TerraMosaic Daily Digest: May 17, 2026
Daily Summary
May 17's geohazard papers are strongest where they convert weak precursor signals or difficult subsurface states into measurable process variables. The landslide-relevant core is compact but substantive: a preferential-flow thermo-hydro-mechanical model resolves freeze-thaw and precipitation controls in a cold-region mine dump, root wood anatomy dates landslide tension-crack evolution, and an ensemble-learning framework predicts direct economic losses across landslides, rockfalls, debris flows, fissures, subsidence, and collapses. These studies move from susceptibility or inventory products toward failure timing, water-routing state, and consequence variables.
The broader set strengthens the forcing and observation layers around those hazards. Urban flood modelling in Niamey and convection-permitting Mediterranean experiments show how hydraulic exposure and extreme precipitation are changing in data-limited or drying regions. AI and remote-sensing papers are most relevant when they serve measurable physical tasks: SARVLM expands semantic radar interpretation, SwAIther-Precip downscales global AI forecasts to kilometer-scale probabilistic precipitation, M2Net improves radar echo extrapolation, and soil-moisture validation on the Tibetan Plateau quantifies the uncertainty in satellite forcing products. Engineering papers expose hidden controls in water-rich tunnels, offshore foundations, fractured rock, CAES caverns, frozen soil-rock mixtures, and bio-cemented sands.
Key Trends
The day is organized around state estimation: preferential flow, crack opening age, direct loss, flood depth, precipitation intensity, radar semantics, freeze-thaw damage, and infrastructure restoration capacity.
- Landslide and geomorphic-hazard studies are shifting from location prediction to mechanism and consequence: mine-dump PF-THM modelling, root-anatomy crack reconstruction, and loss prediction estimate water routing, deformation chronology, and direct economic impact.
- Hydrometeorological inputs are becoming local, probabilistic, and uncertainty-aware: urban flood hydraulics, Mediterranean rainfall experiments, precipitation downscaling, radar extrapolation, snow thresholds, Loess Plateau soil moisture, and Tibetan Plateau validation uncertainty refine hazard forcing.
- Cold-region engineering papers treat freeze-thaw as a structural driver: frozen soil-rock creep, alpine runoff pathways, canal slope aspect, snowmelt thresholds, ice-shelf feedbacks, and bio-cemented sand durability connect phase change to deformation or runoff.
- Remote-sensing and foundation models matter when they map onto hazard observables: SARVLM, ChangeFlow, LDGuid, wildfire-smoke uncertainty, small Earth-data forecasting, and 3D PDE autoencoders target radar semantics, post-event change, smoke severity, sparse monitoring, or physical fields.
- Infrastructure resilience studies are coupling restoration with hidden failure states: shield-tunnel grouting, underwater overburden, caisson seepage, port resilience, sewer failure, road restoration, rocking-block fragility, CAES fracturing, and fractured-rock anchoring translate degradation into operational risk.
Selected Papers
This issue contains 45 selected papers from 1,777 papers analyzed. The leading papers focus on mechanism and consequence rather than simple inventory mapping: thermo-hydro-mechanical mine-dump landslide modelling, root-anatomy reconstruction of landslide tension cracks, direct economic-loss prediction for geological disasters, two-dimensional urban flood hydraulics, and Mediterranean extreme-precipitation intensification. The broader set strengthens the observation and design layer through SAR vision-language modelling, kilometer-scale AI precipitation downscaling, freeze-thaw runoff tracing, frozen soil-rock creep, radar echo extrapolation, snowmelt parameter estimation, ice-shelf melt feedbacks, seismic bedrock reconstruction, tunnel grouting, offshore and port resilience, soil-moisture uncertainty, remote-sensing change detection, and urban network restoration.
1. A novel thermo-hydro-mechanical coupled model for soil-rock mixtures incorporating preferential flow: A case study of Hesigewula mine dump
Core Problem: Cold-region open-pit mine dumps can fail when freeze-thaw, precipitation, preferential flow, and soil-rock heterogeneity interact.
Key Innovation: A dual-porosity PF-THM model represents matrix-fracture water exchange and simulates freeze-thaw and precipitation-driven evolution in the Hesigewula mine dump.
