TerraMosaic Daily Digest: July 24, 2026
Daily Summary
Landslide forecasting is extended from historical susceptibility to non-stationary, slope-specific risk. A probabilistic framework for Lishui propagates climate-conditioned rainfall into shallow-landslide likelihood at individual slopes, while multi-source field observations and machine learning track the evolving instability of an open-pit slope. Together, these studies place changing forcing and measured deformation inside the prediction problem rather than treating them as fixed covariates.
Cascade mechanics receive complementary numerical and experimental constraints. Simulations and physical tests show how sequential landslide-dam failures amplify outburst floods, shifting attention from the stability of an isolated blockage to interactions among multiple impoundments. HyFlood applies a surrogate model to compound coastal flooding, while open-data screening and place-based vulnerability analysis show how limited collective capacity and persistent inequality reorganize the geography of flood risk.
Monitoring studies convert weak or distributed signals into operational diagnostics. Microseismic power-law evolution supports dynamic rockburst-time prediction; three-dimensional point clouds enable automatic discontinuity extraction; and a one-shot vision model identifies tunnel seepage with limited labelled data. Parallel advances in soil-moisture retrieval, Antarctic surface-melt mapping and high-wind GNSS reflectometry strengthen observations of hydrological and cryospheric conditions that precede or compound terrain hazards.
Key Trends
Five methodological shifts connect non-stationary landslide probability, interacting dam failures, compound flooding, precursor dynamics and transferable sensing.
- Climate forcing enters local landslide probability: Non-stationary rainfall is translated into slope-specific failure likelihood rather than appended to a static susceptibility map.
- Dam failures are treated as interacting cascades: Numerical and physical models quantify flood amplification across successive landslide blockages instead of resolving one dam in isolation.
- Compound flooding moves into joint probability space: Surrogate modelling and probabilistic wave-water-level analysis combine coastal drivers under present and rising sea levels.
- Precursors are inferred from evolving signal structure: Deformation histories, microseismic power laws, frictional stick-slip criteria and point-cloud discontinuities support time-dependent instability diagnosis.
- Transfer learning expands environmental observability: Physics-informed retrieval and limited-label vision models recover soil moisture, surface melt, high winds and seepage from heterogeneous sensors.
Selected Papers
The leading papers quantify climate-conditioned shallow-landslide probability, amplification across cascading landslide dams, deformation-based open-pit slope prediction, tectonically inherited rockslide weakness and reservoir-bank rock deterioration. Supporting studies extend the evidence base through rockburst precursors, frictional stick-slip experiments, flood-capacity screening, automatic discontinuity mapping and physics-guided environmental sensing.
1. Assessing rainfall-induced shallow landslides under climate change: A slope-specific probabilistic framework in Lishui, Southeast China
Core Problem: Climate change is driving nonstationary changes in rainfall frequency and intensity, thereby increasing the frequency and severity of rainfall-induced shallow landslides in humid mountainous regions.
Key Innovation: The study develops a novel probabilistic framework that integrates site-specific nonstationary rainfall hazard with physically based slope fragility to quantify shallow landslide probability in a changing climate. Results indicate a significant increase in shallow landslide probability when site-specific, climate-driven nonstationary rainfall is incorporated.
2. Amplification of Outburst Floods During Cascading Landslide Dam Failures: Insights From Numerical and Experimental Modeling
Core Problem: The amplification of outburst floods by successive landslide-dam failures is not adequately explained by models of isolated dam breaching.
Key Innovation: Paired flume experiments and numerical simulations quantify how upstream inflow, dam spacing and breach timing control downstream peak discharge and amplification across a two-dam cascade.
3. Deformation analysis and prediction of an instability area in an open-pit slope in Canada using multi-source field data and machine learning algorithms
Core Problem: Slope deformation is a critical indicator of pit wall stability in open-pit mining, yet its prediction remains challenging due to the combined effects of mining operations and environmental conditions.
Key Innovation: The study investigates the use of machine learning to predict deformation trends in a large open-pit mine in British Columbia, using multi-source field-monitoring data.
