TerraMosaic Daily Digest: Mar 19, 2026
Daily Summary
This March 19, 2026 digest distills 47 selected papers from 1,391 analyzed records. The strongest contributions are process-resolved and directly usable. Debris-flow analysis in the Western Ghats combines corrected stereo DEMs, LiDAR, field data, and RAMMS to reconstruct entrainment and runout. New landslide-warning work stabilizes rock-slope alert levels by fusing creep-distribution indicators with path-dependent state correction. In the Three Gorges Reservoir area, a multi-slide bedding landslide is reconstructed as a staged geological system rather than a single failure surface.
A parallel cluster sharpens underground hazard forecasting. Bayesian tunnel-deformation models, stochastic excavation-risk surrogates, seepage-erosion mechanics, lined-cavern integrity simulations, and caprock destabilization studies all treat hydro-mechanical coupling as the central uncertainty. Hydroclimatic papers show the same methodological shift: extended-range rainfall forecasting, pluvial flood nowcasting, probabilistic flood ensembles, and runoff prediction under extremes are increasingly hybrid or physics-constrained. Across the day, the best AI papers are not generic; they are grounded by explicit mechanics, hydrology, or monitoring architectures.
Key Trends
The main movement today is from descriptive hazard research toward coupled, measurable, and operational geohazard modelling.
- Slope hazards are being recast as time-dependent systems: debris-flow reconstruction, rock-slope warning logic, bedding-landslide evolution, and rockfall inventory-runout modelling now resolve both mechanism and actionable state change.
- Underground instability work is unambiguously hydro-mechanical: tunnels, caverns, caprocks, suffusion, and fractured shales are analysed through coupled flow-deformation frameworks rather than isolated strength metrics.
- Forecast skill is improving through physical discipline: rainfall, flood, and runoff models that embed moisture budgets, process-driven baselines, or adaptive physical constraints now outperform less constrained alternatives.
- Hazard datasets are becoming infrastructure: high-resolution flood precipitation archives, prescribed-fire experiments, standardized earthquake catalogues, snow products, and InSAR thaw-slump inventories are being built for transfer and validation at scale.
- Resilience diagnostics are moving upstream of failure: hydrologic slowing down, drought-conditioned baseflow stability, and irrigation-induced aquifer depletion studies all aim to detect regime degradation before irreversible impacts accumulate.
Selected Papers
This digest features 47 selected papers from 1391 papers analyzed.
1. High‐resolution LiDAR and stereo‐DEMs for debris flow analysis in the Western Ghats, India: Sediment volume computation and run‐out modeling
Core Problem: Post-event debris-flow assessment still suffers from large uncertainty in sediment-volume estimation and runout reconstruction where topography is steep and vegetated.
Key Innovation: Stereo DEMs corrected with dense field observations and high-resolution LiDAR are fused with RAMMS to quantify entrainment, depletion depth, and sub-second runout evolution for the 2020 Pettimudi disaster.
2. Rock slope landslide early-warning level assessment using normal distribution theory and path-dependent effect
Core Problem: Conventional tangential-angle warning methods are unstable in real time because the secondary-creep velocity is difficult to estimate and warning levels oscillate unrealistically.
Key Innovation: Distributional indicators of creep evolution are combined with adaptive weighting, inverse-search updating, and path-dependent state correction to stabilize and operationalize warning-level assessment.
3. Evolution model of bedding landslide with multiple sliding zones in Jurassic strata in the Three Gorges Reservoir area
Core Problem: Multi-sliding-zone bedding landslides remain difficult to interpret because their structures, mineralogical weakening, and long-term evolution are strongly heterogeneous.
Key Innovation: Three-dimensional geological modelling, in situ monitoring, mineralogical evidence, and dating are integrated to reconstruct a multi-stage evolutionary model linking deep-seated deformation, mudstone weakening, and interacting slide bodies.
4. Characterizing geologic and climatic controls on rockfall hazards using an inventory and integrated kinematic and runout model: Skagway, Alaska, USA
Core Problem: Rockfall hazard mapping across steep, forested valleys is limited by the challenge of combining structurally controlled source susceptibility with propagation and seasonal triggering information.
Key Innovation: A 2005-2022 inventory, lidar-informed toppling source analysis, and large RAMMS:Rockfall ensembles jointly resolve geologic source controls, topographic runout sensitivity, and seasonal temperature-linked activity peaks.
5. Prediction of rigid rockfall barrier settlement under dynamic impact using numerical modelling and machine learning methodologies
Core Problem: Rapid design evaluation of rigid rockfall barriers is constrained by the high computational cost of nonlinear impact simulations.
