TerraMosaic Daily Digest: Feb 28, 2026
Daily Summary
This February 28, 2026 digest distills 32 papers from 479 analyzed records (479 deduplicated; 2,139 raw). The highest-confidence contributions are tightly hazard-facing: stochastic 3D slope-failure analysis under spatially variable soils, Transformer-based rainfall time-series integration for multi-type landslide probability, SWOT-enabled monitoring of ice-marginal lakes for GLOF detection, and revised seismic-moment theory clarifying how injection volume governs induced-earthquake occurrence rates.
A second cluster strengthens the mechanics-to-decision pipeline: embankment and foundation behavior under cyclic loading, fault-zone fluid-structure coupling, and infrastructure deformation diagnostics. At the same time, this cycle includes many transferable but non-core AI/remote-sensing studies; these were retained only when the mechanism, uncertainty pathway, and hazard relevance were explicit.
Key Trends
The dominant signal is a shift from static mapping toward process-aware, uncertainty-explicit hazard intelligence.
- Dynamic landslide forecasting is maturing: studies now ingest full rainfall sequences and 3D geotechnical variability instead of relying on static proxies, improving event-specific probability interpretation.
- Satellite hazard observability is expanding: SWOT and Sentinel-1 workflows demonstrate operational value for cryosphere and disturbance surveillance, especially where in situ coverage is sparse.
- Mechanistic geotechnics remains central to deployment: new evidence on soil strength inference, embankment cyclic degradation, and offshore foundation response provides parameters that can be used directly in engineering checks.
- Physics-constrained learning is preferred over black-box accuracy: the strongest AI papers embed structural constraints or interpretable drivers, making transfer to hazard operations more credible.
- Relevance screening is becoming stricter: method papers outside geohazards were down-ranked unless they offered a clear and testable pathway to landslide or multi-hazard practice.
Selected Papers
This digest features 32 selected papers from 479 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. Three-dimensional uncertainty analysis of large deformation slope failure considering varied geometry
Core Problem: Accurately assessing large-deformation slope failure, including sliding volume, runout, and influence distance, while accounting for the coupled effects of soil spatial variability and complex 3D slope geometries.
Key Innovation: A 3D random smoothed particle hydrodynamics (SPH) method, enhanced with Karhunen–Loève (KL) expansion and GPU parallelization, is developed to perform uncertainty analysis of large-deformation slope failure, revealing the statistical characteristics under varied geometries and soil spatial variability.
2. Transformer, more than meets the eye: A deep learning approach to integrate rainfall time-series in multi-type landslide probability modelling
Core Problem: Conventional data-driven methods for landslide susceptibility assessments typically use scalar rainfall representations, limiting their ability to create dynamic, event-specific probability models that account for the continuous nature of rainfall time series and different landslide failure mechanisms.
Key Innovation: Proposed a deep learning approach using a Transformer Neural Network (TNN) coupled with a Dense Neural Network (DNN) to integrate continuous rainfall time series for multi-type landslide probability modeling, achieving strong performance (AUC > 0.90) and providing interpretability via SHAP-based Expected Gradients, setting a foundation for generalized spatiotemporal landslide forecasting.
3. Assessment of SWOT for Monitoring Ice‐Marginal Lake Water Levels in Greenland
Core Problem: The performance of the SWOT satellite for monitoring ice-marginal lakes and detecting Glacial Lake Outburst Floods (GLOFs) was unknown, and previous altimetry missions provided only one-dimensional observations.
Key Innovation: The SWOT mission provides two-dimensional water level observations, showing good agreement with ICESat-2 (0.03m average deviation, 1.43m RMSD). Its higher temporal resolution (less than 21 days) enabled a 100% increase in GLOF detection compared to ICESat-2, making it an excellent data source for monitoring these lakes.
4. Theoretical Maximum and Cumulative Seismic Moment Relationships Confirm that Injection Volume Controls the Occurrence Rate, But Not the Magnitude, of Induced Earthquakes
Core Problem: Classical relationships between cumulative and maximum seismic moment show non-physical anomalies for low b-values, and the precise control of injected volume on the magnitude versus occurrence rate of induced earthquakes was not fully understood.
Key Innovation: New theoretical relationships for seismic moment were derived, showing that injected volume primarily controls seismic nucleation and therefore the seismic rate of occurrence, while magnitude dependence is more tied to the b-value. This has significant implications for earthquake magnitude forecasting for induced seismicity and hazard mitigation.
