TerraMosaic Daily Digest: June 16, 2026
Daily Summary
The June 16 literature is unusually focused: the strongest papers converge on how unstable terrain can be monitored, interpreted, and converted into decision-ready state variables. A Transportation Geotechnics study turns residual strain after shaking into a real-time index for seismic damage and sliding-surface identification in bedrock-overburden slopes. A companion landslide-AI thread is led by LandslideAgent and Multimodal LandslideBench, which pair domain rules with multimodal interpretation to reduce vision-language hallucination in landslide analysis. A graph spatiotemporal attention network extends that direction to displacement prediction, treating irregular monitoring points as a graph rather than forcing them onto regular image grids.
Cryosphere and water hazards form the second axis. A Journal of Hydrology paper reconstructs surge-induced glacial lake outburst floods in the Upper Indus Basin by combining remote sensing, two-dimensional hydraulic modelling, and field geomorphology. Nature and Science Advances papers broaden the same risk horizon: one shows that Antarctic sea-level contribution contains decadal predictive information, another links western US wildfire activity to heatwave windows, and a third argues for AI-remote-sensing synergy as the next architecture for global water-cycle observation. These papers matter because they move hazard assessment from event description toward state estimation, forecast windows, and attribution.
The methodological tail is broad but coherent. Remote-sensing papers target coseismic 3D displacement, mining subsidence, flood mapping, waterlogging severity, wildfire spread tensors, and seismic stabilizing-pile design. GRL and geotechnical papers add conditional climate emulation, autonomous inverse fracture modelling, neural-symbolic constitutive modelling, PINN-FEM tunnel hydro-mechanics, acoustic-emission damage tomography, and excavation uncertainty. The common trend is not another marginal accuracy gain; it is the embedding of physics, geometry, uncertainty, and domain rules into AI systems that must operate under sparse observations and high consequence.
Key Trends
Five movements define the issue: state-aware slope monitoring, agentic landslide interpretation, forecast-window hazard modelling, physics-guided geoscience AI, and probabilistic infrastructure transfer.
- Slope monitoring is becoming state-aware: Residual strain ratios, graph attention over irregular monitoring points, coseismic LiDAR displacement, mining-subsidence data assimilation, and seismic stabilizing-pile force estimation all convert observations into interpretable instability states.
- Agentic remote sensing is entering landslide science: LandslideAgent, Multimodal LandslideBench, and Remote Sensing Agent move beyond single-model segmentation by testing whether domain rules, tool use, and multimodal reasoning can reduce hallucination and improve operational interpretation.
- Cryosphere, flood, and coastal hazards are being linked to forecast windows: Upper Indus GLOFs, Antarctic sea-level predictability, SAR flood mapping, hurricane surge simulation, and water-cycle AI all shift the emphasis from mapped exposure to time-sensitive risk estimation.
- Physics-guided AI is spreading across geoscience: Conditional climate emulators, inverse fracture diffusion models, neural-symbolic constitutive laws, PINN-FEM tunnel analysis, and acoustic-emission tomography share a common aim: constrain learning with process knowledge.
- Infrastructure resilience is moving toward probabilistic transfer: Transfer learning for fragility models, deep-excavation uncertainty analysis, rock-joint reduced-order modelling, and cyclic geogrid-loess interface tests all target reliable extrapolation under limited local data.
Selected Papers
The selected papers emphasize seismic landslide monitoring, landslide-agent benchmarks, graph-based displacement prediction, GLOF reconstruction, SAR flood mapping, coastal surge and sea-level risk, AI-enabled water-cycle observation, remote-sensing agents, coseismic displacement, mining subsidence, permafrost-river carbon feedback, fracture inversion, geotechnical constitutive modelling, tunnel hydro-mechanics, rock instability, and infrastructure fragility transfer. This issue contains 31 selected papers from 2074 papers analyzed.
