TerraMosaic Daily Digest: April 15, 2026
Daily Summary
This April 15, 2026 digest distills 42 selected papers from 1,513 analyzed records. The April 15 digest is led by papers that refuse to stop at static susceptibility surfaces. The strongest landslide studies instead follow consequence, transport, and path dependence. The Hongzhuangzi Gully paper reconstructs a full rainfall-to-landslide-to-channel cascade. The Aratozawa study focuses on why extreme runout was possible on an ultra-gentle slope. The Oregon damability paper moves from individual landslide dams to regional blockage potential. Debris-flow initiation is pushed down to particle-shape effects, while the path-dependent clustering paper argues that landslide timing itself carries predictive structure. Even the machine-learning papers are strongest when they stay attached to a specific failure product: deposition prediction, imbalanced susceptibility mapping, cross-modal detection, and sinkhole segmentation all improve because the target process or geomorphic prior is explicit.
The rest of the set broadens that same logic across floods, cryosphere hazards, and geohazard intelligence. Flood papers are strongest when they quantify a decision variable rather than just produce another map: undercounted agricultural losses, interpretable flood magnitude estimation, multisite cascading flood simulation, urban hydrological vulnerability, and gorge-scale water-level reconstruction all fit that pattern. RTSEvo and the Longyearbyen avalanche paper keep slow and seasonal hazards process-aware. A second axis of progress is the shift toward knowledge-enhanced geohazard reasoning, visible in the CGHD dataset, the geohazard-intelligence review, MiMapper, SeisMoLLM, and multimodal change-understanding work. Overall, this is a day of papers that make hazard models more usable by anchoring them to transport pathways, geomorphic priors, or operational variables.
Key Trends
The best papers today do not just map hazards; they add physically meaningful variables and make the result easier to defend or act on.
- Consequence-aware landslide analysis dominated the strongest papers: Hongzhuangzi Gully, Aratozawa, damability modelling, deposition prediction, and path-dependent clustering all focus on what happens after failure starts rather than on trigger lists alone.
- Geomorphic priors and multi-modal fusion mattered more than generic AI: Cross-modal landslide detection, SinkSAM-Net, CGHD, and the geohazard-intelligence review all gain traction by tying learning systems back to terrain structure or knowledge representation.
- Flood papers stood out when they quantified a decision variable: Agricultural flood loss underestimation, interpretable flood magnitude estimation, stochastic cascading flood simulation, canyon water levels, and urban hydrological vulnerability are useful because they produce something operationally legible.
- Mechanism stayed visible in the better engineering papers: Supported-slope seismic stability, levee failure, mine-dump hydro-mechanics, canal-slope cracking, and avalanche barriers all remain tied to the physical mode of failure or control.
Selected Papers
This digest features 42 selected papers from 1,513 papers analyzed, led by consequence-aware landslide cascades, runout and deposition modelling, and imbalanced or multimodal hazard detection, then widening into flood-risk variables, cryosphere process models, and knowledge-enhanced geohazard intelligence.
1. Cascading hazards triggered by the 7 August 2025 extreme rainfall in Hongzhuangzi Gully, Gansu, China
Core Problem: The 2025 Hongzhuangzi disaster needs reconstruction as a coupled rainfall-to-slope-to-channel cascade rather than as isolated failures.
Key Innovation: The paper traces how extreme rainfall generated linked landslide, debris-flow, and flood impacts in a single gully system, clarifying the full event chain.
2. The damability function: a probabilistic approach to regional landslide dam susceptibility analysis applied to the Oregon Coast Range, USA
Core Problem: Regional landslide-dam assessment still lacks a scalable way to estimate where landslides are most likely to create valley-blocking dams.
Key Innovation: This study introduces a probabilistic damability function and applies it regionally to the Oregon Coast Range to map where dam-forming failures are most plausible.
3. Comprehensive hazard assessment of translational landslides through multi-model integration: a case study of Yanshan Township in the Three Gorges Reservoir area
Core Problem: Translational landslide hazard assessment in the Three Gorges area is still fragmented when different instability indicators are used separately.
Key Innovation: The paper integrates multiple models into one hazard framework to rank translational landslide threat more coherently in Yanshan Township.
4. Rockfall hazard assessment and governing strategy for a cliff feature zone based on the high-resolution online images and terrain analysis
Core Problem: Rockfall management at cliff-feature sites remains weak when monitoring imagery and terrain controls are not analyzed together.
Key Innovation: This study combines high-resolution online imagery with terrain analysis to assess rockfall hazard and propose site-specific control strategies.