2. Reconstruction of tension cracks evolution for geomorphological hazard assessment using wood anatomy of olive and pine roots– example from Monóvar (SE, Spain)
Core Problem: Tension cracks are early indicators of active slope movement, but their formation and widening are hard to date retrospectively.
Key Innovation: Wood-anatomy signals in olive and Aleppo pine roots are used to reconstruct crack opening and widening on a Monovar landslide in southeastern Spain.
3. A geological disaster risk prediction method using susceptibility and ensemble learning algorithms: direct economic loss prediction
Core Problem: Geological disaster risk indicators often lack physical interpretability and do not directly predict losses relevant to planning.
Key Innovation: Exposure, susceptibility, and disaster-causing factors are integrated into a GELP database and modelled with CatBoost, stacking, XGBoost, and baseline learners.
4. 2D hydrodynamic modelling for urban flood hazard assessment: the case of Niamey, Niger
Core Problem: Urban flooding in Niamey is intensified by climate variability, land-use change, and limited drainage infrastructure.
Key Innovation: A 2D hydrodynamic model translates flood routing and depth patterns into spatial hazard assessment for urban risk planning in Niger.
5. Increasing daily precipitation extremes despite declining annual totals in southern Europe: a modeling study on the effects of Mediterranean Sea warming
Core Problem: Southern Europe can experience stronger daily precipitation extremes even as annual totals decline, complicating flood and landslide-trigger assumptions.
Key Innovation: Convection-permitting WRF experiments isolate how Mediterranean Sea warming amplifies heavy-rainfall events through sea-atmosphere-orography interactions.
6. SARVLM: A Vision Language Foundation Model for Semantic Understanding in SAR Imagery
Core Problem: SAR archives are essential for hazard monitoring but lack large multimodal foundation models that connect radar imagery with semantic language descriptions.
Key Innovation: SARVLM-1M builds a million-pair SAR vision-language dataset and uses optical remote sensing as an intermediate bridge for domain transfer training.
7. SwAIther-Precip: Lead-Time-Aware Bias Correction Enables Kilometer-Scale Downscaling of Global AI Precipitation Forecasts over Switzerland
Core Problem: Global AI weather forecasts are too coarse for local hazard applications and suffer from lead-time-dependent bias.
Key Innovation: SwAIther-Precip converts AIFS forecasts into probabilistic kilometer-scale Swiss precipitation fields using lead-time-aware U-Net correction and ensemble calibration.
8. Effects of freeze-thaw processes on sources and transport pathways of runoff in alpine rivers
Core Problem: Freeze-thaw processes are often omitted when quantifying runoff sources in alpine rivers, leaving permafrost-driven hydrologic mechanisms unclear.
Key Innovation: Isotopes, hydrometeorological observations, and Qinghai Lake basin monitoring quantify how soil freeze-thaw shifts runoff sources and transport pathways.
9. A nonlinear viscoelastic-plastic creep damage model of frozen soil-rock mixture under different temperatures and block conditions: Unique sights from the resistivity perspective
Core Problem: Frozen soil-rock mixtures used in subgrade fills deform over time as temperature and block conditions change.
Key Innovation: Creep tests and resistivity monitoring support a nonlinear viscoelastic-plastic damage model for frozen soil-rock mixtures under different temperatures and gradations.
10. M2Net: a multi-scale context fusion and multi-branch memory recall network for radar echo extrapolation
Core Problem: Radar extrapolation models often blur echoes, underestimate intensity, and lose storm-evolution trends.
Key Innovation: M2Net combines multi-scale context fusion with multi-branch memory recall to preserve echo texture, intensity, and spatiotemporal evolution.
11. Uncertainty-Aware Wildfire Smoke Density Classification from Satellite Imagery via CBAM-Augmented EfficientNet with Evidential Deep Learning
Core Problem: Wildfire smoke severity mapping needs both class labels and uncertainty estimates for emergency response and air-quality modelling.
Key Innovation: A CBAM-augmented EfficientNet with evidential learning predicts light, moderate, and heavy smoke density while decomposing epistemic and aleatoric uncertainty.
12. Estimating robust melt factors and temperature thresholds for snow modelling across the Northern Hemisphere
Core Problem: Large-scale hydrological models need robust temperature thresholds and melt factors for snow accumulation and melt beyond gauged basins.
