TerraMosaic Daily Digest: Mar 20, 2026
Daily Summary
This March 20, 2026 digest distills 40 selected papers from 904 analyzed records. The clearest signal is a move from static hazard description toward measurable state evolution. Landslide papers reconstruct failure as a timed process: a frost-heave-driven catastrophic slope collapse in Dingqing, a rainfall-triggered shallow-landslide inventory in Tuscany, an online probabilistic failure-time framework, and a multimodal spatiotemporal prediction model that explicitly separates spatial predisposition from temporal deformation.
A second concentration lies in subsurface and underground instability. Tunnel-face diagnosis, rock-mass fracture estimation ahead of excavation, induced-fault response to CO2 injection, hydraulic-fracture growth from acoustic emissions, blast damage, and cavern pressure limits are all framed through coupled hydro-mechanical or mechanics-guided models. Hydroclimatic studies extend the same logic: extreme precipitation is assessed through changing event frequency and structure, groundwater models are corrected under physical constraints, and cave-flood acoustics, snow-water-equivalent simulation, and seismic hydrology all turn indirect observations into operational signals.
Key Trends
Today's strongest papers do not merely identify hazardous places; they infer how hazardous systems evolve, when they accelerate, and which observables best constrain that evolution.
- Slope hazards are being treated as evolving systems: from frost-heave damage and rainfall-triggered shallow failures to probabilistic failure-time forecasting and deep-slip-surface inversion, state transition matters more than static class labels.
- Rockfall work is becoming mechanically and observationally richer: seismic co-rupture, irregular impact geometry, and topographic amplification are now explicit ingredients in both hazard interpretation and monitoring design.
- Underground instability is now decisively hydro-mechanical: TBM operations, fracture intensity, induced fault activation, blast damage, freeze–thaw retention, and cavern integrity all hinge on coupled flow–stress processes.
- Hydrological prediction is tightening its physical discipline: structural-error-aware groundwater models, event-structure diagnostics for precipitation, and acoustically informed karst flood monitoring all favor interpretability over pure black-box skill.
- Monitoring hardware and data streams are becoming more deployable: low-cost InSAR reflectors, anthropogenic seismic noise, acoustic cave sensing, MWD data, and fused fracture imaging broaden the range of hazard systems that can be tracked continuously.
Selected Papers
This digest features 40 selected papers from 904 papers analyzed.
1. 2025 Dingqing catastrophic landslide induced by frost heaving on high hillslopes in the southeast Tibetan Plateau, China
Core Problem: Frost-heave-triggered catastrophic landslides on high alpine hillslopes remain poorly understood despite the rising exposure of seasonal settlements in permafrost-transition terrain.
Key Innovation: Field investigation, remote sensing, UAV mapping, rock testing, and dynamic simulation are combined to show how freeze-thaw damage, reverse-slope structure, and long-term deformation culminated in the 2025 Dingqing failure.
2. Shallow landslides in Tuscany (Central Italy) triggered by the November 2023 heavy rainfall event
Core Problem: Climate-amplified intense rainfall is increasing shallow landslide losses, but event-specific inventories and infrastructure impacts are still incompletely documented.
Key Innovation: A combined field, satellite, and NDVI-change workflow maps 411 event landslides, links them to rainfall and predisposition factors, and quantifies direct disruption to buildings and road networks.
3. Online probabilistic forecast of landslide failure time via multi-information fusion
Core Problem: Failure-time prediction remains unreliable when acceleration onset, multi-source observations, and competing forecasting models are treated separately.
Key Innovation: The paper builds an ensemble probabilistic framework that fuses changepoint detection, multiple failure-forecast models, and multivariate observations into an online time-to-failure warning workflow.
4. Co-rupture mechanism and progressive failure of seismically triggered rockfalls on high-steep slopes
Core Problem: The co-rupture behavior of structurally controlled rockfalls under seismic excitation lacks direct physical evidence and engineering-scale stability criteria.
