Probabilistic Sediment Transport and Slope Failure Modeling IODP Expeditions 386 and 405 — Japan Trench Frontal Prism
| # | Notebook | Open |
|---|---|---|
| 01 | JCORES VCD Mining and MTD Catalog |  |
| 02 | Stochastic Slope Stability Model | |
| 03 | Parameter Sensitivity Analysis | |
| 04 | Porewater Geochemistry Profiles | |
| 05 | MTD Physical Property Quantification | |
| 06 | XRF Geochemical Lithology Profile | |
| 07 | LWD Physical Property Analysis | |
| 08 | Failure Probability vs Depth | |
| 09 | Core Image Patch Extractor | |
| 10 | Block Size Distribution (Imagery) | |
| 11 | Block Size Distribution (Colorimetry) | |
| 12 | Core Image Texture Analysis | |
| 13 | VCD Interactive Labeler | |
| 14 | Stochastic Forcing Model | |
| 15 | XCT Density Integration | |
| 16 | Resistivity Image Panels | |
| 17 | Structural Mask vs Stress | |
| 18 | Sediment Stability Pipeline | |
| 19 | Pipeline Summary Figures | |
If ProboSed is used in published research, please cite the software:
Castillo, R. (2025). ProboSed: Probabilistic Sediment Transport and Slope Failure Modeling (v1.0.0). Zenodo.https://doi.org/10.5281/zenodo.20669113
And if referencing the underlying research:
Castillo, R. (2027). Slip Happens: Probabilistic Modeling of Submarine Mass Transport Deposits and Sediment Routing in Convergent Margin Basins. PhD Dissertation, The Ohio State University.
A CITATION.cff file is included for automated citation tools.
Deterministic slope-failure models assume a single failure condition, yet natural sedimentary systems are inherently stochastic. Variations in pore pressure, permeability architecture, grain-size distribution, and seismic forcing produce ensembles of possible failure pathways rather than a single deterministic outcome.
ProboSed provides a probabilistic framework for quantifying submarine slope instability and sediment transport at convergent margins. The pipeline connects Visual Core Description (VCD) observations from IODP core material to a physically motivated stochastic model, validated against independent geophysical and geochemical datasets.
The framework is applied to Site C0019 of the Japan Trench frontal prism, where the 2011 Mw 9.0 Tohoku-oki earthquake produced ~50 m of coseismic slip at the plate boundary décollement (Fulton et al., 2013; Chester et al., 2013).
Slope displacement q(t) and fault slip s(t) are modeled as coupled Ornstein-Uhlenbeck stochastic differential equations:
ds = -k_f · s · dt + σ_s · dW_s
dq = (-γq + αs) · dt + σ_q · dW_q
Slope failure is defined as a first-passage problem:
τ = inf{ t > 0 : q(t) > θ(x) }
where the failure threshold θ(x) is derived from the VCD stability index:
θ(x) = clip( x · 2/3, 0, 2 ) x ∈ {0, 1, 2, 3}
Disturbance categories from IODP standardized VCD classification map to a 0–3 stability score:
| Score | Fabric class | Sedimentary interpretation | What to watch for that would make this structural instead | Failure threshold θ |
|---|---|---|---|---|
| 3 | Intact bedding | Undisturbed hemipelagic or turbidite depositional fabric | None — intact bedding is unambiguous | 2.00 |
| 2 | Coherent block | Partially remobilized; coherent blocks with preserved internal stratigraphy suggest early-stage failure or distal MTD | Tilted but unbroken bedding could be fault drag — check for slickensides or calcite veins | 1.33 |
| 1 | Scaly fabric | Pervasive shear fabric consistent with basal shear zone of MTD or active failure surface | High risk of structural misclassification — scaly fabric also diagnostic of fault gouge; needs physical property cross-check | 0.67 |
| 0 | Slurried / MTD | Complete remobilization; homogenized matrix indicates flow or liquefaction | Brecciation from coseismic slip can look identical — check for hydrothermal mineralization or sharp bounding surfaces | 0.00 |
The linear mapping is the maximum-entropy choice given only the constraint that threshold increases monotonically with stability score.