Key Innovation: A Northern Hemisphere SWE dataset and 4,736 station records are used to estimate and evaluate melt factors and temperature thresholds across broad climates.
13. Antarctic ice-shelf basal melt shaped by competing feedbacks
Core Problem: Future sea-level projections remain uncertain because ice-shelf melt feedbacks and forced ocean changes are often separated.
Key Innovation: Circumpolar Antarctic ocean-sea-ice simulations with interactive ice shelves quantify competing melt-driven and externally forced feedbacks.
14. Pore-scale mechanisms of microbially induced carbonate precipitation under high temperature and high pressure: Implications for CO2 geological sealing
Core Problem: CO2 storage security is threatened by leakage through caprock fractures and wellbore pathways under high-temperature and high-pressure conditions.
Key Innovation: A high-temperature and high-pressure microfluidic platform resolves temperature-pressure controls on microbial viability, ureolysis, and calcite precipitation for sealing.
15. A multidisciplinary methodology to reconstruct and assess bedrock effects for seismic risk mitigation: a case study of deep and liquefiable deposits in Italy
Core Problem: Seismic-risk mitigation requires bedrock reconstruction where amplification and liquefaction effects coexist in deep alluvial deposits.
Key Innovation: A multidisciplinary field campaign combines ambient vibration, Vs profiling, geophysical imaging, and 2012 Emilia earthquake evidence to assess bedrock effects.
16. Tadpole: Autoencoders as Foundation Models for 3D PDEs with Online Learning
Core Problem: Large 3D geophysical PDE problems need reusable representations that can transfer across variables, resolutions, and physical systems.
Key Innovation: Tadpole pre-trains autoencoders on online-generated 3D PDE data at large scale and transfers to downstream reconstruction, forecasting, and physical-field tasks.
17. Simple and robust forecasting of spatiotemporally correlated small Earth data with A tabular foundation model
Core Problem: Many geohazard and hydrologic monitoring systems produce short, sparse, but spatially correlated time series that standard deep models overfit.
Key Innovation: A tabular foundation-model strategy characterizes spatiotemporal patterns in small Earth data and reduces global-mean forecasting bias without task-specific deep training.
18. Semiautomated classification of units in volcanic terrain based on three morphometric attributes derived from a digital elevation model (DEM)
Core Problem: Volcanic hazard and resource maps require reproducible landform units rather than subjective photointerpretation alone.
Key Innovation: A semiautomated DEM method classifies volcanic-terrain units from slope, morphology, and roughness, improving consistency in regional geomorphological mapping.
19. Evolving nature-based solutions for urban resilience
Core Problem: Urban nature-based solutions can fail if living systems are treated as static infrastructure under evolving stressors.
Key Innovation: A Science review synthesizes eco-evolutionary dynamics in street trees, rain gardens, green roofs, and wetlands to guide resilient urban design.
20. Assessing port resilience to climate change in Asia: A comparative analysis
Core Problem: Asian ports face typhoons, heavy rain, and climate disruptions that require system-level resilience assessment.
Key Innovation: Objective-oriented Bayesian networks based on expert elicitation, noisy-or modelling, sensitivity analysis, and scenario simulation compare resilience drivers in the Greater Bay Area and Southeast Asia.
21. Effects of slope aspect on the evolution of frozen soil layers in cold regions: A case study of the Hada Mountain water conveyance canal in western Jilin Province
Core Problem: Cold-region canal slopes are controlled by frozen soil layers whose evolution differs between north- and south-facing aspects.
Key Innovation: Ground temperature, geothermal gradient, and freezing-depth monitoring along the Hada Mountain canal quantify aspect-dependent freezing dynamics.
22. Grouting effectiveness in shield tunnels through water-rich sandy pebble strata and its impact on structure
Core Problem: Water-rich sandy pebble strata can dilute grout and weaken shield-tunnel reinforcement, altering ground and structural response.
Key Innovation: Visualized grouting tests, mechanical characterization, microstructural interpretation, and refined numerical modelling quantify diffusion, dispersion, and reinforcement effects.
23. Efficient Prediction of Fracture Surface Morphology and Analysis of Key Controlling Factors in CAES Tunnel-Type Underground Caverns Based on a Machine Learning Surrogate Model
Core Problem: Compressed-air energy storage caverns undergo pressure cycling that can initiate and propagate surrounding-rock cracks.