5. Influence of Soil Particle Shape on Debris Flow Initiation: A Coupled CFD-DEM Study
Core Problem: Debris-flow initiation is still oversimplified when grain-scale particle shape effects are ignored in initiation physics.
Key Innovation: A coupled CFD-DEM framework shows how particle shape changes pore-fluid interaction and the onset conditions for debris-flow mobilization.
6. Long runout of the Aratozawa landslide on an ultra-gentle slope triggered by the 2008 Iwate-Miyagi Nairiku Earthquake
Core Problem: The exceptional runout of the Aratozawa landslide remains difficult to explain using standard slope-angle expectations alone.
Key Innovation: The paper re-examines why the landslide traveled far across an ultra-gentle slope and sharpens the mechanics of long-runout coseismic failures.
7. AI-enhanced landslide deposition prediction: a novel framework integrating discrete element method and generative adversarial networks
Core Problem: Landslide deposition prediction is still too slow for rapid use when it depends only on traditional numerical simulation.
Key Innovation: This framework couples DEM-based simulation with GANs to treat landslide deposition prediction as an image-generation problem with much faster inference.
8. Tree-based and deep learning models for landslide susceptibility assessment with imbalanced data
Core Problem: Landslide susceptibility models often degrade when strong class imbalance is handled naively.
Key Innovation: The study compares tree-based and deep learning approaches explicitly under imbalanced-data conditions and clarifies which modelling choices remain robust.
9. Spatiotemporal clustering for landslide hazard prediction revealed by the path-dependent perspective
Core Problem: Landslide prediction still misses important memory effects when failures are treated as temporally independent events.
Key Innovation: This paper shows that path dependence and spatiotemporal clustering add predictive structure beyond static susceptibility mapping alone.
10. RTSEvo v1.0: a retrogressive thaw slump evolution model
Core Problem: Permafrost-hazard models still lack operational tools for simulating how retrogressive thaw slumps evolve through time.
Key Innovation: RTSEvo provides an explicit process model for thaw-slump growth and retreat, giving cryosphere hazard work a more mechanistic evolution framework.
11. Underestimated agricultural losses due to flooding
Core Problem: Flood-loss accounting still understates agricultural damage, which distorts downstream risk and adaptation planning.
Key Innovation: This study shows that flood-driven agricultural losses are being underestimated and quantifies the scale of that bias.
12. Interpretable feature incorporation machine-learning framework for flood magnitude estimation
Core Problem: Flood magnitude models often gain accuracy at the cost of interpretability, making them harder to trust in practice.
Key Innovation: The paper builds an interpretable ML framework that incorporates explicit predictors rather than treating flood magnitude estimation as a black box.
13. Cross-modal Feature Fusion of Heterogeneous Remote Sensing Data for Improving Landslide Detection
Core Problem: Landslide detection still underuses the complementary signal structure available across heterogeneous remote-sensing modalities.
Key Innovation: This study shows that cross-modal fusion improves landslide detection by combining heterogeneous remote-sensing cues more effectively than single-stream workflows.
14. SinkSAM-Net: Knowledge-Driven Self-Supervised Sinkhole Segmentation Using Topographic Priors and Segment Anything Model
Core Problem: Sinkhole mapping remains data-hungry and brittle when segmentation models are trained without geomorphic priors.
Key Innovation: SinkSAM-Net combines self-supervision, topographic priors, and SAM-style prompting to improve sinkhole segmentation under limited labels.
15. CGHD: Dual-Temporal Dataset of Composite Geological Hazards via Multi-Source Optical Remote Sensing Images
Core Problem: Composite geohazard detection still lacks benchmark datasets that represent hazard change through time rather than single snapshots.
Key Innovation: CGHD provides a dual-temporal multi-source optical remote-sensing dataset designed for composite geological hazard mapping and evaluation.
16. Toward Knowledge-Enhanced Geohazard Intelligence: A Review of Knowledge Graphs and Large Language Models
Core Problem: Geohazard AI workflows are expanding quickly, but their knowledge structures, reasoning limits, and integration paths remain scattered.
Key Innovation: This review synthesizes how knowledge graphs and large language models can be combined for more interpretable geohazard intelligence.
17. Co-Seismic Landslide Detection Combining Multiple Classifiers Based on Weighted Voting: A Case Study of the Jiuzhaigou Earthquake in 2017
Core Problem: Single-classifier landslide mapping can be unstable when applied to complex coseismic damage scenes.
Key Innovation: The paper uses weighted voting across multiple classifiers to improve landslide detection for the Jiuzhaigou earthquake case.