5. Splay Fault Permeability Governs Fluid–Structure Interaction in Accretionary Wedges
Core Problem: The interactions between upper plate deformation and plate interface seismicity in subduction zones are poorly understood, particularly how fluid flow along splay faults modulates upper-plate faulting and influences earthquake hazards.
Key Innovation: Combining field observations from exhumed splay faults with finite element poroelastic models, the study defines two end-member behaviors (impermeable vs. permeable splay faults) that govern fluid flow, rock deformation, and seismicity, providing a framework for understanding prism-scale effects of splay fault permeability on shallow subduction zone deformation and earthquake hazards.
6. Influence of the Loading Stiffness on Sheared Granular Fault Gouge, and Applicability to Slip‐Weakening Theory
Core Problem: The role of loading stiffness on seismic cycles in granular fault simulations and its implications for slip-weakening theory and the idea of an intrinsic friction law for granular gouge layers are not fully understood.
Key Innovation: A numerical model based on the Discrete Element Method demonstrates that loading stiffness strongly influences seismic cycle properties, and the weakening rate in simple linear slip-weakening friction laws is tightly coupled with the loading stiffness, challenging the concept of an intrinsic friction law for granular gouge.
7. Interpretation of strength parameters of clayey soils from shallow to deep penetration using free-fall ball penetrometer with a booster
Core Problem: Accurately interpreting strength parameters of clayey soils (undrained shear strength, rate-dependency, strain-softening) across varying penetration depths using free-fall penetrometers, especially considering complex ball-soil-water interactions and drag resistance.
Key Innovation: An analytical and back-analysis framework, validated by numerical simulations and field tests, is developed to interpret clayey soil strength parameters from free-fall ball penetrometer measurements, accounting for drag resistance and establishing empirical formulas for key factors.
8. Cracking characteristics of rock-anchored beam system during construction of underground powerhouse under high in-situ stress and complex rock mass structures: A case study of Yebatan hydropower station
Core Problem: Research on the cracking behavior of rock-anchored beams in underground powerhouses is limited, particularly under high in-situ stress and complex geological settings, impacting safety and stability.
Key Innovation: Integrated field investigation, monitoring, and numerical simulation to identify local slip along steeply dipping faults under high in-situ stress as the primary cause of extensive cracking in a rock-anchored beam system, leading to the implementation of additional reinforcement.
9. Optimizing compaction of low-grade embankment soils with non-plastic fines under cyclic traffic loading and seasonal moisture variations
Core Problem: The long-term performance of compacted low-grade sandy soils with high fines content, used in road and railway embankments, is insufficiently understood when subjected to combined vehicle-induced cyclic loading and seasonal saturation fluctuations, posing challenges for cost-effective and sustainable construction.
Key Innovation: Investigated the cyclic and post-cyclic behavior of compacted sandy soils, demonstrating how compaction, saturation levels, and fines content influence axial strain, strength, and stiffness, providing practical guidance for optimizing compaction and moisture control to ensure long-term embankment stability.
10. Mechanical response and cumulative displacement prediction of suction bucket foundation in soft clay under vertical cyclic loading
Core Problem: The mechanical response and cumulative displacement of suction bucket foundations in soft clay under extreme environmental vertical cyclic loading, which poses significant risks to stability, are not fully understood or easily predictable across various loading and soil conditions.
Key Innovation: Used numerical methods validated by centrifuge tests to investigate the mechanical response of suction bucket foundations under vertical cyclic loading, studying effects of loading parameters, soil over-consolidation, and bucket length, and developed a simplified prediction method for cumulative displacement based on numerical simulation and superposition theory.
11. Iceberg detection from global open ocean sentinel-1 wave mode SAR data with YOLO deep learning
Core Problem: Underutilization of Sentinel-1 wave mode SAR data for automated and statistically oriented detection and study of km-sized icebergs in the open ocean.
Key Innovation: Develops IByolo, a YOLOv11-adapted deep learning model, for efficient detection and localization of km-sized icebergs from Sentinel-1 WV SAR data, enabling large-scale monitoring and surface area estimations. Validates its performance and applies it to generate a dataset of 80,245 iceberg detections.
12. Quantifying Gas and Thermal Energy Emissions in an Active Geothermal Area: Insights From Le Biancane (Larderello Field, Italy)
Core Problem: A clear understanding of the relative roles of various heat flux components and the mechanisms of mass and energy transfer from depth to the surface in vapor-dominated geothermal fields is needed.
Key Innovation: An integrated study at Le Biancane using soil diffuse CO2 and temperature maps, fumarole analysis, and laboratory measurements demonstrates that heat associated with vapor ascent and condensation is predominantly transferred by conduction in the uppermost soil, offering a clearer understanding of heat exchange dynamics and proposing a simple model for emission estimation.
