Initiated by Dr. Xin Wei, University of Michigan
Ongoing development by the community

TerraMosaic Daily Digest: April 22, 2026

April 22, 2026
TerraMosaic Daily Digest

Daily Summary

This April 22, 2026 digest distills 34 selected papers from 1,805 analyzed records. This set is unusually concrete. Several of the strongest papers replace broad hazard classes with measurable failure-state variables: uncertain landslide source terms become impulse-wave hazard zoning, UAV-derived fissures are ranked by their likelihood of evolving into slip surfaces, tailings liquefaction is expressed through diffusion-controlled localization, avalanche susceptibility is sharpened by modeled snow mechanics, and sinkhole detection is stabilized through synthetic physics-based examples. The scientific gain is not simply better classification. It is a tighter description of what is destabilizing, how that instability propagates, and which observables are credible enough to support intervention.

A second concentration is warming-sensitive infrastructure and flood consequence. Permafrost papers range from probabilistic thaw-settlement envelopes for the Qinghai-Xizang corridor to rail-speed instability thresholds, tower-induced talik growth, thermosyphon retrofits, long-term embankment settlement, and rapid retreat of Arctic retrogressive thaw slumps. Flood work is equally specific but organized differently: one group identifies where exposure is concentrated along the U.S. Gulf and Atlantic coasts, another explains observed damage patterns with interpretable drivers, and another builds systems that can forecast or mitigate inundation in real time. Rockburst and karst studies fit the same overall direction by focusing on operational signals—occurrence time, compound-hazard zoning, and subsurface collapse geometry—rather than generic risk labels.

Key Trends

The most interesting work today is not just predictive. It changes what can be measured, explained, or acted on before failure becomes obvious.

  • Failure is being parameterized, not just localized: Impulse-wave inversion, crack significance scoring, avalanche snow mechanics, tailings liquefaction bandwidth, and permafrost settlement envelopes all focus on the variable that governs escalation instead of merely assigning hazard classes.
  • Cold-region papers read like a connected engineering sequence: They move from thaw-slump acceleration to corridor-scale rail and highway instability, then to foundation diagnosis, embankment cooling, long-term settlement, and parameter inversion.
  • Flood research is organized around decisions rather than footprints: Today's papers identify exposed coastal cities, explain observed damage patterns, build real-time inundation surrogates, and test whether integrated or transferable adaptation frameworks could actually be deployed.
  • Signal quality is treated as a scientific problem in its own right: Synthetic sinkholes, combined-array karst surveys, crack ranking for landslides, and cleaned or timed microseismic signals all reduce the ambiguity that usually weakens warning systems.
  • The AI contributions are strongest when they clarify uncertainty: GAN-assisted inversion, generative multi-hazard simulation, conformal anomaly scoring, climate-model stress testing, and explainable damage mapping make unstable systems easier to inspect rather than simply harder to benchmark against.

Selected Papers

This digest features 34 selected papers from 1,805 papers analyzed, led by state-resolving failure diagnosis, thaw-sensitive infrastructure thresholds, and consequence-aware flood intelligence.

1. Active learning framework for landslide-induced impulse wave risk assessment using GAN-enhanced kriging and adaptive cooperative PSO

Source: Engineering Geology Type: Hazard Modelling Geohazard Type: Landslide-induced impulse waves Relevance: 8/10

Core Problem: Impulse-wave hazard chains remain hard to assess when landslide-source parameters are uncertain and full physics simulation is too expensive for probabilistic shoreline zoning.

Key Innovation: The paper couples GAN-expanded samples, multi-output Kriging, and adaptive cooperative PSO around r.avaflow to invert parameter distributions and produce rapid five-level impulse-wave hazard zoning.

2. Spatial-temporal evolution of landslide cracks revealed by UAV photogrammetry: The FCS-YOLO model for precise crack extraction under complex backgrounds

Source: International Journal of Applied Earth Observation and Geoinformation Type: Detection and Monitoring Geohazard Type: Landslides Relevance: 8/10

Core Problem: Landslide crack monitoring still breaks down where complex loess backgrounds and scale variability hide subtle precursory fissure evolution.

Key Innovation: FCS-YOLO improves UAV crack extraction and pairs it with HRCSI to identify the cracks most likely to evolve into slip surfaces, linking mapped fissures directly to deformation dynamics.

