TerraMosaic Daily Digest: Mar 18, 2026
Daily Summary
This March 18, 2026 digest distills 49 selected papers from 1,274 analyzed records. The strongest studies treat landslides and related geohazards as operational systems rather than isolated case reports. Transfer-learning landslide mapping now achieves cross-region performance with very limited local labels; long-term monitoring of the Jiangdingya landslides resolves how rainfall, multi-level slip surfaces, and anti-slide piles interact after remediation; and new experiments on granular-fluid flows directly partition inertial and viscous controls relevant to debris-flow dynamics.
A second cluster sharpens the treatment of cascading and compound hazards. Flood impacts are decomposed into rainfall, surge, and sea-level-rise contributions for Cyclone Idai; China-wide compound floods are organized by multivariate driver combinations and climate modes; and flash drought onset is linked to abrupt shifts in vertical-motion regimes. Across landslide, liquefaction, wildfire, and seismic applications, the most persuasive AI papers are not generic. They are interpretable, formally verified, or embedded in physically constrained monitoring and modelling workflows.
Key Trends
The day's center of gravity is a shift from descriptive hazard studies toward transferable, physically disciplined, and decision-ready geohazard intelligence.
- Landslide analysis is becoming more portable: compact transfer-learning architectures and integrated field-monitoring systems are reducing the dependence on site-specific data while preserving physical interpretability.
- Compound-process thinking is now central: compound flooding, flash drought, atmospheric rivers, and wildfire-boundary-layer feedbacks are being analysed as interacting hazard chains rather than separate extremes.
- Trustworthy geohazard AI is tightening its standards: formal verification, domain-informed shape constraints, and explainable feature selection are emerging as necessary checks before model deployment.
- Mechanics remains the backbone of resilience work: new studies on pore-pressure generation, anisotropic strength, hydro-mechanical damage, and basin amplification show that constitutive realism still drives the most consequential engineering advances.
- Environmental observations are expanding in scale and utility: SWOT, VIIRS-derived fire extent, mountain DTM correction, and AI-ready snow products are enlarging the measurement infrastructure available for hazard assessment and response.
Selected Papers
This digest features 49 selected papers from 1274 papers analyzed.
1. Reactivation of ancient landslides in complex tectonic settings: Insights from the long-term monitoring of Jiangdingya landslides on the northeastern Tibetan Plateau
Core Problem: Large ancient landslides within active fault zones remain reactivation hazards, yet post-remediation deformation and support performance are rarely constrained by long-duration field evidence.
Key Innovation: It combines SBAS-InSAR, geophysics, deep displacement, pore pressure, pile internal-force monitoring, and 3D simulation to resolve multi-slip-surface hydromechanical reactivation and the distributed loading behavior of anti-slide piles.
2. A CNN–Transformer hybrid network for efficient cross-region landslide detection by transfer learning
Core Problem: Post-event landslide mapping still depends heavily on labeled local data, limiting rapid deployment in newly affected regions.
Key Innovation: The paper introduces LSDFormer and a globally distributed pretraining dataset, showing that transfer learning can deliver high-quality cross-region mapping with very small target-domain label budgets and minute-scale regional inference.
3. Experimental investigation on dynamics and flow resistance of granular-fluid flows
Core Problem: The partitioning of flow resistance in granular-fluid hazards is still poorly understood, especially across changing mixture composition and slope.
Key Innovation: Through controlled experiments and an inertial-viscous framework, the study separates the roles of solid inertial stress and liquid viscous stress and refines the model for nearly viscoplastic flow regimes.
4. Detrending—a new trend in geomorphology? Unravelling erosional processes at the base of submarine landslides
Core Problem: Basal erosion beneath submarine landslides is mechanically important but usually obscured by the overlying slide mass and slope geometry.
Key Innovation: Detrending of high-quality 3D seismic basal surfaces reveals morphometrically distinct scour fields, allowing the authors to distinguish block-tooling erosion from longitudinal-vortex erosion.
5. Nature-based Solutions for landslide risk reduction in informal settlements of Salvador da Bahia, Brazil
Core Problem: The practical effectiveness, affordability, and community acceptability of nature-based landslide mitigation in informal settlements remain uncertain.
Key Innovation: Using expert interviews and resident workshops, the study shows that measures such as reforestation and vetiver grass can rival conventional works while offering substantially lower costs and broader social-ecological benefits.