Key Innovation: Abaqus-generated impact datasets are distilled into interpretable machine-learning surrogates, yielding fast settlement prediction with SHAP-based ranking of the controlling geotechnical parameters.
6. Revisiting collapse sinkholes in urban loess roads affected by dynamic seepage: Theoretical and numerical investigations
Core Problem: The feedback between dynamic loading and seepage in loess roadbeds is poorly represented, obscuring the mechanics of collapse-sinkhole initiation.
Key Innovation: The paper couples boundary-surface soil theory with an implicit finite-element implementation to reproduce sinkhole formation and identify vehicle-load-enhanced seepage as the principal trigger mechanism.
7. Intelligent early warning system for tunnel boring machine jamming under complex geology
Core Problem: TBM jamming is rare but high consequence, and abnormal-event scarcity makes supervised warning systems unreliable.
Key Innovation: An unsupervised U-shaped tabular autoencoder learns normal TBM behavior and uses sliding-window reconstruction deviation to issue robust early warnings for jamming under adverse geology.
8. Deformation risk assessment in braced tunnel excavations: a stochastic physics-based Bayesian Gaussian process model
Core Problem: Deep excavation risk assessment still struggles to combine field observations with physics-based simulation in a way that preserves uncertainty quantification.
Key Innovation: A Bayesian Gaussian-process framework fuses numerical predictions with monitoring data to greatly improve diaphragm-wall deflection prediction and deliver more decision-ready uncertainty bounds.
9. Bayesian updating of analytical model to predict time-dependent large deformation in soft rock tunnel
Core Problem: Time-dependent tunnel deformation in soft rock is hard to forecast because rheological parameters are uncertain and evolve with stress redistribution.
Key Innovation: Field monitoring is assimilated into a viscoelastic-viscoplastic analytical model using Bayesian updating, enabling credible-interval forecasts and a mechanically interpretable plastic-zone evolution signal.
10. CFD-DEM investigation into multi-mode governing equations of underground seepage erosion: Development of suffusion critical equation
Core Problem: Continuum-scale seepage-erosion equations remain poorly founded because laboratory tests do not capture the full diversity of erosion modes.
Key Innovation: CFD-DEM experiments identify four governing erosion modes and derive a suffusion critical equation compatible with continuum theory, helping close the mechanics of underground seepage failure.
11. Physics‐Constrained Network for Enhanced Extended‐Range Precipitation Forecasting in East Asia
Core Problem: Extended-range rainfall forecasting in East Asia remains limited by black-box deep models that lack physical stability and interpretability at long lead times.
Key Innovation: PS2F-Net embeds vertically integrated moisture-flux physics in a hierarchical attention network, improving two-week heavy-rain detection for both monsoonal and typhoon-driven extremes.
12. An attention-based super-resolution model for high-accuracy urban pluvial flood forecasting in coastal megacities
Core Problem: Operational pluvial-flood warning in dense coastal cities requires high-resolution inundation forecasts without the heavy data demands of full fine-grid modelling.
Key Innovation: A coarse-grid hydrodynamic workflow is super-resolved with CBAM-GAN to 10 m inundation maps in under a minute, retaining strong physical consistency even under rainfall beyond the training distribution.
13. A hybrid method coupling physical process-driven model with generative deep learning for probabilistic flood forecasting
Core Problem: Deterministic hydrological models and purely data-driven models each struggle to balance interpretability, forecast skill, and uncertainty quantification for flood warning.
Key Innovation: A process-based XAJ model is coupled with a conditional diffusion model that learns forecast-error structure, producing more accurate and better calibrated probabilistic flood ensembles.
14. Fusing dynamic physical constraints with PINN-xLSTM to enhance accuracy and physical consistency in runoff prediction under extreme hydrological events
Core Problem: Runoff prediction under extreme hydrological events remains vulnerable to abrupt regime shifts and physically inconsistent deep-learning behavior.
Key Innovation: The model introduces dynamic scenario-aware physical constraints into a PINN-xLSTM architecture, sharpening flood-peak behavior while reducing hydrologic-rule violations across basins.
15. A phase field-based comprehensive numerical framework for lined rock caverns under varying structural and geological conditions
Core Problem: Gas tightness in lined rock caverns is controlled by coupled cracking, interface slip, and host-rock heterogeneity that conventional simulation schemes simplify too strongly.
Key Innovation: A phase-field framework unifies crack initiation, propagation, lining plasticity, frictional interfaces, and random material fields to define quantitative safety envelopes for cavern depth and pressure.
