TerraMosaic Daily Digest: Mar 9, 2026
Daily Summary
This March 9, 2026 digest compiles 84 selected papers from 1582 analyzed studies. The strongest slope papers move beyond inventory description: debris-flow mitigation is tested through multistage check-dam design for tailings-dam failure; earthquake-induced landslide assessment improves by fusing Newmark displacement, fault-direction effects, and multi-event inventories; and studies of rock avalanches, loess failure, and sackung clarify how rainfall, fragmentation, and internal damage control onset, mobility, and precursor signals. Geotechnical papers on liquefaction, tunnel stability, and deep excavations extend the same shift toward process-aware prediction in subsurface infrastructure.
Flood, cryosphere, and Earth observation research is becoming more decision-ready. New contributions couple community-based Himalayan warning systems, pluvial-runoff tools, rainfall downscaling, and reservoir-operation frameworks into actionable flood intelligence, while glacier, frost-heave, and frozen-soil studies sharpen the treatment of long-memory environmental change. Coastal wetland stability, subsidence-driven inundation, and offshore storage screening further show that hazard emerges from slow system reorganization as much as from discrete extremes. Across the monitoring stack, foundation models, multisensor fusion, and satellite-video methods are increasingly being adapted to geohazard problems through physical priors, uncertainty control, and hazard-specific targets, moving Earth observation pipelines closer to operational use.
Key Trends
The main trajectory is from descriptive hazard mapping to mechanism-aware slope prediction, operational flood intelligence, and physically anchored Earth observation models.
- Cascading slope hazards are being resolved with explicit mechanics: check-dam performance, earthquake-induced landslide fusion models, off-season rock-avalanche triggers, fragmentation-driven runout, and loess acoustic precursors all tie hazard estimates to failure physics.
- Flood research is moving from mapping to operations: early-warning design, runoff and pluvial tools, rainfall downscaling, dam-break-aware reservoir assessment, and vapor-informed forecasting are built to support decisions rather than post hoc description.
- Geotechnical resilience is becoming time-dependent and data-assimilative: liquefaction transfer learning, tunnel unloading analyses, deep-excavation digital twins, grouting simulations, and soil-water retention models explicitly track evolving subsurface states.
- Cryosphere and frozen-ground studies are converging on long-memory system change: glacier surge dynamics, Thwaites retreat, frost-heave synthesis, ground-surface temperature estimation, and frozen-soil thermodynamics all emphasize hazard emergence under persistent boundary forcing.
- Earth observation foundation models are becoming hazard-specific: current advances add physical priors, uncertainty handling, and exposure-aware targets that move multisensor and satellite-video pipelines closer to operational geohazard analysis.
Selected Papers
This digest features 84 selected papers from 1582 papers 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. Design and effectiveness evaluation of multistage check dams for debris flow of tailings dam failure
Core Problem: Debris flows resulting from tailings dam failures pose a significant threat to downstream communities and infrastructure, necessitating effective mitigation strategies.
Key Innovation: The study designs a multistage check dam based on an arch dam and evaluates its effectiveness in mitigating tailings debris flows using UAV photogrammetry and numerical simulations, demonstrating significant reductions in flow velocity, submergence depth, and impact force on downstream buildings.
2. Earthquake-induced landslides hazard assessment with fused newmark displacement and multi-event inventories
Core Problem: Assessing the hazard of earthquake-induced landslides is crucial for pre-earthquake planning and post-earthquake rescue operations, requiring accurate models that can integrate various factors and data sources.
Key Innovation: The study integrates landslide data from multiple historical earthquakes and introduces a Fault Direction Effect (FDE) factor. It develops coupled Newmark and machine learning models, with the best performing model (improved Newmark coupled with Backpropagation Neural Network(BPNN) incorporating peak ground acceleration (PGA)) showing significant performance improvements in both pre-earthquake prediction and post-earthquake inversion.
3. Mapping the impacts of recurrent floods and windstorms in Bauchi, Nigeria: a hybrid multi-criteria approach using fuzzy-AHP
Core Problem: Bauchi city, Nigeria, experiences recurrent flood and windstorm disasters, with significant disparities in impacts across different neighborhoods.
Key Innovation: Using Fuzzy Analytic Hierarchy Process (AHP) and Inverse Distance Weighting (IDW) spatial interpolation techniques to assess patterns, severity, and exposure impacts of flood and windstorm hazards, revealing the connection between urban planning and urban disaster vulnerability.
4. Triggering mechanisms and dynamics of an off-season rainfall-induced rock avalanche in the Wumeng Mountains, China
Core Problem: Understanding the triggering mechanisms and dynamics of rock avalanches, especially those occurring during unusual conditions like the dry season, is crucial for hazard assessment and mitigation in mountainous regions.
Key Innovation: The study combines field investigations, UAV photogrammetry, laboratory testing, and DEM simulations to analyze a specific rock avalanche event, providing quantitative insights into the roles of lithology, topography, and rainfall in avalanche dynamics. The DEM simulations reconstruct the dynamic evolution of the event, indicating peak velocities and distinct acceleration-deceleration phases.
5. Study on acoustic emission characteristics and failure precursors of Malan loess under uniaxial compression
Core Problem: Traditional displacement monitoring is ineffective in capturing early instability signals of loess collapse, leading to poor disaster warning timeliness.
Key Innovation: The study uses acoustic emission (AE) technology to conduct uniaxial compression tests on Malan loess, identifying reliable AE-based failure precursors (sudden increase in AE ringing count, abrupt b-value drop, and “high-frequency disappearance–low-frequency surge”) for early warning of loess collapse.
