TerraMosaic Daily Digest: Mar 10, 2026
Daily Summary
Research published on March 10, 2026 sharpens geohazard science around one central idea: instability is now being treated as a state-dependent process rather than a static class. The most consequential slope papers resolve how sewage infiltration reorganizes loess collapse, how pedogenesis leaves a mechanical memory in loess-paleosol sequences, and how rainfall, piles, and excavation support alter the geometry of failure. Underground instability is handled with the same precision. Energy-based brittleness metrics, temperature-sensitive rockburst experiments, freeze-front suction measurements, and liquefaction back-analysis all move hazard assessment closer to measurable thresholds and away from purely descriptive susceptibility.
A parallel shift is evident in hydrological and coastal hazards. Coastal flood prediction is being accelerated through lightweight deep-learning surrogates that remain adaptation-aware, while seasonal rainfall and inflow forecasts are being strengthened by ensemble systems and data assimilation. Reservoir-state estimation is improved through SWOT fusion, and sediment forecasting is made more operational by embedding process constraints into machine-learning models. Coastal engineering papers likewise focus on design-ready control of scour, overtopping, and compound seismic-tsunami loading. The most useful Earth observation AI studies are those that attack real monitoring failure modes, such as missing wildfire observations, noisy change maps, or unverified geospatial reasoning, rather than treating hazard analysis as a generic vision benchmark.
Key Trends
Today's literature is organized around mechanistic slope failure, operational water-hazard intelligence, and hazard-specific Earth observation AI.
- Loess instability is being recast as a coupled hydro-mechanical problem: sewage infiltration, pedogenesis, rainfall-driven wetting fronts, and excavation support are all treated as state variables that govern collapse and reinforcement performance.
- Failure assessment in deep and cold-ground engineering is becoming energy- and process-based: BRER, thermal rockburst thresholds, freeze-front suction models, and liquefaction back-analysis describe how damage evolves before visible collapse.
- Coastal hazard research is moving from generic protection to parameterized design rules: seawall stability, scour suppression, revetment roughness, and overtopping-force formulae are increasingly framed for direct infrastructure decisions.
- Flood and sediment forecasting are being made operational through hybrid models: seasonal ensembles, data assimilation, SWOT observation fusion, and physics-informed machine learning improve skill where sparse observations and lead time are the main constraints.
- Earth observation AI matters most when it handles degraded evidence and reasoning fidelity: wildfire reconstruction, robust change detection, and process-supervised geospatial reasoning address the actual weak points of hazard intelligence workflows.
Selected Papers
This digest features 55 selected papers from 1014 papers analyzed across loess failure, slope reinforcement, underground instability, hydroclimatic forecasting, coastal hazards, water-quality risk, post-disaster recovery, and hazard-ready Earth observation methods.
1. Mechanism and failure model of group-occurring loess falls induced by domestic sewage discharge: insight from field investigation and flume experiment
Core Problem: Why prolonged sewage discharge can trigger clustered loess collapse in the absence of extreme rainfall or seismic forcing remains poorly resolved.
Key Innovation: Field observations and flume experiments isolate seepage softening, toe erosion, and progressive retrogression, yielding a trigger-specific failure model for sewage-driven loess collapse.
2. Impact of pedogenesis on the microstructure and mechanical properties of Quaternary loess-paleosol sequences
Core Problem: How pedogenesis and paleoclimate translate into present-day collapsibility and shear-strength contrasts in loess-paleosol sequences is still poorly quantified.
Key Innovation: The study links decarbonization, clay enrichment, and iron-oxide cementation to collapsibility, compressibility, and shear-strength evolution, tying Quaternary climate shifts to modern loess instability.
3. Upper bound limit analysis of pile-reinforced soil slopes before and after rainfall conditions
Core Problem: Existing upper-bound analyses inadequately represent discontinuous slip geometry at piles and the shift to wetting-front-controlled failure after rainfall.
Key Innovation: The work constructs dry and rain-affected failure mechanisms, validates them numerically, and quantifies how rainfall intensity, pile layout, and geometry reshape slope stability.
