TerraMosaic Daily Digest: May 19, 2026
Daily Summary
The May 19 papers shift the landslide signal from event mapping toward climate response, failure material state, and network consequence. A global CNN-based projection of rainfall-induced landslides under CMIP6 scenarios provides the broadest hazard frame, while the Baige and Huangtupo studies resolve two complementary mechanisms: altered-rock strength loss in a tectonic suture zone and cyclic weakening of reservoir slip-zone soil. The road-network and masonry-bridge papers then carry slow-moving landslide deformation into infrastructure prioritisation, using A-DInSAR displacement, virtual damage surveys, and structural defect patterns as operational evidence.
The same process-oriented logic appears in the hydrologic and underground-engineering papers. Flood studies isolate the hydraulic role of culvert blockage, green infrastructure, prompt-guided physics transfer, and gridded precipitation architecture rather than treating rainfall as a single scalar driver. Tunnel, mine, and fractured-rock papers follow stress waves, gas-outburst precursors, seepage parameters, fracture activation, bolt deformation, and sparse-sensor excavation displacement through explicit monitoring or mechanics. Remote-sensing foundation models are relevant where they improve exposure mapping, multimodal data completion, fire prediction, groundwater monitoring, and land-cover context for hazard analysis.
Key Trends
The day is organized around state variables that are directly actionable: future landslide susceptibility, altered-rock strength, slip-zone cyclic strain, road criticality, bridge defects, blockage timing, seepage parameters, and early-warning precursor reliability.
- Landslide science connects climate forcing to material and infrastructure response: global rainfall-induced landslide projection, Baige altered-rock mechanics, Huangtupo cyclic weakening, coseismic GaoFen-2 mapping, road-network prioritisation, and bridge-defect assessment convert slope movement into failure or serviceability variables.
- Flood forecasting is moving from rainfall input to hydraulic causality: culvert blockage, green infrastructure, prompt-guided physics, explainable CNN-LSTM design, Yangtze subseasonal downscaling, and Amazon cutoff analysis identify how routing, slope, infrastructure, and transfer learning produce damaging flows.
- Subsurface engineering studies emphasize monitored degradation paths: rockburst attenuation, dam seepage inversion, coal-gas precursor fusion, excavation RNNs, soft-rock yield bolts, water-rich sand grouting, and true-triaxial fracture activation constrain hidden stress, seepage, and gas states.
- Cold-region, drought, and groundwater papers resolve storage rather than surface symptoms: thaw settlement, contrasting permafrost mechanisms, GRACE water-storage deficits, India drought trends, karst fracture storage, groundwater vulnerability, and regional groundwater levels track where water is stored, lost, or mobilized.
- Remote-sensing and AI methods are strongest when tied to geologic observables: Wildfire-FM, MetaEarth-MM, LandSegmenter, SWOT filtering, GNSS strain, fire-severity mapping, microseismic picking, ultrasonic fracture reconstruction, and discontinuous-coefficient neural operators strengthen exposure, deformation, water-level, fire, seismic, and fracture evidence.
Selected Papers
This issue contains 50 selected papers from 1,965 papers analyzed. The leading papers connect landslide hazard to climate forcing, material weakening, and infrastructure consequence: global rainfall-induced landslide projections, the Baige clay-altered rock failure mechanism, cyclic weakening in Huangtupo slip-zone soil, rapid coseismic landslide mapping, slow-moving landslide road-network prioritisation, and defect-informed bridge risk assessment. The broader set extends the same process-state logic to culvert-blockage flood dynamics, prompt-guided flood forecasting, dam-foundation seepage inversion, green-infrastructure flood mitigation, deep-tunnel rockburst support, coal-gas outburst warning, permafrost thaw settlement, drought storage deficits, groundwater vulnerability, fractured-media flow, microseismic monitoring, and remote-sensing foundation models.
1. Projected response of global rainfall-induced landslides to future climate change
Core Problem: Global trends in future rainfall-induced landslide susceptibility remain uncertain despite intensifying extreme precipitation.
Key Innovation: A global historical landslide catalogue, dynamic and static environmental factors, RF/XGBoost/CNN models, and CMIP6 SSP scenarios are combined to project landslide susceptibility for 2050 and 2100.
2. Formation Mechanism of the Catastrophic Baige Landslide in the Jinsha River, Tibet, China: New Insights from Clay-Altered Rocks in the Tectonic Suture Zone
Core Problem: The 2018 Baige landslide produced a severe Jinsha River disaster chain, but the role of clay-altered rocks in the source zone remained mechanically under-constrained.