4. HyFlood: A Surrogate-Model-Based Framework for Compound Coastal Flooding
Core Problem: However, modeling the joint interaction of waves, storm surge, tides, and rainfall remains computationally demanding, limiting the development of fast and reliable forecast tools.
Key Innovation: Here we present HyFlood, a hybrid statistical-numerical downscaling framework capable of computing and mapping high-resolution compound flood hazards while substantially reducing the computational cost compared with fully process-based hydrodynamic modeling. Validation of the surrogates against the physics-based model outputs demonstrates that HyFlood accurately reproduces daily maxima of spatially distributed flooding depths.
5. Evolution of Mechanical Properties in Single‑Fissure Sandstone Under Periodic Water–Rock Interaction
Core Problem: The deterioration of fissured rock masses in the hydro-fluctuation zone of reservoir bank slopes is significant under periodic reservoir water-level fluctuations, posing a potential threat to slope stability.
Key Innovation: To investigate this deterioration process, a periodic water-rock interaction test scheme was designed using a self-developed YRK-2 rock pressure and soak-air drying equipment. The results show that deterioration is relatively pronounced during the first two water-rock interaction cycles, whereas it becomes less significant during the last four cycles.
6. Role of the tectonic structures on the stability of armored slopes of the anticlinal ridges in Dagestan, North-Eastern Caucasus
Core Problem: Attributing the giant prehistoric Dagestan rockslides to magnitude-7 earthquakes conflicts with regional seismic observations and the accepted seismotectonic model.
Key Innovation: Structural evidence supports an alternative mechanism in which inherited folding deformation and wetter palaeoclimate weakened bedding-controlled slopes, allowing low-magnitude shallow earthquakes to trigger catastrophic failure.
7. Dynamic prediction of rockburst time in deep-buried tunnels integrating microseismic monitoring and power-law risk evolution
Core Problem: Rockbursts in deep-buried tunnels excavated by the drilling and blasting (D&B) method posed a serious threat to construction safety, whereas their occurrence time remained difficult to predict.
Key Innovation: Accordingly, this study proposed a dynamic prediction method for rockburst occurrence time intervals that used multi-parameter microseismic (MS) data and power-law risk evolution.
8. Co‐Occurring Flash and Standard Droughts Amplify Global Fire Behavior
Core Problem: Flash droughts, rapid‐onset dry spells intensified by heat and low humidity, are increasing under climate warming, yet their influence on wildfire behavior remains poorly understood.
Key Innovation: Using global drought and satellite‐based fire data sets (2002–2021), we assess how flash droughts impact fire characteristics as compared to standard droughts.
9. Global Pattern of Flash Drought and Impact‐Relevant Early Warning
Core Problem: Flash drought has devastating impacts on terrestrial ecosystems and challenges the existing drought prediction systems due to its rapid onset and intensification.
Key Innovation: Here, we identify regions prone to flash drought according to detections based on the soil water deficit index, quantify the severity and frequency of flash droughts and their past trends, and assess the performance of the solar‐induced fluorescence rapid change index (SIF‐RCI) as a flash drought early warning for different ecosystems across the globe.
10. Probability-based assessment of wave and water-level conditions under sea-level rise in south-eastern Australian coastal waters
Core Problem: Sea-level rise changes the joint distribution of waves and water levels, but coastal assessments often evaluate these drivers separately.
Key Innovation: A probability-based framework combines wave and water-level conditions across present and future sea levels, resolving how joint hydrodynamic exposure changes between the Southern Ocean, Bass Strait and adjacent coasts.
11. Evaluation Criterion for Stable Sliding to Stick–Slip Transition and Surface Damage Evolution of Granite Fractures Under Coupled Stress–Rate Conditions
Core Problem: The stress-rate conditions governing transition from stable sliding to earthquake-like stick-slip, and the associated fracture-surface damage, remain poorly constrained.
Key Innovation: Friction experiments define a quantitative stick-slip index and transition criterion, identify an optimal slip-rate window near 0.0005-0.001 millimetres per second and link release intensity to diagnostic abrasion and secondary-fracture patterns.