6. A Framework for Modeling Liquefaction-Induced Road Disruptions After Earthquakes: Implications for Emergency Response and Access in the Cascadia Region of North America
Core Problem: Regional emergency planning lacks mechanics-informed estimates of how earthquake-triggered liquefaction translates into road closure and access loss.
Key Innovation: The framework links liquefaction severity, empirically derived fragility relations, and spatially correlated network simulations to estimate closures, service reductions, and hospital access disruption across Cascadia.
7. Harnessing Deep Learning for Dual Gains in S2S‐Scale Soil Moisture Forecasting and Flash Drought Mechanisms
Core Problem: Numerical weather prediction remains weak at sub-seasonal flash-drought forecasting and offers limited insight into evolving drought drivers.
Key Innovation: Interpretable deep-learning ensembles forecast pentad soil-moisture anomalies across the full flash-drought lifecycle and identify water-dominated, energy-dominated, and composite precursor regimes.
8. Climate and impact attribution of compound flooding induced by tropical cyclone Idai in Mozambique
Core Problem: Attributing climate-change effects on tropical-cyclone compound flooding and impacts is difficult, especially in vulnerable and data-scarce regions.
Key Innovation: A storyline modelling chain jointly perturbs rainfall, storm surge, and sea-level rise to show that climate signals can be stronger in realized damage than in flood extent or volume alone.
9. Integrative statistical analysis of compound flood dynamics under climate variability in China’s coastal zone
Core Problem: Compound-flood assessments often isolate drivers instead of jointly analyzing surge, rainfall, discharge, and climate variability.
Key Innovation: The authors combine copulas, LSTM models, and count regression across 391 coastal points to map flood-type seasonality and quantify ENSO and IPO controls on event frequency.
10. 3D physics-based modeling of ground motions and basin amplification in the northern North China Basin: Implications from MW 7.5 earthquakes on the Xiadian fault
Core Problem: Deep sedimentary-basin amplification is not adequately represented in current hazard assessment for the North China Basin.
Key Innovation: High-resolution 3D simulations show strong long-period amplification and prolonged motions tied to basin geometry, revealing where empirical ground-motion models underpredict shaking.
11. Stress-based evaluation of pore pressure generation in sands subjected to undrained cyclic loading
Core Problem: Classical stress-based pore-pressure generation models do not consistently explain how sand properties alter cyclic liquefaction response.
Key Innovation: The study calibrates a revised stress-based model across sands with varying fines, particle shape, void ratio, and loading amplitude, clarifying parameter sensitivity for liquefaction analysis.
12. Formal verification of tree-based machine learning models for lateral spreading
Core Problem: Machine-learning predictors for geotechnical hazards can achieve strong accuracy while still violating basic physical expectations.
Key Innovation: Tree ensembles are encoded in SMT logic so domain-wide monotonicity and safety constraints can be proved or falsified, establishing a verification workflow for lateral-spreading models.
13. Domain-informed explainable boosting machines for trustworthy lateral spread predictions
Core Problem: Even transparent additive models can learn nonphysical response curves when trained on sparse or biased hazard datasets.
Key Innovation: The authors correct learned shape functions with domain knowledge, producing more trustworthy global and local explanations while preserving most predictive skill.
14. Dynamical linkages between planetary boundary layer schemes and wildfire spread processes: a case study using WRF-Fire version 4.6
Core Problem: Wildfire spread simulations remain highly sensitive to how boundary-layer turbulence and fire-atmosphere feedbacks are parameterized.
Key Innovation: A multi-scheme WRF-Fire comparison against station data shows that MYNN3 best captures thermal perturbations, turbulence intensity, and near-surface wind response in mountainous fire conditions.
15. CONFEX: A Database for CONUS Fire EXtent
Core Problem: Consistent, moderate-resolution event-level wildfire extent data remain limited across CONUS and Alaska.
Key Innovation: CONFEX converts VIIRS active-fire detections into clustered fire events with perimeters, ignition locations, centroids, and event dates, creating a reusable fire-extent database for 2012-2024.
16. Sudden Vertical Velocity Regime Shifts Initiate Flash Drought in Puerto Rico
Core Problem: Flash-drought onset in tropical settings lacks robust synoptic-scale predictors that can extend warning lead time.
Key Innovation: The paper links Puerto Rico flash-drought initiation to rapid shifts from upward motion to strong subsidence, pointing to a dynamical precursor with subseasonal predictability value.
17. Distinct Dominant Intraseasonal Oscillation Modes Modulate South China's Persistent Extreme Precipitation in Rainy Season
Core Problem: Persistent extreme rainfall over South China is modulated by intraseasonal variability, but the governing modes and mechanisms are not clearly separated.