1. A strain residual ratio-based framework for real-time seismic damage quantification and sliding surface identification in bedrock-overburden slopes: shaking table validation and engineering implementation
Core Problem: Seismic-induced landslides threaten transportation lifelines, but field monitoring data are rarely converted into quantitative stability indices during or after shaking.
Key Innovation: Defines a strain residual ratio and validates it through shaking-table experiments to quantify seismic damage and infer sliding surfaces in bedrock-overburden slopes.
2. LandslideAgent with Multimodal LandslideBench: A Domain-Rule-Augmented Agent for Autonomous Landslide Identification and Analysis
Core Problem: General-purpose vision-language models can hallucinate or miss geoscientific semantics when interpreting complex landslide scenes.
Key Innovation: Introduces a domain-rule-augmented landslide agent and Multimodal LandslideBench to evaluate autonomous landslide identification, reasoning, and analysis.
3. Hydraulic reconstruction and geomorphological impacts of surge-induced GLOFs in the Upper Indus Basin
Core Problem: Surge-triggered GLOFs in the Upper Indus Basin pose downstream hazards, but their hydraulic propagation and geomorphic effects remain poorly quantified.
Key Innovation: Combines multisource remote sensing, HEC-RAS two-dimensional hydraulic modelling, and field geomorphology to reconstruct flood-wave behavior and impacts.
4. Landslide displacement prediction and interpretability analysis based on a graph spatiotemporal attention network
Core Problem: Landslide monitoring points are spatially irregular, making grid-based deep-learning models poorly matched to the geometry of real displacement observations.
Key Innovation: Uses a graph spatiotemporal attention network to model monitoring-point topology while adding interpretability analysis for landslide early warning.
5. Study on deformation evolution and disturbance characteristics of granular slopes under multi-parameter regulation of material supply
Core Problem: Bedrock disintegration and variable sediment supply can destabilize granular highway slopes, but the relation between supply conditions and deformation evolution is difficult to observe directly.
Key Innovation: Uses a V-shaped physical apparatus and numerical simulation to quantify how material-supply parameters regulate granular-slope deformation and disturbance.
6. Transforming global water cycle observations via synergistic AI and remote sensing
Core Problem: Accelerating water-cycle change and water-related disasters exceed the robustness of region-specific empirical remote-sensing workflows.
Key Innovation: Proposes a bidirectional AI and remote-sensing framework for high-resolution global water-cycle observation under complex terrestrial conditions.
7. Emergent decadal predictability in Antarctic contribution to sea-level rise
Core Problem: Future Antarctic ice-sheet mass loss remains deeply uncertain, limiting the usefulness of sea-level projections for coastal-risk planning.
Key Innovation: Shows that the current Antarctic contribution to sea-level rise contains decadal predictive information across model complexity and emission pathways.
8. Extraction of Detailed 3D Coseismic Displacements in the 2024 Noto Peninsula Earthquake from Airborne LiDAR Data
Core Problem: High-resolution earthquake deformation fields are difficult to recover in three dimensions where ground rupture and localized displacement are spatially complex.
Key Innovation: Uses airborne LiDAR data to extract detailed 3D coseismic displacements for the 2024 Noto Peninsula earthquake.
9. Remote Sensing Agent: Reshaping the Paradigm of Remote Sensing Information Processing
Core Problem: Remote-sensing information processing is increasingly too heterogeneous for fixed, single-task pipelines.
Key Innovation: Frames remote-sensing processing around agentic workflows, with potential relevance to disaster mapping, monitoring, and evidence synthesis.
10. Rapid flood extent mapping and exposure assessment using SAR and machine learning in the Küçük Menderes Basin, Türkiye
Core Problem: Rapid flood response requires cloud-robust inundation mapping and exposure assessment at operational spatial scales.
Key Innovation: Builds a Google Earth Engine workflow combining Sentinel-1 SAR, Sentinel-2 optical baselines, machine learning, and exposure analysis for flood extent mapping.
