: Investigating how topography influences the effect of mycorrhizal type on soil carbon accumulation in subtropical mountainous forests.
Key Innovation: Demonstrates that topography modifies the effect of mycorrhizal type on soil carbon accumulation in subtropical mountainous forests, relevant for understanding soil stability in mountainous terrain.
87. Elevational shifts in diazotrophic communities in subalpine forests: joint effects of temperature and soil properties
Core Problem: Investigating the elevational shifts in diazotrophic communities in subalpine forests and the combined effects of temperature and soil properties.
Key Innovation: Reveals elevational shifts in diazotrophic communities in subalpine forests are jointly determined by temperature and soil properties, contributing to understanding soil conditions in landslide-prone areas.
88. Discrepancies in microbial nitrogen cycling among diverse karst hydrological systems: enhanced nitrate reduction potential in karst conduits
Core Problem: Identifying discrepancies in microbial nitrogen cycling across different karst hydrological systems and understanding the nitrate reduction potential in karst conduits.
Key Innovation: Reveals discrepancies in microbial nitrogen cycling among diverse karst hydrological systems, highlighting enhanced nitrate reduction potential in karst conduits, relevant for understanding karst-related ground deformation.
89. Annual heat budget and seasonal variations in a northern river with ice processes
Core Problem: Understanding the annual heat budget and seasonal variations in northern rivers with significant ice processes.
Key Innovation: Analysis of the annual heat budget and seasonal variations in a northern river, considering ice processes, which is crucial for predicting ice-related flood hazards.
90. Short-term surface settlements induced by EPBM twin tunnelling in saturated sandy soils
Core Problem: Predicting and understanding short-term surface settlements caused by Earth Pressure Balance Machine (EPBM) twin tunneling in saturated sandy soils.
Key Innovation: Investigation of short-term surface settlements induced by EPBM twin tunneling in saturated sandy soils, contributing to understanding ground deformation.
91. Vegetation restoration mitigates meteorological drought on the Loess Plateau
Core Problem: Mitigating meteorological drought, particularly in vulnerable regions like the Loess Plateau.
Key Innovation: Demonstrating that vegetation restoration effectively mitigates meteorological drought on the Loess Plateau.
92. Impact of changing recharge on a sole-source coastal aquifer: multi-model assessment for Virginia’s Eastern Shore
Core Problem: Assessing the impact of changing groundwater recharge on a sole-source coastal aquifer, which can lead to salinization or depletion.
Key Innovation: A multi-model assessment to evaluate these impacts for Virginia’s Eastern Shore.
93. Attributing European runoff changes to climatic drivers under future conditions
Core Problem: Attributing changes in European runoff to specific climatic drivers and projecting these under future conditions.
Key Innovation: Identification and attribution of climatic drivers for European runoff changes.
94. Evolution of streamflow intensity, regional synergistic patterns, and driving mechanisms of the Yangtze River
Core Problem: Analyzing the evolution of streamflow intensity, regional synergistic patterns, and their driving mechanisms in the Yangtze River.
Key Innovation: Identification of the evolution, patterns, and driving mechanisms of streamflow intensity in the Yangtze River.
95. Statistical Postprocessing of Subseasonal Cumulative Precipitation Forecasts Using a Spatial Heterogeneity-aware U-Net
Core Problem: Improving the accuracy of subseasonal cumulative precipitation forecasts through statistical postprocessing, accounting for spatial heterogeneity.
Key Innovation: A spatial heterogeneity-aware U-Net for statistical postprocessing of subseasonal cumulative precipitation forecasts.
96. An LSTM network for joint modeling of streamflow and hydropower generation for run-of-river plants
Core Problem: Developing an LSTM network for jointly modeling streamflow and hydropower generation for run-of-river plants.
Key Innovation: An LSTM network for joint modeling of streamflow and hydropower generation.
97. Active stability and probabilistic estimation of tunnel face in spatially variable and anisotropic soils
Core Problem: Assessing the active stability and probabilistically estimating the failure of tunnel faces in complex spatially variable and anisotropic soils.
Key Innovation: Probabilistic estimation of active tunnel face stability in spatially variable and anisotropic soils.
98. A Global Ensemble Forecast System (GEFS)-based synthetic event set of U.S. tornado outbreaks
Core Problem: Tornado outbreak risk estimates from historical observations are limited by meteorological conditions that have occurred in the historical period, thus not representing the full range of possible outcomes.
Key Innovation: Construction and evaluation of a synthetic event set of U.S. tornado outbreaks using Global Ensemble Forecast System (GEFS) environments and a tornado outbreak index, generating over 200,000 plausible events to estimate 1-in-100-year and 1-in-1000-year event magnitudes and analyze shifts related to ENSO and trends.
99. Technical note: Including hydrologic impact definition in climate projection uncertainty partitioning: a case study of the Central American mid-summer drought
Core Problem: Past changes and future projections of the Central American mid-summer drought (MSD) show strong sensitivity to its definition, raising questions about how to capture uncertainty in projected hydrologic impacts as climate warming continues.
Key Innovation: Characterization of contributions to total uncertainty in MSD projections (using an ensemble of climate models, downscaling methods, and different MSD definitions), revealing that MSD definition contributes least (5%-9%), while climate model variability dominates, providing guidance for water planning and adaptation efforts.
100. Projected impacts of climate change on malaria in Africa
Core Problem: The implications of climate change for malaria eradication in Africa remain poorly resolved, with many studies neglecting interactions between climate, malaria control, socioeconomic environment, and disruption from extreme weather.
Key Innovation: Integration of 25 years of African data on climate, malaria burden/control, socioeconomic factors, and extreme weather into a geotemporal model linked to climate projections, estimating 123 million additional malaria cases and 532,000 additional deaths by 2050, with extreme weather events identified as the primary driver.
101. A hybrid algorithm for human interaction recognition from drone videos: experimental analysis to enhance disaster response and rescue
Core Problem: Efficiently recognizing human interactions from drone videos to improve the effectiveness of disaster response and rescue operations.
Key Innovation: A hybrid algorithm for human interaction recognition from drone videos, experimentally analyzed to enhance disaster response and rescue capabilities.
102. NeuraLSP: An Efficient and Rigorous Neural Left Singular Subspace Preconditioner for Conjugate Gradient Methods
Core Problem: Numerical techniques for solving partial differential equations (PDEs) involve large, sparse linear systems where existing neural preconditioning methods suffer from rank inflation and suboptimal convergence rates, limiting their efficiency and rigor.
Key Innovation: Articulates NeuraLSP, a novel neural preconditioner combined with a new loss metric that leverages the left singular subspace of the system matrix's near-nullspace vectors, compressing spectral information into a fixed low-rank operator, exhibiting theoretical guarantees, empirical robustness to rank inflation, and up to a 53% speedup across diverse PDE families.
103. The Forecast After the Forecast: A Post-Processing Shift in Time Series
Core Problem: Improving the accuracy and uncertainty of deployed time series forecasters without retraining or modifying the backbone model.
Key Innovation: δ-Adapter, a lightweight, architecture-agnostic post-processing method that learns tiny, bounded modules for input nudging and output residual correction, improving accuracy and calibration with negligible compute.
104. CPiRi: Channel Permutation-Invariant Relational Interaction for Multivariate Time Series Forecasting
Core Problem: Current multivariate time series forecasting methods either overfit channel ordering (channel-dependent) or neglect inter-channel dependencies (channel-independent), limiting adaptation and performance under structural and distributional co-drift.
Key Innovation: CPiRi, a channel permutation-invariant framework that couples a spatio-temporal decoupling architecture with a permutation-invariant regularization training strategy, inferring cross-channel structure from data and achieving state-of-the-art results with inductive generalization.
105. AnomalyVFM -- Transforming Vision Foundation Models into Zero-Shot Anomaly Detectors
Core Problem: Vision foundation models (VFMs) lag behind vision-language models (VLMs) in zero-shot anomaly detection due to limited diversity in existing auxiliary anomaly detection datasets and overly shallow VFM adaptation strategies.
Key Innovation: AnomalyVFM, a general and effective framework that transforms any pretrained VFM into a strong zero-shot anomaly detector. It combines a robust three-stage synthetic dataset generation scheme with a parameter-efficient adaptation mechanism (low-rank feature adapters and confidence-weighted pixel loss), substantially outperforming current state-of-the-art methods.
106. bi-modal textual prompt learning for vision-language models in remote sensing
Core Problem: Prompt learning (PL) for vision-language models (VLMs) struggles with transferability to remote sensing (RS) imagery due to unique challenges like multi-label scenes, high intra-class variability, and diverse spatial resolutions, hindering the identification of dominant semantic cues and generalization to novel classes.
Key Innovation: BiMoRS, a lightweight bi-modal prompt learning framework tailored for RS tasks. It employs a frozen image captioning model to extract textual semantic summaries from RS images, which are then fused with high-level visual features from the CLIP encoder. A lightweight cross-attention module conditions a learnable query prompt on this fused representation, achieving consistent performance gains on four RS datasets.
107. A New Dataset and Framework for Robust Road Surface Classification via Camera-IMU Fusion
Core Problem: Existing road surface classification (RSC) techniques often fail to generalize beyond narrow operational conditions due to limited sensing modalities and datasets lacking environmental diversity.
Key Innovation: Introduced a multimodal framework that fuses images and inertial measurements using a lightweight bidirectional cross-attention module and an adaptive gating layer. Also introduced ROAD, a new dataset with diverse real-world, vision-only, and synthetic subsets. The method achieves significant performance improvements and stable performance across challenging visual conditions.
108. The Sound of Noise: Leveraging the Inductive Bias of Pre-trained Audio Transformers for Glitch Identification in LIGO
Core Problem: Transient noise artifacts (glitches) fundamentally limit the sensitivity of gravitational-wave (GW) interferometers and can mimic true astrophysical signals. Current supervised classification methods suffer from a 'label bottleneck' and struggle to generalize to new glitch morphologies.
Key Innovation: Presented a novel cross-domain framework that treats GW strain data as audio, adapting the Audio Spectrogram Transformer (AST) pre-trained on large-scale audio datasets. This approach exploits the strong inductive bias of pre-trained audio models for superior feature extraction, revealing well-separated glitch classes and offering a data-efficient pathway for anomaly detection in GW data.
109. Physics-informed Blind Reconstruction of Dense Fields from Sparse Measurements using Neural Networks with a Differentiable Simulator
Core Problem: Generating dense physical fields from sparse measurements is challenging when spatial statistics or examples of dense fields are unavailable, forcing reliance on synthetic data or less accurate statistical methods.
Key Innovation: A physics-informed blind reconstruction method that integrates an automatically differentiable numerical simulator into the neural network training phase, enabling the generation of dense fields from sparse measurements without prior knowledge of spatial statistics or dense field examples, demonstrating superior results on fluid mechanics problems.
110. Demystifying Prediction Powered Inference
Core Problem: Machine learning predictions are increasingly used to supplement incomplete or costly-to-measure outcomes, but treating them as ground truth introduces bias, while ignoring them wastes valuable information, making principled, valid inference challenging.
Key Innovation: Demystifying Prediction-Powered Inference (PPI) by synthesizing its theoretical foundations, extensions, and diagnostic tools into a unified practical workflow, demonstrating that PPI variants produce tighter confidence intervals while maintaining valid inference through explicit bias correction.
111. VSCOUT: A Hybrid Variational Autoencoder Approach to Outlier Detection in High-Dimensional Retrospective Monitoring
Core Problem: Classical Statistical Process Control (SPC) struggles with high-dimensional, non-Gaussian, and contamination-prone data in retrospective monitoring, leading to distorted baseline estimation and masked anomalies.
Key Innovation: VSCOUT, a distribution-free framework for retrospective monitoring that combines an Automatic Relevance Determination Variational Autoencoder (ARD-VAE) with ensemble-based latent outlier filtering and changepoint detection, followed by a retraining step, achieving superior sensitivity to special-cause structure.
112. Toward Highly Efficient and Private Submodular Maximization via Matrix-Based Acceleration
Core Problem: Submodular function maximization, critical for tasks like sensor placement, is limited by its high computational bottleneck and often lacks privacy guarantees.