Key Innovation: The authors identify two dominant oscillation regimes and show that they intensify extreme precipitation through different moisture-convergence and warm-advection pathways.
18. Seismic risk assessment of urban areas: an integrated multi-scale and multi-refinement framework
Core Problem: Urban seismic risk workflows still struggle to bridge asset-scale analysis, network interactions, and uncertainty propagation across refinement levels.
Key Innovation: This framework models cities as interacting layers of buildings and infrastructure, explicitly carrying uncertainty and cascading effects from asset scale to urban scale.
19. Influence of rubber particle-EICP composite improved loess as a seismic isolation trench on the vibration response of high-voltage transmission towers
Core Problem: Power-transmission towers in loess regions need practical ground-treatment strategies that reduce incoming seismic energy and soil-structure interaction.
Key Innovation: Numerical analyses validated by shaking-table tests show that rubber-particle and EICP-improved loess trenches can markedly reduce tower acceleration and displacement when geometry is optimized.
20. A Framework for Extending Rock Strength Criteria to Anisotropy
Core Problem: Structured and layered rocks exhibit strength anisotropy that conventional isotropic criteria cannot represent consistently.
Key Innovation: The study introduces a structure-tensor formulation that generalizes isotropic strength criteria to anisotropic conditions and reproduces observed orientation-dependent strength behavior.
21. The Failure Mechanism and Damage Evolution of Soft–Hard Interbedded Rocks Under the Influence of Cyclic Dynamic Loads and Moisture Content Based on Acoustic Emission
Core Problem: Moisture and cyclic loading strongly affect interbedded rock failure, but their joint damage evolution remains poorly resolved.
Key Innovation: Acoustic-emission and energy analyses show how increasing moisture shifts failure from brittle toward plastic behavior and accelerates damage accumulation under cyclic loading.
22. Investigation of Microdamage and Seepage Evolution in Sandstone Under Coupled Thermomechanical Loading and Cooling Shocks via NMR
Core Problem: Thermal shock and prestress jointly alter pore evolution and permeability in reservoir rocks, but macro-micro damage coupling remains hard to quantify.
Key Innovation: Low-field NMR tracking across repeated cooling shocks reveals staged pore growth and exponential permeability increase, and supports a corrected macro-micro damage model.
23. Mechanical Behavior and Dual-Threshold Damage Model of Water-Saturated Sandstone Under Hydro-Mechanical (H-M) Coupling
Core Problem: Hydro-mechanical coupling degrades rock stability during excavation, yet existing models do not capture the full nonlinear transition from crack initiation to macro-failure.
Key Innovation: Triaxial tests under varying confining and water pressures support a dual-threshold damage model that links energy dissipation, crack breakthrough, and permeability evolution.
24. Hydro-mechanical-damage modelling of hydraulic fracturing processes in granite with single and double pre-existing flaws
Core Problem: The interaction between pre-existing flaws, seepage, and damage evolution in hydraulic fracturing remains insufficiently coupled in predictive models.
Key Innovation: A fully coupled hydro-mechanical-damage finite-element framework reproduces flaw-controlled initiation, pressure evolution, stress shadowing, and geometry-dependent coalescence in granite.
25. A unified peridynamics-based framework for hydro-mechanical coupling analysis in fractured porous media
Core Problem: Open and sealed fractures create heterogeneous hydro-mechanical behavior that is difficult to capture within conventional continuum models.
Key Innovation: The proposed peridynamic framework couples fracture seepage and deformation through adaptable interface mechanics, enabling efficient simulation of multi-physical response in fractured rock masses.
26. Evaluating the 2020 European Seismic Hazard Model (ESHM20) using ShakeMap-derived ground motion fields in Greece
Core Problem: Regional hazard models need transparent consistency checks against long observational records without overstating formal validation.
Key Innovation: By generating 2,617 ShakeMap fields and comparing exceedance fractions against ESHM20 maps, the study offers a reproducible national-scale screening method for hazard-model behavior.
27. A probabilistic framework for magnitude-based stochastic ground motion simulation using empirical Green's functions
Core Problem: Stochastic ground-motion simulation still faces a trade-off between seismological realism and practical parameter requirements.
Key Innovation: The framework conditions source variability on magnitude and convolves stochastic source functions with empirical Green's functions to generate intensity-measure ensembles with quantified uncertainty.