Key Innovation: A DEM-based failure framework and machine-learning surrogate rapidly predict fracture surface morphology and controlling factors in tunnel-type underground caverns.
24. Characterization of uncertainty in ground-based validation of soil moisture products: A case study of QLB-NET in the Tibetan Plateau
Core Problem: Satellite soil-moisture validation is distorted by ground-sensor error, representativeness error, and pixel geolocation mismatch.
Key Innovation: A QLB-NET case study quantifies validation uncertainty, identifies sub-pixel geolocation shifts, and separates reference-truth error sources.
25. Declining precipitation frequency may increase vegetation recovery time following extreme drought
Core Problem: Ecosystem drought resilience depends not only on total post-drought precipitation but also on how frequently rain returns.
Key Innovation: Global 1982-2022 vegetation recovery analysis links SPEI-defined extreme droughts, NDVI recovery time, and precipitation frequency controls.
26. ChangeFlow -- Latent Rectified Flow for Change Detection in Remote Sensing
Core Problem: Remote-sensing change masks can be context-dependent and ambiguous, while discriminative models return only a single deterministic output.
Key Innovation: ChangeFlow formulates change detection as latent rectified-flow generation, sampling coherent plausible masks while reducing pixel-space generation cost.
27. LDGuid: A Framework for Robust Change Detection via Latent Difference Guidance
Core Problem: Change-detection networks often fail to explicitly encode the semantic differences that matter between pre- and post-event scenes.
Key Innovation: LDGuid learns a bottlenecked latent difference embedding and injects it into U-Net, BIT, and AERNet baselines across multiple benchmark datasets.
28. Quantum Feature Pyramid Gating for Seismic Image Segmentation
Core Problem: Salt-body delineation affects seismic interpretation, velocity modelling, and drilling uncertainty, yet hybrid quantum-classical dense prediction remains under-tested.
Key Innovation: A quantum feature-gating architecture uses parameterized quantum circuits at feature-pyramid fusion points to segment seismic images.
29. Insights on Numerical Damping Formulations Gained from Calibrating Two-Dimensional Ground Response Analyses at Downhole Array Sites
Core Problem: Ground response analyses can overpredict transfer functions when small-strain damping is calibrated only from one-dimensional assumptions.
Key Innovation: Two-dimensional ground response analyses at four downhole array sites test apparent damping formulations for fundamental and higher-mode response.
30. Love number computation within the Ice-sheet and Sea-level System Model (ISSM v4.24)
Core Problem: Sea-level and ice-sheet models need high-resolution solid-Earth response to surface loading, gravity change, bedrock motion, and mantle rheology.
Key Innovation: ISSM v4.24 adds a Love-number solver with high spherical-harmonic truncation and Maxwell, Burgers, and extended Burgers viscoelastic rheologies.
31. An Grided Soil Moisture Profile Data Set Based on an Optimal Land Surface Model With Measured Hydraulic Parameters and Data Assimilation Over the Loess Plateau
Core Problem: Profile soil moisture over the Loess Plateau remains difficult to map because satellites sense shallow layers and land-surface models need better calibration.
Key Innovation: Measured runoff and hydraulic parameters, LAI data assimilation, and Noah-MP modelling produce a gridded soil-moisture profile dataset.
32. Rainwater Regulation Alters Water Partitioning and Vegetation Growth on the Loess Plateau
Core Problem: Rainwater-harvesting interventions can alter runoff, evapotranspiration, storage, and vegetation response across water-limited terrain.
Key Innovation: A Noah-MP implementation of rainwater regulation quantifies how captured surface runoff is redistributed into ET, subsurface runoff, and storage.
33. Asymmetric Spring–Summer Responses of Interannual Dry–Wet Transitions in Eastern Asia and North America Under Global Warming
Core Problem: Warming can intensify abrupt dry-wet transitions that affect drought, flood, and slope-failure forcing.
Key Innovation: SPEI analysis of historical and future spring-summer transitions shows regional asymmetry and more than 50% increases in springtime transition frequency in key regions.
34. Investigation and modeling of seepage-influenced caisson–soil interface monotonic shearing
Core Problem: Storm-induced uplift can create negative pressure beneath suction caissons and alter effective stress along skirt walls.