Key Innovation: An integrated framework that achieves highly efficient and private submodular maximization by introducing a novel matrix-based computation paradigm, approximate data structures for acceleration, and incorporating (epsilon, delta)-DP guarantees.
113. HyResPINNs: A Hybrid Residual Physics-Informed Neural Network Architecture Designed to Balance Expressiveness and Trainability
Core Problem: Physics-informed neural networks (PINNs) need to balance approximation expressiveness and trainability when solving partial differential equations (PDEs).
Key Innovation: Proposes HyResPINNs, a two-level convex-gated architecture that combines trainable, per-block smooth basis functions with trainable sparsity and deep neural networks, along with block gating, to maximize approximation expressiveness and achieve superior accuracy for PDE problems.
114. Sufficient Decision Proxies for Decision-Focused Learning
Core Problem: In decision-focused learning (DFL) for optimization problems under uncertainty, little is known about when specific deterministic problem approximations (decision proxies) are valid for optimal decision-making.
Key Innovation: Investigates problem properties that justify using certain decision proxies in DFL, presents alternative decision proxies that maintain learning task complexity, and demonstrates their effectiveness on continuous and discrete problems with uncertainty in objectives and constraints.
115. Byte Pair Encoding for Efficient Time Series Forecasting
Core Problem: Existing time series tokenization methods are inflexible, generating excessive tokens for simple patterns, leading to substantial computational overhead in time series forecasting.
Key Innovation: Proposes the first pattern-centric tokenization scheme for time series analysis, inspired by byte pair encoding, which merges samples with underlying patterns into tokens for adaptive compression, and introduces conditional decoding for lightweight post-hoc optimization, improving forecasting performance by 36% and boosting efficiency by 1990% on average.
116. Continuous Evolution Pool: Taming Recurring Concept Drift in Online Time Series Forecasting
Core Problem: Recurring concept drift in online time series forecasting poses a dual challenge: mitigating catastrophic forgetting while adhering to strict privacy constraints that prevent retaining historical data.
Key Innovation: Proposes the Continuous Evolution Pool (CEP), a privacy-preserving framework that maintains a dynamic pool of specialized forecasters, using a Retrieval mechanism for concept identification, an Evolution strategy for spawning new forecasters upon distribution shifts, and an Elimination policy for pruning obsolete models, significantly outperforming baselines without accessing historical ground truth.
117. Sparse Equation Matching: A Derivative-Free Learning for General-Order Dynamical Systems
Core Problem: Many existing equation discovery approaches rely on accurate derivative estimation and are limited to first-order dynamical systems, restricting their applicability in real-world scenarios.
Key Innovation: Sparse Equation Matching (SEM) is a unified framework for equation discovery that introduces an integral-based sparse regression approach using Green's functions, enabling derivative-free estimation of differential operators and driving functions in general-order dynamical systems, demonstrating effectiveness in simulations and identifying task-specific brain connectivity patterns from EEG data.
118. Visual Instruction Pretraining for Domain-Specific Foundation Models
Core Problem: The top-down influence of high-level reasoning on the foundational learning of low-level perceptual features in vision foundation models remains underexplored, limiting their effectiveness in domain-specific applications.
Key Innovation: Visual insTruction Pretraining (ViTP), a novel paradigm that directly leverages reasoning to enhance perception by embedding a Vision Transformer within a Vision-Language Model and pretraining it end-to-end with visual instruction data and Visual Robustness Learning (VRL), achieving state-of-the-art performance in remote sensing and medical imaging.
119. Modality-Balanced Collaborative Distillation for Multi-Modal Domain Generalization
Core Problem: Applying Weight Averaging (WA) directly to multi-modal domain generalization (MMDG) is challenging because differences in optimization speed across modalities lead to overfitting to faster-converging ones, hindering effective modality fusion and generalization.
Key Innovation: MBCD, a unified collaborative distillation framework that addresses MMDG challenges by incorporating adaptive modality dropout in the student model, a gradient consistency constraint to align learning signals, and a WA-based teacher for cross-modal distillation, achieving superior accuracy and robustness across diverse unseen domains.
120. Trend-Adjusted Time Series Models with an Application to Gold Price Forecasting
Core Problem: Time series forecasting often struggles with capturing directional movements (trends) and quantitative values simultaneously, and commonly used metrics like MSE/MAE are insufficient for fully assessing performance, especially in volatile data.
Key Innovation: The Trend-Adjusted Time Series (TATS) model is proposed, reframing forecasting as a two-part task: predicting trend with a binary classifier and forecasting quantitative values with models like LSTM/Bi-LSTM, then adjusting forecasts based on predicted trends, achieving significantly lower error and improved trend detection accuracy.
121. Spatiotemporal Semantic V2X Framework for Cooperative Collision Prediction
Core Problem: Intelligent Transportation Systems (ITS) demand real-time collision prediction, but conventional approaches relying on transmitting raw video or high-dimensional sensory data from RSUs to vehicles are impractical due to vehicular communication bandwidth and latency constraints.
Key Innovation: A semantic V2X framework is proposed where RSU-mounted cameras generate spatiotemporal semantic embeddings of future frames using V-JEPA, which are transmitted to vehicles for collision prediction, significantly reducing communication overhead (four orders of magnitude) while improving F1-score by 10%.
122. Fast Converging 3D Gaussian Splatting for 1-Minute Reconstruction
Core Problem: Achieving high-fidelity 3D Gaussian Splatting (3DGS) reconstruction within a strict one-minute time budget is challenging, especially when dealing with noisy SLAM-generated camera poses.
Key Innovation: A two-stage 3DGS reconstruction pipeline combines reverse per-Gaussian parallel optimization, compact forward splatting, load-balanced tiling, anchor-based Neural-Gaussian representation, and global pose refinement for noisy SLAM data, and adapts for accurate COLMAP poses by disabling refinement, reverting to standard 3DGS, and using multi-view consistency-guided splitting, achieving top performance in a competition.
123. GeoDiff3D: Self-Supervised 3D Scene Generation with Geometry-Constrained 2D Diffusion Guidance
Core Problem: Existing 3D scene generation methods are limited by weak structural modeling and heavy reliance on large-scale ground-truth supervision, often producing structural artifacts, geometric inconsistencies, and degraded high-frequency details in complex scenes.
Key Innovation: GeoDiff3D, an efficient self-supervised framework, uses coarse geometry as a structural anchor and a geometry-constrained 2D diffusion model for texture-rich references. It is robust to noisy guidance and introduces voxel-aligned 3D feature aggregation and dual self-supervision to maintain scene coherence and fine details with low computational cost, improving generalization and generation quality.
124. Explainable Temporal Inference for Irregular Multivariate Time Series. A Case Study for Early Prediction of Multidrug Resistance
Core Problem: Challenging 'MTS-to-TS' inference tasks in healthcare due to data irregularity, temporal dependencies, and the need for clinical explainability.
Key Innovation: Novel eXplainable Artificial Intelligence (XAI) methods (IT-SHAP, Hadamard Attention, Causal Conditional Mutual Information) for 'MTS-to-TS' architectures, enabling temporal inference and explainability for early prediction of multidrug resistance.
125. A Multichannel CNN-LSTM-Based Prediction Model for Precipitable Water Vapor in a Region With a Single GNSS Station
Core Problem: Accurate prediction of precipitable water vapor (PWV) is crucial for meteorological applications, but many regions suffer from sparse GNSS station coverage, often with only a single station available.
Key Innovation: A multichannel CNN-LSTM model designed for single-station PWV prediction, integrating surface pressure, weighted mean temperature, and zenith wet delay, achieving superior performance (MAE 0.694 mm, RMS 0.789 mm) compared to GFS and classical LSTM, demonstrating robust prediction capability across seasons and weather conditions.
126. CD-Lamba: Boosting Remote Sensing Change Detection via a Cross-Temporal Locally Adaptive State Space Model
Core Problem: Existing Mamba-based remote sensing change detection (RSCD) approaches struggle to effectively perceive the inherent locality of change regions while maintaining global perception, as features are not continuously distributed within the sequence.
Key Innovation: CD-Lamba, a novel locally adaptive SSM-based approach for RSCD, including a locally adaptive state-space scan (LASS) strategy for locality enhancement, a cross-temporal state-space scan for bitemporal feature fusion, and a window shifting and perception mechanism, achieving state-of-the-art performance on four benchmark datasets with satisfactory efficiency-accuracy tradeoff.
127. Feature-Screened and Structure-Constrained Deep Forest for Unsupervised SAR Image Change Detection
Core Problem: Deep forest-based models for SAR image change detection are generally challenged by noise sensitivity and high feature redundancy, which significantly degrade prediction performance.
Key Innovation: SC-FS-DF (Structure-Constrained and Feature-Screened Deep Forest) for SAR image change detection, proposing a fuzzy multineighborhood information C-means clustering for high-quality pseudo-labels, a feature-screened deep forest (FS-DF) framework with feature importance/redundancy analysis and dropout, and a novel energy function for refining detection maps, verified on five real SAR datasets.
128. Learning Boundary-Aware Semantic Context Network for Remote Sensing Change Detection
Core Problem: Remote sensing change detection is hindered by pseudochanges and inaccurate identification of change boundaries.
Key Innovation: A boundary-guided semantic context network (BSCNet) is proposed, which decouples features, selectively fuses semantic context, and aggregates edge/texture/semantic information to improve change detection accuracy and boundary precision.
129. LESMI: Integrating Linear-Exponential Model, Shapelets, and Multirocket for Wetland Vegetation Inundation Monitoring With Time Series SAR
Core Problem: Accurate monitoring of wetland vegetation inundation is challenging due to complex scattering characteristics and spatial/seasonal heterogeneity.
Key Innovation: A novel approach named LESMI (linear-exponential model, shapelets, and multirocket integration) is proposed to monitor inundation state and temporal changes using SAR backscatter, achieving high accuracy in identifying noninundated, shallow, and deep inundated states.
130. Mamba-CD: Mamba-Based Change Detection Network for Remote Sensing Images With Change Region-Aware Attention and Recursive Context Refinement Mechanism
Core Problem: Existing change detection methods suffer from insufficient global dependency modeling, low computational efficiency, and struggle with complex background interference, leading to imprecise change region boundaries.
Key Innovation: Mamba-CD, a Mamba-based change detection network, is proposed, which uses VMamba as a backbone, a change region-aware attention mechanism with boundary loss, and a gated recursive context refinement module to achieve superior detection performance with lower complexity and more precise boundaries.
131. DCCANet: Dual-Domain With Collaborative Cross-Mapping Attention Network for Remote Sensing Change Detection
Core Problem: Remote sensing change detection faces challenges in effectively suppressing environmental factors and noise interference while achieving efficient global context modeling.
Key Innovation: A dual-domain with collaborative cross-mapping network (DCCANet) is proposed, which utilizes frequency and spatial domain information, a frequency enhancement module, a difference-guided spatial attention mechanism, and a novel cross-pooling mechanism for improved change detection accuracy and global context modeling.
132. Total groundwater quantity management framework for sustainable use: small watershed and AI-based approach
Core Problem: Sustainable groundwater management is critical in the era of climate change, requiring effective regulation and assessment of over-extraction within designated groundwater management units (GMUs).
Key Innovation: The Total Groundwater Quantity Management (TGQM) framework is proposed, which uses a regression tree model to estimate groundwater recharge rates and assesses over-extraction by evaluating the ratio of groundwater usage to development potential for each GMU, establishing management thresholds (60%, 80%, 100%) for caution, warning, and critical stages.
133. Development and application of SWAT-MODFLOW in surface water-groundwater interactions: Current status and future challenges
Core Problem: Understanding surface water and groundwater interactions (SGIs) is critical due to climate change and human activity, and the development and application of coupled models like SWAT-MODFLOW need to be reviewed for their strengths, limitations, and future challenges.
Key Innovation: This article reviews the development and application of the coupled SWAT-MODFLOW model for characterizing SGIs, highlighting its power for simulating hydrological processes, nutrient transport, and climate change impacts, while also discussing uncertainties and the trade-offs between integrated and standalone models for practical applications.