28. Comparative study on water-induced cracking mechanisms and characteristics of compositionally different mudstones: Identification of the rapid damage development stage
Core Problem: Water-softening and cracking in mudstone slopes remain difficult to predict because mineralogical controls on damage evolution are not well resolved.
Key Innovation: By combining swelling tests, acoustic emission, CT, and NMR, the study identifies a critical saturation threshold and links contrasting cracking styles to mineral-composition-driven swelling behavior.
29. A database of objectively identified atmospheric rivers based on a multi-method fusion algorithm
Core Problem: Atmospheric-river identification varies substantially across algorithms, complicating hazard attribution and cross-study comparison.
Key Innovation: A fusion algorithm synthesizes established detection strategies into a global six-hourly database from 1940 to 2024, explicitly designed to reduce identification uncertainty in extreme-event analysis.
30. Forest Impacts on Peak Runoff Revealed by Accounting for the Effects of Climate
Core Problem: Large-sample assessments of land-cover effects on runoff are often confounded by climate and antecedent wetness differences among catchments.
Key Innovation: Grouping catchments by antecedent wetness reveals a clearer forest signal, showing systematically lower peak runoff responses in forested basins than in cropland or grassland basins.
31. Data‐Driven Exploration of Tropical Cyclone's Controllability
Core Problem: Data-driven methods for identifying small perturbations capable of weakening tropical cyclones remain underdeveloped.
Key Innovation: Ensemble Kalman Control is used to identify surface-moisture perturbations that reduce eyewall convection and storm intensity in a realistic cyclone simulation.
32. Observing the tidal pulse of rivers from wide-swath satellite altimetry
Core Problem: The inland propagation of tides through coastal rivers remains poorly mapped globally because gauges are sparse and conventional altimetry undersamples rivers.
Key Innovation: SWOT observations are used to map tidal influence across 3,172 coastal rivers, opening a new basis for monitoring estuarine hazard exposure under sea-level rise and water regulation.
33. Three-dimensional stability analysis of slurry trench in frictional-cohesive soil using the upper bound perturbation method
Core Problem: Three-dimensional trench failure in frictional-cohesive soils is difficult to evaluate with efficient yet accurate design tools.
Key Innovation: A perturbation-based upper-bound method identifies characteristic 3D failure modes and yields a direct design equation for safety-factor estimation.
34. Momentum-based criteria for initiating unbounded sand fluidization induced by pipeline leakage
Core Problem: Existing leakage-induced fluidization models do not treat unconfined failure in gap-graded and partially saturated soils well.
Key Innovation: The paper derives a momentum-based analytical criterion for the minimum leakage flow needed to trigger unbounded sand fluidization across water levels and grading states.
35. Permeability degradation in self-healing rock salt under humidity cycling: experimental and modeling study using multiple characterization techniques
Core Problem: Long-term tightness of salt caverns depends on humidity-driven self-healing in damaged salt, but permeability evolution is difficult to characterize mechanistically.
Key Innovation: Continuous pulse tests, NMR, and CT are fused to identify staged self-healing and to build a permeability model that explicitly incorporates recrystallization-enhanced sealing.
36. Stabilizing gas hydrate-bearing sediments via microbially induced carbonate precipitation: a pore-scale perspective
Core Problem: Mechanical stabilization of hydrate-bearing sediments requires treatments that preserve permeability while remaining effective under hydrate dissociation.
Key Innovation: High-pressure microfluidics reveals new cementation modes, deposition controls, and retention behavior for MICP under hydrate conditions, clarifying its stabilization potential.
37. Intelligent dynamic fire alarm framework for road tunnels using traffic accident prior knowledge and multi-target detection
Core Problem: Tunnel fire monitoring still suffers from high false alarms and missed detections under complex illumination, smoke, and traffic conditions.
Key Innovation: A two-stage system fuses traffic-state recognition, accident cues, and flame-duration logic on top of multi-target detection to deliver faster and more adaptive tunnel-fire alarms.
38. Roadside LiDAR placement optimization in curved tunnels: A framework integrating probabilistic grids and neural radiance models
Core Problem: Sensor deployment in curved tunnels is difficult because visibility is constrained by geometry and dynamic traffic occlusion.
Key Innovation: The study combines probabilistic occupancy grids, neural radiance modelling, and hybrid optimization to derive scenario-specific roadside LiDAR layouts for tunnel perception.
39. Experimental validation of a unified formula for limiting support pressure incorporating cutterhead‒soil interaction on sandy‒gravel composite strata irrespective of stratification sequence
Core Problem: Shield tunnelling in sandy-gravel composite strata lacks a unified support-pressure formulation that spans different layering orders.
