GeoDFN

Geologically Consistent Discrete Fracture Network Generator

Generate ensembles of geologically plausible fracture networks from outcrop statistics - ready for flow and transport simulations.

Publication: Kamel Targhi et al. (2025) - From outcrop observations to dynamic simulations: an efficient workflow for generating ensembles of geologically plausible fracture networks and assessing their impact on flow and transport - Geoenergy

---

Demo

GeoDFN_GUI.mp4

---

What is GeoDFN?

Fractures are ubiquitous in geological formations and can significantly influence heat and mass transfer in geothermal systems, groundwater management, and CO₂ storage. GeoDFN provides a computationally efficient workflow that bridges outcrop observations and dynamic flow simulations.

GeoDFN uses a hybrid mechanical-statistical approach: fracture lengths, orientations, and spacings are sampled from probability distributions fitted to field data, while fracture placement follows a mechanical rule - fractures are inserted sequentially (longest first), and each new fracture must respect the stress shadow (buffer zone) of existing fractures. This prevents unrealistically close spacing and produces fracture patterns that are geologically consistent without the high computational cost of full fracture growth simulations.

The result is an ensemble of equiprobable fracture networks that honour the statistical properties of the outcrop and can be fed directly into multi-purpose flow and transport simulators such as MRST, DuMux, DARTS, PorePy, and OpenCSMP.

It ships in two forms - a desktop app for interactive use, and a Python library for scripting and large-scale ensemble generation.

---

Get Started

Option A - Python package

For scripting, batch generation, or integration into simulation pipelines, install GeoDFN directly via pip:

pip install GeoDFN

Below is an example:

import numpy as np
from GeoDFN.Classes.DFNGenerator import DFNGenerator

gen = DFNGenerator(
    domainLengthX=300, domainLengthY=600,
    sets=[{
        'I': 0.01,
        'fractureLengthPDF': 'Log-Normal',
        'fractureLengthPDFParams': {'mu': 2.4, 'sigma': 0.73, 'Lmin': 2.59, 'Lmax': 57.48},
        'spatialDistributionPDF': 'Power-law',
        'spatialDistributionPDFParams': {'alpha': 0.51, 'min distance': 1, 'max distance': 600},
        'orientationDistributionPDF': 'Von-Mises',
        'orientationDistributionPDFParams': {'kappa': 8.55, 'loc': 1.4,
                                              'thetaMin': np.radians(30), 'thetaMax': np.radians(120)},
        'bufferZone': {'method': 'constant', 'constant': 1.4},
    }],
    apertureCalculationParameters={'method': 'subLinear', 'scalingCoefficient': 0.001, 'scalingExponent': 0.5},
    DFNName='my_dfn',
    numOfRealizations=10,
)

print(f"{len(gen.realizations)} realizations generated")

Option B - Desktop App (no Python required)

The easiest way to get started without any installation:

1. Download the application
2. Unzip the file
3. Double-click GeoDFN.exe - your browser opens with the interface

Option C - Run the GUI locally

If you have Python installed, you can run the interface directly:

pip install GeoDFN
streamlit run app.py

Your browser will open automatically with the full interface.

---

GUI or Python Package?

	Desktop GUI	Python Package
Best for	Visual exploration, parameter tuning, quick results	Large ensembles, batch generation, simulation pipelines
Setup	Download & double-click	pip install GeoDFN
Realizations	Interactive, one run at a time	Hundreds to thousands, fully automated

For uncertainty quantification studies and generating training data for machine-learning emulators, we recommend the Python Package. It allows full automation over the parameter space that defines the fracture network ensemble.

---

Capabilities

Feature	Options
Fracture length	Log-Normal · Power-law · Exponential · Constant
Orientation	Von-Mises · Uniform · Constant
Spatial distribution	Power-law · Log-Normal · Uniform
Stress shadow (buffer zone)	Constant width · Linear scaling with fracture length
Aperture model	Constant · Sub-linear scaling · Barton-Bandis · Lepillier
Stress correction	Multi-azimuth stress-dependent aperture
Output	Coordinates · Apertures · Statistics · Stereonets · Visualizations

---

Output

Each run writes results to DFNs/<name>/:

DFNs/<name>/
├── fractureCoordinates/       # Start/end (x, y) of each fracture
├── aperture/                  # Aperture values
├── fractureSet/               # Full fracture list per set
├── orientationStereographic/  # Stereonet plots
├── outputPropertiesTotal/     # Network statistics
└── pics/                      # DFN visualizations

---

For Developers

Installation, tests, and building the desktop app

Install with dev dependencies

pip install -e ".[dev]"

Run tests

pytest tests/

Build the desktop app

pip install pyinstaller
python -m PyInstaller geodfn.spec --noconfirm

Distributable is generated in dist/GeoDFN/. Share the entire folder - users double-click GeoDFN.exe.

Project structure

GeoDFN/
├── Classes/
│   ├── DFNGenerator.py                      # Random-seed generator
│   ├── DFNGeneratorWithSeed.py              # Fixed seed-point generator
│   ├── DFNGeneratorWithSeedAndExclusion.py  # Generator with exclusion zones
│   ├── _validation.py                       # Input validation
│   ├── fractureLengthPDFs.py
│   ├── orientationPDFs.py
│   ├── spatialDistributionPDFs.py
│   ├── apertureCalculator.py
│   └── bufferZoneCalculator.py
├── Example-BrazilRandomSeeds.py
├── Example-BrazilFixedSeeds.py
├── Example-BrazilFixedSeedsAndExclusion.py
├── Example-BrazilAperture.py
├── Examples.ipynb
└── PercolationAnalysis.ipynb
app.py          # Streamlit GUI
launcher.py     # Desktop app entry point
geodfn.spec     # PyInstaller build spec

---

Citation

If you use GeoDFN in your research, please cite:

Kamel Targhi, E., et al. "From outcrop observations to dynamic simulations: an efficient workflow for generating ensembles of geologically plausible fracture networks and assessing their impact on flow and transport." Geoenergy 3.1 (2025): geoenergy2025-028.

---

License

MIT License - © 2025 Elahe Kamel Targhi. See LICENSE for details.