TerraMosaic Daily Digest: Mar 11, 2026
Daily Summary
Research published on March 11, 2026 converges on a more explicit treatment of instability as a measurable process. The strongest papers do not stop at hazard labels. They quantify how colluvial landslides reorganize through full-field deformation, how fault permeability and rupture behavior emerge from grain crushing, pore collapse, and fluid migration, and how earthquake-induced landslide hazard changes once multiple ground-motion measures are admitted into probabilistic analysis. Even in engineered settings, the same shift is visible: slope and wall monitoring in a quarry-derived urban park, retaining-wall deformation forecasting, dam-break discharge prediction, and mining-induced fault activation all move toward mechanistic indicators rather than black-box extrapolation.
Hydrological and coastal work is equally notable for turning sparse observations into operational state estimates. SAR reclassification of Arctic ice now accounts for under-ice salinity; GRACE storage anomalies are downscaled for flood monitoring; drought diagnostics are tied to vegetation and terrestrial water storage rather than rainfall alone; and karst recharge pathways are resolved during rainstorms with radon tracing. Coastal studies sharpen the treatment of tsunami debris, low-crested defense stability, saltmarsh sediment exchange, and delta morphodynamics. The most useful computational contributions are those that directly improve hazard intelligence under imperfect data, including remote-sensing foundation-model selection, cloud-contaminated image reconstruction, SAR oil-spill segmentation, and dynamics-informed extreme-event prediction.
Key Trends
Today's literature is defined by process-aware instability diagnostics, sparse-observation hydrology, and Earth observation methods that are becoming operational rather than merely generic.
- Full-field and multiscale state variables are replacing coarse hazard descriptors: deformation entropy, microseismic voxelization, permeability evolution, and dense-array tomography are being used to identify how instability actually develops.
- Probabilistic hazard frameworks are becoming both broader and more interpretable: earthquake-induced landslide analysis, dam-break discharge forecasting, tsunami debris transport, and retaining-wall deformation prediction now expose their controlling variables more clearly.
- Hydrological monitoring is shifting toward indirect but mechanism-aware state reconstruction: GRACE downscaling, SAR ice-regime correction, radon tracing, and chloride-source apportionment all infer flood or drought state from sparse, heterogeneous observations.
- Compound coastal and cryosphere hazards are being treated as coupled physical systems: tsunami loading, debris transport, saltmarsh sediment budgets, delta evolution, and permafrost degradation are analyzed through interacting hydraulic, thermal, and geomorphic processes.
- Earth observation AI is most valuable where it removes workflow bottlenecks: model selection, cloud removal, oil-spill detection, and dynamics-informed forecasting matter because they improve monitoring reliability under missing or noisy data.
Selected Papers
This digest features 35 selected papers from 975 papers analyzed across slope instability, earthquake processes, dam and tsunami hazards, drought and flood monitoring, cryosphere change, and hazard-ready Earth observation methods.
1. Monitoring of slopes, rock faces and masonry walls in a 19th century public park: the example of the Buttes Chaumont Park (Paris, France)
Core Problem: Long-lived landslide, rockfall, and sinkhole hazards in a quarry-derived urban park require consistent multiscale surveillance and interpretation.
Key Innovation: A four-tier geotechnical monitoring scheme combines field inspection, tacheometry, manual gauges, and automated extensometers to refine hazard mapping and resolve displacement-meteorology links.
2. Instability evolutionary characteristic of colluvial landslide based on the 3D deformation field and curvature Shannon entropy in physical model test
Core Problem: Instability transitions in colluvial landslides remain difficult to diagnose quantitatively from evolving full-field deformation.
Key Innovation: Curvature-Shannon-entropy analysis of 3D deformation fields in physical models separates instability stages and links curvature evolution to velocity and internal stress redistribution.
3. Probabilistic Hazard Analysis of Earthquake-Induced Landslides considering Multiple Ground Motion Intensity measures
Core Problem: Performance-based landslide hazard analysis usually relies on too few ground-motion intensity measures, leaving displacement forecasts unnecessarily uncertain.