Key Innovation: Model tests support a single face-stability equation that incorporates stratification ratio and cutterhead-soil friction, clarifying failure stages and settlement sensitivity.
40. Effects of gravel isolation layers on evaporation-driven water-salt migration and bearing behavior of saline soil subgrades
Core Problem: Evaporation-driven salt migration degrades saline-soil subgrades and undermines bearing capacity in arid transport corridors.
Key Innovation: Laboratory model tests show that gravel isolation layers create a capillary barrier that sharply reduces salt accumulation and identifies a salinity range of peak structural risk.
41. Quantifying moisture effects on the dynamic friction coefficient of compacted crushed rock class IV material and its implications for corrugation of unsealed roads
Core Problem: The moisture dependence of dynamic friction in unsealed-road materials is rarely measured directly, limiting predictive models of corrugation and deterioration.
Key Innovation: Pendulum tests across drying-wetting states yield an exponential friction-moisture relationship that links suction-driven strengthening to road-surface behavior.
42. Detection of Shallow Tectonic Tremors in Hyuga‐nada, Nankai Trough, Using the Newly Established N‐Net OBS Network
Core Problem: Shallow tremor activity along complex subduction interfaces is poorly observed offshore, limiting inference on stress transfer and megathrust preparation.
Key Innovation: Machine-learning detection on the new N-net ocean-bottom network reveals tremor episodes linked to nearby earthquakes, migration behavior, and fluid-rich plate-boundary segments.
43. Multi-method investigation of overlying strata stability and hazard mitigation in ultra-shallow-buried isolated mining faces with burned-out zones
Core Problem: Ultra-shallow isolated mining faces with burned-out zones combine brittle failure, fractured overburden, and water-conducting pathways in ways that are difficult to stabilize safely.
Key Innovation: FLAC3D, UDEC, and field bolt monitoring are integrated to trace evolving strata geometry, stress concentration, and water-conducting fractures, enabling targeted support and drainage measures.
44. An explainable correction and fusion framework for global bare-earth DTM generation in mountain areas
Core Problem: Vegetation and building biases degrade mountain DEMs precisely where terrain accuracy matters most for hazard analysis and process modelling.
Key Innovation: The authors pair explainable feature selection, a CNN-Transformer correction model, and ICESat-2-guided fusion to improve bare-earth terrain reconstruction in rugged mountains.
45. A framework for in‐stream jams formed by snow avalanche‐delivered large wood
Core Problem: Large-wood behavior in avalanche-prone drainages is poorly systematized despite its effect on channel blockage and geomorphic disturbance.
Key Innovation: The paper defines blanket-jam and transport-jam end members and links their geometry to avalanche frequency, slope roughness, and valley morphology.
46. A novel framework for analysing river planform characteristics in data‐scarce regions using remote sensing data
Core Problem: Data-scarce braided rivers lack integrated metrics for diagnosing morphological dynamics relevant to river management and hazard exposure.
Key Innovation: A Sentinel-2-based framework combines channel count, width, wetted area, migration area, and water occurrence to define seasonal and annual river-state diagnostics.
47. A time series deep learning algorithm for soil moisture retrieval from Sentinel-1 data using Transformer
Core Problem: Short-interval change-detection methods for Sentinel-1 soil-moisture retrieval break down when vegetation changes rapidly.
Key Innovation: A Transformer-based annual time-series model leverages seasonal vegetation evolution instead of assuming static cover, improving 200 m soil-moisture retrieval over cropland and grassland.
48. ChinaAI-FSC: a comprehensive AI-ready MODIS fractional snow cover dataset for China (2000–2022)
Core Problem: Long-duration, standardized snow-cover training data are scarce, limiting reproducible AI-based cryosphere monitoring.
Key Innovation: ChinaAI-FSC provides a large, quality-controlled, AI-ready fractional snow-cover sample library with harmonized labels, features, and benchmarking protocols for snow and hydrological applications.
49. Variational and Monte Carlo Methods for Bayesian Inversion of Dynamic Subsurface Flow Simulations Using Seismic and Fluid Pressure Data
Core Problem: Bayesian inversion of dynamic subsurface flow remains computationally expensive, especially when jointly constrained by seismic and pressure observations.
Key Innovation: The paper benchmarks variational and Monte Carlo inference schemes for CO2-storage-style inverse problems and shows that physically structured variational inference can balance posterior quality with tractable cost.