Model parameters are informed by published IODP proceedings:
| Parameter | Value | Physical basis |
|---|---|---|
| σ_q = 0.6 | Slope noise amplitude | Frontal prism Vp = 1550–1750 m/s, resistivity = 0.5–1.8 Ω·m (Exp 405 LWD, Unit I) |
| γ = 1.0 | Slope damping | Moderate restoring force; higher values (1.5–2.0) for seismically strengthened hemipelagic sediment (Exp 386) |
| α = 0.5 | Fault-slope coupling | Pc' = 17 MPa overconsolidation supports non-zero coupling (Exp 343, Valdez et al. 2015) |
| slip_mag = 3.0 | Mainshock impulse | Scaled from ~50 m coseismic slip at C0019 décollement (Fulton et al. 2013) |
ProboSed/
├── notebooks/
│ ├── 01_labeling.ipynb JCORES VCD mining and MTD catalog
│ ├── 02_toy_model.ipynb Stochastic slope stability simulation
│ ├── 03_sensitivity.ipynb Parameter sensitivity analysis
│ ├── 04_geochem.ipynb Porewater geochemistry profiles
│ ├── 05_mtd_quantification.ipynb MTD physical property quantification
│ ├── 06_xrf_profile.ipynb XRF geochemical lithology profile
│ ├── 07_lwd_analysis.ipynb LWD physical property analysis
│ ├── 08_failure_probability.ipynb Failure probability vs depth
│ ├── 09_patch_extractor.ipynb Core image patch extraction
│ ├── 10_block_size_imagery.ipynb Block size distribution from imagery
│ ├── 11_block_size_colorimetry.ipynb Block size distribution from L* profiles
│ ├── 12_texture_analysis.ipynb GLCM texture analysis of core images
│ ├── 13_vcd_labeler.ipynb Interactive VCD labeling widget
│ ├── 14_stochastic_forcing.ipynb Stochastic forcing model with SRCMOD data
│ ├── 15_xct_density.ipynb XCT density integration per MTD section
│ ├── 16_resistivity_panels.ipynb RAB resistivity image panels
│ ├── 17_structural_stress.ipynb Structural mask vs borehole breakout stress
│ ├── 18_stability_pipeline.ipynb Sediment stability pipeline (legacy)
│ └── 19_summary_figures.ipynb Pipeline workflow and summary figures
│
├── slope/
│ ├── __init__.py
│ ├── stability.py Core physics — OU model, Lyapunov, sensitivity
│ ├── toy_model.py Standalone simulation script
│ └── sensitivity_run.py Sensitivity analysis script
│
├── core_ml/
│ ├── __init__.py
│ └── labeler.py JCORESMiner and VCDLabeler
│
├── geochem/
│ ├── __init__.py
│ └── geochem_analysis.py Porewater and headspace gas depth profiles
│
├── transport/
│ ├── __init__.py
│ └── agents.py Agent-based sediment transport ensemble
│
├── utils/
│ ├── __init__.py
│ └── patcher.py Core image patch utility
│
├── requirements.txt
├── pyproject.toml
├── CITATION.cff
├── LICENSE
└── README.md
All notebooks are designed to run in Google Colab without local installation. Each notebook imports from the slope/, core_ml/, geochem/, transport/, and utils/ modules — the .py files are the source of truth and the notebooks are interactive wrappers.
| Notebook | Purpose | Key outputs |
|---|---|---|
| 01_labeling | Extracts stability scores from JCORES VCD PDFs using standardized IODP disturbance terminology. Identifies MTD boundaries as contiguous intervals where stability ≤ 1. Supports all five Site C0019 holes; outputs are tagged by hole. | C0019J_VCD_stability_log.csv, C0019J_MTD_catalog.csv, C0019J_stability_profile.png |
| 02_toy_model | Runs the coupled OU slope stability ensemble under slope-only and fault-coupled forcing. Produces trajectory figures, failure probability comparison, and Lyapunov exponent estimate. | Three dissertation figures, failure probability, Lyapunov exponent |
| 03_sensitivity | One-at-a-time parameter sensitivity analysis across physically motivated ranges. Tests robustness of the qualitative result across α, σ_q, slip_mag, γ, and θ. | sensitivity_results.csv, sensitivity figure, robustness table |
| 04_geochem | Produces five-panel porewater and headspace gas depth profiles for C0019J and C0019M. Overlays MTD intervals from notebook 01 when the catalog CSV is available. | Three geochemistry figures |
| 05_mtd_quantification | Statistical comparison of resistivity, Vp, and porosity inside and outside VCD-derived MTD intervals. Validates that the stability index captures real mechanical contrasts. | C0019J_MTD_statistics.csv, C0019J_MTD_properties.csv, depth profile and box plot figures |
| 06_xrf_profile | Identifies the depth of the mafic geochemical discontinuity (SiO2 < 50 wt%) and produces the integrated four-panel physical/structural profile. | C0019J_XRF_profile.png, C0019J_integrated_profile.png |