134. Sealing of subsea tunnels: experimental investigation of groundwater chemistry and flow velocities for engineering applications
Core Problem: The sealing efficiency of grouting subsea tunnels under various groundwater environments (chemistry, flow velocities) remains unclear, hindering effective risk mitigation for water and mud inrush.
Key Innovation: This study experimentally investigates grouting under groundwater environments, revealing that sealing efficiency weakens in acidic/alkaline conditions and with increased flow speed, but superfine cement-sodium silicate grout exhibits excellent resistance, proposing a practical grout material selection principle validated by the Haicang subsea tunnel.
135. Modelling the mechanical behaviour of natural gas hydrate-bearing sediments using a transfer learning neural network
Core Problem: Accurately understanding and predicting the mechanical behavior of natural gas hydrate-bearing sediments (GHBS) is challenging due to complex theoretical models, difficult parameterization, and limited experimental data for data-driven approaches, hindering safe exploitation.
Key Innovation: Developed a data-driven model using transfer learning, integrating theoretical constitutive models and experimental data, to accurately predict the mechanical behavior of various GHBS samples under different conditions, demonstrating improved accuracy and generalization compared to conventional methods.
136. Analysis of Wellbore Wall Stress for SC-CO<sub>2</sub> Horizontal Well Fracturing in Bedded Rocks and the Influence Mechanism of Bedding Structure: Development of a Fracture Initiation Criterion and a Theoretical Prediction Model for Initiation Pressure in SC-CO<sub>2</sub> Fracturing
Core Problem: The absence of reliable theoretical guidance for SC-CO2 fracturing in bedded rocks, which exhibits different fracture propagation behavior compared to water-based fracturing due to bedding structure and SC-CO2 properties, impedes its application.
Key Innovation: Developed a novel tensile failure criterion ("Y-L criterion") and a theoretical prediction model for fracture initiation pressure in SC-CO2 horizontal well fracturing in bedded rocks, incorporating bedding structure and SC-CO2 fluid characteristics, and validated it experimentally, providing theoretical guidance for fracturing design.
137. Fully Coupled Adsorption-Hydro-Mechanical Model for Wellbore Stability Analysis in Non-Saturated Shale Gas Reservoirs Using Dual-Pore Dual-Permeability Model
Core Problem: Predicting wellbore stability in shale gas reservoirs is challenging due to their dual-pore, dual-permeability characteristics and non-saturated water-gas phases, which are often neglected in existing models, leading to inaccurate predictions.
Key Innovation: Developed a fully coupled dual-pore dual-permeability two-phase fluid hydro-mechanical model, incorporating adsorption-diffusion and fracture-matrix cross-flow, to accurately assess wellbore stability in non-saturated shale gas reservoirs, revealing that neglecting these complexities leads to conservative drilling fluid density designs and that higher matrix permeability/initial water saturation increases instability risk.
138. An implicit feature learning approach for automatic digital twinning of three-dimensional subsurface stratigraphy from limited boreholes
Core Problem: Automatically creating accurate three-dimensional digital twins of subsurface stratigraphy from limited borehole data is a significant challenge.
Key Innovation: Proposed an implicit feature learning approach for automatic digital twinning of three-dimensional subsurface stratigraphy, designed to work effectively even with limited borehole data.
139. Predicting permissible soil stress from soil resistivities and physical parameters: case study of soils in Fokoué Urban Center, West-Cameroon
Core Problem: Accurately predicting permissible soil stress, especially in specific urban centers, from readily available soil resistivities and physical parameters is a practical challenge.
Key Innovation: Developed a method to predict permissible soil stress from soil resistivities and physical parameters, demonstrated through a case study in Fokoué Urban Center, West-Cameroon.
140. A simple TDR waveform analysis for estimating volumetric water content in marine clays
Core Problem: Accurately and simply estimating volumetric water content in marine clays is important for geotechnical applications.
Key Innovation: Proposed a simple TDR waveform analysis method for estimating volumetric water content specifically in marine clays.
141. Machine learning approaches to predicting the uniaxial compressive strength of granite from image-derived mineralogical features
Core Problem: Predicting the uniaxial compressive strength (UCS) of granite, a critical rock mechanical property, from image-derived mineralogical features can be challenging.
Key Innovation: Applied machine learning approaches to predict the uniaxial compressive strength (UCS) of granite using image-derived mineralogical features.
142. Propagation of porous slurry veins in marine hydrate reservoirs via dual-enhanced fracture grouting
Core Problem: Effectively propagating porous slurry veins in marine hydrate reservoirs for stabilization or resource recovery is a complex challenge.
Key Innovation: Investigated the propagation of porous slurry veins in marine hydrate reservoirs using a dual-enhanced fracture grouting technique.
143. Engineering geological classification of gravelly deposits based on enhanced CPT
Core Problem: Accurately classifying gravelly deposits for engineering geological purposes can be challenging, especially with conventional methods.
Key Innovation: Developed an engineering geological classification method for gravelly deposits based on enhanced Cone Penetration Test (CPT) data.
144. A unified van Genuchten-type water retention model for compacted bentonite
Core Problem: Developing a unified water retention model for compacted bentonite that accurately describes its behavior across various conditions is important for geotechnical applications.
Key Innovation: Proposed a unified van Genuchten-type water retention model specifically for compacted bentonite.
145. Symbolic regression-based prediction of coefficient of permeability for granular soils
Core Problem: Accurately predicting the coefficient of permeability for granular soils, a critical geotechnical parameter, can be complex.
Key Innovation: Developed a symbolic regression-based method for predicting the coefficient of permeability for granular soils.
146. Spatial-X fusion for multi-source satellite imageries
Core Problem: Effectively fusing multi-source satellite imageries to enhance data utility.
Key Innovation: A "Spatial-X fusion" method for multi-source satellite imageries.
147. Multispectral airborne laser scanning for tree species classification: A benchmark of machine learning and deep learning algorithms
Core Problem: Evaluating the performance of different ML/DL algorithms for tree species classification using multispectral airborne laser scanning.
Key Innovation: A benchmark study comparing machine learning and deep learning algorithms for tree species classification using multispectral airborne laser scanning.
148. RegScorer: Learning to select the best transformation of point cloud registration
Core Problem: Improving the selection of optimal transformations for point cloud registration.
Key Innovation: RegScorer, a learning-based method to select the best transformation for point cloud registration.
149. Effects of CO<sub>2</sub>-water-rock interactions on the fracture performance of transversely isotropic shale: Transition from strengthening to weakening
Core Problem: Understanding how CO2-water-rock interactions influence the fracture performance of transversely isotropic shale, leading to strengthening or weakening.
Key Innovation: Demonstrating the transition from strengthening to weakening in transversely isotropic shale's fracture performance due to CO2-water-rock interactions, relevant for rock failure mechanisms.
150. Addressing class imbalance extends the performance frontier of classification–regression satellite-gauge precipitation fusion
Core Problem: Improving the performance of satellite-gauge precipitation fusion, particularly when dealing with class imbalance in data.
Key Innovation: A method to extend the performance of classification-regression satellite-gauge precipitation fusion by addressing class imbalance.
151. Diverse methods of incorporating physics into neural networks: A comprehensive review
Core Problem: Reviewing and categorizing various methods for integrating physical laws and constraints into neural networks.
Key Innovation: A comprehensive review and classification of diverse methods for incorporating physics into neural networks.
152. Evaluating the Performance of Uni‐ and Multivariate Bias Correction Techniques: Challenges in Preserving Temporal and Dependence Structures
Core Problem: Global Climate Models (GCMs) have systematic biases and coarse resolution, and existing bias correction (BC) techniques often distort temporal and dependence structures, limiting their direct application in climate studies and downstream impact assessments.
Key Innovation: A comprehensive evaluation of uni- and multivariate bias correction techniques (QM, dOTC, R2D2, MBCn) for daily precipitation and temperature, assessing their performance in correcting univariate, multivariate, and temporal features, and preserving inter-variable dependencies, offering practical insights for climate applications.
153. Zooming in: SCREAM at 100 m using regional refinement over the San Francisco Bay Area
Core Problem: Pushing global climate models to large-eddy simulation (LES) scales over complex terrain has remained a major challenge, limiting their ability to realistically capture fine-scale atmospheric processes.
Key Innovation: First known implementation of a global model (SCREAM) at 100m horizontal resolution using a regionally refined mesh (RRM) over the San Francisco Bay Area, demonstrating stable LES-scale performance, realistic capture of topography and coastal processes, and substantial improvements in near-surface atmospheric variable biases.
154. Seismic and mineralogical evidence for an iron-rich mega–ultralow-velocity zone beneath Hawai’i
Core Problem: Lack of detailed understanding of deep Earth structures, specifically beneath Hawai'i, which are crucial for understanding geological processes.
Key Innovation: Provision of seismic and mineralogical evidence for an iron-rich mega–ultralow-velocity zone beneath Hawai’i, contributing to understanding Earth's interior.
155. A Source-Free Approach for Domain Adaptation via Multiview Image Transformation and Latent Space Consistency
Core Problem: Existing domain adaptation methods are computationally expensive, requiring source domain data, adversarial training, or complex pseudo-labeling, hindering knowledge transfer when data distributions differ.
Key Innovation: A novel source-free domain adaptation method that learns domain-invariant features directly from the target domain by enforcing consistency between multiple augmented views in the latent space, eliminating the need for source-target alignment or pseudo-label refinement.
156. CURVE: Learning Causality-Inspired Invariant Representations for Robust Scene Understanding via Uncertainty-Guided Regularization
Core Problem: Scene graphs often overfit to spurious correlations, severely hindering out-of-distribution generalization for robust scene understanding.
Key Innovation: CURVE, a causality-inspired framework that integrates variational uncertainty modeling with uncertainty-guided structural regularization to suppress high-variance, environment-specific relations, promoting a sparse and domain-stable topology for robust scene understanding.
157. FedRD: Reducing Divergences for Generalized Federated Learning via Heterogeneity-aware Parameter Guidance
Core Problem: Generalizing federated learning models to unseen clients under heterogeneous data is crucial, but faces challenges of Optimization Divergence and Performance Divergence.
Key Innovation: FedRD, a heterogeneity-aware federated learning algorithm that collaboratively utilizes parameter-guided global generalization aggregation and local debiased classification to reduce divergences, aiming for an optimal global model for participating and unseen clients.
158. PatchFormer: A Patch-Based Time Series Foundation Model with Hierarchical Masked Reconstruction and Cross-Domain Transfer Learning for Zero-Shot Multi-Horizon Forecasting
Core Problem: Many existing time series forecasting approaches require domain-specific feature engineering and substantial labeled data for each task, limiting their generalization and efficiency across diverse applications.
Key Innovation: Introduced PatchFormer, a patch-based time series foundation model that uses hierarchical masked reconstruction for self-supervised pretraining and lightweight adapters for efficient transfer. It achieves state-of-the-art zero-shot multi-horizon forecasting across 24 benchmark datasets, reducing mean squared error by 27.3% relative to strong baselines with 94% less task-specific training data.
159. Cross-Country Learning for National Infectious Disease Forecasting Using European Data
Core Problem: National infectious disease forecasting is often limited by the length and variability of single-country historical data, restricting machine learning model performance.
Key Innovation: Investigating a cross-country learning approach where a single model is trained on time series data from multiple countries, demonstrating consistent improvements in multi-step forecasting performance for COVID-19 compared to models trained solely on national data.
160. Comparing Time-Series Analysis Approaches Utilized in Research Papers to Forecast COVID-19 Cases in Africa: A Literature Review
Core Problem: A lack of comprehensive comparison of various time-series analysis approaches used for forecasting COVID-19 cases in Africa, hindering insights into their effectiveness and limitations.
Key Innovation: A literature review that systematically compares time-series analysis approaches utilized in research papers to forecast COVID-19 cases in Africa, highlighting different methodologies, their effectiveness, and limitations.
161. UDEEP: Edge-based Computer Vision for In-Situ Underwater Crayfish and Plastic Detection
Core Problem: Invasive signal crayfish and plastic pollution have detrimental impacts on aquatic ecosystems, requiring effective in-situ detection methods to safeguard native species and water quality.
Key Innovation: UDEEP, a Cognitive Edge Device (CED) computing platform for in-situ underwater crayfish and plastic detection, which evaluates YOLO variants on publicly available datasets, achieving high detection accuracy (YOLOv5s mAP@0.5 of 0.90).