6. Fragmentation induced mobility and runout mechanism of large rock avalanche: A case study on the Tibetan Plateau, China
Core Problem: The mechanisms underlying the hypermobility of large rock avalanches are not fully understood, limiting the ability to predict their runout distances and mitigate their hazards.
Key Innovation: The study combines field surveys, morphological analysis, sedimentological analysis, and numerical simulations to investigate the Tagharma rock avalanche, highlighting the enhanced mobility from fragmentation and basal interaction, and the critical role of initial rock mass structure in controlling fragmentation and deposit morphology.
7. HURRI-GAN: A Novel Approach for Hurricane Bias-Correction Beyond Gauge Stations using Generative Adversarial Networks
Core Problem: Physical simulation models for hurricane forecasting require significant computational resources and time, especially at high resolutions, hindering near real-time demands of emergency responders. Systemic errors in these models also need correction.
Key Innovation: HURRI-GAN, an AI-driven approach using time series generative adversarial networks (TimeGAN), augments physical simulation model results to compensate for systemic errors and extrapolate bias corrections beyond gauge station locations, reducing the necessary mesh size and runtime without loss in forecasting accuracy.
8. On the coupled geometrical-mechanical origin of the earthquake b-value in fault networks
Core Problem: The underlying physical origin of the earthquake b-value, which quantifies the relative frequency of small versus large events, remains poorly understood, particularly the roles of geometrical versus mechanical controls.
Key Innovation: Analytical and numerical models demonstrate that the b-value emerges from the power-law scaling of fault rupture area together with the scaling of slip magnitude, providing a physically grounded interpretation of earthquake b-values.
9. N-Tree Diffusion for Long-Horizon Wildfire Risk Forecasting
Core Problem: Long-horizon wildfire risk forecasting requires generating probabilistic spatial fields under sparse event supervision while maintaining computational efficiency across multiple prediction horizons.
Key Innovation: N-Tree Diffusion (NT-Diffusion), a hierarchical diffusion model designed for long-horizon wildfire risk forecasting that shares early denoising stages and branches at later levels, reducing redundant sampling and improving accuracy and inference cost.
10. Post-Disaster Affected Area Segmentation with a Vision Transformer (ViT)-based EVAP Model using Sentinel-2 and Formosat-5 Imagery
Core Problem: Refining disaster-affected area segmentation from remote sensing imagery is crucial for supporting and enhancing disaster response efforts, but accurate ground truth is often unavailable.
Key Innovation: A vision transformer (ViT)-based deep learning framework is proposed to refine disaster-affected area segmentation from Sentinel-2 and Formosat-5 imagery, using PCA-based feature space analysis and a confidence index to expand labels and train encoder-decoder models with multi-band inputs.
11. Surging Dynamics of the ZhongFeng Glacier, Western Kunlun Mountains
Core Problem: Understanding the surge dynamics of glaciers in the Western Kunlun Mountains, a region with a high concentration of surge-type glaciers.
Key Innovation: Analysis of the 2021–2023 surge of the eastern branch of ZhongFeng Glacier (ZFG) and review of the 2003–2004 surge of its western branch, utilizing multisource digital elevation models, Landsat, Sentinel-2, and meteorological data to reveal surge initiation during summer, peak flow velocities, and ice mass gains, concluding that both branches are influenced by hydrological mechanisms and topographic differences.
12. Integrating SMART principles in flood early warning system design in the Himalayas
Core Problem: Extreme rainfall events increase vulnerability to flash floods in mountainous regions, and there is a need for effective, community-led urban flood early warning systems (EWS) in data-scarce areas.
Key Innovation: The study demonstrates how low-cost, community-engaged hydrometeorological monitoring can improve urban flood early warning in the Lesser Himalayas, using a network of water-level sensors and rain gauges to capture spatial variability in rainfall.
13. Accumulation-based Runoff and Pluvial Flood Estimation Tool (AccRo v.1.0)
Core Problem: Pluvial flood forecasting and disaster management require spatially distributed inundation depth and overland flow quantities, which are often derived from computationally demanding 2D hydrodynamic models, limiting real-time forecasting capabilities.
Key Innovation: The AccRo model is a computationally efficient method to derive maximum inundation depth, flow velocity, and specific discharge of a flood event at larger spatial scales, based on an improved flow accumulation method.
14. Deformation monitoring and risk assessment of ultra-high voltage transmission channel based on Sentinel-1 and RADARSAT-2 multi-source InSAR data
Core Problem: UHV transmission lines are threatened by complex terrain, geological disasters, and human activities, requiring high-precision deformation monitoring and risk identification.
Key Innovation: Integration of Sentinel-1A and RADARSAT-2 data with SBAS-InSAR technology and Getis-Ord Gi* statistics to perform time-series deformation analysis and assess subsidence hot spots, enabling rapid identification of high-risk sections in UHV transmission corridors.
15. Study on impulse wave propagation attenuation generated by the failure of large deposits in a reservoir of the Lancang River
Core Problem: Impulse waves generated by the failure of large deposits in reservoirs pose a direct threat to dam safety, necessitating a thorough understanding of their propagation and attenuation characteristics.
Key Innovation: The study uses centrifuge tests, large-scale physical simulation tests, and numerical simulations based on fluid-solid coupling to investigate the propagation attenuation characteristics of impulse waves generated by reservoir deposit failure, analyzing the influence of river channel geometry and identifying key factors affecting wave attenuation.
16. Sackung at Bald Eagle ridge, central Colorado: An updated interpretation of ridge-spreading movement, structures, and mechanisms from 50 years (1975–2025) of U.S. Geological Survey research
Core Problem: Slow gravitational failures of mountain peaks and ridges (sackungen) are poorly understood, limiting the ability to predict and mitigate their potential hazards.