16. Microscopic damage and permeability evolution of prefabricated fractured shale under hydro-mechanical coupling after heat treatment
Core Problem: The coupled influence of heat treatment, confining pressure, and pre-existing fractures on shale pore-network evolution remains incompletely characterized.
Key Innovation: CT-based pore reconstruction reveals how temperature-dependent fracture-network reorganization controls porosity and permeability trajectories under hydro-mechanical loading.
17. Destabilization of shale caprock by cold CO2 injection: Implications to caprock integrity for geological carbon storage
Core Problem: Caprock failure during geological carbon storage is often treated too simply, despite multiple thermo-hydro-mechanical mechanisms acting during cold CO2 injection.
Key Innovation: The study identifies stress arching, caprock deformation, and undrained cooling as distinct destabilization pathways and shows how faulting regime governs their combined integrity impact.
18. Permeability of caprock in aquifer underground gas storage under impact disturbance: Experimental study and predictive modeling
Core Problem: Dynamic disturbance can sharply alter caprock permeability, but post-impact transport evolution in saturated sealing rocks is poorly quantified.
Key Innovation: Impact tests, CT imaging, and dual-porosity modelling are combined to predict permeability-growth factors and show how saturation amplifies damage while liquid films suppress absolute permeability.
19. A novel nonlocal damage model for fluid-driven mixed-mode fracture propagation in poroelastic medium
Core Problem: Hydraulic fracture models that neglect mixed-mode damage misrepresent both required pressure and evolving fracture geometry in poroelastic media.
Key Innovation: A gradient-enhanced nonlocal damage framework independently tracks tensile and shear driving forces, demonstrating why mixed-mode formulations are necessary for predictive fluid-driven fracture modelling.
20. Numerical fragility assessment of monopile bridges subjected to different subduction-related earthquake mechanisms, considering soil-structure interaction
Core Problem: Bridge fragility in subduction settings is sensitive to rupture mechanism, yet this source dependence is rarely quantified with full soil-structure interaction.
Key Innovation: Cloud-based fragility analysis on nonlinear bridge models and real Chilean records distinguishes how crustal, interplate, and intraplate events shift damage probability across bearings and columns.
21. Integrating GNSS and Hydrological Data to Understand Seasonal Microseismicity at La Soufrière de Guadeloupe
Core Problem: Seasonal modulation of volcano-tectonic microseismicity is difficult to separate from endogenous unrest at active hydrothermal volcanoes.
Key Innovation: Five years of seismicity, GNSS strain, rainfall, and inferred lake-level dynamics are fused in a poroelastic interpretation that attributes much of the seasonal signal to pore-pressure forcing.
22. Critical Slowing Down Reveals Hydrologic Resilience Loss Across Amazon Sub‐Basins
Core Problem: Large-basin resilience loss in the Amazon is hard to diagnose before severe hydrologic disruption becomes obvious.
Key Innovation: A multi-metric critical-slowing-down framework applied to precipitation, evapotranspiration, soil moisture, and streamflow isolates coherent sub-basin signals of declining recovery capacity.
23. Dynamical linkages between planetary boundary layer schemes and wildfire spread processes: a case study using WRF-Fire version 4.6
Core Problem: Boundary-layer parameterization remains a major source of uncertainty in coupled wildfire-atmosphere simulations.
Key Innovation: A scheme intercomparison against station observations during the Jinyun Mountain wildfire shows how turbulence representation alters fire-driven wind response and identifies MYNN3 as the most responsive option.
24. InSAR Deformation Analysis and Response Mechanism of Retrogressive Thaw Slumps in Permafrost Region of the Qinghai–Tibet Engineering Corridor
Core Problem: The spatiotemporal evolution and forcing hierarchy of thaw slumps along critical high-altitude infrastructure corridors remain insufficiently resolved.
Key Innovation: SBAS-InSAR, optical remote sensing, and causal-network analysis identify 238 thaw slumps and separate the aspect-dependent roles of precipitation, land-surface temperature, and vegetation in stepwise subsidence.
25. StageIV-IRC: a high-resolution dataset of extreme orographic Quantitative Precipitation Estimates (QPE) constrained to water budget closure for historical floods in the Appalachian Mountains
Core Problem: Historical flood analysis in complex terrain is hindered by coarse and biased precipitation estimates that fail to close the event-scale water budget.
Key Innovation: StageIV-IRC corrects radar precipitation using physically based inverse rainfall correction and initial-condition adjustment, yielding a 250 m event-scale QPE resource for 215 Appalachian flood-producing storms.