4. The mechanical behavior of pipe umbrellas in loess tunnels
Core Problem: Pipe-umbrella support in deep loess tunnels lacks a transparent analytical framework for stiffness, grouting benefit, and excavation-step optimization.
Key Innovation: An analytical Pasternak-foundation model with load-transfer metrics identifies excavation footage and grouting conditions that maximize advance-support efficiency.
5. Pile-soil arching mechanism acting on bridge piles in a soil slope
Core Problem: Inclined pile-soil arching in sloping ground remains poorly constrained, leaving lateral pile loading and arching-depth predictions uncertain.
Key Innovation: The paper develops an inclined arching model that predicts arching depth, stress transfer, and the validity range of soil-arching effects for bridge piles on slopes.
6. An index for evaluating rock failure using energy release and rock brittleness
Core Problem: Conventional brittleness indices indicate susceptibility but not the likely severity and spatial extent of brittle rock failure.
Key Innovation: The proposed BRER metric combines brittleness with confinement-sensitive energy release and is validated against acoustic and microseismic observations from deep hard-rock settings.
7. Influence of Thermal Damage on the Energy Accumulation and Release Characteristics of Sandstone during Impact Rockburst
Core Problem: The temperature window that maximizes elastic-energy storage and violent release in sandstone rockburst remains uncertain.
Key Innovation: True-triaxial impact tests identify a 200-300 °C critical range with peak pre-failure energy accumulation and the fastest post-peak release.
8. Experimental investigation and modeling of coupled thermal–hydraulic responses in seasonally frozen weathered sandstone under unidirectional freezing
Core Problem: Freeze-front migration, unfrozen-water redistribution, and suction growth in highly erodible weathered sandstone are still poorly represented.
Key Innovation: Column-freezing experiments and saturation-aware constitutive fits quantify three-zone thermal structure and freezing-driven moisture redistribution in weathered sandstone.
9. Seismic performance of four adjacent buildings affected by severe liquefaction in İskenderun during the 2023 M7.8 Pazarcık earthquake
Core Problem: The coupled soil-structure mechanism behind clustered building settlement on liquefied reclaimed ground required physically consistent back-analysis.
Key Innovation: Calibrated nonlinear effective-stress site and structural models reproduce the observed punching settlement during the 2023 Pazarcik earthquake.
10. Seismic and tsunami-induced stability assessment of seawalls against passive failure under varying water levels
Core Problem: Stability assessment of waterfront retaining walls rarely integrates variable groundwater, unsaturated backfill, and compound tsunami-seismic loading.
Key Innovation: An upper-bound anti-sliding solution with generalized effective stress and suction effects enables lifecycle stability evaluation under compound loading.
11. Climate adaptation-aware flood prediction for coastal cities using Deep Learning
Core Problem: City-scale coastal flood scenario modeling under sea-level rise and shoreline adaptation remains too slow for iterative planning.
Key Innovation: A lightweight CNN surrogate predicts flood-depth fields across adaptation and sea-level scenarios and transfers between Abu Dhabi and San Francisco.
12. Monitoring Reservoir Storage Using SWOT Satellite Observations and a Reservoir Operation Model
Core Problem: Reservoir storage monitoring remains weak where water level and inundated area are not observed synchronously.
Key Innovation: The study fuses simultaneous SWOT elevation-area observations with an operations model to maintain low storage error while filling substantial temporal gaps.
13. Assessing the ability of the ECMWF seasonal prediction model to forecast extreme September–November rainfall events over Equatorial Africa
Core Problem: Seasonal prediction skill for September-November extreme rainfall over Equatorial Africa remains uneven and its circulation controls underdiagnosed.
Key Innovation: The evaluation shows that SEAS5.1 captures key rainfall teleconnections and identifies where September initializations yield operationally useful warning skill.
14. Droughts in South East Europe (SEE): current picture, tendencies and impact
Core Problem: Drought knowledge across Southeast Europe remains fragmented across methods, indicators, and national evidence bases.