Key Innovation: Long-term measurements (50 years) of slope movement at Bald Eagle ridge provide insights into sackung formation and evolution, revealing the importance of basal-slip surfaces, normal faulting, and flexural toppling in the failure mechanism.
17. Gravelly soil liquefaction prediction via transfer learning from sandy soil using Bayesian networks
Core Problem: Earthquake-induced liquefaction of gravelly soils threatens infrastructure stability, but prediction accuracy is severely constrained by data scarcity.
Key Innovation: A multisource Bayesian network transfer learning (MS-BNTL) method using sandy soil data as the source domain and gravelly soil data as the target domain is proposed. The method identifies and transfers similar seismic causal subnetworks between domains while avoiding negative transfer from mismatched knowledge.
18. Unloading-induced failure mechanisms of layered phyllite and the influence on tunnel stability
Core Problem: Layered phyllite presents engineering challenges during tunnel construction, including rock mass collapse and asymmetric structural failure, particularly under unloading conditions.
Key Innovation: Integration of laboratory experiments, numerical simulations, and field monitoring to investigate the unloading mechanical properties of layered phyllite, applying the findings to real-world engineering projects.
19. Data-mechanism hybrid-driven digital twin for spatiotemporal prediction of multiple evolving risk in deep excavation
Core Problem: Accurately forecasting the joint spatiotemporal evolution of multi-source risks in deep excavation is critical for safe and efficient project delivery, but existing data-driven methods often lack physical interpretability.
Key Innovation: Develops a data–mechanism hybrid digital twin framework that couples the physical excavation system with graph-based spatiotemporal prediction, using a Temporal Heterogeneous Graph Attention Network (THGAN) to predict the spatiotemporal evolution of multi-source risks.
20. Quantification method for comprehensive flood control benefits of an earth-rockfill dam reservoir considering dam break effects
Core Problem: Quantifying the flood control benefits (FCB) of earth-rockfill dam reservoirs is critical for evaluating investment rationality, especially considering dam break effects (DBE).
Key Innovation: Proposed methods for determining the critical overtopping flood that causes DBE, calculating the single dam break effects (SDBE) and the average DBE, and quantifying comprehensive flood control benefits (CFCB) by integrating DBE and FCB.
21. How Does Fluid Exchange Between Pores in Unsaturated Porous Media?
Core Problem: Tracking fluid exchange among distinct pore domains in unsaturated media remains difficult because pore connectivity evolves with saturation and conventional NMR interpretations break down.
Key Innovation: Coupled micro-CT-derived pore-network characterization with NMR-response simulation to show how declining saturation drives tortuous and ultimately disconnected flow pathways, enabling pore-coupling diagnosis in vadose materials.
22. Small Target Detection Based on Mask-Enhanced Attention Fusion of Visible and Infrared Remote Sensing Images
Core Problem: Detecting small targets in remote sensing images is challenging due to weak textures and complex backgrounds, hindering high-precision detection with general algorithms.
Key Innovation: ESM-YOLO+, a lightweight visible infrared fusion network with a Mask-Enhanced Attention Fusion (MEAF) module and training-time Structural Representation (SR) enhancement for improved small-target detection.
23. Physics-Guided VLM Priors for All-Cloud Removal
Core Problem: Existing cloud removal pipelines separate thin-cloud correction from thick-cloud reconstruction, requiring explicit cloud-type decisions and often leading to error accumulation and discontinuities.
Key Innovation: A Physical-VLM All-Cloud Removal (PhyVLM-CR) approach integrates the semantic capability of a Vision-Language Model (VLM) into a physical restoration model, achieving high-fidelity unified cloud removal via adaptive weighting.
24. Designing probabilistic AI monsoon forecasts to inform agricultural decision-making
Core Problem: Farmers need reliable monsoon forecasts to make informed decisions, but current forecasts often lack the necessary skill and are not tailored to individual farmer's needs.
Key Innovation: Develops a decision-theory framework and AI-driven system for tailoring probabilistic monsoon forecasts to farmers' needs, blending AI weather models with a statistical model of farmer expectations.
25. The direct spectral element method for the calculation of synthetic seismograms in self-gravitating, spherically symmetric planets
Core Problem: Calculating synthetic seismograms in self-gravitating, spherically symmetric planets, especially with complex fluid stratification, requires accurate and efficient methods.
Key Innovation: Implementation of the direct solution method (DSM) using radial spectral elements with a displacement formulation throughout, allowing for full accounting of self-gravitation and arbitrary fluid stratification. Code is benchmarked against existing methods with excellent agreement.
26. Prithvi-EO-2.0: A Versatile Multi-Temporal Foundation Model for Earth Observation Applications
Core Problem: Need for improved geospatial foundation models for Earth observation tasks, including disaster response.
Key Innovation: Prithvi-EO-2.0, a geospatial foundation model trained on multi-temporal satellite data with temporal and location embeddings, demonstrating enhanced performance across various geospatial tasks, including disaster response.
27. SPEX: A Vision-Language Model for Land Cover Extraction on Spectral Remote Sensing Images
Core Problem: Spectral information is underutilized in vision-language models for pixel-level interpretation, resulting in suboptimal performance in multispectral remote sensing scenarios for land cover extraction.
Key Innovation: SPEX, a multimodal LLM designed for instruction-driven land cover extraction, leveraging a vision-language instruction-following dataset named SPIE that encodes spectral priors of land-cover objects into textual attributes.
28. A quasi-manifold-based probabilistic method for real-time interpretation of interbedded strata from sparse boreholes
Core Problem: Delineating the spatial distribution of interbedded strata from sparse boreholes is challenging due to complex spatial features and significant variability, impacting coastal infrastructure resilience.