13. Dual-amplitude weighted transition dispersion entropy and its application in feature extraction of ship-radiated noise
Core Problem: Dispersion entropy methods for time-series analysis do not jointly encode absolute amplitude and adjacent-amplitude transitions, which limits robustness for complex signal characterization.
Key Innovation: Introduced dual-amplitude weighted transition dispersion entropy (DWTDE/MDWTDE), coupling transition-structure weighting with dual-amplitude coding to improve feature discrimination in noisy ship-radiated signals.
14. A Spatiotemporal Causal Model for Revealing Developmental Changes in Infants' Brain Effective Connectivity Networks During the First Year of Life
Core Problem: Infant brain studies during the first year still rely mainly on functional connectivity, while extracting stable effective-causality structure from spatiotemporal data remains difficult.
Key Innovation: Proposed a spatiotemporal Granger-causality framework with sparsity constraints to recover effective connectivity trajectories, improving causal interpretability in infant neurodevelopment data.
15. A Multichannel CNN for Global Empirical Ionospheric Modeling
Core Problem: Classical empirical ionospheric models are constrained by fixed functional forms and underperform during disturbed space-weather periods, reducing GNSS correction accuracy.
Key Innovation: Developed a multichannel CNN empirical ionosphere model that jointly assimilates solar-geometric and geomagnetic drivers, substantially improving global TEC prediction stability and precision.
16. Mapping water content dynamics in SAT systems using 3D electrical tomography
Core Problem: Three-dimensional infiltration and retention pathways in soil-aquifer treatment systems are difficult to resolve in situ, hindering optimization of recharge strategies and reactive media design.
Key Innovation: Applied cross-hole 3D electrical tomography to quantify transient water-content fields under pulsed versus continuous recharge, revealing preferential flow, retention enhancement, and biofilm-growth controls.
17. Critical station identification framework combining layered computation with cross-layer fusion for urban public transport networks
Core Problem: Critical-node identification in multimodal public transport systems remains biased by single-layer assumptions and poor treatment of intra-day structural variability.
Key Innovation: Built a multilayer, multi-period network framework with Transport Gravity Centrality (TGC) to fuse within-mode function and cross-mode influence for more stable critical-station discovery.
18. Hydrological response to climate changes and ecological restoration in karst region of Southwest China
Core Problem: Runoff response in karst basins under concurrent climate forcing and ecological restoration remains difficult to quantify because lithology and terrain controls are often underrepresented.
Key Innovation: Integrated karst-specific controls (including slope and soil texture) into a Budyko-based attribution framework, deriving a transferable relationship between karst extent and runoff elasticity.
19. Limited Passive Lithospheric Underthrusting and Localized Crustal Thickening Beneath the Qilian Shan, Northeastern Tibetan Plateau: Evidence From Receiver Function Imaging
Core Problem: The mode of lithospheric deformation beneath the Qilian Shan and the mechanisms of Tibetan Plateau growth, particularly the role of passive underthrusting, remain subjects of debate.
Key Innovation: Using common conversion point stacking with P and S receiver functions from a seismic array, the study provides images of crustal and upper mantle discontinuities, revealing significant crustal thickening and a continuous southward-dipping lithosphere-asthenosphere boundary, supporting a model of passive underthrusting of the Asian Plate influenced by the Alxa block.
20. Indo–Western Pacific Tropical Heating Anomalies Regulate the Cross‐Pacific Atmospheric River Highways During Boreal Summer
Core Problem: Cross-Pacific atmospheric-river pathways in boreal summer are poorly constrained by operational predictors, limiting seasonal flood-risk anticipation.
Key Innovation: Identified coordinated Indo–Western Pacific heating anomalies as physically grounded predictors of AR pathway shifts and landfall tendencies, extending useful lead information across the Pacific rim.
21. Rethinking the Filling Process of Mega‐Dam Cascades
Core Problem: Reservoir filling in mega-dam cascades is rarely analyzed as a system-level risk phase, despite major downstream hydrological consequences.
Key Innovation: Reconstructed multi-reservoir filling trajectories and quantified electricity–hydrology trade-offs, showing how coordinated filling can reduce downstream alteration without collapsing generation benefits.
22. An integrated SAR denoising–enhancement–detection pipeline for robust oriented ship recognition
Core Problem: Maritime SAR ship detection pipelines are vulnerable to compounded noise, blur, and resolution loss when denoising, enhancement, and detection are optimized independently.