Key Innovation: Sand-steel shear tests under seepage-equivalent stress histories quantify interface strain-softening and friction-angle response during caisson uplift.
35. A multi-basis spacing algorithm for modeling directional uncertainty in in situ block size distribution of rock masses with non-persistent joints
Core Problem: Rock-mass stability estimates depend on in situ block-size distributions that must reflect non-persistent joint geometry and directional uncertainty.
Key Innovation: A multi-basis spacing algorithm models fracture-network hierarchy and directional uncertainty to estimate emergent block-size distributions.
36. A prediction model for minimum overburden thickness for underwater shield tunnel crossing structural planes
Core Problem: Minimum overburden thickness for underwater shield tunnels is hard to estimate when the tunnel crosses structural planes.
Key Innovation: Discrete-element models parameterized by laboratory rock and structural-plane tests evaluate overburden requirements for tunnels crossing discontinuities.
37. Mechanical responses of plate-monopile hybrid foundation under cyclic load with different patterns
Core Problem: Offshore wind foundations must withstand bidirectional and asymmetric cyclic load histories.
Key Innovation: Finite-element models compare plate-monopile hybrid and conventional monopile responses under multiple cyclic loading patterns and amplitudes.
38. Experimental Study on Shear Anisotropy and Anchoring Mechanism of Fractured Rock Masses
Core Problem: Bolt design for fractured rock needs to account for persistence, aperture, offset, and dip-angle controls on shear-compression response.
Key Innovation: Direct shear tests and ImageJ fracture-network quantification identify fracture persistence as the dominant strength-degradation factor and clarify bolt reinforcement mechanisms.
39. Eco-friendly wheat straw-reinforced MICP for sand stabilization: Enhancing solidification uniformity and modeling freeze–thaw pore damage
Core Problem: MICP treatment for sandy-land stabilization is limited by non-uniform cementation and uncertain freeze-thaw durability.
Key Innovation: Wheat-straw-reinforced MICP increases calcite precipitation, improves solidification uniformity, and models pore damage after freeze-thaw cycling.
40. From sources to risks: A multi-method approach to groundwater hydrogeochemical evolution in the Hotan Oasis of arid inland river basin, Northwest China
Core Problem: Arid oasis groundwater management is complicated by seasonal hydrology and spatially heterogeneous geochemical sources.
Key Innovation: SOM, PMF, hydrochemical diagrams, isotope tracing, and health-risk assessment quantify groundwater source-process-risk pathways in the Hotan Oasis.
41. Experimentally informed limit states and fragility curves of rocking blocks via shaking table tests
Core Problem: Rocking-block vulnerability assessments need repeatable experimental data that capture response variability across geometry and ground motions.
Key Innovation: More than 300 shaking-table tests support experimentally informed limit states and fragility curves for free-standing rocking blocks.
42. Implications of varying thermal regimes on the thermo-mechanical and fracturing characteristics of granite
Core Problem: Granite in geothermal reservoirs experiences thermal shock, progressive heating, and cyclic stimulation that change strength and fracture networks.
Key Innovation: Laboratory heating scenarios to 500 C compare mechanical degradation and microcrack evolution under direct, sequential, and cyclic thermal regimes.
43. Enhancing lifecycle resilience of urban road networks under uncertainty: A two-stage stochastic programming approach for integrated reinforcement and restoration
Core Problem: Extreme disasters can fragment urban road networks, requiring joint decisions on pre-disaster reinforcement and post-disaster restoration.
Key Innovation: A two-stage stochastic programming model optimizes lifecycle resilience under uncertainty in damaged-link numbers, locations, and disaster intensity.
44. Deep learning-based failure rate prediction of sewer networks with uncertainty quantification under climatic variations: A Hong Kong case study
Core Problem: Sewer failure prediction must combine pipe aging, climate, regional characteristics, and uncertainty for urban resilience planning.
Key Innovation: Hong Kong sewer inspection and climate datasets support deep-learning failure-rate models with Monte Carlo dropout and kernel-density uncertainty quantification.
45. Prediction of urban stormwater runoff quality in data-deficient areas using a semi-supervised machine learning framework
Core Problem: Stormwater quality models are limited where rainfall-event pollutant labels are scarce.
Key Innovation: A semi-supervised random forest framework combines labeled and unlabeled rainfall-event data to improve event mean concentration prediction in urban catchments.