Key Innovation: Proposes a quasi-manifold learning approach to stochastically interpret high-dimensional features by traversing an embedded manifold, enabling the construction of geological cross-sections and 3D domains with quantified stratigraphic uncertainty.
29. Understanding the combined mental health impacts of flooding and COVID-19 in Hue City, Central Vietnam
Core Problem: There is a lack of understanding regarding the combined impact of co-occurring disasters, such as floods and COVID-19, on mental health, particularly in developing countries.
Key Innovation: The study uses face-to-face interviews and regression models to assess the combined mental health impacts of flooding and COVID-19 in Hue City, Vietnam, identifying specific stressors and their effects.
30. Geospatial assessment of United States tropical cyclone disaster risk
Core Problem: Understanding the spatial patterns of tropical cyclone (TC) hazard risk, population exposure, societal vulnerability, and community resilience is critical for effective disaster management and mitigation.
Key Innovation: The study develops TC hazard and disaster indices combining measures of hazard risk, exposure, vulnerability, resilience, and disaster risk to identify counties most prone to TC disasters, highlighting the importance of inland flash flooding and providing insights for targeted mitigation strategies.
31. Experimental Study on Dynamic Shear Properties of Infilled Joints Under Dynamic Shearing: Insights from Normal Stress and Shear Rate
Core Problem: Insufficient exploration of the dynamic shear behavior of infilled joints under impact load, crucial for geotechnical engineering design and maintenance.
Key Innovation: Impact-induced direct shear tests on infilled planar granite joints using a biaxial Hopkinson pressure bar system, investigating the effect of normal stress and shear rate on dynamic shear characteristics and grain comminution.
32. Frost heave in soils: Research status and perspectives
Core Problem: Frost heave is a multi-disciplinary problem involving geotechnical engineering, hydrology, environmental science, and climatology, with key research challenges remaining.
Key Innovation: Provides an overview of experimental, theoretical, and numerical developments in frost heave, identifying key research challenges and proposing future studies focus on frost heave in coarse-grained soils and its engineering implications.
33. Plant community structure and wetland stability: A biogeomorphic perspective for tidal creek dynamics in the Danish Wadden Sea
Core Problem: Understanding the interplay between plant assemblages, soil conditions, and sedimentary/erosional dynamics in natural wetland ecosystems is crucial for coastal stability, but long-term in situ studies are lacking.
Key Innovation: Long-term observational investigation of geomorphological transformations and plant community structure across tidal channels, highlighting the role of Atriplex portulacoides in mediating plant species diversity and driving geomorphological modifications.
34. Projected seawater inundation induced by land subsidence will weaken the coastal carbon sink effect
Core Problem: Climate change-driven sea-level rise (SLR) and anthropogenic land subsidence (LS) pose significant challenges to coastal sustainability, and their synergistic impact on coastal carbon sink effect remains unclear.
Key Innovation: Model-based projections suggest that the combined effects of SLR and LS could inundate between 4.47% and 11.58% of the YRD terrestrial area by 2035, resulting in a 2.22% to 7.07% decline in regional carbon sequestration capacity.
35. Mechanical properties of novel prefabricated inverted arch in NATM tunnels: Insights from numerical experiment and in-situ tests
Core Problem: Improving the mechanical properties of inverted arches and enhancing the stability of the foundation surrounding rock in NATM tunnels.
Key Innovation: Proposes a prefabricated inverted arch construction technology and demonstrates its effectiveness through in-situ tests and numerical simulations, showing reduced settlement and convergence compared to cast-in-place inverted arches.
36. A coupled material point method radial basis functions for free-surface water waves by large deformations in bottom topography
Core Problem: Simulating free-surface water waves generated by large deformations in bottom topography, such as those caused by landslides or coastal erosion, requires accurately capturing the complex interplay between solid and fluid domains, which is challenging for traditional numerical methods.
Key Innovation: The study proposes a hybrid numerical framework that couples the Material Point Method (MPM) with Radial Basis Functions (RBF) to simulate free-surface water waves generated by large deformations in bottom topography, enabling accurate resolution of complex soil deformations and fluid flow with a stable wet/dry treatment for scenarios involving near-dry beds.
37. Recent Observations of Thwaites Glacier, West Antarctica Are Consistent With High Rates of Loss in Next 50 Years
Core Problem: Thwaites Glacier is experiencing accelerating mass loss, and accurate projections of future mass loss are needed to assess sea-level rise risks.
Key Innovation: Applied transient calibration to ice-sheet models using time-varying velocity and surface elevation data, demonstrating that surface-elevation-constrained models provide the most realistic representation of observed dynamical changes and project the largest future mass losses.
38. California Temperature Since 1520 CE Shows Interactions in Extremes of Heat, Drought, and Fire
Core Problem: Summer maximum temperatures in the Sierra Nevada have risen rapidly, and placing this warming in a long-term context is needed to understand ecological vulnerability.
Key Innovation: Developed a 504-year reconstruction of growing-season Tmax from tree-ring data, showing that the 20th-21st centuries were the warmest of the past five and that post-1980 warming and compound extremes mark a new era of temperature-driven ecological vulnerability.
39. Residual NAPL Intermittent Expulsion With Increasing Freeze–Thaw Cycles in Saturated Porous Media: Findings From a 2.5‐D Microfluidic Platform
Core Problem: Freeze-thaw remobilization of residual non-aqueous phase liquids in saturated porous media is poorly resolved at pore scale, limiting contaminant prediction in cold regions.