12. Integrating capacity deficit into open data flood risk screening in a data-scarce informal settlement in Yaoundé, Cameroon
Core Problem: Integrated flood risk assessment in informal settlements is hindered by scarce hydrometric, exposure, and socioeconomic data, limiting the ability to identify and target the highest-risk locations for disaster risk reduction.
Key Innovation: The authors present an open-data, proxy-based framework for the Ngousso quarter (Yaoundé, Cameroon) that maps hazard, exposure, household vulnerability, and collective capacity on a 100 m grid, and explicitly represents limited collective capacity as risk-increasing capacity deficit (1 − C). Adding (1 − C) increases continuous risk in 88% of grid cells and raises 28% of cells to a higher risk class.
13. A place-based framework for assessing flood-related socio-economic vulnerability in the Yazoo-Mississippi Delta
Core Problem: The Yazoo-Mississippi Delta (YMD) is characterized by persistent riverine flood exposure alongside long-standing socio-economic inequality and uneven adaptive capacity.
Key Innovation: The study examines three spatial dimensions relevant to flood-related social vulnerability in the Delta: socio-economic susceptibility, lack of adaptive capacity, and the threat of riverine flooding. Results indicate that elevated socio-economic susceptibility is concentrated in portions of Tunica, Bolivar, Leflore, and Holmes Counties, with additional clusters in Sunflower, Humphreys, Sharkey, and Quitman Counties.
14. DOWSER: A DINO-based One-shot Water Seepage Recognition network for TBM tunnel inspection
Core Problem: The industrial deployment of supervised models for tunnel damage detection is limited by the high cost of data collection and pixel-level annotation.
Key Innovation: To address this issue, we propose DOWSER (DINO-based One-shot Water Seepage Recognition), a framework designed to achieve high-precision seepage segmentation with only a single annotated sample.
15. Impact of rising groundwater level on uplift behavior of subway stations considering soil–structure interface friction: Centrifuge and numerical investigation
Core Problem: The continuous rebound of groundwater level in Beijing poses a significant challenge to the anti-flotation safety of existing subway stations, which were originally designed under historically lower groundwater level conditions.
Key Innovation: The study investigates the uplift response of a typical cut-and-cover subway station under progressive groundwater level rise, with particular emphasis on the role of soil-structure interface friction. Results indicate that interface friction coefficient significantly influences both structural uplift and ground surface heave.
16. Seismic asymmetric response of a concrete-faced rockfill dam on deep overburden: insights from large-scale shaking table tests
Core Problem: Deep overburden and reservoir loading create poorly resolved asymmetric seismic demands on concrete-faced rockfill dams.
Key Innovation: Large-scale shaking-table tests combine acceleration and pore-pressure measurements with transfer-function and Hilbert-Huang analyses, revealing upstream-biased response, overburden liquefaction, cross-dam seepage and downstream-slope cracking.
17. Rapid and automatic identification of rock discontinuities from 3D point clouds using supervoxel-based segmentation
Core Problem: Manual discontinuity surveys and conventional point-cloud workflows are too slow for rapid, spatially complete rock-mass characterization.
Key Innovation: A supervoxel workflow combines constrained region growing, multilevel merging and saliency-ranked DBSCAN clustering, matching reference surveys while processing millions of points within seconds.
18. A scattering model-informed deep transfer learning framework for surface soil moisture retrieval
Core Problem: However, most existing ML-based methods rely solely on publicly available auxiliary data and reference SSM for nonlinear modeling, which lacks the integration of physical mechanisms.
Key Innovation: To address these challenges, this study proposed a deep transfer learning framework integrating scattering models with various transfer strategies. It achieved a Pearson correlation coefficient (R) ranging from 0.627 to 0.909 and an unbiased root mean square error (ubRMSE) between 0.051 and 0.058 m3/m3 in cross-validation under dependent scenario.
19. SARMelt: MGLANet-driven high-resolution surface melt mapping of circum-Antarctic ice shelves from Sentinel-1 imagery during 2015–2023
Core Problem: However, existing passive and active microwave products, while valuable for long term monitoring, remain inadequate for resolving fine-scale melt features over heterogeneous ice shelf surfaces.