162. Provably Efficient RL under Episode-Wise Safety in Constrained MDPs with Linear Function Approximation
Core Problem: Theoretical results for reinforcement learning (RL) in constrained Markov decision processes (CMDPs) with function approximation, especially with episode-wise safety guarantees, remain scarce.
Key Innovation: Proposes an RL algorithm for linear CMDPs that achieves $ ilde{\mathcal{O}}(\sqrt{K})$ regret with an episode-wise zero-violation guarantee, is computationally efficient, and converges to the optimal policy under specific model set structures.
163. AdaSCALE: Adaptive Scaling for OOD Detection
Core Problem: State-of-the-art out-of-distribution (OOD) detection methods using activation shaping apply a static percentile threshold, leading to suboptimal in-distribution (ID)-OOD separability.
Key Innovation: Proposes AdaSCALE, an adaptive scaling procedure that dynamically adjusts the percentile threshold for OOD detection based on a sample's estimated OOD likelihood, leveraging the observation that OOD samples show more pronounced activation shifts under minor perturbation, resulting in significantly improved OOD detection performance.
164. Smart Exploration in Reinforcement Learning using Bounded Uncertainty Models
Core Problem: Reinforcement learning (RL) often requires large amounts of data to learn an optimal policy, and efficient exploration in uncertain environments remains a challenge.
Key Innovation: Introduces BUMEX (Bounded Uncertainty Model-based Exploration), an RL strategy that incorporates prior model knowledge (a model set containing the true transition kernel and reward function) to guide exploration by optimizing over this set to obtain Q-function bounds, providing theoretical convergence guarantees and accelerating learning.
165. JAFAR: Jack up Any Feature at Any Resolution
Core Problem: Foundation Vision Encoders produce low-resolution spatial features, necessitating efficient and flexible feature upsampling to generate high-resolution modalities for downstream dense vision tasks.
Key Innovation: JAFAR is a lightweight, flexible, attention-based feature upsampler that enhances spatial resolution from any Foundation Vision Encoder to an arbitrary target resolution, promoting semantic alignment between high-resolution queries and low-resolution keys via Spatial Feature Transform (SFT) modulation, and generalizes well to significantly higher output scales without high-resolution supervision.
166. WaterFlow: Explicit Physics-Prior Rectified Flow for Underwater Saliency Mask Generation
Core Problem: Underwater Salient Object Detection (USOD) faces significant challenges from underwater image quality degradation and domain gaps, with existing methods often ignoring or failing to fully exploit the valuable information contained in underwater physical imaging principles.
Key Innovation: WaterFlow, a rectified flow-based framework for USOD that innovatively incorporates underwater physical imaging information as explicit explicit priors directly into network training and introduces temporal dimension modeling, significantly enhancing salient object identification.
167. GCL-OT: Graph Contrastive Learning with Optimal Transport for Heterophilic Text-Attributed Graphs
Core Problem: Existing structure-text contrastive learning methods for text-attributed graphs rely on homophily assumptions and hard optimization, limiting their applicability to heterophilic graphs and leading to suboptimal alignment due to mixed, noisy, and missing semantic correlations.
Key Innovation: GCL-OT, a novel graph contrastive learning framework with optimal transport, equipped with tailored mechanisms for multi-granular heterophily (RealSoftMax-based similarity estimator, prompt-based filter, and OT-guided soft supervision) to achieve flexible and bidirectional alignment and enhance learning of latent homophily.
168. Stochastic Voronoi Ensembles for Anomaly Detection
Core Problem: Existing anomaly detection methods struggle with datasets exhibiting varying local densities, with distance-based methods missing local anomalies and density-based approaches requiring careful parameter selection and incurring quadratic time complexity.
Key Innovation: SVEAD (Stochastic Voronoi Ensembles Anomaly Detector) constructs ensemble random Voronoi diagrams and scores points by normalized cell-relative distances weighted by local scale, achieving linear time complexity, constant space complexity, and outperforming 12 state-of-the-art approaches on 45 datasets.
169. Simplicity is Key: An Unsupervised Pretraining Approach for Sparse Radio Channels
Core Problem: State-of-the-art unsupervised representation learning for wireless channel state information (CSI) often takes little or no account of domain-specific knowledge, forcing models to learn well-known concepts solely from data.
Key Innovation: Sparse pretrained Radio Transformer (SpaRTran), a hybrid method based on compressed sensing for wireless channels, builds around a wireless channel model to constrain optimization and induce a strong inductive bias, cutting positioning error by up to 28% and increasing top-1 codebook selection accuracy for beamforming by 26 percentage points.
170. Online Conformal Model Selection for Nonstationary Time Series
Core Problem: Classical model selection methods fail for nonstationary time series, which are common in real-world dynamic environments.
Key Innovation: MPS (Model Prediction Set), a novel framework combining conformal inference with model confidence sets for online model selection in nonstationary time series, adaptively selecting models and providing a confidence set for the best model.
171. Efficient Group Lasso Regularized Rank Regression with Data-Driven Parameter Determination
Core Problem: High-dimensional regression methods based on least-squares are unreliable in the presence of heavy-tailed noise and outliers.
Key Innovation: A robust group lasso regularized rank regression method using a non-smooth Wilcoxon score objective, incorporating a data-driven, simulation-based tuning rule, and solved efficiently by a proximal augmented Lagrangian method, demonstrating improved robustness and scalability.
172. Adversary-Aware Private Inference over Wireless Channels
Core Problem: AI-based sensing at wireless edge devices, while beneficial for applications like environmental monitoring, faces privacy risks as sensitive personal data can be reconstructed from transmitted features, and existing differential privacy mechanisms don't protect individual features.
Key Innovation: A novel framework for privacy-preserving AI-based sensing where edge devices apply transformations to extracted features before transmission to a model server, addressing the protection of individual features against privacy violations.
173. QUASAR: An Evolutionary Algorithm to Accelerate High-Dimensional Optimization
Core Problem: High-dimensional numerical optimization of complex, non-differentiable problems is challenging due to the curse of dimensionality.
Key Innovation: QUASAR (Quasi-Adaptive Search with Asymptotic Reinitialization), an evolutionary algorithm that enhances Differential Evolution with quasi-adaptive mechanisms (probabilistic mutation, rank-based crossover, decaying covariance reinitializations) to accelerate convergence and improve solution quality in high-dimensional optimization.
174. Communication-Avoiding Linear Algebraic Kernel K-Means on GPUs
Core Problem: Kernel K-means, while capable of handling non-linearly separable clusters, is computationally and memory intensive, limiting its scalability to large datasets on single GPUs.
Key Innovation: A suite of distributed-memory parallel algorithms for large-scale Kernel K-means clustering on multi-GPU systems, mapping components to communication-efficient distributed linear algebra primitives, enabling efficient clustering of million-scale datasets and achieving significant speedups.
175. A novel thin floating plate formulation in SPH: Extension to a three dimensional Applied Element Method framework
Core Problem: Developing an advanced numerical formulation for thin floating plates within a 3D Applied Element Method framework.
Key Innovation: A novel thin floating plate formulation using SPH, extended to a three-dimensional Applied Element Method framework.
176. Predicting Traumatic Brain Injury Post-Trauma Using Temporal Attention on Sleep-Wake Data
Core Problem: Accurately classifying Traumatic Brain Injury (TBI) and identifying the optimal time window for detecting sleep/wake changes post-trauma.
Key Innovation: A deep learning model with temporal attention applied to longitudinal sleep/wake data for early identification of TBI.
177. Granularity-Inconsistent Transformer for Unsupervised Hyperspectral Anomaly Detection
Core Problem: Existing deep models for hyperspectral anomaly detection (HAD) often fail to capture the inherent spatial characteristics of images, leading to semantic and structural information loss during background reconstruction.
Key Innovation: The Granularity-Inconsistent Transformer (GIFormer) for unsupervised HAD, which leverages spatial-spectral interaction, applies patch-level anomaly elimination masks in the encoder for background reconstruction, and uses pixel-level fine-grained receptive fields in the decoder, outperforming existing techniques.
178. SFPNet: Self-Learning Small Object Detection for Large-Scale Remote Sensing Images
Core Problem: Small object detection in large-scale remote sensing images encounters problems such as information loss, noise characteristic interference, and disturbance of the bounding box, making it challenging compared to generic object detection.
Key Innovation: SFPNet, a self-learning feature point network for small object detection, which captures geometric information using feature points, enhances representation with a feature points orientation transformation module, and improves learning with a positive and negative sample evaluation and balancing strategy, achieving over 1.3% improvement in accuracy on benchmark datasets.
179. Adaptive Adversarial Cross-Domain Segmentation Network for High-Resolution Remote Sensing Images
Core Problem: Convolutional neural networks (CNNs) for semantic segmentation of high-resolution remote sensing images are ineffective when training and test data distributions differ (e.g., due to imaging modalities or geographic locations), especially for small-scale objects.
Key Innovation: An adaptive adversarial cross-domain segmentation network for HRSIs, featuring a feature discrepancy module to locate small-scale objects and a scale consistency module with a dynamic self-training strategy to align feature distributions, outperforming state-of-the-art models on cross-domain segmentation tasks.
180. IE-TV: Structural Information Enhanced 3-D Weighted Correlated Total Variation for Hyperspectral Image Denoising
Core Problem: Hyperspectral image (HSI) denoising is challenging due to the difficulty in recovering complex structures under mixed-noise, with existing methods often failing to preserve fine textures or exacerbating spectral over-smoothing.
Key Innovation: Structural Information Enhanced 3D Weighted Correlated Total Variation (IE-TV) for HSI denoising, which constructs a weighted CTV regularization term and learns a structural information enhancement matrix to impose constraints on the gradient nuclear norm, balancing detail preservation and global structure restoration, outperforming state-of-the-art methods.
181. Dual-Student Self-Distillation for Hyperspectral Image Classification With Noisy Labels
Core Problem: Deep-learning-based hyperspectral image classification (HSIC) performance degrades significantly when training data contain noisy labels, an unavoidable problem in real-world applications that can cause error accumulation.
Key Innovation: Dual-Student Self-Distillation (DSSD) framework for robust HSIC with noisy labels, synergizing dual-student learning (cross-update for mutual training with clean samples) and teacher-free self-distillation (generating confidence-calibrated soft targets and semantically aware nontarget labels), significantly outperforming state-of-the-art methods across noise levels.
182. Cross-Modality Fusion of Visible Light, Infrared, and SAR Images Under Few-Shot Conditions for Target Recognition
Core Problem: Multimodal image fusion recognition faces challenges including data heterogeneity and feature redundancy, particularly under few-shot conditions, limiting target information complementarity and accuracy.
Key Innovation: A multisource heterogeneous image fusion recognition method for few-shot scenarios, proposing a cross-modal sampling module, an image quality assessment module for adaptive weighting, an intra-modal bidirectional guided cross-attention module, and a stepwise fusion strategy, demonstrating significant advantages on few-shot datasets.
183. Enhancing Hyperspectral Image Classification With Spectral–Spatial Convolutional FourierKAN Transformer
Core Problem: Transformer architectures for hyperspectral image (HSI) classification, despite their self-attention mechanisms, face critical challenges in mathematical interpretability and sensitivity to spectral redundancy.
Key Innovation: The Spectral-Spatial Convolutional FourierKAN Transformer (S2CFKAT) framework for HSI classification, combining a mixed convolutions module for feature extraction with a FourierKAN transformer architecture that synergizes multihead self-attention and learnable spectral-spatial activation functions, demonstrating superior generalization and robustness on small and unbalanced training samples.
184. A Stepwise BRDF-Based Land Surface Reflectance Model Over Rugged Terrain
Core Problem: Land surface reflectance, an essential variable of remote sensing, is significantly affected by topography and atmosphere, leading to large errors in quantitative remote sensing reflectance retrieval in mountainous regions.
Key Innovation: A stepwise bidirectional reflectance distribution function based land surface reflectance (SBLSR) model that decouples atmospheric and terrain influences, leveraging four-stream radiative transfer theory and mountainous radiative transfer principles, showing great agreement with validation datasets (RMSE 0.0014 simulated, 0.0242 field).