Key Innovation: Built a 2.5-D microfluidic platform that directly visualizes water-ice-NAPL interactions and defines an expulsion factor showing why early freeze-thaw cycles favor contaminant mobilization before snap-off and compression suppress it.
40. Joint 3D Gravity and Magnetic Inversion via Rectified Flow and Ginzburg-Landau Guidance
Core Problem: The ill-posed nature of jointly reconstructing subsurface densities from magnetic and gravitational data, and the limitations of deterministic algorithms in capturing the distribution of possible solutions.
Key Innovation: A novel framework that reframes 3D gravity and magnetic joint inversion as a rectified flow on the Noddyverse dataset, using a Ginzburg-Landau (GL) regularizer for ore identification and a guidance methodology based on GL theory.
41. DLRMamba: Distilling Low-Rank Mamba for Edge Multispectral Fusion Object Detection
Core Problem: Current State Space Models (SSMs) like Mamba suffer from parameter redundancy, hindering deployment on resource-constrained hardware for multispectral fusion object detection in maritime surveillance and remote sensing.
Key Innovation: Low-Rank Two-Dimensional Selective Structured State Space Model (Low-Rank SS2D) and Structure-Aware Distillation strategy to reduce computational complexity and memory footprint while preserving high-fidelity spatial modeling.
42. SIGMAE: A Spectral-Index-Guided Foundation Model for Multispectral Remote Sensing
Core Problem: Applying MAE to multispectral remote sensing images is challenging due to complex backgrounds, indistinct targets, and the lack of semantic guidance during masking.
Key Innovation: SIGMAE, a Spectral Index-Guided MAE approach that incorporates domain-specific spectral indices to guide dynamic token masking toward informative regions using Semantic Saliency-Guided Dynamic Token Masking (SSDTM).
43. FedEU: Evidential Uncertainty-Driven Federated Fine-Tuning of Vision Foundation Models for Remote Sensing Image Segmentation
Core Problem: Dynamic adaptation of pretrained models to heterogeneous client data in federated RSIS increases update uncertainty and compromises the reliability of collaborative optimization.
Key Innovation: FedEU, a federated optimization framework for fine-tuning RSIS models driven by evidential uncertainty, using personalized evidential uncertainty modeling and a Top-k uncertainty-guided weighting (TUW) strategy.
44. SiamGM: Siamese Geometry-Aware and Motion-Guided Network for Real-Time Satellite Video Object Tracking
Core Problem: Single object tracking in satellite videos is challenged by small target size, blurred background, large aspect ratio changes, and frequent visual occlusions, causing appearance-based trackers to accumulate errors and lose targets.
Key Innovation: SiamGM is a geometry-aware and motion-guided Siamese network that introduces an Inter-Frame Graph Attention (IFGA) module and an Aspect Ratio-Constrained Label Assignment (LA) method to mitigate spatial ambiguities, and a Motion Vector-Guided Online Tracking Optimization method to utilize historical trajectory information.
45. Geometric Knowledge-Assisted Federated Dual Knowledge Distillation Approach Towards Remote Sensing Satellite Imagery
Core Problem: Analyzing remote sensing satellite imagery (RSSI) using federated learning (FL) faces challenges due to the large scale and data heterogeneity of images collected from multiple satellites.
Key Innovation: A Geometric Knowledge-Guided Federated Dual Knowledge Distillation (GK-FedDKD) framework that distills a teacher encoder from multiple student encoders, generates global geometric knowledge (GGK) for local embedding augmentation, and uses a novel loss function and multi-prototype generation pipeline to stabilize training.
46. BuildMamba: A Visual State-Space Based Model for Multi-Task Building Segmentation and Height Estimation from Satellite Images
Core Problem: Accurate building segmentation and height estimation from single-view RGB satellite imagery are ill-posed due to structural variability and the high computational cost of global context modeling, leading to boundary bleeding and underestimation of high-rise structures.
Key Innovation: BuildMamba, a unified multi-task framework exploiting the linear-time global modeling of visual state-space models, with a Mamba Attention Module, a Spatial-Aware Mamba-FPN, and a Mask-Aware Height Refinement module to improve performance and robustness.
47. DMFS: differentiable modeling for frozen soil thermodynamic characteristics
Core Problem: Predicting frozen soil behavior in cold regions requires robust thermodynamic constitutive relationships, but estimating these relationships is challenging and requires expensive testing.
Key Innovation: Introduction of a novel differentiable modeling approach (DMFS) to infer key constitutive relationships governing heat transfer in frozen soils using observed temperature data. DMFS uses physically constrained neural networks to represent constitutive relationships and optimizes them through gradient backpropagation.
48. Model for predicting soil suction via tensiometer cavitation curves: accounting for contact interface resistance and dynamic process of water flow
Core Problem: Accurate measurement of high soil matric suction is limited by water cavitation in tensiometers, leading to cavitation curves instead of equilibrium data. Inverting soil suction from these curves is essential.
Key Innovation: Development of a modified tensiometer response model based on the Green–Ampt framework, incorporating time-dependent hydraulic processes at the ceramic cup–soil interface to represent evolving contact resistance. The model enables superior soil suction prediction through tensiometer response curves, even from incomplete cavitation curves.
49. An Estimate of Effective Resolution for CYGNSS Surface Water Detection Revealed by Comparison With Multisensor Optical Data
Core Problem: The paper addresses the problem of determining the minimal detectable size of a water body within the GNSS-R signal footprint, as the signal saturates even at relatively low quantities of surface water.