Key Innovation: The paper combines literature synthesis with CDI and EDID evidence to identify regional drought patterns, blind spots, and knowledge asymmetries across nine countries.
15. The impact of High-Intensity earthquakes on sediment transport characteristics during flood events: A Case study of rivers in the Longmenshan region
Core Problem: The long tail of major earthquake impacts on flood-time sediment transfer at basin scale remains insufficiently quantified.
Key Innovation: Long records reveal a post-2013 regime shift in flood-event sediment load and quantify the growing contribution of Longmenshan tributaries to Three Gorges sediment inflow.
16. Physics-informed machine learning with embedded sediment rating curve constraints for high-fidelity multi-lead-time forecast of suspended sediment concentration
Core Problem: Standard empirical or machine-learning sediment forecasts degrade under hysteretic, high-flow, and multiscale suspended-sediment dynamics.
Key Innovation: Regime-aware sediment rating curves are embedded into model training, preserving flood hysteresis and maintaining strong skill across lead times up to 14 days.
17. A data-assimilated SEAS5 forecasting framework for seasonal hydropower inflows in a snow-dominated basin
Core Problem: Seasonal inflow forecasts in snow-dominated basins are limited by weak snow-state initialization under nonstationary climate conditions.
Key Innovation: Variational data assimilation, HBV, and SEAS5 ensembles are coupled to improve 1-3 month predictions of inflow, snow cover, and snow water equivalent.
18. Quantifying climate-induced cascading effects on runoff in a cold region using a glacier-enhanced Budyko framework
Core Problem: Conventional Budyko attribution cannot separate direct climate forcing from glacier-mediated and land-surface-mediated cascading runoff responses.
Key Innovation: The framework extends Budyko with glacier mass balance and ridge regression to partition runoff change into direct climate, cascading climate, and human components.
19. Evaluating channel morphological dynamics and identifying sediment hotspots in Tamor River, Central Himalaya, Nepal
Core Problem: Long-term channel migration and sediment-hotspot evolution in Himalayan rivers remain difficult to quantify consistently over decades.
Key Innovation: Multi-temporal satellite analysis resolves reach-scale erosion, deposition, and centerline migration, identifying the most unstable interval and downstream hotspot accumulation.
20. Holocene fan systems in the Alpine Rhine basin as records of sedimentary pulses during a warming climate
Core Problem: The controls on large Holocene fan construction in Alpine headwaters remain poorly constrained, limiting interpretation of sediment-pulse archives.
Key Innovation: Sedimentology and cosmogenic denudation evidence link fan aggradation to post-glacial sediment pulses, base-level rise, and the downstream Flims rock avalanche.
21. Combined effects of altered hydrodynamics and skew angles on scour around a complex cap pile group foundation
Core Problem: Engineering-altered hydrodynamics and skewed approach flow can intensify scour around complex bridge foundations, but their joint effect is poorly quantified.
Key Innovation: Tidal-current simulation and flume testing show how project-driven velocity increases and skew angle deepen and widen scour pits around cap-pile groups.
22. Effect of pile corner rounding on local scour around offshore wind turbine foundations in sands using CFD–DEM
Core Problem: The scour-reduction benefit of rounding pile corners is known qualitatively but not optimized quantitatively.
Key Innovation: CFD-DEM simulations identify how rounded corners reorganize near-bed flow and locate an optimal R/D ratio for minimizing equilibrium scour depth.
23. In-situ cementation of riprap for scour protection around offshore monopile foundations: from prototype modeling to field implementation
Core Problem: Conventional riprap protection around offshore monopiles degrades under coupled wave-current loading and sustained local scour.
Key Innovation: The study develops and field-validates in-situ cemented riprap that preserves internal structural continuity and sharply outperforms adjacent conventional protections.
24. Solitary wave overtopping impact on a car at a sloped seawall crest
Core Problem: Vehicle-scale hazard at seawall crests under overtopping waves is poorly quantified despite rising coastal exposure.