18. Machine Learning in Slope Stability: A Review with Implications for Landslide Hazard Assessment
Core Problem: Machine learning in slope stability has proliferated faster than coherent evaluation of what is genuinely useful for landslide practice.
Key Innovation: This review consolidates the main ML directions in slope stability and connects them back to landslide hazard assessment needs.
19. Landslide Susceptibility Assessment Using AHP, Frequency Ratio, and LSI Models: Understanding Topographical Controls in Hanang District, Tanzania
Core Problem: Regional susceptibility mapping is still informative when it isolates the terrain controls that dominate slope failure in poorly characterized areas.
Key Innovation: This study compares AHP, frequency-ratio, and LSI approaches to clarify topographic controls on landsliding in Hanang District.
20. Landslide Susceptibility Assessment Based on a Quantitative Continuous Model: A Case Study of Wanzhou
Core Problem: Many susceptibility models remain class-based and discontinuous, making it harder to interpret gradual spatial change in hazard.
Key Innovation: The paper applies a quantitative continuous model to Wanzhou and emphasizes smoothly varying landslide susceptibility rather than coarse zonation.
21. Canal slope stability under the coupling effect of desiccation crack dynamics, hydro-climatic seasonality, and water level fluctuations
Core Problem: Canal-slope failure is still mischaracterized when crack evolution, seasonality, and water-level fluctuations are analyzed separately.
Key Innovation: This study treats canal slope stability as a coupled hydro-climatic and crack-dynamics problem and resolves the combined effect on failure propensity.
22. Geohazard Assessment of Historic Chalk Cavity Collapses in Aleppo, Syria
Core Problem: Historic cavity-collapse hazards remain under-assessed where subsurface voids and urban legacy conditions are poorly documented.
Key Innovation: The Aleppo case assembles historical evidence and hazard analysis to clarify the urban geohazard posed by chalk cavity collapses.
23. Present‐Day Kinematics and Seismic Hazards in Turkey: Insights From a TVR‐Optimized Block Model and High‐Resolution GNSS Strain Rates
Core Problem: Turkey’s seismic hazard remains hard to interpret without jointly resolving block motion, slip partitioning, and distributed strain.
Key Innovation: A TVR-optimized block model and high-resolution GNSS strain analysis provide a clearer picture of present-day kinematics and moment accumulation.
24. Effect of Earthquake and Hydrostatic Water Pressure on the Seismic Stability of Slopes Supported by Mechanically Stabilized Earth Retaining Walls
Core Problem: Seismic stability of supported slopes is often simplified by neglecting how hydrostatic loading changes earthquake response.
Key Innovation: The paper quantifies how earthquake forcing and hydrostatic water pressure interact in slopes supported by mechanically stabilized retaining walls.
25. Numerical Modeling of Levee Failure Mechanisms by Integrating Seepage and Stability Processes
Core Problem: Levee-failure assessment remains incomplete when seepage and stability are modeled in separate stages.
Key Innovation: This work integrates seepage and stability into one levee-failure modelling framework to capture coupled breach mechanisms more realistically.
26. Multi-scale investigation of the hydro-mechanical behavior of open-pit mine dump soils regulated by polyacrylamide and fly ash
Core Problem: Open-pit dump slopes remain vulnerable under intense rainfall because hydraulic regulation and mechanical reinforcement are rarely optimized together.
Key Innovation: The study shows how polyacrylamide and fly ash jointly alter infiltration and strength across scales in mine-dump soils.
27. A GNN routing module is all you need for LSTM Rainfall–Runoff models
Core Problem: Rainfall-runoff models still lose physical structure when basin connectivity is treated too crudely.
Key Innovation: This paper shows that adding a graph-based routing module can materially improve LSTM rainfall-runoff performance by representing flow connectivity explicitly.
28. An Attention-Based Stochastic Simulator for Multisite Extremes to Evaluate Nonstationary, Cascading Flood Risk
Core Problem: Cascading flood risk is difficult to evaluate when multisite extremes are simulated without flexible nonstationary dependence structures.
Key Innovation: The paper introduces an attention-based stochastic simulator to generate multisite extremes for nonstationary, cascading flood-risk analysis.
29. SeisMoLLM: Advancing Seismic Monitoring via Cross‐Modal Transfer With Pretrained Large Language Model
Core Problem: Seismic monitoring models still suffer from task fragmentation and limited pretraining compared with other AI domains.
Key Innovation: SeisMoLLM transfers pretrained language-model sequence knowledge into seismic monitoring and improves performance across multiple seismic tasks.