185. FedC-DAC: A Federated Clustering With Dynamic Aggregation and Calibration Method for SAR Image Target Recognition
Core Problem: Federated learning (FL) for multisensor synthetic aperture radar (SAR) target recognition faces challenges from heterogeneous data distributions across clients and hierarchical heterogeneity among sensors, which can induce local model drift.
Key Innovation: FedC-DAC, a clustered FL framework that explicitly captures and utilizes multilevel heterogeneity by integrating Gaussian-mixture-model-guided soft grouping, intracluster dynamic aggregation, and cross-cluster calibration, demonstrating consistent gains in Accuracy, Kappa, and F1 on SAR benchmark datasets under strong heterogeneity.
186. GAN-Guided Unsupervised, Wrapper Feature Selector for PolSAR Image
Core Problem: PolSAR data interpretation tasks are challenged by high computing burden and negative effects from a mass of features, requiring effective unsupervised feature selection (UFS) to eliminate redundancy and retrieve discriminative features.
Key Innovation: WGAN-GP-UFS, a novel unsupervised feature selection algorithm for PolSAR images using a Wasserstein generative adversarial network with gradient penalty, featuring a hybrid architecture for feature subset search (subspace-dividing and SFSSK) and a Wasserstein distance-based loss function for subset evaluation, outperforming 16 state-of-the-art UFS algorithms.
187. FFCA-UNet: Feature Fusion and Cross-Attention Mechanism for Remote Sensing Image Semantic Segmentation
Core Problem: High-resolution remote sensing (RS) image segmentation remains challenging due to large-scale variations, the presence of small and sparsely distributed objects, and the high visual similarity between land-cover categories.
Key Innovation: FFCA-UNet, a hybrid CNN–transformer architecture that integrates an efficient multiscale feature fusion module (EMFM) for spatial consistency and a CNN-guided cross-attention module (CGCAM) for globally consistent reasoning, achieving state-of-the-art performance (80.43% MIoU on Vaihingen, 64.00% MIoU on Potsdam) on RS benchmarks.
188. MDPNet: Multimodal Diffusion Prior Guided Self-Text Attention Network for Remote Sensing Semantic Segmentation
Core Problem: Accurate semantic segmentation of remote sensing images is crucial but challenging due to the heterogeneity of multimodal data and limited target attribute representation, especially for fine-grained boundary features.
Key Innovation: Multimodal Diffusion Prior guided Self-Text Attention Network (MDPNet), introducing a denoising diffusion probabilistic model for robust structural priors, a multimodal prior feature guidance module for cross-modal fusion, and a self-text attention mechanism for category-related textual cues, setting a new state-of-the-art on ISPRS Potsdam and Vaihingen datasets.
189. Relational Structure-Aware Mamba Network for Semantic Segmentation of Remote Sensing Images
Core Problem: Accurate semantic segmentation of high-resolution remote sensing images requires strong structural awareness and precise boundary delineation, which remain challenging due to complex spatial patterns and semantic ambiguities.
Key Innovation: ReSMamba, a relational structure-aware segmentation framework based on linear state-space modeling, introducing a structure-aware topology Mamba block, a structure-aware semantic assembly module, and a directional connectivity prediction module, consistently outperforming state-of-the-art methods on ISPRS Potsdam, Vaihingen, and LoveDA datasets.
190. ASENet: Thin Cloud Removal Network for Complex Scenes via Atmospheric Scattering Modeling and Feedback Enhancement
Core Problem: Thin clouds in optical remote sensing pose a significant challenge for cloud removal in complex scenes due to their high brightness and spectral similarity to bright man-made objects, leading to suboptimal image recovery.
Key Innovation: Atmospheric Scattering-driven Recovery Enhancement Network (ASENet), a novel network integrating atmospheric scattering modeling with a multilevel feedback enhancement mechanism, featuring a feature fusion residual dehazing generator and a spatial detail enhanced discriminator, outperforming state-of-the-art methods on three benchmark datasets for thin cloud removal.
191. Few-Shot Object Detection on Remote Sensing Images Based on Decoupled Training, Contrastive Learning, and Self-Training
Core Problem: Few-shot object detection (FSOD) in remote sensing imagery faces critical challenges from limited labeled instances, complicated backgrounds, and multiscale objects, leading to large numbers of false alarms and miss detections.
Key Innovation: DeCL-Det, a FSOD framework that applies self-training to generate high-quality pseudoannotations, introduces an auxiliary network to mitigate label noise, proposes a gradient-decoupled framework (GCFPN) for multiscale feature learning, and incorporates a supervised contrastive learning head, performing better than several existing approaches on the DIOR dataset.
192. Heterogeneous RFI Mitigation in Image-Domain via Subimage Segmentation and Local Frequency Feature Analysis
Core Problem: Heterogeneous radio frequency interference (RFI) in spaceborne synthetic aperture radar (SAR) images (e.g., Sentinel-1) degrades image quality and increases the complexity of interference detection and mitigation.
Key Innovation: A heterogeneous RFI mitigation method based on subimage segmentation and local spectral features analysis, which divides the original single look complex image into multiple subimages to enhance interference features in the range frequency domain, performs spectral analysis for detection/mitigation, and reconstructs the image, effectively mitigating RFI artifacts in various scenarios.
193. PyramidMamba: An Effective Hyperspectral Remote Sensing Image Target Detection Network
Core Problem: Lack of prior knowledge and efficient feature extraction are challenging issues in target detection for hyperspectral remote sensing images.
Key Innovation: A network named PyramidMamba is proposed, which uses spectral data augmentation, a Mamba residual module, and pyramid wavelet transform for long-term dependency and multiscale feature extraction, achieving higher detection accuracy and computational efficiency.
194. Using Agricultural Production to Evaluate the Novel Method of Reducing the Uncertainty in GNSS-R Soil Moisture Retrieval
Core Problem: Existing GNSS-R soil moisture retrieval methods lack uncertainty quantification, and their application potential for agricultural production has not been thoroughly evaluated.
Key Innovation: A natural gradient boosting (NGBoost)-based framework is proposed to quantify and reduce uncertainty in GNSS-R soil moisture retrieval, and the retrieved SM is evaluated using GPP and ET, showing improved accuracy and better reflection of agricultural production.
195. Thick Cloud Removal Using Cloud Mask-Guided SAR-Optical Fusion Network
Core Problem: Thick cloud cover in optical imagery leads to significant surface information loss, and existing SAR-optical fusion methods for cloud removal struggle with cross-modal feature alignment and coherent SAR noise.
Key Innovation: PSF-CR, a novel cloud-mask-guided, global perception-enhanced SAR-optical fusion network, is proposed, which incorporates a gated attention module for noise suppression, weighted L1 and composite spatial-frequency losses, and frequency-domain attention modules for efficient cross-modal alignment and full-scene reconstruction.
196. HVTC-GAN: A High-Level Vision Task Cooperative GAN for SAR-to-Optical Translation via Semantic Segmentation
Core Problem: SAR-to-optical translation (S2OT) studies often prioritize visual quality over practical applicability to downstream tasks like land-cover classification, and struggle to preserve structural information from SAR.
Key Innovation: A high-level vision task-coordinated S2OT framework (HVTC-GAN) is proposed, which integrates semantic segmentation loss to guide the network towards task-relevant features, introduces SSIM loss and SAR-derived semantic maps as auxiliary inputs to preserve structural information, and uses identity loss to align distributions, improving both translation fidelity and downstream land-cover classification.
197. Drained shear behavior of coral sand reinforced with recycled tyre strips
Core Problem: Coral sand's inherent brittleness and high crushability pose substantial engineering challenges, and there's a need for sustainable methods to enhance its mechanical properties.
Key Innovation: This study examines the enhancement of coral sand reinforced with recycled tyre strips through triaxial shear tests, demonstrating that reinforcement reduces particle crushing, increases peak shear strength (22–52%), stabilizes volumetric strains, and provides an optimized framework for coral sand reinforcement in coastal infrastructure.
198. Mechanisms of compacted soil deterioration in biopolymer-treated sands under cyclic wetting and drying
Core Problem: The long-term stability and durability of biopolymer-treated soils under cyclic drying and wetting conditions are not well understood, lacking standardized assessment tests.
Key Innovation: This study examines the performance and deterioration mechanisms of Xanthan Gum-treated sandy soils under cyclic drying/wetting, revealing that low hydraulic conductivity creates unique moisture distribution, and rheological/FTIR investigations indicate a reduction in pore fluid yield stress and biological degradation of the biopolymer, contributing to failure.
199. Investigation on the role of waste water-based drilling geopolymer in improving the properties of loess: an experimental insight
Core Problem: Loess is generally unsuitable as a landfill cover material due to its loose structure, high porosity, and poor cohesion, requiring effective improvement methods.
Key Innovation: This study investigates the role of waste water-based drilling geopolymer (WWDG) in improving loess properties, specifically reducing gas permeability (GP), through triaxial GP tests and microstructural analysis, demonstrating that WWDG significantly decreases GP by filling pores and altering pore structure, showing potential for waste reutilization and loess improvement.
200. Calibration of a hybrid bedding-plane DEM model for shale specimens using image-based bedding plane data
Core Problem: Previous research often oversimplifies shale bedding planes, neglecting the anisotropic effects of varied orientations, leading to inaccurate predictions of its complex mechanical behavior.
Key Innovation: This study develops a comprehensive methodology for establishing and calibrating a realistic hybrid bedding-plane DEM model for shale specimens, incorporating accurate image-based bedding plane representations (log-normal distribution of lengths), and demonstrating that the calibrated UDEC model accurately replicates stress-strain behaviors and failure modes across bedding plane orientations.
201. Dynamic evolution mechanism of shale structures under fluid-rock interaction based on fractal theory
Core Problem: The shale damage mechanism caused by long-term contact between water-based drilling fluid and the borehole wall, particularly the dynamic evolution of shale pore structures, needs systematic investigation.
Key Innovation: This study systematically investigates the dynamic evolution of shale pore structures at different hydration times through fixed-point observation experiments, revealing that hydration promotes mineral dissolution, increases pore diameter and number, and exhibits clear fractal characteristics, with the fractal dimension increasing over time, providing a scientific foundation for wellbore instability.
202. Characteristics of wind-sand flow in longitudinal slope embankment section of desert highway
Core Problem: Understanding the wind-sand flow response regularity in longitudinal slope sections of desert highways is crucial for selecting optimal road slopes and mitigating sand accumulation.
Key Innovation: This study uses CFD numerical simulation to analyze wind-sand flow field distribution characteristics in different longitudinal slope sections of a desert highway, revealing that embankment height and longitudinal slope significantly influence low-wind-speed zones, sand particle transport, and accumulation patterns, providing a scientific basis for slope selection.
203. Tensile strength and failure modes of bio-cemented grains: insights from DEM simulations
Core Problem: While MICP (microbially induced carbonate precipitation) enhances soil geotechnical properties, the particle-scale responses and fundamental insights into the nature of cementitious bonds and their failure modes remain underexplored.
Key Innovation: This study uses DEM simulations to explore the pull-out response of MICP-treated grains under tensile loading, developing a particle generation algorithm to simulate crystal nucleation and growth, and analyzing various CaCO₃ precipitation patterns and failure characteristics, indicating that tensile failure can occur in debonding and internal failure modes depending on cementation content, crystal morphology, and spatial distribution.
204. Conditioning and prediction methods for the permeability of the slurry mixed sandy soil
Core Problem: Excessive soil permeability during EPB TBM operations in sandy strata can induce screw conveyor spewing, and the selection of bentonite slurry parameters for conditioning is mainly based on experience.
Key Innovation: This study researches the permeability characteristics of sandy soils conditioned with bentonite slurry, conducting experiments by adjusting slurry concentration, injection ratio, and air pressure, establishing a permeability prediction model using regression analysis and least squares methods, and providing reasonable slurry parameter ranges for different sandy soils.
205. Analytical model and numerical simulation for stiffened deep cement mixing piles under vertical load
Core Problem: Accurately describing the load–deformation characteristics of stiffened deep cement mixing (SDCM) piles in both isolated piles and pile groups, and simplifying their computational analysis, remains a challenge.