26. A field dataset from replicated prescribed fire experiments on wildland fire behaviour and fire–atmosphere interactions
Core Problem: Process-level fire-atmosphere datasets remain sparse at the scale needed to bridge laboratory combustion studies and operational fire modelling.
Key Innovation: Thirty-five replicated prescribed burns provide synchronized fuel, wind, temperature, radiative power, gas, and LiDAR measurements, creating a benchmark dataset for model evaluation in wildfire science.
27. Assessing the ability of the ECMWF seasonal prediction model to forecast extreme September–November rainfall events over Equatorial Africa
Core Problem: Operational services in Equatorial Africa need clearer evidence on how well seasonal models capture the circulation drivers of extreme SON rainfall.
Key Innovation: A 43-year evaluation shows where ECMWF-SEAS5.1 reproduces key teleconnections and moisture-flux patterns, clarifying both the skill and the amplitude biases of extreme-rainfall forecasts.
28. A probabilistic retention model for fine sand infiltration in granular filters based on experimental and numerical insights
Core Problem: Filter remediation for eroded gap-graded soils requires a predictive description of how injected fines migrate and are retained within a coarser skeleton.
Key Innovation: Experiments and numerical simulations support a micro-mechanical probabilistic model for infiltration distance that is tied directly to constriction-size distributions in the receiving granular medium.
29. Comparison of volumetric threshold strain, pore-pressure threshold strain and stiffness-degradation threshold strain of saturated coral sand under complex cyclic loading
Core Problem: Threshold-strain definitions used in liquefaction and cyclic-response analysis are often treated as interchangeable despite different physical meanings.
Key Innovation: Systematic cyclic tests on saturated coral sand show that volumetric, pore-pressure, and stiffness-degradation thresholds diverge strongly with relative density and material type.
30. Deep multistrutted excavation-induced soil stress relief and modulus attenuation behind retaining wall considering soil arching effect
Core Problem: Soil-arching-driven stress redistribution behind retaining walls remains computationally demanding and hard to represent in practical deep-excavation assessment.
Key Innovation: A stress-modulus attenuation model calibrated from multiple field cases captures soil stress relief, earth pressure evolution, and shear-wave-velocity change with low computational cost.
31. Water scarcity indicator based on GRACE derived total water storage for fast water scarcity monitoring
Core Problem: Existing global water-scarcity indicators are often model-heavy, data-intensive, and difficult to reproduce quickly.
Key Innovation: A GRACE-based total-water-storage indicator reproduces the population and area exposure patterns of several blue-water scarcity metrics while using a simpler and more holistic storage framework.
32. Transient snow line altitudes of glaciers in the European Alps from multi-mission remote sensing data (2000–2025)
Core Problem: Long glacier snow-line records are difficult to maintain continuously across sensors and changing illumination conditions.
Key Innovation: An automated multi-mission segmentation workflow produces roughly 200,000 glacier snow-line observations for the European Alps, revealing a strong regional snow-line rise since 2000.
33. Global Trends in Day‐Night Compound Thermal Extremes in Lakes
Core Problem: Day-night compound thermal extremes in lakes remain under-characterized despite their persistence and ecological significance.
Key Innovation: Hourly global lake-temperature records show that sustained day-night compound extremes are far more frequent and persistent than daytime-only or nighttime-only events and are intensifying under warming.
34. Geomorphic adjustment and reach-scale sensitivity of the highly braided Kosi River: A dynamic systems approach to resilience-based management
Core Problem: Braided-river management needs spatially explicit identification of reaches that are most sensitive to threshold exceedance and channel migration.
Key Innovation: Cloud-based hydro-morphological analysis over three decades identifies temporally scale-dependent sensitivity classes and converts them into a resilience-oriented management framework for the Kosi River.
35. A century of channel morphological adjustments in the western Bengal Basin: insights from the Mayurakshi River in Eastern India
Core Problem: Long-term river adjustment under the combined pressure of dams, floods, mining, and land-use change is rarely reconstructed with phase-specific morphodynamic metrics.
Key Innovation: Historical maps and remote sensing are synthesized into five adjustment phases, revealing how dams and later flood sequences reorganized planform, active bars, and geomorphic status.
36. Modelling the geomorphological evolution of coastal wetlands under rising sea levels: A reduced dimensional and multi-temporal evaluation
Core Problem: Decision makers need sea-level-rise projections for coastal wetlands, but it is unclear when reduced-complexity models are sufficient.
Key Innovation: Multi-temporal evaluation shows that simple zero-dimensional models can match observed decadal wetland response as effectively as more complex alternatives when sediment supply remains adequate.