Key Innovation: Shaking-table tests and particle-flow simulations reveal a dynamic damage-feedback cycle and identify rear-edge fissure depth as a critical control on seismically triggered rockfall stability.
5. Landslide spatial prediction distinguishing spatial constraints and temporal trends
Core Problem: Conventional susceptibility models confound long-term spatial predisposition with short-term deformation trends, weakening their operational value.
Key Innovation: A graph-neural-network and Transformer fusion framework explicitly separates spatial constraints from InSAR-derived temporal dynamics, improving multimodal landslide prediction across mapping units.
6. Integrating InSAR surface deformation with sparse borehole data for three-dimensional inversion of deep-seated slip surfaces: Application to the Jungong landslide, Upper Yellow River
Core Problem: Mass-conservation-based inversion of deep-seated slip surfaces remains ill-posed when only surface deformation is available.
Key Innovation: The DSIM framework injects sparse borehole control into InSAR-driven inversion to recover a more physically consistent 3D slip-surface geometry and quantify river-blocking hazard under extreme failure.
7. Multi-field coupling mechanisms and chain-type failure evolution of mining-rainfall-induced landslides in karst mountain regions: Insights from physical modeling and 3DEC simulations
Core Problem: Mining-rainfall-induced landslides in karst terrain evolve through interacting structural and hydraulic pathways that are rarely resolved as a unified failure chain.
Key Innovation: Physical similarity experiments and 3DEC simulations quantify how cavity-guided fractures, rainfall-driven pore-pressure rise, and repeated mining reactivate instability through a staged disaster chain.
8. Interpretable susceptibility mapping of loess landslides integrating soil parameters using machine learning and SHAP: a case study from the loess plateau in western Shanxi, China
Core Problem: Machine-learning susceptibility models in loess terrain often omit regional soil parameters and remain physically opaque.
Key Innovation: Regional cohesion, friction, permeability, and collapsibility are integrated with environmental predictors, and weighted multi-model SHAP exposes both dominant controls and the physical meaning of the added soil variables.
9. Characterizing irregularly shaped rockfall dynamics using seismic signals: Geometric source effects combined with deep learning-driven strategy
Core Problem: Quantitative inversion of rockfall dynamics from seismic records is hindered by contact-geometry effects and nonlinear wave-propagation distortion.
Key Innovation: A hybrid physics-AI framework links irregular impact geometry to waveform characteristics through visco-elastoplastic source simulation and a CVAE propagation model, enabling mechanics-guided rockfall inversion.
10. Dynamic response of anti-dip thin-layered rock slopes under strong-motion earthquakes: Insights from shaking table tests and discrete element simulations
Core Problem: Anti-dip thin-layered rock slopes are highly vulnerable to toppling, but their dynamic thresholds and amplification patterns remain insufficiently quantified.
Key Innovation: Shaking-table and discrete-element analyses establish cracking and destabilization thresholds, quantify crest amplification, and show how slope height and bedding dip intensify seismic toppling risk.
11. Time-varying reliability analysis of a reservoir bank slope considering creep behavior and sequential Bayesian updating
Core Problem: Reservoir-bank stability evolves through creep and water effects, yet probabilistic risk frameworks often fail to update mechanical parameters consistently as new monitoring arrives.
Key Innovation: A sequential Bayesian-updating workflow ingests multi-point deformation time series to infer time-dependent mechanical parameters and track evolving slope failure probability.
12. Effect of high-frequency microseismicity on shear strength of interlayer structural planes of bedding rock landslides
Core Problem: Long-term slope degradation under high-frequency microseismicity is poorly characterized because most studies focus only on strong earthquake loading.
Key Innovation: Cyclic-loading analysis shows a three-stage evolution from short-term strengthening to fatigue damage and rapid deterioration, linking repeated microseismicity directly to structural-plane weakening.