| 07_lwd_analysis | Loads the LWD LAS file from Site C0019H and runs Mann-Whitney U comparisons of resistivity and Vp between MTD and stable matrix intervals. | Statistical comparison table |
| 08_failure_probability | Monte Carlo failure probability versus depth using vane shear strength and MAD data. Uses quantile-based depth binning to accommodate sparse sampling. | Failure probability depth profile |
| 09_patch_extractor | Extracts sliding-window image patches from ruler-free core section photographs, classified as MTD or intact using the MTD catalog. | Patch archive (MTD/intact subdirectories), patch_manifest.csv |
| 10_block_size_imagery | Detects block boundaries in core photographs using a combined brightness and lateral-variance signal. Fits a power law to the cumulative block size distribution. | C0019J_block_sizes_imagery.csv, distribution figures |
| 11_block_size_colorimetry | Block boundary detection from MSCL colorimetry L* profiles. Power law fit to test scale-invariant fragmentation. | Block size distribution, power law fit |
| 12_texture_analysis | GLCM texture metrics (contrast, correlation, energy, homogeneity), orientation statistics, and autocorrelation length from core image patches. | Texture metric depth profiles |
| 13_vcd_labeler | Interactive ipywidgets labeling interface for core image patches. Records lithology, disturbance classification, confidence, and notes to CSV. | C0019J_VCD_labels.csv |
| 14_stochastic_forcing | Two-component SDE model with optional SRCMOD Tohoku finite-fault slip-rate forcing. Compares response-only and forced ensemble failure probabilities. | Ensemble figures, s_forcing.csv |
| 15_xct_density | DICOM CT Hounsfield unit profiles per MTD section from XCT scan data. | Density profiles by MTD interval |
| 16_resistivity_panels | RAB resistivity image panel extraction for MTD-9 and MTD-10 from Hole C0019H. | Resistivity image panels |
| 17_structural_stress | Structural overprint flags versus Conin et al. JFAST/JTRACK breakout stress data. | Structural-stress comparison figure |
| 18_stability_pipeline | Early VCD lexicon parser, geotechnical data merge, and stability profile figure generation (legacy pipeline). | Stability depth profile |
| 19_summary_figures | Graphviz workflow diagram and two-panel mechanical regime depth profile. | vcd_pipeline.png, regime figure |
Clone the repository:
git clone https://github.com/rocknrene/ProboSed.git
cd ProboSed
pip install -r requirements.txtOr install as a package:
pip install probosedfrom slope.stability import run_ensemble, calculate_lyapunov, threshold_from_vcd
# Run fault-coupled ensemble
# Parameters informed by IODP Expeditions 343, 405, 386
q, s, p_fail, transported = run_ensemble(
N_paths = 10_000,
gamma = 1.0, # slope damping
sigma_q = 0.6, # slope noise (weak frontal prism material)
alpha = 0.5, # fault-slope coupling
slip_mag = 3.0, # mainshock impulse (~50 m coseismic slip, scaled)
threshold = 1.0, # failure threshold
mainshock_t = 5.0,
)
print(f"Failure probability: {p_fail:.3f}")
# Estimate Lyapunov exponent
lyapunov, log_growth = calculate_lyapunov(q, dt=0.01, warmup_fraction=0.01)
print(f"Lyapunov exponent: {lyapunov:.4f} per unit time")
# Map a VCD stability score to a failure threshold
theta = threshold_from_vcd(2) # coherent block -> theta = 1.33
print(f"Threshold (score 2): {theta:.2f}")from core_ml.labeler import JCORESMiner
miner = JCORESMiner('405-C0019J_VCD.pdf', 'Summary_C0019J.xlsx')
backbone = miner.build_backbone()
vcd_df = miner.extract(backbone)
mtds = JCORESMiner.score_to_mtd_catalog(vcd_df, stability_threshold=1)
print(f"MTD intervals identified: {len(mtds)}")
print(mtds)from geochem.geochem_analysis import load_iw, load_gc, plot_C0019J
import pandas as pd
iw_j = load_iw('SummarySheet-IW-Hole_Exp405_C0019J_250312.xlsx')
gc_j = load_gc('SummarySheet-GC_200_160206_Exp405_C0019J.xlsm', iw_j)
# optional: overlay MTD boundaries from the VCD pipeline
mtd_catalog = pd.read_csv('C0019J_MTD_catalog.csv')
plot_C0019J(iw_j, gc_j, 'C0019J_geochemistry_profiles.png',
mtd_catalog=mtd_catalog)from transport.agents import SedimentAgentModel, forcing_from_slope, calculate_clast_distribution
# compute physically motivated forcing from slope geometry
# Japan Trench frontal prism: ~8 degree slope, moderate pore pressure
forcing = forcing_from_slope(
slope_angle_deg = 8.0, # degrees
pore_pressure_ratio = 0.35, # lambda — moderate overpressure
)