Key Innovation: This study presents a theoretical framework for SDCM piles under vertical load, employing the equivalent pier method to transform piles/pile groups into a unified pier representation, establishing a theoretical model based on the load transfer method to characterize vertical bearing performance, and adopting exponential and elastic-perfectly plastic models for interfaces, with results highly consistent with previous data.
206. On the irreversible free energy and stress-dilatancy relationship for granular materials
Core Problem: Existing stress-dilatancy energy equations for granular materials are incomplete, failing to fully account for the irreversible free energy stored during plastic deformation, leading to an incomplete understanding of macroscopic stress-dilatancy relationships.
Key Innovation: Investigated the irreversible free energy components in granular materials using DEM, proposing a new stress-dilatancy equation that incorporates both frictional dissipation and an irreversible free energy function, which accurately describes macroscopic stress-dilatancy, especially back-hook patterns for dense materials.
207. Investigation of Rock Fracturing During Cut Blasting with a Large Empty Hole Under Non-hydrostatic Confining Stress: Insights for Optimizing Blast Parameters
Core Problem: Non-hydrostatic confining stress in deep mining significantly constrains blast-induced rock fracturing during cut blasting, leading to suboptimal outcomes and hindering the formation of cut cavities.
Key Innovation: Investigated rock fracture behavior during cut blasting with a large empty hole under non-hydrostatic confining stress using finite-element models, revealing anisotropy and diminished fracturing efficacy, and proposed an optimization strategy for blast parameters and an empirical formula for predicting cavity volume.
208. A Study on the Bidirectional Coupling Mechanism Between Multiple Fracture Propagation and Proppant Transport in Complex Natural Fracture Systems
Core Problem: Multi-cluster hydraulic fracturing in naturally fractured reservoirs faces significant challenges due to the strong, bidirectionally coupled interaction between multiple fracture propagation and proppant transport, leading to non-uniform proppant placement and numerical instabilities.
Key Innovation: Developed a bidirectionally coupled numerical method for multi-fracture propagation and proppant transport in complex naturally fractured systems, integrating hydraulic-natural fracture interactions and proppant feedback, and validated it to reveal the strong coupling, the influence of fluid viscosity on fracture morphology and proppant-carrying capacity, and the impact of intersection angle on conductivity.
209. Fabrication and Testing of Transversely Isotropic Rock-Like Materials Prepared Using Self-Compacting Mortar
Core Problem: Accurately studying strength anisotropy in rock engineering, particularly in sedimentary and metamorphic rocks, is challenging due to the difficulty in obtaining representative natural samples and controlling their properties for experimental investigation.
Key Innovation: Developed a methodology to fabricate and test transversely isotropic rock-like materials using self-compacting cement mortar, validating their similarity to natural rock and characterizing their anisotropic mechanical behavior under various loading conditions, and evaluating the applicability of existing failure criteria.
210. Active and passive co-observations from a spaceborne lidar: Retrieving surface reflectance and aerosol optical thickness using ICESat-2 signal and noise data
Core Problem: Retrieving surface reflectance and aerosol optical thickness from spaceborne lidar data.
Key Innovation: A method using active and passive co-observations from a spaceborne lidar (ICESat-2 signal and noise data) to retrieve surface reflectance and aerosol optical thickness.
211. Reconstructing all-weather remotely sensed air temperature via a kernel-based temporal filling and bias correction (KTF-BC) framework
Core Problem: Reconstructing all-weather remotely sensed air temperature accurately.
Key Innovation: A kernel-based temporal filling and bias correction (KTF-BC) framework for reconstructing all-weather remotely sensed air temperature.
212. A novel land surface temperature retrieval method using channel correlation for atmospheric parameter modeling from SDGSAT-1 data
Core Problem: Developing an improved method for land surface temperature retrieval from satellite data.
Key Innovation: A novel land surface temperature retrieval method using channel correlation for atmospheric parameter modeling from SDGSAT-1 data.
213. AgriFM: A multi-source temporal remote sensing foundation model for Agriculture mapping
Core Problem: Developing a robust and versatile model for agriculture mapping using diverse remote sensing data.
Key Innovation: AgriFM, a multi-source temporal remote sensing foundation model specifically designed for agriculture mapping.
214. Hyperspectral BRDF based on UAV measurements can characterize optical properties of flat desert surfaces: A comprehensive comparison with laboratory and satellite data
Core Problem: Characterizing the optical properties of flat desert surfaces using hyperspectral data.
Key Innovation: Demonstrating that hyperspectral BRDF based on UAV measurements can effectively characterize optical properties of flat desert surfaces, validated against laboratory and satellite data.
215. Integrating classifier transfer and sample transfer strategies for in-season crop mapping based on sample weighting technique
Core Problem: Improving the accuracy and efficiency of in-season crop mapping using remote sensing.
Key Innovation: An integrated approach combining classifier transfer and sample transfer strategies with a sample weighting technique for in-season crop mapping.
216. Retrieving forest LAI from Landsat via 3D look-up table generated by realistic LiDAR scenes
Core Problem: Improving the accuracy of forest LAI retrieval from Landsat data.
Key Innovation: A method for retrieving forest LAI from Landsat data using a 3D look-up table generated by realistic LiDAR scenes.
217. A novel hybrid approach for mapping global surface solar radiation with DSCOVR/EPIC: Combining deep learning with physical algorithm
Core Problem: Accurately mapping global surface solar radiation using satellite data.
Key Innovation: A novel hybrid approach combining deep learning with a physical algorithm for mapping global surface solar radiation using DSCOVR/EPIC data.
218. Highest quality remote sensing reflectance database compiled from 20+ years of MODIS-aqua measurements
Core Problem: Providing a high-quality, long-term remote sensing reflectance database.
Key Innovation: Compilation of the highest quality remote sensing reflectance database from over 20 years of MODIS-aqua measurements.
219. Varying sensitivities of RED-NIR-based vegetation indices to the input reflectance affect the detected long-term trends
Core Problem: Understanding how varying sensitivities of RED-NIR-based vegetation indices to input reflectance affect detected long-term trends.
Key Innovation: Analysis of the varying sensitivities of RED-NIR-based vegetation indices to input reflectance and their impact on long-term trend detection.
220. An SW-TES hybrid algorithm for retrieving mountainous land surface temperature from high-resolution thermal infrared remote sensing data
Core Problem: Accurately retrieving land surface temperature in mountainous regions from high-resolution thermal infrared remote sensing data.
Key Innovation: An SW-TES hybrid algorithm developed for retrieving mountainous land surface temperature from high-resolution thermal infrared remote sensing data.
221. Uncovering spatial process heterogeneity from graph-based deep spatial regression
Core Problem: Effectively uncovering and modeling spatial process heterogeneity in spatial data.
Key Innovation: A graph-based deep spatial regression approach to uncover spatial process heterogeneity.
222. Hyperspectral indicators of vegetation vitality across scales: From trees to forests
Core Problem: Identifying and applying hyperspectral indicators to assess vegetation vitality across different scales, from individual trees to entire forests.
Key Innovation: Development and application of hyperspectral indicators for assessing vegetation vitality across multiple scales, with potential implications for wildfire risk.
223. Temporally dense 100-m land surface temperature retrieval via attention-based super-resolution deep learning
Core Problem: Retrieving high-resolution (100-m) and temporally dense land surface temperature data.
Key Innovation: An attention-based super-resolution deep learning framework for retrieving temporally dense 100-m land surface temperature, a parameter relevant to various environmental processes.
224. Effects of confining stress on crack propagation and energy evolution during rock indentation: Insights from 2D-DEM simulations and implications for mechanized mining
Core Problem: Understanding the effects of confining stress on crack propagation and energy evolution during rock indentation, with implications for mechanized mining.
Key Innovation: Insights from 2D-DEM simulations on the effects of confining stress on crack propagation and energy evolution during rock indentation, relevant for rock failure mechanisms.
225. Mechanical and fracture behavior of red sandstone under confining pressure and laser irradiation using CT scanning and uniaxial compression
Core Problem: Investigating the mechanical and fracture behavior of red sandstone under combined confining pressure and laser irradiation.
Key Innovation: An experimental study using CT scanning and uniaxial compression to analyze the mechanical and fracture behavior of red sandstone under confining pressure and laser irradiation, contributing to rock failure understanding.
226. Spatial heterogeneity and controlling mechanisms of multi-interface groundwater recharge in karst critical zone
Core Problem: Understanding the spatial heterogeneity and controlling mechanisms of multi-interface groundwater recharge in karst environments.
Key Innovation: Characterizing these complex recharge dynamics within the karst critical zone.
227. A hybrid Penman-Monteith and machine learning model for simulating evapotranspiration and its components
Core Problem: Accurately simulating evapotranspiration and its various components.
Key Innovation: A hybrid model combining the Penman-Monteith equation with machine learning techniques for improved simulation.
228. From empirical to physical constraints: Revisiting the structure of monthly water balance models with global evaluation
Core Problem: Revisiting the structure of monthly water balance models by incorporating physical constraints and evaluating them globally.
Key Innovation: A revised structure for monthly water balance models incorporating physical constraints and global evaluation.
229. Genetic landform classification of the Tibetan Plateau plains using multisource data
Core Problem: Lack of large-scale automated landform genetic classification methods for complex geomorphic regions like the Tibetan Plateau, which integrate morphology and genesis.
Key Innovation: Development of an automated, multisource data-integrated classification method for plain genetic types on the Tibetan Plateau, achieving high accuracy and providing a transferable framework for complex geomorphic regions.
230. Measuring exposure of agriculture to observed temperature change
Core Problem: Lack of simple indicators to quantify the exposure of agriculture (rural populations, land, crops, livestock) at regional and global levels to observed temperature change thresholds (ΔT > 1.5 °C and ΔT > 2.0 °C).
Key Innovation: Computation of simple indicators of agricultural exposure to temperature change thresholds using FAO statistics, revealing significant increases in exposure of rural populations, land, and specific crops and livestock globally and regionally since 1995, highlighting the urgency for adaptation strategies.
231. Funding cuts could put research into emerging threats to lung health at risk
Core Problem: Emerging environmental threats to lung health, such as wildfire smoke, spore-spread fungal diseases, and microplastics, are on the rise, but US government support for respiratory research and policy is being slashed.
Key Innovation: Highlights the critical risk posed by funding cuts to research into emerging environmental threats to lung health, including wildfire smoke, emphasizing the need for sustained support to address these rising hazards.
232. ProFlow: Zero-Shot Physics-Consistent Sampling via Proximal Flow Guidance
Core Problem: Inferring physical fields from sparse observations while strictly satisfying partial differential equations (PDEs) using deep generative models, without costly retraining or disrupting the prior.
Key Innovation: ProFlow, a proximal guidance framework for zero-shot physics-consistent sampling, which alternates between projecting flow predictions onto physically/observationally consistent sets and mapping the state back to the generative trajectory, achieving superior consistency and accuracy.
233. Reversible Efficient Diffusion for Image Fusion
Core Problem: Diffusion models for image fusion suffer from detail loss due to noise error accumulation and computational inefficiency when incorporating explicit supervision.
Key Innovation: Reversible Efficient Diffusion (RED) model, an explicitly supervised training framework that inherits diffusion models' generative capability while avoiding distribution estimation, addressing detail loss and computational efficiency in image fusion.
234. Robust SDE Parameter Estimation Under Missing Time Information Setting
Core Problem: Parameter estimation for SDEs typically relies on accurately timestamped observational sequences, but existing methods fail when temporal ordering information is missing or corrupted.
Key Innovation: A novel framework that simultaneously reconstructs temporal information and estimates SDE parameters by exploiting asymmetries between forward and backward processes, deriving a score-matching criterion to infer temporal order.
235. TPGDiff: Hierarchical Triple-Prior Guided Diffusion for Image Restoration
Core Problem: Existing all-in-one image restoration methods struggle to reconstruct content in severely degraded regions and integrating semantic information into shallow layers of diffusion models often disrupts spatial structures.
Key Innovation: Triple-Prior Guided Diffusion (TPGDiff) network, which incorporates degradation priors throughout the diffusion trajectory, structural priors into shallow layers, and semantic priors into deep layers, enabling hierarchical and complementary prior guidance for unified image restoration.