Key Innovation: The paper examines the sensitivity of the CYGNSS mission by comparing it with high-resolution surface water extent data, revealing that the lower limit of water area detectable by CYGNSS was 0.62 km² within a 0.05° grid cell, suggesting a probabilistic interpretation of CYGNSS observations for synergistic use with other remote sensing technologies. This could be useful in monitoring inundation related to landslides.
50. Critical evaluation of strong ground motions in Izmir and implications for future earthquake simulation results
Core Problem: Seismic hazard assessment in Izmir, Turkey, requires accurate ground motion prediction, but current code practices may not adequately account for local site effects.
Key Innovation: The study calibrates a 1D site response model using the 2020 Sisam earthquake data and generates possible future events to compare with the Turkish Earthquake Code (TEC), providing concrete recommendations about local site modification factors.
51. Drying of partially saturated granular materials: a combined study using X-ray tomography and lattice Boltzmann modeling
Core Problem: Understanding the hydro-mechanics of partially saturated soils is crucial due to their sensitivity to seasonal changes, requiring recognition of the correlation between soil microstructure and water retention characteristics.
Key Innovation: Combined X-ray tomography and lattice Boltzmann modeling (LBM) to study the drying of partially saturated granular materials, capturing microscopic characteristics (fluid distribution, cluster coordination) and deriving macroscopic strength indicators (cohesive strength, Bishop effective stress parameter).
52. Mechanical Response and Failure Mechanism of Rock with Holes Under Multiple Stress Gradients
Core Problem: Rock failures in underground engineering due to heterogeneous stress gradients from repeated stress adjustments.
Key Innovation: Novel experimental method simulating continuous multiple stress gradients in rock specimens, analyzed using AE, DIC, and numerical simulations to identify instability mechanisms.
53. Geotechnical Particle Finite Element Simulation of Vibratory Probe Compaction in Structured Loess
Core Problem: Insufficient understanding of improvement mechanisms in structured loess under dynamic penetration and large deformation during vibratory probe compaction.
Key Innovation: Numerical investigation using the Geotechnical Particle Finite Element Method with an elasto-plastic constitutive model to simulate the dynamic behavior of structured loess during vibratory probe compaction.
54. Post-wildfire recovery of forest vegetation, soil, and hydrological responses: A review
Core Problem: Post-wildfire recovery is often described separately for vegetation, soils, and hydrology, obscuring the coupled controls that determine recovery trajectories and duration.
Key Innovation: Synthesized post-fire evidence into a stage-based framework linking initial forest condition, fire regime, and coupled vegetation-soil-hydrology feedbacks to recovery pathways and timescales.
55. A tailored deep learning method to improve spatial rainfall downscaling
Core Problem: Hazard-relevant rainfall downscaling must recover both extreme intensity and storm structure, yet training data are often sparse and inconsistent across scales.
Key Innovation: Combined spatial correction of training fields with a lightweight RM-ResNet that jointly learns rainfall occurrence and magnitude, improving 8 km to 1 km downscaling of extremes and storm centers.
56. Leveraging water vapor to extend forecast horizons for forecast-informed reservoir operations: a vapor-precipitation-streamflow three-line defense
Core Problem: Current forecast-informed reservoir operations (FIRO) schemes often overlook precipitable water vapor (PWV), which can provide early signals of precipitation, limiting the forecast horizon and potentially reducing the efficiency of floodwater utilization.
Key Innovation: The study proposes a vapor-precipitation-streamflow (VPS) FIRO scheme that leverages PWV to extend forecast horizons by deriving potential precipitation (PP) from water vapor, enabling earlier prerelease operations and reducing the volume of spilled water compared to traditional precipitation-streamflow (PS) FIRO schemes.
57. Mineral failure behavior and erosion in sandstone by abrasive waterjet using material point method
Core Problem: Limited understanding of the damage behavior of abrasives and high-pressure water on sandstone minerals during abrasive water jet (AWJ) application in underground engineering.
Key Innovation: Using the material point method (MPM), the study elucidates the stress transmission characteristics of the AWJ and the minerals’ damage behavior, explaining the AWJ’s rock-breaking mechanism, including stress concentration, shear failure, and the water wedge effect.
58. A Novel Dual‐Clustering Approach for Identifying Hydrological Response Patterns From Catchment Characteristics and Environmental Changes
Core Problem: Few studies have systematically linked catchment characteristics, environmental changes, and hydrological responses.
Key Innovation: Proposing a novel dual-clustering approach for identifying hydrological response patterns by constructing a catchment characteristic indicator system and introducing dynamic similarity indicators.
59. Physics-Informed Diffusion Model for Generating Synthetic Extreme Rare Weather Events Data
Core Problem: Data scarcity in developing robust ML models for detecting rapidly intensifying tropical cyclones, especially rare Category 4-equivalent events.
Key Innovation: A physics-informed diffusion model based on the Context-UNet architecture to generate synthetic, multi-spectral satellite imagery of extreme weather events, conditioned on critical atmospheric parameters.
60. LEPA: Learning Geometric Equivariance in Satellite Remote Sensing Data with a Predictive Architecture
Core Problem: Geometric mismatches between user-defined areas and precomputed embeddings in geospatial foundation models lead to unreliable latent-space interpolation.
Key Innovation: A Learned Equivariance-Predicting Architecture (LEPA) that conditions a predictor on geometric augmentations to directly predict the transformed embedding, enabling accurate geometric adjustment without re-encoding. Could be useful for landslide mapping and monitoring.
61. Enhancing Unregistered Hyperspectral Image Super-Resolution via Unmixing-based Abundance Fusion Learning
Core Problem: Improving the spatial resolution of hyperspectral imagery using unregistered high-resolution reference images is challenging due to misalignment and difficulty in learning super-resolution models.