37. Multi-scale factors controlling fine sediment interstitial storage along a river continuum with contrasted bedload transport conditions
Core Problem: Fine-sediment clogging in altered gravel-bed rivers is controlled by multiple scales of forcing that are rarely disentangled in a systematic field design.
Key Innovation: A three-year Rhône survey separates station-scale hydraulic controls from reach-type effects and shows how armored reaches retain ecologically problematic interstitial fine-sediment loads.
38. Assessing infiltration dynamics using integrated hydrogeophysical monitoring in a managed aquifer recharge pond
Core Problem: Managed aquifer recharge efficiency is limited by poorly resolved infiltration pathways, evolving saturation patterns, and clogging development.
Key Innovation: Time-lapse DCIP, GPR, and hydrological monitoring reveal lateral high-permeability pathways, localized mounding, and progressive clogging, supporting adaptive inflow control in recharge ponds.
39. Interplay of Tectonics and Topography Facilitated Sudden Dyke Intrusion in 2022 at São Jorge Island, Azores
Core Problem: Why magma ascent localizes beneath islands within a broad extensional plate-boundary zone remains uncertain in hotspot-rift settings.
Key Innovation: A 3D finite-element model shows how tectonic stretching and topographic loading together create stress conditions that favored the rapid 2022 dyke intrusion beneath São Jorge.
40. Integrated and standardised earthquake catalogue of Indonesia (88 AD–2024)
Core Problem: Indonesia's seismic records are fragmented across historical, national, and international catalogues with inconsistent standards and magnitude reporting.
Key Innovation: The study harmonizes these sources into a standardized long-duration earthquake catalogue designed for direct use in seismic-hazard assessment and tectonic analysis.
41. Changes in irrigation practices may deplete aquifers faster and more severely than meteorological droughts: A numerical modeling approach
Core Problem: Climate adaptation in irrigated agriculture can unintentionally intensify aquifer depletion when shifts in irrigation practice reduce return-flow recharge.
Key Innovation: A calibrated 3D groundwater model shows that switching from surface irrigation to drip irrigation can erode storage and groundwater-dependent ecosystems more severely than a two-year meteorological drought.
42. Vegetation restoration enhances baseflow stability during hydrological droughts on the Loess Plateau, China
Core Problem: The effect of large-scale vegetation restoration on baseflow stability during drought is still debated in ecologically fragile basins.
Key Innovation: Long streamflow and land-use records across six catchments show that post-2000 forest and grassland expansion increased baseflow stability and buffered drought-period hydrological variability.
43. Remobilisation of fine sediment from chalk stream gravel beds under flushing flows: A flume experiment
Core Problem: Fine-sediment cleanout from chalk-stream gravel beds is poorly constrained, limiting management targets for ecologically sensitive streams.
Key Innovation: Flume tests quantify how bed shear stress governs cleanout depth and reveal why existing remobilisation models overpredict flushing in chalk-stream settings.
44. Assessing the global performance of a parsimonious soil temperature model for frozen ground prediction
Core Problem: Frozen-ground representation in hydrological models is often too computationally expensive for broad operational deployment.
Key Innovation: A parsimonious model driven only by air temperature and snow cover achieves strong global frozen-ground classification skill, offering a lightweight option for distributed hydrological applications.
45. A novel adaptive soil moisture retrieval method via stacked ensemble learning and a local Bayesian dynamic algorithm
Core Problem: High-resolution soil-moisture inversion remains difficult across heterogeneous soils, environments, and topography.
Key Innovation: A dynamically weighted stacking ensemble with Bayesian optimization improves multilayer soil-moisture retrieval while explicitly balancing error, complexity, and stability.
46. Predicting surface soil moisture of northern peatlands from hyper- and multispectral satellite data
Core Problem: Peatland hydrological monitoring requires moisture estimates that are sensitive enough to detect shifts in carbon-critical wetness conditions.
Key Innovation: Wavelet-transformed hyperspectral EnMAP data outperform multispectral alternatives for peatland surface-moisture retrieval, showing the value of spectral richness over spatial resolution.
47. ChinaAI-FSC: a comprehensive AI-ready MODIS fractional snow cover dataset for China (2000–2022)
Core Problem: Long-duration, standardized snow-cover training data remain a bottleneck for reproducible AI-based cryosphere monitoring.
Key Innovation: ChinaAI-FSC assembles a large, quality-controlled sample library with harmonized labels, features, and evaluation protocols for snow-cover mapping and cryosphere-hydrological modelling.