Key Innovation: The proposed framework couples vector probabilistic seismic hazard analysis with landslide displacement modelling so multiple intensity measures can be used in probabilistic hazard curves.
4. From black box to physical insight: An explainable machine learning framework for dam break forecasting validated by numerical and physical tests
Core Problem: Accurate dam-break discharge forecasting often comes at the cost of physical interpretability, obscuring breach-evolution controls.
Key Innovation: A physics-guided explainable learning framework identifies the dominant breach controls and validates the learned discharge relations against both numerical simulations and large-scale physical tests.
5. A probabilistic model of tsunami debris transport using stochastic reflection angle in collision dynamics
Core Problem: Stochastic collision behavior makes tsunami-driven debris trajectories and dispersion difficult to predict reliably.
Key Innovation: The model injects probabilistic reflection-angle perturbations into collision dynamics and reproduces laboratory debris dispersion and travel timing across density regimes.
6. 2D Experimental Study on Tsunami-like Flow Hydrodynamics and their effect on the Hydraulic Stability of Cubipod Homogeneous Low-Crested Structures
Core Problem: Hydraulic stability of low-crested Cubipod structures under tsunami-like bores remains insufficiently constrained for design.
Key Innovation: Flume experiments calibrate tsunami-like flow generation and derive semiempirical relations for bore properties and Cubipod hydraulic stability under tsunami attack.
7. Thermal stability analysis of permafrost subgrade based on deep learning and numerical simulation
Core Problem: Permafrost degradation threatens transport corridors, but thermal-field prediction under complex boundary conditions remains computationally difficult.
Key Innovation: A Bayesian-optimized temporal transformer trained against numerical simulations improves prediction of permafrost subgrade thermal evolution on the Qinghai-Tibet Plateau.
8. Seismic behavior of adjacent underground structures in liquefiable ground
Core Problem: Adjacent underground structures interact strongly in liquefiable soils, yet arrangement-dependent seismic response is not well resolved.
Key Innovation: Validated dynamic simulations quantify how different structural layouts alter uplift, horizontal deformation, and bending demand in liquefiable ground.
9. An integrated spatiotemporal framework for identification and early warning of mining-induced fault activation
Core Problem: Hazardous time windows and locations of mining-triggered fault activation are difficult to identify before dynamic failure occurs.
Key Innovation: An octree-based multiscale microseismic framework fuses clustering, b-value, energy, and activity indicators into a spatiotemporal warning index for fault activation.
10. Spatio-Temporal Forecasting of Retaining Wall Deformation: Mitigating Error Accumulation via Multi-Resolution ConvLSTM Stacking Ensemble
Core Problem: Long-horizon retaining-wall deformation forecasts in staged excavation suffer from rapid error accumulation.
Key Innovation: A multi-resolution ConvLSTM stacking ensemble exploits multiple temporal scales to stabilize long-horizon deformation forecasts for retaining structures.
11. Improving SAR‐Based Classification of Arctic Lake, Bay and Lagoon Ice by Accounting for Under Ice Water Salinity
Core Problem: Bedfast versus floating ice classification in Arctic water bodies is systematically biased when under-ice salinity is ignored.
Key Innovation: Field salinity data, seasonal Sentinel-1 observations, and topographic context are integrated to correct ice-regime mapping in Arctic lakes, bays, and lagoons.
12. A semi-supervised LSTM framework for spatiotemporal downscaling of GRACE-derived terrestrial water storage anomalies to improve flood monitoring in the Yarlung Tsangpo River basin
Core Problem: GRACE-derived terrestrial water storage is too coarse and intermittent for practical basin-scale flood surveillance.
Key Innovation: A semi-supervised LSTM downscaling framework fuses multiple GRACE products and reconstructs daily high-resolution storage anomalies for improved flood monitoring.
13. Response of hydrological drought to vegetation change in China
Core Problem: The role of large-scale vegetation greening in shaping hydrological drought evolution across China remains insufficiently quantified.
Key Innovation: Scenario experiments with observed versus detrended leaf-area forcing show that vegetation change materially accelerates the national drying trend.