Key Innovation: An unmixing-based fusion framework is proposed that decouples spatial-spectral information and uses a coarse-to-fine deformable aggregation module and spatial-channel modulated fusion to enhance super-resolution performance.
62. Prediction of Steady-State Flow through Porous Media Using Machine Learning Models
Core Problem: Traditional CFD methods for solving flow through porous media are computationally expensive for large and complex geometries.
Key Innovation: Developing a machine-learning framework using convolutional autoencoder, U-Net, and Fourier Neural Operator models to predict steady-state flow through porous media, with FNO outperforming other models.
63. Generative Adversarial Regression (GAR): Learning Conditional Risk Scenarios
Core Problem: Learning conditional risk scenarios for risk management, ensuring robustness across various policies.
Key Innovation: A Generative Adversarial Regression (GAR) framework that trains generators to match the policy-induced risk of real data, using a minimax formulation to ensure robustness across a broad class of policies.
64. Semantic Segmentation of Land Cover in Remote Sensing Images Based on Multiscale Feature Adaptive Enhancement
Core Problem: Existing land cover semantic segmentation methods struggle to balance local details and global dependencies due to the varying scales of objects in high-resolution remote sensing images.
Key Innovation: The MFESegNet architecture, featuring adaptive grouping gated convolution, dual attention efficient transformer, global-local feature fusion, and multiscale feature attention fusion modules, along with a deformable scanning strategy mamba (DSSM) decoder, enhances multiscale feature representation and adaptive context modeling for improved land cover semantic segmentation.
65. FSINet: A Robust Feature Separation and Integration Network for Multiscale SAR Object Detection
Core Problem: SAR image object detection is impacted by complex noise, intricate backgrounds, and multiscale objects, hindering accurate and reliable detection.
Key Innovation: The Feature Separation and Integration Network (FSINet) mitigates noise interference and improves multiscale object detection accuracy through a feature separation module, an adaptive feature integration module, and a shape normalized Wasserstein distance loss.
66. Identification of hydro-meteorological drivers for forest low greenness events in Europe
Core Problem: Understanding the effects of heatwaves and droughts on forests is crucial, especially with climate change projecting more frequent extreme hydro-meteorological events.
Key Innovation: The study presents a novel, large-scale analysis of forest browning drivers across Europe, using a random forest modeling framework to identify the most relevant hydro-meteorological predictors of low NDVI events.
67. Reconstructing nineteenth-century Danube river water levels with transformer-based computer vision
Core Problem: There is a need to rescue analogue hydrometric records to enable long-term analyses of extremes and regulation impacts.
Key Innovation: The study presents a semi-automated workflow combining document pre-processing, dewarping, transformer-based line extraction, and pixel-to-curve calibration to convert nineteenth-century Danube gauge charts into daily water-level series.
68. Multidecadal reconstruction of terrestrial water storage changes by combining pre-GRACE satellite observations and climate data
Core Problem: Climate model evaluation and climate change attribution require multi-decadal terrestrial water storage anomaly (TWSA) time series, which are limited by the relatively short duration of GRACE and GRACE-FO missions.
Key Innovation: The study combines regression approaches with large-scale time-variable gravity information from geodetic satellite laser ranging (SLR) and Doppler Orbitography by Radiopositioning Integrated on Satellite (DORIS) tracking to reconstruct GRACE-like TWSA from 1984 onward.
69. Multi-system analysis of offshore geologic carbon storage: a review of open-source data science solutions
Core Problem: Offshore geologic carbon storage is expanding faster than the analytical workflows needed to screen seabed instability, infrastructure exposure, and multi-system operational risk.
Key Innovation: Reviewed open-source data-science tools that couple geological, oceanographic, engineered, and regulatory dimensions, framing offshore storage assessment as an integrated multi-system screening problem.
70. Experimental and Numerical Simulation Investigation of Diffusion and Sealing Mechanism of Polymer Slurry in Rough Fracture Network Under Flowing Water Grouting
Core Problem: Complicated diffusion behavior of expansive polymer slurry within fracture networks in grouting repair engineering.
Key Innovation: Numerical model and physical experiments simulating polymer slurry diffusion in rough fracture networks, analyzing sealing rate and diffusion pressure, and revealing the impact of fracture roughness and intersection angles on sealing efficiency.
71. Deciphering solute transport in crack networks-clay matrix systems: Insights from experiments and numerical modeling
Core Problem: Robust numerical methods for simulating solute seepage in fractured soils at the field scale are lacking, hindering accurate prediction of contamination and risk assessment in cracked porous media.
Key Innovation: The study establishes a validated numerical framework integrating Voronoi-based DFN modeling with Richards' equation and the Advection-Dispersion Equation (ADE) to simulate solute transport in cracked soil systems, clarifying the impact of crack features on preferential flow and diffusion mechanisms.
72. Microplate dynamics through the Wilson cycle: Insights from modelling and observations
Core Problem: The dynamic behavior and feedback mechanisms of microplates across different stages of the Wilson Cycle remain poorly constrained.
Key Innovation: The study synthesizes insights from numerical modeling and geological observations, focusing on the roles of microplates in divergent, convergent, and intraplate tectonic settings, highlighting rheological strength as the primary control on their dynamic behavior.
73. Influences of map resolution, quality and interpretation on fault network topology
Core Problem: Fault-network topology is widely used to infer seismic hazard and fluid connectivity, but its sensitivity to map resolution, data quality, and interpretive choices is often underestimated.
Key Innovation: Compared fault maps across scales to show how node statistics and apparent connectivity shift with resolution and mapping practice, establishing topology itself as a diagnostic of map reliability.