3. Fully coupled flow-liquefaction in tailings storage facilities: Induced shear bandwidth on partially drained triggers

Source: Computers and Geotechnics Type: Concepts & Mechanisms Geohazard Type: Tailings dam liquefaction Relevance: 8/10

Core Problem: Predicting partially drained liquefaction in tailings dams has long been weakened by mesh-dependent strain localization and poorly resolved diffusion effects.

Key Innovation: The study shows that fully coupled hydro-mechanical formulation introduces a physical diffusion length scale, regularizes localization, and yields scalable criteria for liquefaction triggering in TSFs.

4. Integrating snowpack mechanical properties into snow avalanche susceptibility mapping in continental dry–cold mountain regions

Source: Geomatics, Nat. Haz. & Risk Type: Hazard Modelling Geohazard Type: Snow avalanches Relevance: 8/10

Core Problem: Avalanche susceptibility maps rarely include snowpack mechanical variables, so they miss the instability processes that actually control dry-cold avalanche release.

Key Innovation: The paper integrates WRF-derived snow shear strength and density into machine-learning susceptibility mapping and uses SHAP to show how weak snowpack states reorganize avalanche-prone terrain.

5. Physics-informed synthetic data and transformer-based segmentation for sinkhole detection in railway LiDAR point clouds

Source: ISPRS J. Photogrammetry Type: Detection and Monitoring Geohazard Type: Sinkholes Relevance: 8/10

Core Problem: Railway sinkhole detection suffers from sparse real examples and high false positives in LiDAR-based monitoring.

Key Innovation: This framework combines physics-informed synthetic sinkhole generation with a SuperPoint Transformer and geometry-aware refinement to achieve operationally robust 3D sinkhole segmentation.

6. A tale of two coasts: Unveiling US Gulf and Atlantic coastal cities at high flood risk

Source: Science Advances Type: Risk Assessment Geohazard Type: Floods Relevance: 8/10

Core Problem: Large-scale coastal flood assessment often identifies hazardous zones without quantifying where people and infrastructure are concentrated under different damage regimes.

Key Innovation: This Science Advances study uses historical flood-damage data and geospatial AI to rank high-risk Gulf and Atlantic coastal cities, estimate exposed populations and buildings, and isolate the dominant drivers of general versus extreme flood damage.

7. Prediction of rockburst occurrence time: Implications from microseismic monitoring and deep learning

Source: JRMGE Type: Early Warning Geohazard Type: Rockbursts Relevance: 8/10

Core Problem: Deep-tunnel rockburst warning remains limited because monitoring systems rarely predict occurrence time rather than only intensity or probability.

Key Innovation: The study fuses waveform imagery and temporal amplitude evolution through CNN-BiLSTM with cross-attention to estimate rockburst timing hours in advance from microseismic data.

8. Identifying zones of rockburst-collapse compound hazards in deep tunnels

Source: JRMGE Type: Risk Assessment Geohazard Type: Rockburst-collapse compound hazards Relevance: 7/10

Core Problem: Compound rockburst-collapse zones are harder to anticipate than single hazards because fractured and intact rock interact under concentrated stress ahead of excavation.

Key Innovation: The paper builds a practical indicator from TSP-derived P-wave velocity, GSI, BQ, and strength-stress ratios to delineate segment-scale compound hazard zones before excavation reaches them.

9. Electrical resistivity tomography with combined arrays for shallow karst collapse hazard identification in Fuchuan, Guangxi, China

Source: Frontiers in Earth Science Type: Detection and Monitoring Geohazard Type: Karst collapse Relevance: 7/10

Core Problem: Shallow covered karst collapse is difficult to zonate when single-array geophysics resolves either vertical layering or lateral boundaries but not both.

Key Innovation: The study shows that combined Wenner-Schlumberger ERT improves anomaly resolution and delineates validated karst-collapse hazard zones for engineering management.

10. Probabilistic predictions of the impacts of climate warming on permafrost stability along the Qinghai-Xizang Railway, China

Source: Cold Regions Sci. & Tech. Type: Hazard Modelling Geohazard Type: Permafrost instability Relevance: 7/10

Core Problem: Long-horizon railway risk in warming permafrost terrain remains uncertain because climate, soil-ice, and settlement uncertainty are rarely propagated together.

Key Innovation: The paper couples downscaled climate projections, Monte Carlo temperature ensembles, and hydrothermal-ground models to derive confidence-bounded thaw settlement and stability classes along the corridor.