14. Combined hydro-meteorological drought assessment of Ganga-Brahmaputra Basin: insights of the control of total water storage anomaly in drought occurrence
Core Problem: Precipitation-only drought analyses miss the role of terrestrial water storage depletion in the Ganga-Brahmaputra basin.
Key Innovation: A combined drought index merges precipitation and GRACE storage anomalies to partition climatic and storage controls on basin-scale drought occurrence.
15. Estimating Net Sediment Fluxes in Tidal Systems Using Sporadic Data Sets: Implications for Using Remote Sensing to Assess Saltmarsh Resilience
Core Problem: Asynchronous observations of suspended sediment and water flux limit scalable sediment-budget estimation for saltmarsh resilience assessment.
Key Innovation: Phase folding and Monte Carlo analysis show that sporadic observations spanning enough tidal cycles can still distinguish marsh sediment sources from sinks.
16. Control of channel geometry on centennial morphological evolution of deltas: A numerical simulation perspective
Core Problem: The centennial influence of channel geometry on delta evolution is hard to isolate from the broader fluvial-sedimentary system.
Key Innovation: Numerical simulations quantify how channel geometry steers long-term delta morphological trajectories and emergent planform behavior.
17. Morphodynamic differences of vegetated and unvegetated mid-bars in the Jingjiang reach in the middle Yangtze River over 1960–2018
Core Problem: Long-term morphodynamic differences between vegetated and unvegetated mid-channel bars remain poorly resolved in large rivers.
Key Innovation: Multi-decadal analysis of the Jingjiang reach shows how vegetation alters mid-bar stability, geometry, and evolutionary pathways.
18. Latent Data Assimilation for Efficient and Accurate Groundwater Modeling
Core Problem: High-dimensional groundwater data assimilation remains computationally expensive and sensitive to uncertain error structure.
Key Innovation: A latent-space data-assimilation workflow built on dimensionality reduction preserves forecast accuracy while enabling efficient updates from sparse observations.
19. Two New Analytical Models for Three‐Dimensional Transport in a Confined Aquifer With a Permeable Reactive Barrier: A New Adsorptive‐Reactive Robin Matching Condition
Core Problem: Analytical permeable-reactive-barrier transport models inadequately represent adsorption and reaction across the barrier interface.
Key Innovation: New three-dimensional analytical solutions and an adsorptive-reactive Robin matching condition improve contaminant-transport representation in aquifer-PRB systems.
20. Radon tracing to identify water sources in karst springs during rainstorm events
Core Problem: Rapid storm-event source partitioning in karst systems remains difficult under natural conditions.
Key Innovation: High-frequency radon tracing and end-member mixing resolve how different recharge sources contribute to karst spring discharge during rainstorms.
21. Elucidating seasonal pathways and sources of chloride to freshwater streams across urban and rural catchments
Core Problem: Seasonal chloride pathways and source attribution remain weakly resolved across contrasting urban and rural catchments.
Key Innovation: Hydrograph separation, mixing analysis, and loading estimates show pervasive road-salt dominance but distinct seasonal export timing between urban and rural settings.
22. Microstructural control of permeability, electrical resistivity, and seismic velocity and their relationships in sheared rock fractures under normal stress
Core Problem: Geophysical proxies for fracture permeability remain difficult to interpret mechanistically in sheared rock systems.
Key Innovation: Coupled simulations show how aperture, contact area, roughness, and shear displacement jointly govern permeability, resistivity, and seismic velocity.
23. A Semi‐Analytical Approach to Model Borehole Tube Waves for Interpretation of Downhole Seismic Data in Fault Zones
Core Problem: Full poroelastic simulation of borehole tube-wave generation is too expensive for routine interpretation of fault-zone downhole seismic data.
Key Innovation: A semi-analytical propagator-matrix formulation captures impedance contrasts, infiltration, and borehole irregularities while remaining consistent with poroelastic theory.
24. The Influence of Grain Crushing and Pore Collapse on the Formation of Faults
Core Problem: The coupled roles of grain crushing and pore collapse in localized fault-core formation remain poorly understood.
Key Innovation: A breakage-mechanics model with Cosserat enrichment links shear-band compaction, porosity evolution, and permeability reduction during fault development.