30. Decoding the Delta: Unifying Remote Sensing Change Detection and Understanding with Multimodal Large Language Models
Core Problem: Remote-sensing change detection and change interpretation are still often handled by separate systems.
Key Innovation: This work uses multimodal large language models to connect change detection with higher-level scene understanding in one framework.
31. MiMapper: A Cloud-Based Multi-Hazard Mapping Tool for Nepal
Core Problem: Operational hazard mapping remains hard to scale where analysts lack shared tools for visualizing multiple hazards in one interface.
Key Innovation: MiMapper provides a cloud-based mapping environment for Nepal that consolidates multi-hazard information into a more accessible workflow.
32. LiDAR-Based Delineation and Classification of Alluvial and High-Angle Fans for Regional Post-Wildfire Geohazard Assessment in Colorado, USA
Core Problem: Regional post-wildfire assessment still needs terrain units that distinguish where debris-rich runoff and fan hazards are likely to organize.
Key Innovation: This study uses LiDAR to delineate and classify fan systems relevant to post-wildfire geohazard assessment in Colorado.
33. Mass Movements in Wetlands: An Analysis of a Typical Amazon Delta-Estuary Environment
Core Problem: Mass movements in wetland and delta-estuary settings remain less documented than their upland counterparts.
Key Innovation: The paper characterizes how mass movements operate in a representative Amazon delta-estuary wetland environment.
34. The Snow Avalanches That Hit Longyearbyen in 2015 and 2017 Led to Better Forecasts and Physical Barriers
Core Problem: Avalanche risk reduction is most informative when post-disaster lessons are tied directly to forecast improvement and structural mitigation.
Key Innovation: The Longyearbyen case shows how major snow-avalanche events reshaped both forecasting practice and barrier design.
35. Agricultural Drought Hazard Using Satellite-Based Indices for Drought Risk Mapping in Koel River Basin (India) Through Geospatial Technologies
Core Problem: Satellite-based drought mapping remains important where agricultural drought must be assessed in data-limited basins.
Key Innovation: This study uses geospatial drought indices to map agricultural drought hazard across the Koel River Basin.
36. Integrated Approach to Assessing Spatial Susceptibility to Flooding in the Upper Mono Basin Valley in Togo: Local Perceptions and Multi-Criteria Risk Analysis
Core Problem: Flood susceptibility studies are stronger when formal spatial analysis is combined with how local communities perceive hazard.
Key Innovation: The paper integrates local perceptions with multi-criteria analysis to assess flood susceptibility in the Upper Mono Basin Valley.
37. Hydrological Vulnerability and Flood Risk: Mexico City Study Case
Core Problem: Urban flood studies often stop at exposure without clarifying the hydrological vulnerability structure that produces risk.
Key Innovation: This Mexico City case links hydrological vulnerability directly to urban flood-risk interpretation.
38. Multi-Technique Data Fusion for Obtaining High-Resolution 3D Models of Narrow Gorges and Canyons to Determine Water Level in Flooding Events
Core Problem: Flood response in narrow gorges remains difficult to reconstruct without high-resolution geometry for water-level estimation.
Key Innovation: This study fuses multiple data sources to build detailed 3D gorge and canyon models for more reliable flood water-level analysis.
39. Impacts of Extreme Storms in Surface Water Resources, Systems, and Infrastructure—Evidence from Storm Daniel (2023) in Greece
Core Problem: Storm-impact analysis remains more useful when it follows how extreme events disrupt water resources and infrastructure together.
Key Innovation: Using Storm Daniel in Greece, the paper documents how extreme storms propagated through surface-water systems and infrastructure impacts.
40. Ground-Motion Dataset for Shallow Earthquakes in Colombia
Core Problem: Regional seismic hazard work remains constrained where validated strong-motion datasets are sparse or fragmented.
Key Innovation: This dataset provides a curated ground-motion resource for shallow Colombian earthquakes that can support hazard modelling and benchmarking.
41. Landslide Occurrence and Mitigation Strategies: Exploring Community Perception in Kivu Catchment of Rwanda
Core Problem: Landslide mitigation planning is weaker when community perception and recurring slope hazard are treated separately.
Key Innovation: The paper connects landslide occurrence with local mitigation perceptions in the Kivu Catchment to clarify the human side of risk reduction.
42. Investigating Soil Properties at Landslide Locations in the Eastern Cape Province, South Africa
Core Problem: Basic material-property constraints are still missing for many landslide-prone regions, weakening process interpretation and susceptibility work.
Key Innovation: This study documents soil-property patterns at landslide sites in the Eastern Cape to support more grounded slope-failure analysis.