11. Instability of railway tracks in permafrost regions, targeted for failure under modern climate conditions, with operational vibrations of trains

Source: Cold Regions Sci. & Tech. Type: Hazard Modelling Geohazard Type: Permafrost infrastructure instability Relevance: 7/10

Core Problem: Track safety in thawing permafrost cannot be evaluated from climate or vibration loads alone because failure emerges from their coupled action.

Key Innovation: The study links THM degradation to realistic train loading and identifies operational speed thresholds at which thaw-weakened subgrades cross unacceptable deformation and ride-stability limits.

12. δHT4P: A differentiable physical modeling framework for thermal evolution of permafrost

Source: Computers and Geotechnics Type: Hazard Modelling Geohazard Type: Permafrost instability Relevance: 7/10

Core Problem: Permafrost forecasting remains limited when thermodynamic parameters are fixed rather than inferred from sparse observations.

Key Innovation: This differentiable framework embeds a CNN inside the heat equation to learn depth-dependent constitutive properties while supporting both temperature prediction and parameter inversion across Arctic boreholes.

13. Tower foundation-induced through Talik characteristics in warm permafrost terrains: The case of power lines in Jagdaqi, Northeast China

Source: Engineering Geology Type: Detection and Monitoring Geohazard Type: Permafrost degradation Relevance: 7/10

Core Problem: Infrastructure-induced talik development beneath foundations is still weakly characterized in warm permafrost terrains, limiting transmission-line risk assessment.

Key Innovation: Field surveys and ERT reveal how adjacent tower foundations accelerate deep through-talik formation in wetland permafrost and show that charred-debris backfilling materially suppresses thaw growth.

14. Laboratory study on a new enhancing method for an in-service crushed-rock embankment underlain by permafrost

Source: Cold Regions Sci. & Tech. Type: Mitigation Geohazard Type: Permafrost infrastructure instability Relevance: 7/10

Core Problem: Aging crushed-rock embankments lose cooling capacity over time, but non-excavation retrofit options for active corridors remain limited.

Key Innovation: The paper demonstrates that thermosyphon insertion can restore convective cooling and increase cold-season heat release in in-service permafrost embankments without excavation.

15. An integrated computational framework for high-dimensional parameter optimization in coupled hydro-thermal permafrost modelling

Source: Cold Regions Sci. & Tech. Type: Hazard Modelling Geohazard Type: Permafrost instability Relevance: 7/10

Core Problem: Coupled hydro-thermal permafrost models remain hard to calibrate because vertical heterogeneity introduces many interacting parameters.

Key Innovation: This framework automates GEOtop-based high-dimensional parameter identification and sensitivity analysis, producing reproducible cold-region calibration at borehole scale.

16. Retrogressive thaw slumps of Novaya Sibir’ Island (East Siberian Sea): activity increase under unique ground ice conditions

Source: Cold Regions Sci. & Tech. Type: Detection and Monitoring Geohazard Type: Thermokarst hazards Relevance: 7/10

Core Problem: High-Arctic thaw disturbances are expanding rapidly, but detailed site-scale evidence remains sparse for the ice-rich settings where retrogressive thaw slumps can accelerate most strongly.

Key Innovation: Using 16 years of high-resolution remote sensing, the paper quantifies exceptionally fast headwall retreat rates and links them to massive ground ice and rising thawing degree-days in a rapidly warming coastal Arctic setting.

17. Resilience assessment of an integrated green-grey-blue-control system for urban flood risk mitigation under changing climate and urbanization

Source: Journal of Hydrology Type: Mitigation Geohazard Type: Floods Relevance: 7/10

Core Problem: Single-class flood interventions are increasingly insufficient under compounding urbanization and extreme rainfall.

Key Innovation: The paper proposes and optimizes an integrated green-grey-blue-control system that combines structural and non-structural regulation and sharply reduces inundation and economic losses across design storms.

18. Designing an inundation monitoring and real-time urban flood forecasting system: a synthetic study

Source: Journal of Hydrology Type: Early Warning Geohazard Type: Urban floods Relevance: 7/10

Core Problem: Real-time urban inundation forecasting remains constrained by the computational cost of physics-based models and sparse observations.

Key Innovation: This synthetic framework uses optimized sensor placement and a spatiotemporal ViT surrogate to reconstruct inundation fields and deliver 1–12 h forecasts in seconds.