25. Seismicity and Detailed Fault Zone Structure Beneath Haicheng Seismic Zone, NE China: Implications for the Fluid‐Induced Fault Reactivation
Core Problem: The seismogenic mechanism of the 1975 Haicheng earthquake has remained poorly resolved despite its historical importance.
Key Innovation: Dense-array seismicity relocation and crustal imaging reveal conjugate fault geometry and a fluid-rich middle-crust reservoir consistent with fluid-induced fault reactivation.
26. Exceptionally Elongated Strike‐Slip Rupture Caused by the 2025 MW 7.8 Myanmar Earthquake
Core Problem: Controls on exceptionally elongated strike-slip rupture remain poorly understood even for geometrically simple faults.
Key Innovation: Geodetic and teleseismic inversions show bilateral rupture with record aspect ratio and link the extreme elongation to fault connectivity and segmentation.
27. Dynamic restrengthening and fault heterogeneity explain megathrust earthquake complexity
Core Problem: The preexisting conditions that generate multistage megathrust rupture complexity remain uncertain.
Key Innovation: Three-dimensional dynamic rupture ensembles show that rapid restrengthening plus data-informed heterogeneity can reproduce Tohoku-like rupture complexity and trench slip.
28. EarthquakeNPP: A Benchmark for Earthquake Forecasting with Neural Point Processes
Core Problem: Neural point-process earthquake forecasting lacks a leak-free benchmark grounded in current seismological evaluation practice.
Key Innovation: EarthquakeNPP standardizes catalogs, ETAS baselines, and evaluation metrics into a stronger benchmark for neural earthquake forecasters.
29. Site effects in Managua and their relationship with the damage observed during the December 23, 1972 earthquake: a retrospective analysis
Core Problem: The contribution of local site effects to the severe 1972 Managua damage pattern required modern instrumental reassessment.
Key Innovation: Ambient-noise and strong-motion HVSR analyses confirm strong site amplification and soil-structure resonance in the zones of greatest historical damage.
30. REMSA: Foundation Model Selection for Remote Sensing via a Constraint-Aware Agent
Core Problem: Selecting a suitable remote-sensing foundation model remains difficult because task requirements and deployment constraints are poorly organized.
Key Innovation: REMSA pairs a structured RSFM database with a constraint-aware agent to rank candidate models and justify the selection process.
31. How To Embed Matters: Evaluation of EO Embedding Design Choices
Core Problem: Earth-observation pipelines increasingly rely on reusable embeddings, but the impact of embedding design choices is still undercharacterized.
Key Innovation: A systematic benchmark shows how backbone, representation depth, aggregation, and combination strategies affect downstream EO task quality and compression.
32. Causally regularized full-resolution channel attention and frequency refinement for cloud removal in satellite imagery
Core Problem: Cloud removal methods for optical-satellite imagery still lose fine detail and mis-handle high-frequency structure under thick cloud cover.
Key Innovation: A causal-regularized optical-SAR fusion model with full-resolution attention and frequency refinement improves cloud removal fidelity in satellite scenes.
33. LiDAR remote sensing meets weak supervision: Concepts, methods, and perspectives
Core Problem: LiDAR interpretation and inversion remain bottlenecked by scarce labels and field measurements across many tasks.
Key Innovation: This review unifies weak-supervision paradigms for LiDAR remote sensing and clarifies how pseudo-labeling, consistency, and domain transfer address LiDAR-specific constraints.
34. OilSAM2: Memory-Augmented SAM2 for Scalable SAR Oil Spill Detection
Core Problem: Unordered SAR oil-spill scenes break the temporal-memory assumptions used by existing SAM-based segmentation frameworks.
Key Innovation: OilSAM2 introduces hierarchical memory and structure-semantic update rules so cross-scene SAR information can be reused without temporal drift.
35. Dynamics-Informed Deep Learning for Predicting Extreme Events
Core Problem: Rare extreme events in chaotic systems require mechanism-aware precursors rather than purely statistical predictors.
Key Innovation: The framework derives instability-aware precursors from state snapshots and feeds them to a Transformer for longer-lead extreme-event prediction.