Key Innovation: Experiments and simulations yield a predictive relation for peak horizontal force on vehicles as a function of wave and placement parameters.
25. Wave overtopping performance of mild slope rock revetments enhanced with additional armour layer thickness
Core Problem: Overtopping behavior of mild-slope rock revetments with thicker armour layers remains underconstrained for design under higher water levels.
Key Innovation: Physical model tests extend the EurOtop evidence base and derive practical roughness-factor ranges for mild-slope revetments with an added armour layer.
26. Robust Wildfire Forecasting under Partial Observability: From Reconstruction to Prediction
Core Problem: Satellite wildfire forecasts fail when smoke, clouds, and sensor gaps corrupt the observations available at deployment.
Key Innovation: A two-stage framework reconstructs plausible fire states before forecasting spread, largely restoring next-day skill under severe masking.
27. DFPF-Net: Dynamically Focused Progressive Fusion Network for Remote Sensing Change Detection
Core Problem: Bi-temporal change detection remains vulnerable to pseudo-changes and local noise from illumination and scene variability.
Key Innovation: Progressive transformer fusion and dynamic focus modules suppress both global pseudo-change and local edge noise in change maps.
28. OmniEarth: A Benchmark for Evaluating Vision-Language Models in Geospatial Tasks
Core Problem: Earth-observation vision-language models still lack a rigorous benchmark that jointly tests perception, reasoning, and robustness under realistic geospatial tasks.
Key Innovation: OmniEarth introduces a multi-source benchmark spanning 28 tasks and demonstrates persistent gaps in visually grounded geospatial reasoning.
29. GeoSolver: Scaling Test-Time Reasoning in Remote Sensing with Fine-Grained Process Supervision
Core Problem: Remote-sensing chain-of-thought systems still produce intermediate reasoning steps that are not visually faithful.
Key Innovation: A process reward model and tree-structured reinforcement-learning scheme improve stepwise remote-sensing reasoning and test-time scaling.
30. Seepage characteristics and mechanisms at clayey sand-structure interface using NMR technology
Core Problem: The staged evolution and microscale controls of seepage erosion along soil-structure interfaces in earth-rock dams remain poorly characterized.
Key Innovation: Visual seepage testing plus NMR resolves three erosion stages and quantifies how clay content, compaction, and loading duration shift critical gradients.
31. Characteristics and formation mechanism of axial chain rockbursts in deep-buried tunnels traversed by stiff rock vein through gneiss strata
Core Problem: The origin of axial chain rockbursts, prolonged and repetitive failures in deep tunnels, remains poorly understood.
Key Innovation: Field surveys, microseismic monitoring, and true-triaxial tests reveal a vein-controlled three-stage incubation and a stress-energy mechanism for chain bursting.
32. Drivers and compounding factors of temporal sedimentation rate variations over the past 100 years in a human‐impacted mountainous catchment, Korea
Core Problem: Century-scale sedimentation change in mountainous catchments reflects interacting land-use, precipitation, and disturbance signals that are hard to separate.
Key Innovation: Lead-210-dated cores and land-use reconstruction identify distinct depositional regimes and show how channel modification, rainfall, and earthquakes can compound.
33. Model‐Based Interpretation of Solute Exports and Carbon Partitioning During Shale Weathering in a Mountainous Hillslope
Core Problem: Coupled hydrological and geochemical controls on shale-weathering solute export remain difficult to quantify under transient mountain conditions.
Key Innovation: A multidimensional process model partitions the roles of snowmelt, oxygen access, mineral reactions, and slope drainage in carbon and solute export.
34. Evaluating Evapotranspiration Simulation Performance in 30 Conceptual Hydrological Models: Insights Into ET Representation Across Diverse Climates
Core Problem: Hydrological model intercomparison has emphasized streamflow while leaving evapotranspiration representation under climate variability under-tested.
Key Innovation: The cross-catchment evaluation shows that the most reliable evapotranspiration formulations depend strongly on climatic regime.