19. Assessing the potential of implementing room for the river in Taiwan to enhance flood resilience

Source: IJDRR Type: Mitigation Geohazard Type: Floods Relevance: 7/10

Core Problem: Translating flood-resilience concepts between governance systems is difficult even when hydraulic benefits are plausible.

Key Innovation: The study couples dyke-relocation simulations with governance diagnosis to evaluate how a Room for the River approach could be adapted to Taiwan rather than copied mechanically.

20. Simulating spatial multi-hazards with generative deep learning

Source: NHESS Type: Hazard Modelling Geohazard Type: Multi-hazards Relevance: 7/10

Core Problem: Sampling realistic spatially dependent multi-hazard extremes remains difficult for conventional high-dimensional statistical models.

Key Innovation: The paper blends generative adversarial networks with extreme value theory to simulate spatial co-occurrence of wind, rainfall, and pressure extremes while preserving multivariate extremal dependence.

21. Flood Susceptibility Mapping and Runoff Modeling in the Upper Baishuijiang River Basin, China

Source: Remote Sensing (MDPI) Type: Hazard Modelling Geohazard Type: Floods Relevance: 7/10

Core Problem: Mountain flood susceptibility studies often map prone areas statistically without tying them back to runoff processes.

Key Innovation: This study links PSO-MaxEnt susceptibility mapping to HEC-HMS runoff simulation to show that flood-prone corridors follow physically interpretable infiltration and flow-concentration controls.

22. An Explainable Machine Learning Framework for Flood Damage Mapping Using Remote Sensing and Ground-Based Data: Application to the Basilicata Ionian Coast (Italy)

Source: Remote Sensing (MDPI) Type: Risk Assessment Geohazard Type: Flood damage Relevance: 7/10

Core Problem: Hydraulic hazard maps do not reliably reproduce observed flood damage patterns or reveal which exposure variables are actually driving loss.

Key Innovation: The paper uses observed damage records, multi-source geospatial factors, XGBoost, and SHAP to map flood-damage susceptibility and expose threshold-like hazard-exposure controls in a Mediterranean coastal setting.

23. Assessing the Robustness of Climate Foundation Models under No-Analog Distribution Shifts

Source: ArXiv (Geo/RS/AI) Type: Concepts & Mechanisms Geohazard Type: General geohazard intelligence Relevance: 7/10

Core Problem: Climate foundation models may look accurate in-distribution while failing under forced states outside the historical analog range.

Key Innovation: The paper benchmarks ClimaX and baselines in a historical-only training regime to expose the accuracy-stability trade-off under temporal extrapolation and cross-scenario forcing shifts.

24. Adaptive Conformal Anomaly Detection with Time Series Foundation Models for Signal Monitoring

Source: ArXiv (Geo/RS/AI) Type: Detection and Monitoring Geohazard Type: General geohazard monitoring Relevance: 7/10

Core Problem: Foundation-model monitoring pipelines still lack calibrated anomaly scores that remain interpretable under distribution shift.

Key Innovation: This method adds adaptive conformal prediction on top of pretrained time-series models to produce p-value-like anomaly scores with stable false-alarm control and no fine-tuning.

25. Application of Data Processing Methods and Optimization Algorithms Based on Micro-seismic Monitoring in Short-Term Rockburst Prediction

Source: Rock Mech. & Rock Eng. Type: Early Warning Geohazard Type: Rockbursts Relevance: 7/10

Core Problem: Poor-quality microseismic datasets often limit short-term rockburst prediction more than model choice does.

Key Innovation: The paper builds a preprocessing-and-optimization workflow—imputation, outlier filtering, class balancing, and GWO-tuned ensembles—that lifts short-term rockburst warning accuracy and exposes key precursor features.

26. Strength reduction search algorithms for slope stability analysis using spectral finite element method: From ‘brittle’ to ‘ductile’ failure

Source: Computers and Geotechnics Type: Hazard Modelling Geohazard Type: Slope instability Relevance: 6/10

Core Problem: Three-dimensional slope-stability calculations remain expensive and can misidentify failure thresholds when brittle and ductile cases are treated the same way.

Key Innovation: The study combines spectral FEM with adaptive strength-reduction search and ductility indices to cut computation cost while preserving reliable factor-of-safety detection across failure styles.