13. Injection Induced Seismicity in Complex Fault Zone Architecture
Core Problem: Induced seismicity in realistic fault networks cannot be predicted reliably from single-fault assumptions because pore pressure, aseismic slip, and elastic interactions coevolve.
Key Innovation: Multi-cycle simulations in complex fault architectures show how stress transfer and confining stress organize emergent seismicity patterns, offering more realistic guidance for safer injection protocols.
14. Monitoring Near‐Surface Changes Using Anthropogenic Seismic Vibrations
Core Problem: Near-surface hydrological change is difficult to monitor continuously at high temporal resolution using conventional geophysical methods alone.
Key Innovation: Railway- and wind-turbine-generated vibrations are used to retrieve attenuation changes that respond more strongly than velocity to rainfall and pumping, establishing seismic attenuation as a practical hydrological proxy.
15. Widespread Co‐Location of Less Frequent and More Intense Daily Precipitation Over Land
Core Problem: It remains unclear where declining wet-day frequency and rising rainfall intensity are co-located, even though that combination directly affects runoff, drought, crops, and fire danger.
Key Innovation: Observation- and model-based analysis shows that fewer-but-stronger daily rainfall events are widespread across land surfaces, clarifying an important pathway by which warming reshapes hydrological risk.
16. Mean Biases Dominate CMIP6 Model Deficiencies in Simulating Heavy Rainfall During Monsoon Intraseasonal Oscillation
Core Problem: Climate models still misrepresent heavy-rainfall episodes during the monsoon intraseasonal oscillation because core moisture-transport biases remain unresolved.
Key Innovation: The study traces failed heavy-rainfall simulation to biased mean zonal winds and deficient low-level moisture advection, offering a concrete route for improving monsoon hazard modelling.
17. Impact of an Earthquake Event on Submarine Groundwater Discharge in Jeju Island, Korea
Core Problem: The coastal hydrologic consequences of earthquakes are rarely quantified directly, especially for submarine groundwater discharge and nutrient fluxes.
Key Innovation: Field observations after the 2021 Jeju earthquake show a tenfold short-term increase in submarine groundwater discharge, demonstrating that moderate seismic events can strongly perturb coastal hydrogeology.
18. A Data‐Driven Approach Coupled With Physical Constraints to Improve Groundwater Models With Structural Error
Core Problem: Data-driven correction of structurally biased groundwater models often improves fit at the expense of violating physical laws such as mass conservation.
Key Innovation: A new likelihood formulation couples data-driven error correction with explicit physical constraints, improving predictive accuracy while sharply reducing mass-balance violations and overfitting.
19. Increased Dependency on Extreme Precipitation in a Warmer Climate
Core Problem: It is not enough to know that extreme rainfall is intensifying; water systems also depend on how much of annual precipitation becomes concentrated in extremes.
Key Innovation: The Extreme Precipitation Dependency Index shows that many land regions could receive a much larger share of annual rainfall from extremes under warming, sharpening implications for water management and rain-fed agriculture.
20. Observation‐Constrained Physical Snow Water Equivalent Simulations Using a Physics‐Guided Machine Learning Approach
Core Problem: Physical snow models often misrepresent snow water equivalent anomalies and timing, especially in complex terrain.
Key Innovation: A hybrid CLM-LSTM framework constrained by observations substantially improves daily SWE magnitude, timing, and anomaly estimation, showing how physics-guided learning can serve as a snow emulator.
21. Towards the use of acoustic monitoring in karst networks to infer hydrological dynamics during flood events
Core Problem: Flood propagation in cave systems is difficult to observe because sensors are vulnerable to the very events they are meant to capture.
Key Innovation: Multiparametric cave monitoring reveals a plunger-chamber air-water mechanism and identifies repetitive low-frequency sounds that can precede flood rise, opening a path to cave-specific flood alerts.