Key Innovation: Field surveys, large in-situ direct shear tests, and back-analysis link alteration degree, block proportion, heterogeneous slip-zone strength, and tectonic-suture-zone controls to failure initiation.
3. Investigation of cyclic weakening and deformation behavior of slip zone soil in the Huangtupo landslide under reservoir water level fluctuations
Core Problem: Reservoir water-level cycling can weaken slip-zone soils and drive step-like landslide deformation.
Key Innovation: Stress- and strain-controlled cyclic triaxial tests, SEM observations, and numerical simulations quantify cyclic weakening and deformation mechanisms in Huangtupo slip-zone soil.
4. Rapid co-seismic landslide mapping using object-based change detection by bi-temporal GaoFen-2 in Tibet, China near Nepal
Core Problem: Emergency response after earthquakes requires rapid and accurate landslide delineation over complex mountain terrain.
Key Innovation: Object-based change detection with bi-temporal GaoFen-2 imagery targets fast regional coseismic landslide mapping in Tibet near Nepal.
5. Risk-mitigation priorities for road networks exposed to slow-moving landslides at municipal scale
Core Problem: Mountain road networks exposed to slow-moving landslides need mitigation priorities under budget and information constraints.
Key Innovation: A municipal-scale framework integrates landslide susceptibility, A-DInSAR displacement, virtual-survey damage severity, and road-network criticality to rank intervention priorities.
6. A Defect-Informed Smart Prioritisation Framework for Landslide Risk Assessment of Masonry Arch Bridges
Core Problem: Bridge landslide-risk screening often separates geomorphic hazard from structural response, underestimating progressive slope-structure interaction.
Key Innovation: Inspection-derived crack and deformation patterns are integrated with conventional landslide descriptors to refine prioritisation for masonry arch bridges.
7. Culvert blockages in 2D-hydrodynamic flash flood modeling: quantifying the impact on flood dynamics and mitigation strategies
Core Problem: Culvert blockages can strongly alter flood depths and flow paths but are commonly omitted from two-dimensional flood models.
Key Innovation: A dynamic blockage module in TELEMAC-2D uses water-level thresholds and multiple blockage scenarios to quantify local flood amplification and mitigation options in a German catchment.
8. A prompt-guided differentiable physics model for flood forecasting in data-scarce basins
Core Problem: Flood models trained on data-rich basins often transfer poorly and lack explicit physical constraints.
Key Innovation: A prompt-guided differentiable physics model combines static-temporal pretraining masks, dual-path prompts, physical-parameter priors, and hydrologic observations to improve transferable flood prediction.
9. Neural operator-based multi-objective optimization for spatiotemporal data-informed seepage parameter identification: a case study on deep overburden in a dam foundation
Core Problem: Dam-foundation seepage parameter identification is biased when monitoring data are sparse, spatially incomplete, or treated at single time points.
Key Innovation: A DeepONet surrogate and improved multi-objective coati optimization fuse time-varying water levels and multi-point seepage-pressure observations for spatiotemporal parameter inversion.
10. The potential of green infrastructure in urban pluvial flood mitigation – a scenario-based modelling study in Berlin
Core Problem: Urban surface sealing increases runoff and flood damage during short-duration heavy rainfall.
Key Innovation: Scenario-based SWMM simulations in Berlin compare bioretention, green roofs, pervious pavement, conventional drainage, flood depths, and building damage under design storms.
11. A Novel Wave-Absorbing Support System for Mitigating Rock Burst in Deep-Buried Hard Rock Tunnels
Core Problem: Stress waves can trigger rockbursts away from excavation faces in deep-buried hard-rock tunnels.
Key Innovation: A wave-absorbing support concept uses low-impedance anchorage materials, grouting, and shotcrete to attenuate stress-wave propagation in surrounding rock.
12. Online early warning of coal and gas outbursts through reliability-constrained fusion of acoustic emission, electromagnetic radiation, and gas concentration
Core Problem: Coal and gas outburst warning is distorted by noisy, shifting monitoring signals and changing coupling among precursors.
Key Innovation: Reliability-constrained fusion combines acoustic emission, electromagnetic radiation, gas concentration, robust denoising, dynamic lag estimation, and correlation-confidence rules.
13. Debris Transport and Impact Dynamics under Dam-Break Flow: Effects of Initial Conditions and Bathymetric Transitions
Core Problem: Floating debris entrainment and impact loads under dam-break flows remain stochastic and sensitive to initial and bathymetric conditions.
Key Innovation: Instrumented debris models, force sensors, computer vision, and 135 controlled trials quantify how impoundment depth, debris placement, submergence, release angle, and slope-to-berm transitions control impact dynamics.