35. Satellite Solutions: Facing Chlorophyll‐a Retrieval in Small Mountain Lakes in the Sierra Nevada, Spain
Core Problem: Small mountain lakes are difficult to monitor from space because adjacency effects and scale mismatch degrade chlorophyll retrieval.
Key Innovation: The comparison of Sentinel-2, Planet, and WorldView-3 shows that carefully chosen green and near-infrared predictors can outperform expectations based solely on resolution.
36. Integrating Satellite Data Into Meso‐Scale Habitat Modeling for Non‐Perennial Rivers and Streams
Core Problem: Flow intermittency and ponding duration in non-perennial rivers are rarely resolved with enough temporal detail to assess ecological hazard.
Key Innovation: More than 1700 satellite images are combined with meso-scale habitat simulation to quantify drying-state transitions and habitat vulnerability.
37. Longitudinal Mean Velocity and Turbulent Kinetic Energy Within an Emergent Canopy in Nonuniform Flows
Core Problem: Turbulence growth in vegetated nonuniform flows is not captured well by uniform-flow assumptions.
Key Innovation: Flume experiments and a new analytical model quantify how water-surface gradient reshapes velocity-TKE scaling inside emergent canopies.
38. Numerical experiments of cloud seeding for mitigating localization of heavy rainfall: a case study of Mesoscale Convective System in Japan
Core Problem: It remains uncertain whether cloud seeding can disperse localized extreme rainfall in mesoscale convective systems rather than intensify it.
Key Innovation: Numerical experiments show that mid-upper-tropospheric overseeding can suppress graupel growth in the core and shift rainfall downwind.
39. Reducing risk together: moving towards a more holistic approach to multi-hazard and multi-risk assessment and management
Core Problem: Multi-hazard risk governance still lacks consistent concepts, scenario tools, and operational integration of equity and warning systems.
Key Innovation: The paper distills MYRIAD-EU lessons into a concrete agenda for dynamic multi-risk evidence, usable tools, equitable adaptation, and multi-hazard early warning.
40. Exogenous inputs and biological consumption regulate nitrate concentrations and fluxes in a river-groundwater-reservoir continuum
Core Problem: Nitrate dynamics across river-groundwater-reservoir systems remain difficult to apportion among loading, transport, and in-system transformation.
Key Innovation: Stable isotopes and Bayesian mixing separate dominant manure and sewage sources and show how transport and biogeochemical controls shift seasonally.
41. Wind-water energy characteristics and sediment transport prediction in sandy coarse sand basin
Core Problem: Wind-water composite erosion in the coarse-sand Yellow River region is hard to predict because surface types alter the operative energy balance.
Key Innovation: An energy-based erosion and sediment-transport model differentiates behavior across sand-covered bedrock, sand-covered loess, and loess erosion zones.
42. Chloride source dynamics under dry–wet conditions in a watershed with imbalanced chloride input–output: Insights from tracer-aided modeling
Core Problem: Chloride-source shifts across dry-wet transitions remain dynamically unresolved in salinizing watersheds.
Key Innovation: A tracer-aided StorAge Selection model separates atmospheric, interflow, and groundwater contributions and shows the strongest source reorganization during transition periods.
43. A Scalable Potential of Alpha Earth and Multisensor Datasets for Assessing Mangrove Intactness and Degradation Using Deep and Machine Learning Algorithms
Core Problem: Mangrove degradation monitoring at scale remains difficult despite increasing multisensor Earth observation availability.
Key Innovation: The study shows that Alpha Earth-style embeddings outperform conventional multisensor combinations for mapping intactness and degradation over 2017-2024.
44. Multisatellite Scheduling Method Based on a Historical Cloud Coverage at Regional Scale
Core Problem: Regional optical-satellite scheduling wastes opportunities when historical cloud climatology is ignored.
Key Innovation: A cloud-aware optimization framework maximizes observation success fraction under a target success-probability constraint.