22. An adaptive decomposition-denoising and temporal context fusion framework for multi-station water-level forecasting
Core Problem: Water-level forecasting across station networks remains vulnerable to noise, nonstationarity, and multi-step error accumulation.
Key Innovation: A decomposition-denoising and temporal-context fusion architecture improves both short- and medium-range multi-station water-level prediction, with the biggest gains at longer lead times that matter for flood warning.
23. Spatiotemporal evolution and hyetograph changes of global extreme precipitation
Core Problem: Flood risk depends not only on event intensity but on how precipitation is distributed before and after the peak, yet global hyetograph evolution remains underexplored.
Key Innovation: Global analysis of annual maximum 3-hour events shows declining temporal concentration and increasing total event rainfall in many regions, implying more flood-prone precipitation structures despite falling peak intensity in some areas.
24. Global cases of groundwater recovery after interventions
Core Problem: Groundwater depletion is often treated as effectively irreversible in highly stressed basins, obscuring what interventions can actually restore storage.
Key Innovation: This synthesis compiles global recovery cases to show that well-designed interventions can stabilize or reverse depletion, offering an evidence base for groundwater resilience planning.
25. Peat fires contribute disproportionately to Siberian fire carbon emissions
Core Problem: Carbon losses and persistence of peat fires in Siberia remain undercounted, weakening assessments of permafrost and fire-climate feedback risk.
Key Innovation: Thirty-meter burn-area and peatland maps show that peat fires occupy a smaller burned fraction but dominate carbon emissions, with overwintering fires and extreme weather emerging as key amplifiers.
26. Extreme precipitation reshapes nutrient flows and balance in North America’s largest river basin
Core Problem: Extreme rainfall does more than increase total nutrient export; it can reorganize nutrient ratios in ways that alter downstream ecological hazards.
Key Innovation: An integrated data-model framework for the Mississippi Basin shows that extreme precipitation preferentially boosts phosphorus export and shifts N:P balances toward conditions that can reshape bloom risk.
27. A low-cost passive corner reflector design for SAR and InSAR applications
Core Problem: Dense InSAR monitoring campaigns are often limited by the cost, weight, and installation drawbacks of standard metallic corner reflectors.
Key Innovation: Polycarbonate reflectors covered with aluminium tape achieve near-standard Sentinel-1 visibility at far lower cost and weight, including successful deployment on an active landslide site.
28. A novel soil moisture retrieval method via combining radiative transfer model and machine learning
Core Problem: Large-scale soil-moisture retrieval needs more interpretable relationships between satellite observations and moisture state than most black-box models provide.
Key Innovation: A radiative-transfer-informed Kolmogorov–Arnold Network derives explicit retrieval expressions from brightness temperature, surface temperature, and vegetation optical depth, improving interpretability without sacrificing global skill.
29. Precise diagnosis of mixed-face ground through distributed cutter working status monitoring data in slurry TBM
Core Problem: Sparse preconstruction boreholes often fail to capture the actual mixed-ground conditions encountered by slurry TBMs.
Key Innovation: Distributed cutter rotation-speed monitoring is translated into a face-diagnosis method that identifies mixed ground types in near real time and helps tune excavation parameters.
30. Data-driven estimation of surrounding rock areal fracture intensity with semi-supervised learning using measurement-while-drilling data
Core Problem: Fracture intensity ahead of the tunnel face is difficult to estimate with sufficient lead time using conventional geological investigation alone.
Key Innovation: A semi-supervised deep-learning framework links MWD, blasting, and construction data to areal fracture intensity, enabling ahead-of-face risk estimation with limited labeled observations.
31. Fault activity responses to CO2 injection revealed by similarity modeling
Core Problem: CO2 storage safety depends on fault behavior during injection, yet injection-induced fault response is difficult to observe directly at scale.
Key Innovation: Large physical similarity experiments with strain and temperature monitoring show that pore-pressure increase and preferential migration along faults are the dominant controls on fault activation during CO2 injection.