27. A simple three-dimensional model to describe the viscoplastic behavior of soils

Source: Computers and Geotechnics Type: Concepts & Mechanisms Geohazard Type: Slope deformation and geotechnical failure Relevance: 6/10

Core Problem: Viscoplastic soil behavior is still hard to bring into routine 3D analysis because many constitutive models are cumbersome to calibrate and implement.

Key Innovation: This paper introduces a simpler 3D structured-soil viscoplastic model with laboratory-calibrable parameters and a finite-element-ready formulation for time-dependent geotechnical analysis.

28. Prediction of excavation damaged zone depth of chamber groups by damage initiation and spalling limit: Numerical simulation and machine learning algorithms

Source: JRMGE Type: Hazard Modelling Geohazard Type: Rock mass failure Relevance: 6/10

Core Problem: Chamber-group stability assessment is slow when excavation damaged zone depth must be resolved repeatedly with full numerical simulation.

Key Innovation: The paper combines DISL-based FLAC3D simulations with Bayesian-optimized XGBoost to predict HDZ and EDZ depths accurately and reveal the geometric controls on damage growth.

29. An enhanced three-dimensional reduced-order track model for predicting differential settlement in railway transition zones on stratified soils

Source: Transportation Geotechnics Type: Hazard Modelling Geohazard Type: Ground settlement Relevance: 6/10

Core Problem: Long-term settlement in railway transition zones is difficult to simulate because dynamic vehicle-track interaction and subgrade evolution operate together over large time spans.

Key Innovation: The paper builds a reduced-order 3D framework that couples dynamic contact forces and layered subgrade response to predict where differential settlement localizes near transitions.

30. Long-term embankment deformation of the Qinghai-Tibet Highway in permafrost regions

Source: Cold Regions Sci. & Tech. Type: Detection and Monitoring Geohazard Type: Permafrost instability Relevance: 6/10

Core Problem: Highway maintenance in permafrost zones still lacks long-duration field evidence tying thaw evolution to persistent embankment settlement patterns.

Key Innovation: The study uses multi-decadal monitoring from six sites to quantify thaw rates, settlement rates, and asymmetric deformation modes that track permafrost degradation beneath the highway.

31. A numerical method for frost heaving damage in open fractured rock mass under unidirectional freezing

Source: Computers and Geotechnics Type: Hazard Modelling Geohazard Type: Frost-heave damage Relevance: 6/10

Core Problem: Frost-heave damage in fractured rock remains difficult to predict because crack propagation, phase change, and water migration are tightly coupled.

Key Innovation: The paper develops a CZM-based numerical method that reproduces frost-heaving pressure evolution and crack growth, then uses it to analyze tunnel-scale cold-region damage.

32. Effect of bedrock outcrops and underground fissures on soil loss, flow hydraulics and hydrological connectivity characteristics of karst slopes in alpine canyon area

Source: Journal of Hydrology Type: Concepts & Mechanisms Geohazard Type: Erosion and karst slope instability Relevance: 6/10

Core Problem: Karst erosion models still underresolve how bedrock exposure and fissure connectivity reorganize runoff and soil-loss pathways.

Key Innovation: Rainfall simulations plus path analysis identify critical bedrock-fissure combinations that intensify hydraulic forcing and local erosion hotspots on karst slopes.

33. Integrating 10Be analyses and an empirical erosion model to unveil catchment-scale landscape and sediment dynamics in a tectonically active Mediterranean area (Calabria, southern Italy)

Source: Geomorphology Type: Concepts & Mechanisms Geohazard Type: Erosion and sediment hazards Relevance: 5/10

Core Problem: Source-to-sink hazard interpretation is weakened when millennial denudation and modern sediment export are studied on separate timescales.

Key Innovation: The paper combines cosmogenic 10Be with an empirical erosion model to show how tectonic uplift, storage, and connectivity decouple hillslope denudation from channel incision and coastal sediment delivery.

34. Stability of EICP–ASKG-reinforced aeolian sand along desert highways under the combined effects of plant germination and wind erosion

Source: Transportation Geotechnics Type: Mitigation Geohazard Type: Wind erosion and embankment instability Relevance: 5/10

Core Problem: Bio-cemented sand reinforcement for desert infrastructure is rarely evaluated under the combined disturbance of vegetation emergence and ongoing wind erosion.

Key Innovation: The study tests an EICP–ASKG reinforcement scheme under coupled germination and erosion effects to assess whether treated aeolian sand can retain stability along desert highways.