Key Innovation: To address these limitations, we developed a Multiscale Global-Local Attention Network (MGLANet) for circum-Antarctic surface melt mapping. Within this framework, LSAM focuses on local fine-scale structure extraction, GLA models long-range contextual dependencies to characterize macro-scale melt distribution, and MCIM improves multiscale feature fusion across decoder levels.
20. PLGF-NET: A Physics-Guided Local-Global fusion network for GNSS-R high wind speed retrieval
Core Problem: Accurate global sea surface wind speed retrieval under high-wind conditions (> 15 m/s) remains a persistent bottleneck for spaceborne Global Navigation Satellite System Reflectometry (GNSS-R).
Key Innovation: To transcend these limitations, this study proposes the Physics-Guided Local-Global Fusion Network (PLGF-NET), establishing a novel physics-data dual-driven paradigm. Furthermore, geophysical consistency is guaranteed through a selective regularization mechanism governed by Sample Prediction Difficulty (SPD).
21. Introducing a new methodology for applying field experiments to evaluate gully evolution processes
Core Problem: Direct observation of gully-headcut growth under controlled runoff is scarce because existing studies rely heavily on flumes, historical imagery and post-event surveys.
Key Innovation: A portable Marriott-tube apparatus delivers controlled field discharges from 0.5 to 20 litres per second while continuous imaging resolves runoff velocity and headcut erosion across successive flow steps.
22. Modeling Environmentally Driven Seasonal Moisture Migration and Ground Movements in Expansive Clays
Core Problem: The study describes the formulation of a numerical model for simulating environmentally driven one-dimensional (1D) ground movements of expansive clay.
Key Innovation: The authors synthesize and modify highly nonlinear constitutive relationships for (1) hysteretic soil water retention; (2) reversible soil shrinkage and expansion of clay; and (3) hydraulic conductivity, explicitly incorporating desiccation cracks through a multidomain framework and assuming a critical surface crack depth. The results showed very reasonable first-order agreement with the measured data and highlight the potential of the proposed formulation.
23. A Hausdorff-guided deep learning approach for tracking the motion of rotating Arctic ice floes
Core Problem: Rotation and drift in the marginal ice zone degrade conventional feature matching, limiting reliable estimates of ice-floe motion.
Key Innovation: A Hausdorff-guided workflow combines geometric rotation search with learned feature matching, improving the efficiency of optimal-angle estimation by a reported factor of 14.4 over deep-learning-only baselines.
24. Numerical simulation of wave evolution within rigid vegetation using a porous media approach
Core Problem: Coastal vegetation is widely recognized as a nature-based solution for wave attenuation, yet most macroscopic numerical models rely on drag coefficients with limited applicability beyond the laboratory conditions from which they were derived.
Key Innovation: To address this limitation, a porous-media-based numerical model was developed by incorporating an extended Darcy-Forchheimer resistance formulation into the momentum equation to represent vegetation-induced momentum loss. The overall calibration RMSE is 0.0037 m (mean case-wise NRMSE = 4.0%), and the independent validation errors remain on the order of 10−3 m.
25. Designing a novel multi-scale and multi-factor framework for interval-valued significant wave height forecasting
Core Problem: However, existing studies focus solely on point-valued forecasts, which cannot fully characterize the fluctuation range and potential extreme risks of sea states within a given period.
Key Innovation: Therefore, this study designs a novel multi-scale and multi-factor framework for interval-valued SWH forecasting. For each decomposed mode, multivariate relevance vector machine (MVRVM) with a sparse multiple-input multiple-output structure is adopted to jointly predict its lower and upper bounds.
26. SFMF-DETR: A Spatial-Frequency Collaborative Enhancement and Multiscale Feature Fusion Framework for Small Object Detection
Core Problem: However, existing end-to-end frameworks face performance bottlenecks due to inadequate high-frequency detail preservation, fine-grained feature dilution during multiscale fusion, and spatial distortion from traditional upsampling.
Key Innovation: To address these issues, this article proposes SFMF-DETR, a spatial-frequency collaborative enhancement and multiscale feature fusion framework for small object detection. Experimental results on the VisDrone2019-DET, AI-TOD, and RSOD datasets demonstrate that SFMF-DETR achieves significant gains while maintaining real-time efficiency.