32. Deep learning-based temporal inference of hydraulic fracture length and branching using acoustic emission
Core Problem: Acoustic-emission monitoring records fracture evolution richly but translating those signals into incremental fracture geometry remains difficult.
Key Innovation: A spatiotemporal AE network predicts both cumulative fracture length and branching from historical and near-future AE patterns, providing a more direct route from monitoring signals to fracture-state estimates.
33. Prediction of rockburst risk induced by mine tremor using ensemble learning techniques
Core Problem: Rockburst early warning in deep mines is still hampered by class imbalance, qualitative microseismic interpretation, and weak adaptability to evolving tremor states.
Key Innovation: Bayesian-optimized ensemble models trained on sliding-window microseismic datasets improve high-risk classification and identify the most informative tremor indicators for dynamic rockburst warning.
34. Seismic behavior of pile group-supported bridges in liquefiable sloping ground covered with crusts: Insights from experimental observations
Core Problem: Bridge piles in liquefiable sloping ground with crust layers experience complex kinematic and inertial interactions that are poorly captured in simplified design assumptions.
Key Innovation: Large shaking-table experiments resolve how liquefaction stage, lateral spreading, and pile-group shadowing redistribute demand between leading and trailing piles during seismic loading.
35. A unified soil–water retention curve model for unsaturated soils considering freeze–thaw cycle effects
Core Problem: Freeze-thaw-driven pore evolution alters soil-water retention in ways that conventional SWRC models do not capture well.
Key Innovation: A unified bimodal retention-curve model links void-ratio evolution during freeze-thaw cycling to hydraulic behavior, improving both fit and physical interpretability in seasonally frozen soils.
36. Experimental study and mechanistic understanding of the characteristics of bentonite slurry infiltrating into unsaturated soil
Core Problem: Filter-cake formation and slurry loss in unsaturated ground are not governed the same way as in saturated TBM conditions, complicating tunnel-face stability assessment.
Key Innovation: Dedicated infiltration tests show how saturation controls mud spurt, pore-pressure evolution, and transition time to filter-cake formation, clarifying slurry behavior in unsaturated soils.
37. An improved constitutive model for predicting blast-induced damage zones and surrounding rock pressure
Core Problem: Ignoring tension-compression coupling reduces the reliability of damage-zone predictions around deep tunnel blasts.
Key Innovation: A strain-rate-dependent coupled damage model differentiates crushing and tensile fracture zones and shows how increasing in-situ stress can either suppress or amplify blast-induced damage.
38. A multi-device data fusion method for 3D rock fracture reconstruction: Development, comparison, and implications for hydro-mechanical simulations
Core Problem: No single imaging device can capture fracture void geometry accurately enough for reliable hydro-mechanical simulation.
Key Innovation: A local equivalent volume fusion method merges CT and surface scanning to reconstruct aperture fields that reproduce observed closure and seepage behavior more faithfully than single-device approaches.
39. Evaluating the permeability evolution in a granite fracture by short- and long-term flow-through tests and X-ray CT imaging
Core Problem: Permeability evolution in fractured granite under coupled thermal, hydraulic, mechanical, and chemical forcing is difficult to tie directly to internal structural change.
Key Innovation: Repeated CT imaging during short- and long-term flow-through tests shows how contact evolution and pressure dissolution drive progressive fracture sealing and conductivity loss.
40. Analytical solution for compressed air energy storage (CAES) caverns in deep elastic-brittle-plastic rock masses from construction to operation: Determination of operating gas storage pressure
Core Problem: Safe CAES operation depends on path-dependent elastoplastic rock response, yet pressure limits and lining demand remain hard to evaluate analytically across construction and operation phases.
Key Innovation: An analytical framework tracks historical stress effects across excavation, storage, and release stages to derive surrounding-rock response and operating-pressure limits for deep CAES caverns.