14. Landslide Research: State of the Art and Innovations
Core Problem: Landslide research spans diverse mechanisms, environments, triggers, and mitigation needs that are difficult to consolidate across subfields.
Key Innovation: A GeoHazards special-issue synthesis frames contemporary landslide research around characterization, triggering, risk prevention, mitigation, and innovation across continental and underwater settings.
15. Patterns and Prediction of Thaw Settlement and Thaw Compression in Permafrost
Core Problem: Permafrost foundations can deform through both thaw settlement and thaw-induced compression, with contributions controlled by soil structure and loading.
Key Innovation: Laboratory tests define thaw-settlement and thaw-compression coefficients, introduce a thaw-proportion index, and identify load-porosity conditions for significant compressive deformation.
16. Modeling Circulation‐Precipitation Overlapped Co‐Evolution for Improving Subseasonal Prediction in the Yangtze River Basin
Core Problem: Severe precipitation in the Yangtze River Basin requires skillful subseasonal prediction for flood preparedness.
Key Innovation: An overlapping circulation-precipitation co-evolution model combines observed recent evolution with predictable S2S circulation components to improve rainfall-anomaly forecasts.
17. Floodplain slope organizes chute and neck meander cutoffs across rivers in the Amazon basin
Core Problem: The frequency and type of Amazon river cutoffs remain difficult to explain across a basin-scale slope gradient.
Key Innovation: Satellite images from 1984-2021 identify 1,132 cutoffs and show that floodplain slope organizes neck versus chute cutoff processes.
18. Reliability evaluation of pseudo-samples from model-free and model-based approaches in wildfire susceptibility mapping
Core Problem: Wildfire susceptibility models can be unstable when true event samples are limited and pseudo-sample construction is uncertain.
Key Innovation: Model-free and model-based pseudo-sample strategies are evaluated to test reliability in data-limited wildfire susceptibility mapping.
19. Digital Landscapes: Assessing Fire Severity and Its Drivers Using Remote Sensing and Google Earth Engine Based on dNBR and NPP Indicators
Core Problem: Mediterranean forest fire severity is difficult to characterize in data-scarce landscapes.
Key Innovation: Landsat-8 dNBR, NPP, topography, and Google Earth Engine are used to map fire severity and driver variability in northern Lebanon.
20. Precise Contemporary Crustal Strain and Rotation Rates Derived from GNSS Measurements in the Pamir–Tian Shan Region
Core Problem: Contemporary strain models for the Pamir-Tian Shan differ because of uneven GNSS distribution and analytical choices.
Key Innovation: Updated GNSS velocities referenced to stable Eurasia estimate strain, shear, dilatation, and rotation fields with numerical reliability validation.
21. A WSDI framework describing GRACE terrestrial water storage deficit for monitoring and tracking the drought process and spatial pattern in the Loess Plateau, China
Core Problem: Surface drought indices do not fully capture water-storage deficits across groundwater, soil moisture, canopy water, and snow.
Key Innovation: A multidimensional WSDI framework based on GRACE tracks meteorological-to-groundwater drought propagation and spatial patterns in the Wei River Basin.
22. Mechanistic dichotomy in permafrost thaw between Alaska and the Qinghai-Tibet Plateau
Core Problem: Permafrost projections are limited by uncertainty in whether thaw controls transfer across regions.
Key Innovation: Multi-source 2000-2020 data, machine learning, and causal inference reveal different linear and nonlinear controls on permafrost thaw in Alaska and the Qinghai-Tibet Plateau.
23. Towards more accurate streamflow and flood prediction via explainable CNN-LSTM: leveraging ERA5 gridded precipitation and hydrological observations through input-architecture co-design
Core Problem: Sparse upstream observations and heterogeneous gridded rainfall make source-sink representation difficult in flood forecasting.
Key Innovation: A 3D-CNN-LSTM architecture fuses ERA5 gridded precipitation with hydrological observations and uses explainability to guide input-architecture co-design.
24. Sediment hysteresis loop variability in watersheds during rainfall-runoff erosion: effects of internal watershed controls revealed by process-based modeling
Core Problem: Sediment hysteresis loops vary strongly between rainfall-runoff events, limiting their reliability as simple erosion indicators.
Key Innovation: Process-based simulations isolate how particle-size distribution, microtopography, raindrop impact, and rainfall duration control sediment loop form.
25. Incorporating a precipitation factor into the soil vulnerability index
Core Problem: The Soil Vulnerability Index omits rainfall intensity and seasonality, limiting transferability across erosion climates.