11. A Variational Data Assimilation Framework for Mining Subsidence Reconstruction from Heterogeneous D-InSAR and TLS Observations
Core Problem: Mining-induced ground deformation is observed by heterogeneous sensors whose spatial coverage, resolution, and uncertainty differ.
Key Innovation: Uses variational data assimilation to reconstruct mining subsidence from combined D-InSAR and terrestrial laser scanning observations.
12. 80 Years of research on tsunamigenic earthquakes in the Makran subduction zone (1945-2025): a review - part C: tsunami hazard and risk assessments
Core Problem: The Makran subduction zone remains a high-consequence but methodologically fragmented tsunami-risk setting.
Key Innovation: Synthesizes probabilistic and deterministic tsunami hazard and risk assessment methods developed for the Makran subduction zone over eight decades.
13. Heatwaves enable wildfire activity in the western United States
Core Problem: The role of heatwaves in controlling wildfire occurrence and growth remains less resolved than their direct human and ecological impacts.
Key Innovation: Shows that a large fraction of western US burned area occurs during and immediately after heatwaves, linking compound heat-fire hazard windows to fuel and meteorological conditions.
14. Interactive Climate Projection via Conditional Generative AI
Core Problem: Century-scale Earth-system simulations are too expensive to densely sample policy-relevant forcing pathways.
Key Innovation: Develops a conditional generative emulator for rapid projections of temperature, precipitation, and sea-surface height under arbitrary CO2 forcing pathways.
15. Physics-Supervised Autonomous Inverse Fracture Modeling via Generative Artificial Intelligence
Core Problem: Fracture networks control groundwater flow and subsurface transport, yet remain difficult to infer from sparse and noisy observations.
Key Innovation: Introduces GenFrac, a physics-supervised denoising-diffusion inverse model for autonomous fracture-network reconstruction.
16. Adaptive weighting of InSAR and GNSS data for coseismic deformation and slip inversion of the MW 6.4 Meinong earthquake, Taiwan
Core Problem: Joint InSAR-GNSS slip inversion is sensitive to empirical data weighting, which can distort coseismic deformation estimates.
Key Innovation: Applies Helmert variance component estimation to objectively weight InSAR and GNSS observations for 3D deformation reconstruction and slip inversion.
17. Hurricane surge prediction using a statistical-deterministic modeling framework: application to St Mary Parish, Louisiana, USA
Core Problem: Climate change and sea-level rise require surge-risk methods that can sample many plausible future hurricane events.
Key Innovation: Links a synthetic hurricane-track simulator with high-resolution hydrodynamic modelling to estimate future hurricane surge hazards.
18. Bridging Data Gaps in Structural Fragility Modeling through Transfer Learning: Methodology and Case Studies
Core Problem: Fragility models often fail under scarce target labels, domain shift, and class imbalance.
Key Innovation: Compares instance-based, parameter-based, hierarchical Bayesian, and multi-source transfer-learning strategies for interpretable fragility adaptation.
19. Rock weathering can counteract river CO2 emissions induced by permafrost thaw
Core Problem: Permafrost thaw releases organic carbon to rivers, but mineral weathering may partly offset emissions through inorganic carbon reactions.
Key Innovation: Combines CO2 emission and weathering evidence to quantify how biological and geological carbon cycles interact in thawing permafrost rivers.
20. PINN-FEM-based approach to hydro-mechanical analysis of tunnels in anisotropic strata
Core Problem: Tunnel stability in anisotropic strata depends on coupled seepage and mechanical fields that are expensive to resolve with conventional numerical schemes.
Key Innovation: Combines physics-informed neural networks with finite-element analysis for hydro-mechanical modelling of tunnels in anisotropic ground.
21. Damage Tomography and Instability Prediction of Rock Under Tensile Loading Based on Passive Acoustic Emission Monitoring
Core Problem: Brittle rock failure offers few acoustic-emission events and short precursory windows, limiting early-warning reliability.