# run grain transport ensemble
model = SedimentAgentModel(
settling_velocity = 0.02, # m/s — fine silt, D50 ~20 microns
drag_coeff = 0.5, # natural irregular grains
)
results = model.run(n_agents=10_000, n_steps=500, forcing=forcing)
model.summary(results)
# compute clast distribution (proxy for MTD deposit thickness profile)
counts, edges, centers = calculate_clast_distribution(results['final_positions'])from utils.patcher import slice_core_image, batch_slice
# slice a single core section scan into 256×256 patches
result = slice_core_image(
image_path = 'C0019J_section_14K_1.tif',
output_folder = 'patches/14K_1/',
patch_size = 256,
overlap = 0,
)
print(f"{result['n_patches']} patches written to {result['output_folder']}")
# or process a whole folder of core scans at once
batch = batch_slice(
input_folder = 'core_scans/',
output_root = 'patches/',
patch_size = 256,
)
print(f"{batch['total_patches']} total patches from {batch['n_images']} images")The stability index is validated against five independent datasets, each measured by different instruments with different error structures:
| Dataset | Physical basis | Expected signature at MTD intervals |
|---|---|---|
| P-wave velocity (Vp) | Acoustic impedance — sensitive to porosity and cementation | Lower Vp in weak, disaggregated material |
| Electrical resistivity | Pore fluid connectivity | Lower resistivity in high-porosity remobilized zones |
| Porewater geochemistry | Fluid advection through permeable failure surfaces | SO4, Ca, and alkalinity anomalies co-located with MTD boundaries |
| Borehole breakouts | In-situ stress field from wellbore deformation | Stress reorientation at mechanical discontinuities |
| Lithostratigraphic column | Sedimentary facies from hand-specimen description | Chaotic or structureless intervals at modeled failure zones |
Convergence of anomalies across independent datasets at the same stratigraphic horizons provides evidence that the stability index captures real mechanical contrasts rather than artifacts of any single measurement method.
- v0.1 — Probabilistic slope-failure framework ✓
- v0.2 — JCORES VCD extraction pipeline ✓
- v0.3 — Porewater geochemistry integration ✓
- v0.4 — Parameter sensitivity analysis ✓
- v0.5 — Agent-based sediment transport ensemble ✓
- v0.6 — Core image patch extraction utility ✓
- v0.7 — Block size distribution from core imagery ✓
- v0.8 — GLCM texture analysis pipeline ✓
- v0.9 — LWD and XRF geochemical integration ✓
- v0.10 — Interactive VCD labeling widget ✓
- v1.0 — Integrated MTD simulation framework
ProboSed is designed for research in:
- Submarine landslides and mass-transport deposits
- Earthquake-triggered sediment transport at convergent margins
- Probabilistic slope stability modeling
- Marine sediment routing systems
- IODP core data integration with geophysical observations
Development uses core and logging data from the Japan Trench margin, IODP Expeditions 386 and 405, Site C0019.
Contributions are welcome. Please open an issue to discuss proposed changes before submitting a pull request.
MIT License — see LICENSE for details.
Development of ProboSed is part of doctoral research at The Ohio State University, School of Earth Sciences.
Advisors Dr. Brendan Crowell — Co-Advisor Dr. Jill Leonard-Pingel — Co-Advisor
Dissertation Committee Dr. Cole · Dr. Keep · Dr. Krissek
Field Science Dr. Christine Regalla — Co-Chief Scientist, IODP Expedition 405
Data This work uses data from the International Ocean Discovery Program (IODP), Expeditions 386 and 405, Japan Trench margin.
Chester, F.M., et al. (2013). Structure and composition of the plate-boundary slip zone for the 2011 Tohoku-oki earthquake. Science, 342(6163), 1208–1211.
Fulton, P.M., et al. (2013). Low coseismic friction on the Tohoku-oki fault determined from temperature measurements. Science, 342(6163), 1214–1217.
Valdez, R.D., et al. (2015). Data report: permeability and consolidation behavior of sediments from the northern Japan Trench subduction zone, IODP Site C0019. Proc. IODP, 343/343T.
Expedition 405 Scientists (2025). Site C0019. In Proc. IODP, 405.
Alstott, J., Bullmore, E., & Plenz, D. (2014). powerlaw: A Python package for analysis of heavy-tailed distributions. PLOS ONE, 9(1), e85777.
Haralick, R.M., Shanmugam, K., & Dinstein, I. (1973). Textural features for image classification. IEEE Transactions on Systems, Man, and Cybernetics, 3(6), 610–621.