74. Characteristics and influencing factors of soil moisture memory across mainland China
Core Problem: Large-scale estimates of soil moisture memory (SMM) based on in situ observations remain scarce, and recently proposed SMM metrics have not yet been comprehensively evaluated using extensive ground measurements.
Key Innovation: Comprehensive assessment of SMM over mainland China based on a large amount of in situ soil moisture observations, revealing its spatial and depth-dependent characteristics and key controlling factors.
75. Experimental study of suspended sediment transport at river confluence
Core Problem: River confluences are critical nodes in fluvial networks, significantly influencing flow dynamics, sediment transport, and overall river morphology, but the suspended sediment transport at river confluence and its dependence on complex hydro-morphology is rarely studied.
Key Innovation: Investigation of the flow and sediment transport dynamics within a 90° river confluence model under varying discharge ratios, with detailed measurements of 3D velocity, turbulence, and suspended sediment concentration in a laboratory flume.
76. Quantifying the attribution of ecohydrological degradation: a comparative deep learning approach in a changing environment
Core Problem: Disentangling the impacts of anthropogenic pressures versus climatic variability on ecohydrological degradation in intensively managed basins is critical for sustainable water management, but traditional methods often fail to capture the dynamic, multi-decadal evolution of these impacts.
Key Innovation: The study employs a Gated Recurrent Unit (GRU) deep learning framework to reconstruct the pristine hydrological regime and generate a counterfactual scenario, enabling a dynamic attribution analysis that reveals distinct phases of human impact, including flood peak amplification due to mining and flood suppression due to dam regulation.
77. Spectral Gaps and Spatial Priors: Studying Hyperspectral Downstream Adaptation Using TerraMind
Core Problem: Geospatial Foundation Models lack native support for Hyperspectral Imaging (HSI), hindering their adaptability to HSI downstream tasks.
Key Innovation: This study investigates the adaptability of TerraMind to HSI downstream tasks without HSI-specific pretraining, comparing Naive Band Selection and Spectral Response Function grouping strategies.
78. Land Surface Temperature Trends Over Central and Southern Europe: Derivation and Analyses of Long-Term (1986–2018) Monthly Maxima
Core Problem: The paper aims to monitor long-term land surface temperature (LST) time series and analyze their anomalies and trends to understand spatial patterns of global warming in Europe.
Key Innovation: The paper derives and analyzes monthly maximum LST trends over central and southern Europe at 1 km² resolution from 1986–2018, finding significant positive LST trends, particularly in eastern Europe, and analyzing these trends with respect to different land cover classes and elevation ranges. This could be used to understand the impact of climate change on landslide occurrence.
79. Text-Guided Contrastive Mamba for Hyperspectral Image Classification
Core Problem: The paper addresses the challenges in hyperspectral image (HSI) classification due to high spectral complexity and the need to model long-range dependencies.
Key Innovation: The paper proposes text-guided contrastive mamba (Mamba-TGC), a novel text-guided contrastive learning framework for HSI classification that integrates a Mamba encoder to model spectral and spatial patch embeddings, aligning these features with category-level textual descriptions through contrastive loss, demonstrating improved performance in hyperspectral representation learning. This could be used to improve the accuracy of landslide detection.
80. A Season-Adaptive Machine Learning Framework for Estimating Ground Surface Temperature in Northeast China
Core Problem: The paper addresses the challenge of acquiring GST data with high spatio-temporal continuity and accuracy over large scales due to limitations of in-situ measurements and remote sensing technologies.
Key Innovation: The paper develops robust machine learning models to simulate GST dynamics during growing and snowing seasons in Northeast China, revealing significant seasonal variations in model performance and identifying air temperature and snow cover as primary factors driving variations in GST. This could be used to understand the impact of climate change on landslide occurrence.
81. Surface Water Extraction by Multidimensional Feature Fusion of Sentinel-1/2 and Random Forest: A Five-Year Analysis in Jiangsu, China
Core Problem: Surface water extraction in regions with complex terrain suffers from low accuracy and misclassification.
Key Innovation: A high-precision water extraction framework based on the Random Forest classifier integrates multipolarization scattering features from Sentinel-1, multispectral water indices from Sentinel-2, and slope constraints from DEM, improving surface water boundary delineation.
82. Improving thermodynamic nudging in the E3SM Atmosphere Model version 2 (EAMv2): strategy and hindcast skills on weather systems
Core Problem: Improper application of nudging techniques in atmospheric simulations can distort physical processes and introduce spurious biases.
Key Innovation: An improved nudging implementation that applies vertically modulated tendencies to thermodynamic variables enhances hindcast skill in EAMv2, improving agreement with ERA5 without degrading the hydrological cycle.
83. Seismic image constraints on the geodynamic evolution of the Hegenshan Suture (Eastern CAOB)
Core Problem: The absence of high-resolution, lithospheric-scale constraints on accretionary architectures hinders a comprehensive understanding of tectonic evolution in the Central Asian Orogenic Belt (CAOB).
Key Innovation: Analysis of two high-resolution deep seismic reflection profiles revealing fine lithospheric accretionary structures, depicting multistage northward oceanic subduction relics and tectonic crocodile structures.
84. From small to mega river basins−adaptation of a lumped hydrological model for runoff simulation
Core Problem: Accurate runoff simulation across diverse river basin scales is crucial for water resource management, but existing hydrological models often struggle with scale adaptability, particularly in representing baseflow generation in perennial rivers and humid conditions.
Key Innovation: The study updates the CAWM IV model by incorporating a groundwater reservoir and soil-groundwater exchange threshold, enabling robust runoff simulation across a wide range of basin scales, including mega basins, with performance comparable to or better than established models like GR4J and HEC-HMS.