Key Innovation: Thirty-year SWAT+ simulations across 79 watersheds show rainfall erosivity and winter precipitation thresholds that improve SVI interpretation.
26. Multiscale discrete fracture-matrix method for simulating groundwater flow in fractured geological media
Core Problem: Fractured media require groundwater-flow models that balance explicit fracture detail with computational efficiency.
Key Innovation: A multiscale discrete fracture-matrix method combines explicit, embedded, and implicit fracture representations using coupling finite elements.
27. Three-dimensional reconstruction method for ultrasonic detection of rock-interaction fractures based on Optuna-1D CNN
Core Problem: Laboratory rock-failure studies need accurate 3D reconstruction of interacting fractures from indirect measurements.
Key Innovation: An Optuna-optimized 1D CNN maps ultrasonic signals to voxel-level fracture distributions using a numerical-experimental red sandstone dataset.
28. Effect of unloading paths on fracture activation: Insights from true triaxial unloading tests
Core Problem: Underground excavation can activate pre-existing fractures, but the effect of true-triaxial unloading paths is unclear.
Key Innovation: True-triaxial unloading tests on fractured granite identify which stress paths activate fractures versus break intact rock.
29. Anchorage mechanical behavior of interface-shear-stress-dispersing yield bolt for soft rock tunnel
Core Problem: Soft-rock tunnel stability requires support systems that maintain anchorage while accommodating progressive interface damage.
Key Innovation: An interface-shear-stress-dispersing yield bolt with compressible layers is evaluated through theory and field pull-out tests.
30. Automated monitoring of deep excavation-induced horizontal displacements using sparse sensor measurements: A recurrent neural network approach
Core Problem: Dense automatic inclinometer arrays are costly, limiting continuous monitoring of deep excavation deformation.
Key Innovation: LSTM, GRU, and BiLSTM models predict horizontal displacements from sparse field sensor measurements in two deep excavations.
31. Does Your Wildfire Prediction Model Actually Work, or Just Score Well?
Core Problem: Earth foundation models are rarely pretrained or evaluated specifically for sparse wildfire forecasting.
Key Innovation: WILDFIRE-FM is paired with fixed-contract evaluation checks that isolate matching-rule effects and feature-transfer reliability.
32. MetaEarth-MM: Unified Multimodal Remote Sensing Image Generation with Scene-centered Joint Modeling
Core Problem: Hazard mapping often lacks complete paired observations across optical, SAR, elevation, and other remote-sensing modalities.
Key Innovation: MetaEarth-MM introduces scene-centered joint modelling for unified paired generation and any-to-any translation across five remote-sensing modalities.
33. LandSegmenter: Towards a flexible foundation model for Land Use and Land Cover mapping
Core Problem: LULC mapping models usually depend on fixed modalities and class taxonomies, limiting transfer to new exposure and vulnerability settings.
Key Innovation: LandSegmenter uses a large weakly labelled multimodal dataset and a flexible foundation-model framework to support adaptable LULC mapping.
34. Fault Volume Digital Twin to Reproduce the Full Slip Spectrum, Scaling, and Statistical Laws
Core Problem: Fault models struggle to reproduce the full observed spectrum of seismic and aseismic slip without ad hoc frictional heterogeneity.
Key Innovation: A 2D shear-zone digital twin with off-fault damage-zone fractures and rate-state friction reproduces slip dynamics, scaling, and statistical laws.
35. Prediction of groundwater vulnerability in the Ganges Delta using GIS-based statistical techniques, self-organizing map (SOM) and wavelet analysis: issues and challenges
Core Problem: Groundwater quality degradation in deltaic settings requires vulnerability models that handle many controlling variables and spatial structure.
Key Innovation: GIS statistics, SOM, wavelet analysis, and 355 groundwater samples are used to predict vulnerability controls in the Ganges Delta.
36. Rock fracture characteristics regulate water storage and seasonal tree water uptake in karst
Core Problem: Karst water storage and plant water uptake depend on heterogeneous fractures that are difficult to quantify.
Key Innovation: Stable isotopes, MixSIAR, and hydrometric monitoring separate soil and rock water contributions and residence times by fracture characteristics.
37. Estimating Regional Groundwater Level by Combining Satellite, Model, and Large-Sample Observations Inputs
Core Problem: Regional groundwater levels are hard to estimate because storage and aquifer conditions are spatially heterogeneous.
Key Innovation: Large observation samples, GLDAS, GRACE, and Random Forest modelling estimate groundwater levels across the Yellow River Basin.
38. A Large-Scale Evaluation of SWOT-Derived Water Surface Elevations: Precision Drivers and Strategies to Enhance Data Availability
Core Problem: SWOT water-surface elevations need systematic quality evaluation across diverse lake geometries and observation conditions.