Key Innovation: Proposed an end-to-end SAR analytics chain (SCUNet + DPSR + oriented YOLO) that jointly improves image quality and oriented detection robustness in operationally degraded scenes.
23. A physics-informed Temporal–Spatial gated Kolmogorov–Arnold network for real-time response prediction of floating structures
Core Problem: High-fidelity simulations of floating-structure dynamics are too slow for real-time risk monitoring during rapidly evolving metocean loading.
Key Innovation: Developed a physics-informed deep architecture (P-DeepGKAN) constrained by governing-response relations, enabling fast and accurate real-time prediction of structural responses under coupled wave–wind forcing.
24. A path planning method for USV in maritime SAR missions based on improved deep Q-network
Core Problem: USV route planning for maritime search-and-rescue is degraded by drifting targets and coupled wind-current uncertainty.
Key Innovation: Introduced a multimodal deep Q-network with dynamic target-probability evolution, improving adaptive routing efficiency under uncertain marine drift conditions.
25. Constructing a SWOT Internal Wave Dataset Using Deep Learning
Core Problem: The need for an automatic, multi-region framework and a high-quality dataset for detecting internal waves using high-resolution satellite observations (SWOT) to understand their role in energy transfer and mixing.
Key Innovation: Development of SWOT_IWD, a deep learning-based framework for automatic internal wave recognition, and construction of a comprehensive SWOT internal wave detection dataset covering 13 global regions, achieving 91.21% accuracy and demonstrating potential for multi-source internal wave tracking.
26. A benchmark dataset of water levels and waves for SWOT validation in the St. Lawrence Estuary and Saguenay Fjord, Quebec, Canada
Core Problem: The need for a comprehensive, high-quality in situ dataset of water levels and wave parameters to validate the SWOT satellite mission's products and support hydrodynamic modeling in complex estuarine environments like the St. Lawrence Estuary and Saguenay Fjord.
Key Innovation: Creation of an evolving, publicly accessible benchmark dataset integrating water level measurements from 18 pressure transducers and 13 GNSS-IR sensors, along with three directional wave buoys, providing unprecedented spatial coverage and rigorous quality control for SWOT validation and estuarine research.
27. Counting trees together in the EU
Core Problem: Cross-border forest monitoring in Europe is fragmented, limiting consistent detection of disturbance pressures and coordinated management response.
Key Innovation: Highlights an EU-level harmonization pathway for interoperable forest monitoring, emphasizing shared protocols and scalable continental coverage.
28. Noise-resilient exceptional point sensing with immunity to undesired perturbations
Core Problem: Exceptional-point sensors offer high nominal sensitivity but are easily destabilized by parasitic perturbations and cavity imperfections.
Key Innovation: Demonstrated non-resonant exceptional-point sensing in periodic microwave metamaterials with substantially improved immunity to undesired perturbations while retaining ultrasensitive response.
29. Reservoir prediction methods under sparse well conditions in offshore fields: perspectives and challenges
Core Problem: Offshore reservoir prediction under sparse-well constraints suffers from multi-scale data gaps and uncertain geological priors.
Key Innovation: Synthesized multimodal workflows that combine seismic attributes, stochastic geological modeling, and AI-based inversion to improve prediction reliability under sparse supervision.
30. General control of uplift gradients and rock erodibility in formation and evolution of main drainage divides
Core Problem: The initial formation and subsequent evolution of main drainage divides, particularly how uplift gradients and rock erodibility interact to control these processes, remain subjects of ongoing debate.
Key Innovation: Landscape evolution modeling is employed to simulate the formation of main drainage divides on uplift ramps, demonstrating how uplift gradients and rock erodibility systematically influence the propagation of divides and topographic growth over time.
31. European forest disturbance alerting using Sentinel-1
Core Problem: Continental near-real-time forest disturbance detection requires cloud-independent sensing while controlling false alarms from phenology and freeze-thaw effects.
Key Innovation: Built an operational Sentinel-1 alerting pipeline that fuses radar signal change with temperature and forest-type constraints for robust pan-European disturbance detection.
32. Decreasing soil water repellency during infiltration: Linking time-dependent contact angle to a physically meaningful scaling parameter
Core Problem: Traditional infiltration models fail to capture convex-shaped infiltration curves in water-repellent soils, and the physical meaning of newly introduced scaling parameters (αWR) for decreasing repellency is unknown.
Key Innovation: Demonstrated that the parameter αWR in modified Green-Ampt models for water-repellent soils can be physically interpreted as the rate of change in soil water suction head corresponding to the time-dependent contact angle, and that this contact angle can serve as a direct model input to improve simulation accuracy.