Key Innovation: Develops passive acoustic-emission damage tomography to locate evolving damage and predict instability under tensile loading.
22. A neural-symbolic thermodynamic framework for explicit constitutive modeling of geomaterials
Core Problem: Classical geomaterial constitutive models can be difficult to generalize and interpret under complex loading paths.
Key Innovation: Uses a neural-symbolic thermodynamic framework to derive explicit, interpretable constitutive models for geomaterials.
23. Rainfall-Stratified Explainable Machine Learning for Quantifying Nonlinear Drivers of Waterlogging Severity: A Case Study in Shanghai, China
Core Problem: Urban waterlogging severity depends on nonlinear drivers whose effects vary with rainfall regime.
Key Innovation: Uses rainfall stratification and explainable machine learning to quantify nonlinear controls on waterlogging severity in Shanghai.
24. WildfireCube: A Dense Spatiotemporal Tensor to Support Multi-Regime Wildfire Spread Modeling at 30 m/3 h Resolution
Core Problem: Wildfire spread models need dense, high-resolution spatiotemporal data that can represent multiple fire regimes.
Key Innovation: Introduces a 30 m, 3-hour spatiotemporal tensor designed to support wildfire spread modelling across regimes.
25. Hydraulic pressure initiates ice lens formation and growth in freezing coarse-grained soil
Core Problem: Coarse-grained soils are often considered less frost-susceptible, yet hydraulic pressure can still drive ice-lens formation and frost heave.
Key Innovation: Clarifies how hydraulic pressure initiates and sustains ice-lens growth in freezing coarse-grained soil.
26. Lateral force on stabilizing piles under seismic condition considering stress redistribution in vertical and horizontal plane
Core Problem: Stabilizing-pile design under earthquake loading depends on lateral forces that are altered by soil arching and stress redistribution in both vertical and horizontal planes.
Key Innovation: Derives a pseudostatic formulation for lateral force on stabilizing piles under seismic conditions and verifies it against experimental and numerical results.
27. A reduced-order constitutive model for rock joints capturing continuous shear stiffness evolution
Core Problem: Rock-joint shear models often represent pre- and post-peak behavior with discontinuous transitions and limited numerical efficiency.
Key Innovation: Develops a reduced-order constitutive model that captures continuous shear-stiffness evolution in rock joints.
28. CNN Enhanced Random Finite Element Analysis of Lateral Wall Movements in Deep Excavations
Core Problem: Deep excavations in spatially variable soils require uncertainty-aware prediction of lateral wall movements.
Key Innovation: Combines convolutional neural networks with random finite-element analysis to estimate excavation-induced wall displacement under spatial variability.
29. Steel Pipe Shrinkable Energy-Absorbing Bolt/Cable for Deep Underground Engineering: Design, Testing, and Performance Evaluation
Core Problem: Deep underground engineering faces large deformation, soft-rock squeezing, and rockburst hazards that can exceed conventional support capacity.
Key Innovation: Designs and tests a shrinkable energy-absorbing bolt/cable system for high-deformation underground support.
30. Cyclic and Post-Cyclic Shear Behavior of the Geogrid-Loess Interface under Direct Shear Loading
Core Problem: The cyclic response of geogrid-loess interfaces affects the seismic performance of reinforced soil structures in loess regions.
Key Innovation: Uses cyclic and post-cyclic direct shear tests to quantify interface behavior under seismic-type loading.
31. Exploring the Influence of Equatorial Waves on a Record-Breaking Extreme Precipitation Event in Central Sahel: Insights From Convection Permitting Simulations
Core Problem: Record-breaking Sahel precipitation events remain difficult to attribute to interacting tropical wave dynamics.
Key Innovation: Uses convection-permitting simulations to isolate how equatorial waves shaped an extreme precipitation event linked to severe flooding.