Key Innovation: Field comparisons for 132 Brazilian lakes identify precision drivers and a SQRTL filtering strategy that increases usable observations.
39. Excavation-induced stress release and ground improvement depth in normally consolidated rock, transitional rock, and soil media
Core Problem: Excavation-induced stress release governs deformation beneath excavation bases in soils, transitional rocks, and rock masses.
Key Innovation: A simplified analytical OCR-based stress framework estimates preliminary ground-improvement depth envelopes for normally consolidated geomaterials.
40. Research on the Lightweight DAMS-Net and Transfer Learning Methods for Microseismic P-Wave Arrival Time Picking
Core Problem: Microseismic P-wave picking in mines is degraded by strong noise, complex waveforms, and limited data quality.
Key Innovation: A lightweight DAMS-Net with transfer learning improves robust P-wave arrival picking for mine safety and rock-engineering monitoring.
41. Capturing brittle-to-ductile transitions in porous rocks via meso-scale informed constitutive modelling
Core Problem: Porous rocks can fail through localized or diffuse modes depending on confinement, which conventional models struggle to reproduce.
Key Innovation: A two-scale constitutive framework couples a meso-scale localization band with surrounding inelastic bulk response.
42. Experimental Investigation of Bentonite Slurry Infiltration through Porous Media: Governing Factors and Applications in Tunnel Boring Machine Drilling
Core Problem: Bentonite slurry infiltration controls TBM face stability but its fast initial penetration depends on interacting material and pressure factors.
Key Innovation: Instrumented infiltration tests isolate the effects of slurry yield stress, viscosity, sand type, pressure, and setup geometry on pore-pressure response.
43. Digital image correlation for mechanical damage assessment in geogrids: Experimental characterization and 3D numerical modelling of soil-geogrid interaction
Core Problem: Mechanical damage to geogrids affects reinforcement performance but is poorly captured by conventional tests.
Key Innovation: DIC full-field strain measurements and 3D finite-element modelling characterize damage-induced localization in soil-geogrid interaction.
44. Seismic responses of soil–twin tunnels system subjected to Rayleigh wave propagation
Core Problem: Twin tunnel systems can amplify ground and lining response under surface-wave excitation.
Key Innovation: Time-domain finite-element simulations examine tunnel spacing, burial depth, and seismic frequency effects on Rayleigh-wave-induced amplification.
45. Strain-dependent shear modulus and damping ratio of unsaturated non-plastic silt under varying matric suctions
Core Problem: Unsaturated silt dynamic properties change with matric suction and strain level, affecting site-response assessment.
Key Innovation: Bender element and resonant column-torsional shear tests quantify strain-dependent shear modulus and damping across suction states.
46. Diffusion of polymer permeation grouting in water-rich sand layers: Mathematical model and experimental validation
Core Problem: Polymer grouting in water-rich sand must account for slurry diffusion under moving groundwater.
Key Innovation: A Darcy-Hagen-Poiseuille mathematical model and laboratory validation describe polymer permeation under flow velocity, grouting pressure, and sand-layer controls.
47. Three-dimensional time-dependent fracturing model for hard rock involving stress-induced anisotropic cracks
Core Problem: Hard-rock time-dependent failure is governed by evolving excavation-induced microcracks.
Key Innovation: A 3D tensile-shear fracturing model captures the full time-dependent crack evolution process under stress-induced anisotropy.
48. Walsh-Hadamard Neural Operators for Solving PDEs with Discontinuous Coefficients
Core Problem: Fourier neural operators struggle with sharp interfaces and discontinuous coefficients common in hydrogeologic and geomechanical PDEs.
Key Innovation: A Walsh-Hadamard neural operator uses rectangular-wave spectral bases to solve Darcy flow, heat conduction, and Burgers problems with discontinuities.
49. Multi-scale spatiotempral characterization and trend analysis of meteorological and geographical characteristics of drought in India’s arid and semi-arid regions
Core Problem: Drought evolution across hot and cold arid sub-climatic zones in India is insufficiently resolved.
Key Innovation: SPI and SPEI at multiple time scales from 1981-2020 quantify drought duration, frequency, intensity, and trends across vulnerable subregions.
50. Impact of blast design parameters on rock fragmentation in sub-level caving: A multivariate regression approach
Core Problem: Fragmentation controls ore flow, recovery, dilution, and operational risk in sublevel caving mines.
Key Innovation: Mechanistic blasting simulations and multivariate regression relate explosive density and burden to fragmentation outcomes under complex broken-ground conditions.