236. OSDEnhancer: Taming Real-World Space-Time Video Super-Resolution with One-Step Diffusion
Core Problem: Existing STVSR methods struggle in real-world scenarios with complex unknown degradations and maintaining temporal coherence, as they predominantly address spatiotemporal upsampling under simplified degradation assumptions.
Key Innovation: OSDEnhancer, a framework for real-world STVSR using an efficient one-step diffusion process, which initializes spatiotemporal structures, trains temporal refinement and spatial enhancement mixture of experts (TR-SE MoE), and uses a bidirectional deformable VAE decoder.
237. RAW-Flow: Advancing RGB-to-RAW Image Reconstruction with Deterministic Latent Flow Matching
Core Problem: Existing learning-based RGB-to-RAW reconstruction methods struggle with detail inconsistency and color deviation due to the ill-posed nature of inverse ISP and information loss in quantized RGB images.
Key Innovation: RAW-Flow, a novel framework that reformulates RGB-to-RAW reconstruction as a deterministic latent transport problem using flow matching, enhanced by a cross-scale context guidance module and a dual-domain latent autoencoder, achieving accurate reconstruction.
238. Explainable deep learning reveals the physical mechanisms behind the turbulent kinetic energy equation
Core Problem: Understanding the physical mechanisms governing turbulent kinetic energy transport, which are complex and not fully revealed by classical coherent structures.
Key Innovation: Used an explainable deep learning (XDL) model based on SHapley Additive exPlanations (SHAP) to identify and percolate high-importance structures for the evolution of turbulent kinetic energy budget terms in a turbulent channel flow. Revealed that important structures are predominantly near-wall and associated with sweep-type events, and dissipation is the dominant organizing mechanism of near-wall turbulence.
239. FLOL: Fast Baselines for Real-World Low-Light Enhancement
Core Problem: Current deep learning solutions for low-light image enhancement (LLIE) struggle with efficiency and robustness for real-world scenarios (e.g., scenes with noise, saturated pixels).
Key Innovation: Proposes FLOL, a lightweight neural network that combines image processing in the frequency and spatial domains, achieving fast and robust low-light enhancement comparable to state-of-the-art methods, with real-time processing capabilities for high-resolution images.
240. SeedVR2: One-Step Video Restoration via Diffusion Adversarial Post-Training
Core Problem: Diffusion-based video restoration (VR) models have high computational costs, and extending one-step distillation approaches to high-resolution VR in real-world settings remains challenging.
Key Innovation: SeedVR2 is a one-step diffusion-based VR model that performs adversarial training against real data, incorporating an adaptive window attention mechanism for high-resolution VR and a series of effective losses (including a novel feature matching loss) to achieve comparable or better performance than existing VR approaches in a single step.
241. PLANA3R: Zero-shot Metric Planar 3D Reconstruction via Feed-Forward Planar Splatting
Core Problem: Metric 3D reconstruction of indoor scenes, particularly from unposed two-view images, is challenging, and prior feedforward methods require explicit 3D plane annotations during training.
Key Innovation: PLANA3R, a pose-free framework for zero-shot metric Planar 3D Reconstruction from unposed two-view images, which uses Vision Transformers to extract sparse planar primitives and supervise geometry learning via planar splatting without explicit 3D plane annotations, enabling scalable training and strong generalization.
242. Self-induced stochastic resonance: A physics-informed machine learning approach
Core Problem: Modeling and predicting Self-induced Stochastic Resonance (SISR) in stochastic excitable systems (like the FitzHugh-Nagumo neuron) is challenging, and traditional methods can be computationally intensive.
Key Innovation: A physics-informed machine learning framework that embeds governing stochastic differential equations and SISR-asymptotic timescale-matching constraints into a Physics-Informed Neural Network (PINN) based on a Noise-Augmented State Predictor architecture, accurately predicting spike-train coherence with improved accuracy and generalization.
243. Weakly supervised framework for wildlife detection and counting in challenging Arctic environments: a case study on caribou (Rangifer tarandus)
Core Problem: Scalable and accurate monitoring of caribou populations from aerial imagery is challenging due to severe background heterogeneity, dominant empty terrain, small/occluded targets, and wide variation in density and scale, making manual interpretation labor-intensive and error-prone.
Key Innovation: A weakly supervised patch-level pretraining based on a detection network's architecture (HerdNet) is proposed, learning from empty vs. non-empty labels to produce early knowledge, enhancing detection and counting accuracy on multi-herd imagery and independent test sets, enabling reliable mapping of animal-containing regions.
244. Some Robustness Properties of Label Cleaning
Core Problem: Standard machine learning procedures using raw, noisy labels lack robustness, especially when models are mis-specified or surrogate losses are used.
Key Innovation: Demonstrating that learning procedures relying on aggregated (cleaned) labels achieve superior robustness properties, including stronger risk consistency guarantees and convergence to optimal classifiers even with mis-specified losses, by refining noisy signals.
245. The Second Skin: A Wearable Sensor Suite That Enables Real-Time Human Biomechanics Tracking Through Deep Learning
Core Problem: Obtaining precise, real-time, task-independent, and user-independent joint state estimation for human lower-body kinematics and dynamics.
Key Innovation: A wearable sensor suite (IMUs and pressure insoles) combined with deep learning for real-time, generalizable human biomechanics tracking.
246. Knowledge-Augmented Patient Network Embedding-Based Dynamic Model Selection for Predictive Analysis of Pediatric Drug-Induced Liver Injury
Core Problem: Developing robust machine learning frameworks for EHR-based predictive tasks, addressing data limitations, patient diversity, and complex event mechanisms.
Key Innovation: The Knowledge-augmented Patient Network embedding-based Dynamic model Selection (KPNDS) framework, integrating graph ML, knowledge augmentation, and meta-learning for individualized risk prediction.
247. A NMF-Based Non-Euclidean Adaptive Feature Extraction Scheme for Limb Motion Pattern Decoding in Pattern Recognition System
Core Problem: Low decoding performance and lack of robustness to data drift in existing feature extraction techniques for EMG-based limb motion pattern recognition.
Key Innovation: A NMF-based non-Euclidean adaptive feature extraction scheme that operates unsupervised, reduces data drift, and aligns data distributions for robust limb motion pattern decoding.
248. Detection of Phalaenopsis Fusarium Wilt From Hyperspectral VNIR/SWIR Imagery Using EE-CSPNet
Core Problem: Fusarium wilt poses a significant threat to Phalaenopsis orchid cultivation, leading to substantial economic losses, and requires accurate, nondestructive early-stage disease detection.
Key Innovation: EE-CSPNet, a lightweight deep learning architecture for hyperspectral disease detection in VNIR/SWIR, enabling end-to-end learning of full-band spectral features, integrating blueprint separable convolutions, cross-stage partial fusion, and a novel entropy-enhanced attention module, achieving high accuracy (98.12% VNIR, 97.64% SWIR) with reduced parameters.
249. Sentinel-2 Multispectral Imagery Case-I Water Semianalytical Bathymetry Retrieval Model Assisted by Satellite-Derived Pixel Substrate Spectrum
Core Problem: Accurate bathymetry retrieval in shallow Case-I waters using semianalytical models is challenging due to many unknown parameters and difficulty in obtaining substrate spectra.
Key Innovation: A Case-I water semianalytical bathymetry retrieval model assisted by a pixel substrate spectrum (SBM-P) is proposed, which obtains individual pixel substrate spectra and reparametrizes the SA model, showing improved accuracy over fixed substrate spectrum models.
250. Range Image Reconstruction From GM-APD LiDAR Data Based on Neyman–Pearson Criterion
Core Problem: Geiger-mode avalanche photodiode LiDAR suffers from inaccurate range reconstruction under low signal-to-background ratios (SBRs) and sparse echoes.
Key Innovation: A novel framework is proposed that integrates a global histogram for enhanced photon collection, a Neyman–Pearson criterion-based binary hypothesis testing model for adaptive target interval determination and noise removal, and matched filtering for high-fidelity range image reconstruction.
251. Research on the rock Leeb hardness testing methods and rock classification
Core Problem: The lack of standardized procedures for Leeb hardness testing in rock engineering leads to inaccurate design parameters and increased risks.
Key Innovation: This study systematically explores effective methods for rock Leeb hardness testing, analyzing factors like test number, practices, moisture, roughness, and sample size, proposing a modified empirical formula, and categorizing rocks into four hardness classes, suggesting Leeb hardness as a supplementary parameter for rock mass classification.
252. Comparison of strain and temperature fields between Micro-NPR and PR anchor rods under uniaxial tension
Core Problem: Existing research on anchor rods predominantly focuses on mechanical performance, with limited attention to the coupled evolution of strain and temperature fields during tensile deformation, hindering a comprehensive understanding of their synergistic response.
Key Innovation: This study systematically investigates the evolution and correlation of strain and temperature fields in micro-NPR and ordinary PR anchor rods under uniaxial tension using DIC and IRT, revealing distinct strain localization patterns and consistent two-phase evolution in standard deviation curves for both fields, demonstrating a synergistic mechanical-thermal response.
253. Identification of the critical state line of sands using coupled shear-volumetric strain paths
Core Problem: Identifying the critical state line of sands in $$e-p'$$ space is arduous, often requiring large strains that are challenging to achieve in triaxial testing without sample distortion or equipment limitations.
Key Innovation: This paper demonstrates how the critical state line in $$e-p'$$ space can be indirectly derived from triaxial compression tests with coupled shear-volumetric strain paths (not requiring axial strain beyond 15%), based on the observation that post-phase transformation and post-peak stress ratio instability points form a single curve, which can also be used for calibrating double surface plasticity model parameters.
254. A three-stage framework for stand-level automated stem volume estimation in temperate forests using Mobile laser scanning
Core Problem: Automating stand-level stem volume estimation in temperate forests.
Key Innovation: A three-stage framework for automated stand-level stem volume estimation in temperate forests using Mobile laser scanning.
255. Characterizing diurnal variability in power plant carbon emissions in Asia: A top-down estimation approach constrained by geostationary NO<sub>2</sub> and OCO-3 CO<sub>2</sub> observations
Core Problem: Characterizing diurnal variability in power plant carbon emissions in Asia.
Key Innovation: A top-down estimation approach constrained by geostationary NO2 and OCO-3 CO2 observations to characterize diurnal variability in power plant carbon emissions.
256. Airborne and spaceborne imaging spectroscopy capture belowground microbial communities and physicochemical characteristics in invaded grasslands
Core Problem: Capturing belowground microbial communities and physicochemical characteristics in invaded grasslands using remote sensing.
Key Innovation: Utilizing airborne and spaceborne imaging spectroscopy to capture belowground microbial communities and physicochemical characteristics in invaded grasslands.
257. Global analysis of nitrogen dioxide and formaldehyde column densities from the Pandora global network: Variability and implications for satellite validation
Core Problem: Analyzing global variability of nitrogen dioxide and formaldehyde column densities and their implications for satellite validation.
Key Innovation: A global analysis of NO2 and formaldehyde column densities from the Pandora network, providing insights into variability and implications for satellite validation.
258. Generating an annual 30 m rice cover product for monsoon Asia (2018–2023) using harmonized Landsat and Sentinel-2 data and the NASA-IBM geospatial foundation model
Core Problem: Generating a high-resolution annual rice cover product for monsoon Asia.
Key Innovation: Generating an annual 30m rice cover product for monsoon Asia (2018–2023) by harmonizing Landsat and Sentinel-2 data and leveraging the NASA-IBM geospatial foundation model.
259. Leaf fall witnessed by night-time light: A first attempt to detect urban leaf fall dates using satellite nighttime light data
Core Problem: Detecting urban leaf fall dates using remote sensing.
Key Innovation: A novel approach using satellite nighttime light data to detect urban leaf fall dates.
260. Why do classification models go wrong? The importance of adaptations and acclimations in driving landscape-level spectral variation in Fremont cottonwood
Core Problem: Understanding the causes of errors in classification models due to spectral variation in vegetation.
Key Innovation: Investigating the importance of adaptations and acclimations in driving landscape-level spectral variation in Fremont cottonwood to understand classification model failures.