27. Three-Dimensional Fine Geological Modeling Utilizing a Novel Implicit Neural Network to Constrain Mesh Partitioning Algorithm From GPR Data
Core Problem: Limited by technologies such as the accuracy of detection data and three-dimensional (3-D) geological modeling, the refined management of geological diseases in coastal power plants is still difficult to meet the actual needs at present.
Key Innovation: The authors propose a 3-D fine geological modeling method for coastal power plants based on ground penetrating radar technology. First, a new implicit neural network (GeoINet) is designed to uniformly encode the spatial coordinates and derived features, automatically learn the nonlinear features of the extracted layer data, predict the layer data of the unknown area, and transform the underground concealed geological interface into an interactive and high-precision 3-D visualization.
28. A Dual-Branch Fusion Framework for Height Estimation From a Single Satellite Image
Core Problem: Height estimation from a single satellite image is a critical task in remote sensing image analysis.
Key Innovation: Addressing the challenges posed by multisize objects and complex spatial relationships in satellite images, this article proposes a dual-branch fusion framework for height estimation from a single satellite image. Through the deterministic transformation based on statistical priors, the height distribution is mapped to a standardized latent space to achieve absolute height estimation.
29. An improved grout bag–secondary grouting (GBSG) method with time-dependent viscosity modeling for hydrodynamic sealing in karst groundwater conduits
Core Problem: Large-scale karst conduits pose persistent challenges for underground sealing because of strong groundwater flow, highly irregular geometries, and severe grout loss.
Key Innovation: In this study, an innovative grout bag-secondary grouting (GBSG) methodology is introduced. Overall, by developing a new numerical viscosity model capable of capturing time-dependent grout rheology, this study enables the optimization of a novel grouting method for large-scale karst conduits and provides a generalizable numerical framework for process innovation and optimization of non-Newtonian grouting in complex hydrogeological environments.
30. Role of coarse particle morphology and relative density in suffusion of gap-graded soils: experimental insights
Core Problem: Suffusion is a major internal instability process that can affect the hydraulic stability and subsequent shear response of gap-graded soils used in road subgrades, embankment fills, and levees.
Key Innovation: The study investigates the coupled effects of coarse particle morphology and relative density on suffusion behaviour and the mechanical response after suffusion of internally unstable binary gap-graded soils. Controlled suffusion tests were performed using an improved triaxial suffusion apparatus, followed by gradation analyses of different specimen layers and undrained triaxial compression tests.
31. Diagnosing Submesoscale Divergence From Sea Surface Height
Core Problem: The Surface Water and Ocean Topography satellite mission now delivers global sea surface height (SSH) observations at scales fine enough to resolve submesoscale eddies (<50 km).
Key Innovation: Here, we present a new dynamical framework that diagnoses ageostrophic currents and, in particular, divergent motions directly from SSH. The framework is trained and validated using a high‐resolution numerical simulation of a western boundary current system, where submesoscale eddies are the most energetic.
32. PND-Net: A Unified Framework With Attention Fusion and Dynamic Feature Mining for Visible-Infrared Object Detection
Core Problem: Currently, dual-stream fusion architectures incorporating transformers are challenged by the suppression of detailed features and insufficient mining of common features across modalities.
Key Innovation: This article proposes PND-Net, a framework that includes the pixel-wise attention (PWA), neighborhood interaction fusion (NIF), and dynamic feature mining (DFM) modules, which effectively enhance details in each modality and improve feature fusion from local to global, while collaboratively mining common and unique features of each modality.
33. Embedding geological realizations in global sensitivity analysis: a Morris-based screening approach for complex aquifer systems
Core Problem: Global sensitivity analysis (GSA) is increasingly used for interpreting complex hydrogeological models and identifying the relative importance of different sources of uncertainty.
Key Innovation: In this study, we propose an extension of the Morris’ Elementary Effects Test to incorporate geological uncertainty represented through multiple geostatistical facies realizations in a GSA framework.