45. Integrating Deep Learning and Distance‐Based Clustering to Optimize the Field Scale In Situ Uranium Leaching System in Heterogeneous Reservoirs
Core Problem: Geological heterogeneity and costly reactive-transport simulations hinder optimization of in situ leaching under contamination risk.
Key Innovation: A deep-learning proxy and distance-based clustering strategy reduce simulation redundancy while optimizing well-control strategies.
46. A Pseudomultitask Neural Network Classification Model for Cropland Mapping in Mountainous Areas Using High-Resolution Remote Sensing Images
Core Problem: In mountainous terrain, boundary fidelity in high-resolution land-surface mapping is degraded by scale effects and complex topography.
Key Innovation: A pseudomultitask segmentation design fuses mask, edge, and distance cues to improve boundary representation.
47. AMD-DSMNet: Asymmetric Multidirectional Convolutional Attention and Dynamic Spatial Modulation Network for Cropland Change Detection
Core Problem: High-resolution change detection still suffers from boundary ambiguity, pseudo-change, and broken edge continuity.
Key Innovation: Directional attention and dynamic spatial modulation preserve topology while suppressing irrelevant change.
48. Small-Vehicle Change Detection in UAV Imagery via Physics-Aware Spatiotemporal Cues and Reproducible Evaluation
Core Problem: Subtle, small-object change detection in UAV imagery remains difficult under low contrast and limited temporal signal.
Key Innovation: Physics-aware spatiotemporal cues after bitemporal alignment improve weak-target change detection in UAV scenes.
49. HY-1E COCTS2 Near-Real-Time Skin Sea Surface Temperature Retrieval
Core Problem: Higher-resolution, near-real-time sea-surface temperature retrieval is needed for coastal and marine hazard monitoring.
Key Innovation: The HY-1E COCTS2 workflow achieves sub-degree performance against in situ and SLSTR references in near-real-time SST retrieval.
50. SinGeo: Unlock Single Model's Potential for Robust Cross-View Geo-Localization
Core Problem: Cross-view geolocalization models often collapse under unseen field-of-view and orientation conditions.
Key Innovation: Dual discriminative learning and curriculum training allow a single model to remain robust across view variations without extra geometry modules.
51. GeoAlignCLIP: Enhancing Fine-Grained Vision-Language Alignment in Remote Sensing via Multi-Granular Consistency Learning
Core Problem: Remote-sensing vision-language models remain weak at fine-grained regional interpretation because they rely too heavily on global alignment.
Key Innovation: Multi-granular semantic alignment and a new fine-grained dataset improve region-text correspondence in remote-sensing imagery.
52. Measuring spatial urban recovery and expansion in tsunami-affected areas by satellite remote sensing and machine learning approach
Core Problem: Long-term urban recovery after tsunami impact is difficult to monitor consistently at city scale.
Key Innovation: Time-series Landsat and random-forest regression reconstruct two decades of building dynamics in Banda Aceh and resolve uneven recovery trajectories.
53. A satellite-based approach for estimating runoff and river discharge in the Pan-Arctic region from 2003 to 2022
Core Problem: Daily runoff and discharge remain poorly observed across ungauged Arctic basins where snow and storage dynamics complicate estimation.
Key Innovation: Multiple satellite observables are integrated into a conceptual model to reconstruct daily Pan-Arctic runoff and discharge from 2003-2022.
54. Impact of Density Gradients and Momentum Ratios on Streamwise Circulation at River Confluences
Core Problem: Density contrast and momentum ratio jointly control river-confluence mixing, but their combined effect on streamwise circulation lacks predictive formulation.
Key Innovation: Eddy-resolved simulations derive a predictive relation for dominant circulation regimes and downstream mixing behavior.
55. Use of Gaussian mixture model for standardized precipitation index in monsoon region
Core Problem: Single-distribution SPI formulations can misrepresent multi-peaked precipitation regimes in monsoon climates.
Key Innovation: Kernel-density screening with selective Gaussian-mixture fitting improves SPI estimation across South Korea's diverse monsoon precipitation patterns.