261. RPI-GMM: A novel structure-based and phenology-independent algorithm for mapping latest 10-m resolution national-level rubber plantations
Core Problem: Developing an accurate and robust method for mapping rubber plantations at high resolution.
Key Innovation: RPI-GMM, a novel structure-based and phenology-independent algorithm for mapping 10-m resolution national-level rubber plantations.
262. Quantification of phytoplankton primary production from space: A revisit based on high-frequency observations with the aid of Himawari-8/AHI
Core Problem: Improving the quantification of phytoplankton primary production from space using high-frequency observations.
Key Innovation: A revisit of phytoplankton primary production quantification from space, leveraging high-frequency observations from Himawari-8/AHI.
263. Determinants of L-band backscatter in dry tropical ecosystems: Implications for biomass mapping
Core Problem: Understanding the factors influencing L-band backscatter in dry tropical ecosystems for improved biomass mapping.
Key Innovation: Identification of determinants of L-band backscatter in dry tropical ecosystems and their implications for biomass mapping.
264. Large-scale tree-level mapping of forest structure including species type with remote sensing data and ground measurements
Core Problem: Achieving large-scale, tree-level mapping of forest structure and species type.
Key Innovation: A method for large-scale tree-level mapping of forest structure, including species type, by integrating remote sensing data and ground measurements.
265. Comparing the performance of different hyperspectral satellite imaging spectroscopy in mapping methane point-source emissions
Core Problem: Evaluating the performance of various hyperspectral satellite imaging spectroscopy systems for mapping methane point-source emissions.
Key Innovation: A comparative analysis of different hyperspectral satellite imaging spectroscopy systems for mapping methane point-source emissions.
266. A kernel-driven BRDF model by accounting for urban building structures: Model development and preliminary application with satellite data
Core Problem: Developing a BRDF model that accurately accounts for urban building structures.
Key Innovation: A kernel-driven BRDF model that incorporates urban building structures, with preliminary application using satellite data.
267. Emulation-based self-supervised SIF retrieval in the O<sub>2</sub>-A absorption band with HyPlant
Core Problem: Developing an efficient and accurate method for retrieving SIF from hyperspectral data.
Key Innovation: An emulation-based self-supervised method for SIF retrieval in the O2-A absorption band using HyPlant data.
268. Development of an interference mitigation chlorophyll index for mitigating soil and canopy dependence to improve vegetation chlorophyll content monitoring
Core Problem: Improving the accuracy of vegetation chlorophyll content monitoring by mitigating soil and canopy dependence in chlorophyll indices.
Key Innovation: Development of an interference mitigation chlorophyll index to improve vegetation chlorophyll content monitoring by reducing soil and canopy dependence.
269. Hyperspectral OCI/PACE observations of the Atlantic <em>Sargassum</em>
Core Problem: Characterizing Atlantic Sargassum using hyperspectral satellite observations.
Key Innovation: Utilizing hyperspectral OCI/PACE observations to study Atlantic Sargassum.
270. Retrieval of global surface phytoplankton community structure using a minimal set of predictors
Core Problem: Retrieving global surface phytoplankton community structure efficiently from remote sensing data.
Key Innovation: A method for retrieving global surface phytoplankton community structure using a minimal set of predictors from remote sensing data.
271. Estimation of sea surface foam coverage and effective foam layer thickness from satellite microwave measurements
Core Problem: Accurately estimating sea surface foam coverage and thickness from satellite data.
Key Innovation: A method for estimating sea surface foam coverage and effective foam layer thickness from satellite microwave measurements.
272. Edge effects on forest dynamics in China from 2000 to 2020: Evidence from satellite remote sensing
Core Problem: Quantifying and understanding edge effects on forest dynamics in China.
Key Innovation: Providing evidence of edge effects on forest dynamics in China from 2000 to 2020 using satellite remote sensing.
273. Weak supervision makes strong details: fine-grained object recognition in remote sensing images via regional diffusion with VLM
Core Problem: Improving fine-grained object recognition in remote sensing images, especially with limited strong supervision.
Key Innovation: A regional diffusion approach with Vision-Language Models (VLM) and weak supervision to achieve strong detail recognition in remote sensing images.
274. A weakly supervised approach for large-scale agricultural parcel extraction from VHR imagery via foundation models and adaptive noise correction
Core Problem: Efficient and accurate large-scale agricultural parcel extraction from VHR imagery, particularly with weak supervision and noise.
Key Innovation: A weakly supervised approach leveraging foundation models and adaptive noise correction for robust large-scale agricultural parcel extraction from VHR imagery.
275. Knowledge distillation with spatial semantic enhancement for remote sensing object detection
Core Problem: Improving remote sensing object detection performance, especially for smaller models or in resource-constrained environments.
Key Innovation: A knowledge distillation framework with spatial semantic enhancement to improve remote sensing object detection.
276. AMS-Former: Adaptive multi-scale transformer for multi-modal image matching
Core Problem: Improving multi-modal image matching accuracy and robustness, especially across different scales.
Key Innovation: AMS-Former, an adaptive multi-scale transformer network designed for robust multi-modal image matching.
277. Progressive uncertainty-guided network for binary segmentation in high-resolution remote sensing imagery
Core Problem: Improving binary segmentation accuracy in high-resolution remote sensing imagery, especially in challenging areas.
Key Innovation: A progressive uncertainty-guided network designed for robust binary segmentation in high-resolution remote sensing imagery.
278. Beyond synthetic scenarios: Weakly-supervised super-resolution for spatiotemporally misaligned remote sensing images
Core Problem: Performing super-resolution on real-world, spatiotemporally misaligned remote sensing images without relying solely on synthetic scenarios.
Key Innovation: A weakly-supervised super-resolution method specifically designed for spatiotemporally misaligned remote sensing images, moving beyond synthetic data.
279. MARSNet: A Mamba-driven adaptive framework for robust multisource remote sensing image matching in noisy environments
Core Problem: Achieving robust multisource remote sensing image matching, especially in noisy environments.
Key Innovation: MARSNet, a Mamba-driven adaptive framework for robust multisource remote sensing image matching in noisy environments.
280. RSMT: Robust stereo matching training with geometric correction, clean pixel selection and loss weighting
Core Problem: Improving the robustness and accuracy of stereo matching, particularly in challenging scenarios.
Key Innovation: RSMT, a robust stereo matching training approach incorporating geometric correction, clean pixel selection, and loss weighting.
281. Two-stage offline knowledge distillation for onboard registration of multispectral satellite images
Core Problem: Efficient and accurate onboard registration of multispectral satellite images, especially for resource-constrained satellite platforms.
Key Innovation: A two-stage offline knowledge distillation method for improving onboard registration of multispectral satellite images.
282. Towards High spatial resolution and fine-grained fidelity depth reconstruction of single-photon LiDAR with context-aware spatiotemporal modeling
Core Problem: Achieving high spatial resolution and fine-grained fidelity depth reconstruction from single-photon LiDAR data.
Key Innovation: A context-aware spatiotemporal modeling approach for high spatial resolution and fine-grained fidelity depth reconstruction of single-photon LiDAR data.
283. A laboratory-based spectrometer intercomparison for the measurement of snow spectra
Core Problem: Ensuring accuracy and comparability of snow spectral measurements across different laboratory spectrometers.
Key Innovation: A laboratory-based intercomparison study to evaluate and improve the accuracy of snow spectra measurements using various spectrometers, relevant for snow/ice hazard monitoring.
284. Experimental insights into water flow path and relative permeability in porous media through visual microfluidic technology
Core Problem: Understanding water flow paths and relative permeability in porous media.
Key Innovation: Experimental insights gained through the application of visual microfluidic technology.
285. Cover crop roots redesign the pore structure and organic matter composition to alter hydraulic properties of the rhizosphere
Core Problem: Understanding how cover crop roots influence the hydraulic properties of the rhizosphere.
Key Innovation: Demonstrating that cover crop roots redesign pore structure and organic matter composition, thereby altering hydraulic properties.
286. Hidden contribution of canopy interception to afforestation-driven evapotranspiration enhancement
Core Problem: Quantifying the contribution of canopy interception to the overall enhancement of evapotranspiration caused by afforestation.
Key Innovation: Revealing the previously hidden contribution of canopy interception to afforestation-driven evapotranspiration enhancement.
287. Irrigation reshaping groundwater dynamics to regulate salinity threshold and transport
Core Problem: Understanding how irrigation practices alter groundwater dynamics and influence salinity thresholds and transport.
Key Innovation: Investigating the specific ways irrigation reshapes groundwater dynamics to regulate salinity.
288. Evapotranspiration dominates vegetation cooling in drylands under hydrological limitations
Core Problem: Identifying the dominant mechanism for vegetation cooling in dryland environments, especially under hydrological limitations.
Key Innovation: Demonstrating that evapotranspiration is the primary driver of vegetation cooling in drylands under these conditions.
289. Significance of ancient artesian fresh groundwater below the playa of a hypersaline terminal lake of hemispheric significance
Core Problem: Understanding the significance and characteristics of ancient artesian fresh groundwater systems located beneath hypersaline terminal lake playas.
Key Innovation: Characterizing this ancient groundwater system and its broader implications.
290. Integrated multi-dimensional framework for water conservation capacity evaluation and attribution in the Yellow River water conservation area
Core Problem: Evaluating and attributing water conservation capacity in the Yellow River water conservation area.
Key Innovation: An integrated multi-dimensional framework developed for comprehensive evaluation and attribution.
291. The interaction between vegetation greenness and hydro-climatic factors in China
Core Problem: Understanding the complex interactions between vegetation greenness and various hydro-climatic factors across China.
Key Innovation: Analyzing these interactions to reveal spatial and temporal patterns in China.
292. Markov-chain Monte Carlo estimation of aquifer parameters, non-linear well losses, and probable costs of water extraction
Core Problem: Estimating aquifer parameters, non-linear well losses, and the probable costs associated with water extraction.
Key Innovation: Utilizing Markov-chain Monte Carlo estimation for a robust assessment of these hydrological and economic factors.
293. Differentiable parameter learning of reservoir operation modules
Core Problem: Developing a method for differentiable parameter learning in reservoir operation modules.
Key Innovation: A differentiable parameter learning approach for reservoir operation modules.
294. A situated proposal for a grounded approach to socio-hydrology
Core Problem: Proposing a situated and grounded approach to the field of socio-hydrology.
Key Innovation: A situated proposal for a grounded approach to socio-hydrology.
295. A graph-based algorithm for the Continuous-Projection Embedded Discrete Fracture Model (CpEDFM-U) to simulate two-phase flows in naturally fractured porous media using the MPFA-D method on general unstructured meshes
Core Problem: Developing a graph-based algorithm for the CpEDFM-U to simulate two-phase flows in naturally fractured porous media using the MPFA-D method on unstructured meshes.
Key Innovation: A graph-based algorithm (CpEDFM-U with MPFA-D) for simulating two-phase flows in fractured porous media.
296. Enhancing the estimation of soil water content using a resistive heater in the dual-probe distributed temperature sensing approach
Core Problem: Improving the accuracy of soil water content estimation using distributed temperature sensing.
Key Innovation: Enhancement of soil water content estimation using a resistive heater in the dual-probe distributed temperature sensing approach.
297. Monitoring dry snow metamorphism using 4D tomography across 20 experimental conditions
Core Problem: Refined observations of the temporal evolution of snow microstructure are scarce and often limited in terms of explored snow evolution conditions, hindering improved understanding and modeling of snow metamorphism.
Key Innovation: Development of a snow-metamorphism cell for continuous control of thermal boundary conditions during X-ray tomography, used to conduct 20 experiments and provide a unique 4D dataset of dry snow microstructure evolution over 7 days, suitable for investigating local processes and computing physical properties.
298. Multimodal learning with next-token prediction for large multimodal models
Core Problem: Developing a unified algorithm that can learn from and generate across modalities (text, images, video) has been a fundamental challenge in AI, with next-token prediction's extension to multimodal domains remaining limited.
Key Innovation: Introduction of Emu3, a family of multimodal models trained solely with next-token prediction, which equals the performance of well-established task-specific models across perception and generation, matching flagship systems without diffusion or compositional architectures, and demonstrating coherent video generation and interleaved vision-language generation.