Mineye-Terranigma Documentation

Overview

Mineye-Terranigma is a research project for probabilistic geological modeling and geophysical inversion. The project combines state-of-the-art implicit geological modeling with Bayesian inference techniques.

Core Capabilities:

  • 3D Geological Modeling: Build implicit 3D models from structural data using GemPy

  • Geophysical Forward Modeling: Compute gravity and magnetic responses from geological models

  • Probabilistic Inference: Quantify uncertainty using Pyro and PyTorch

  • Satellite Image Segmentation: Bayesian classification of remote sensing data

  • Joint Inversion: Integrate multiple data sources for improved geological interpretations

Getting Started

Basic Examples

Foundational workflows for 3D geological modeling and geophysical forward modeling. These examples introduce core concepts using real data from the Tharsis mining district in Spain.

Simple Geological Model - Tharsis Region

Simple Geological Model - Tharsis Region

Complex Geological Model - Tharsis Region

Complex Geological Model - Tharsis Region

Forward Gravity Modeling - Tharsis Region

Forward Gravity Modeling - Tharsis Region

Probabilistic Modeling

Advanced examples demonstrating uncertainty quantification, error propagation, and Bayesian inversion techniques for geological and geophysical applications.

Error Propagation in Geological Models

Error Propagation in Geological Models

Error Propagation for Dip Angles

Error Propagation for Dip Angles

Bayesian Gravity Inversion: Complete Workflow

Bayesian Gravity Inversion: Complete Workflow

Bayesian Magnetic Inversion: TMI Inversion Workflow

Bayesian Magnetic Inversion: TMI Inversion Workflow

Bayesian EnMap Inversion: Categorical Likelihood and Ordinal Probabilities

Bayesian EnMap Inversion: Categorical Likelihood and Ordinal Probabilities

Bayesian Joint Inversion: Gravity and EnMap

Bayesian Joint Inversion: Gravity and EnMap

Hyperspectral Segmentation

Workflows for lithological segmentation of EnMap hyperspectral data using Bayesian inference. These examples cover feature extraction, preprocessing, and model comparison.

EnMap Lithological Segmentation

EnMap Lithological Segmentation

EnMap Data Extraction and Point Sampling

EnMap Data Extraction and Point Sampling

EnMap Likelihood and Model Comparison

EnMap Likelihood and Model Comparison

Key Features

Bayesian Inference with Pyro

Geological Modeling with GemPy

The project uses PyTorch and Pyro for probabilistic programming, enabling:

  • Variational Inference (VI): Fast approximate posterior estimation

  • Hamiltonian Monte Carlo (HMC/NUTS): Accurate sampling for complex posteriors

  • GPU Acceleration: Scalable inference for large-scale inversions

Building on the GemPy framework, the project supports:

  • Implicit Surface Modeling: Continuous 3D geological surfaces from sparse data

  • Structural Complexity: Faults, unconformities, and erosive contacts

  • Forward Modeling: Gravity and magnetic field computations

  • Uncertainty Quantification: Probabilistic geological interpretations

API Reference

Detailed documentation of the Mineye-Terranigma Python API.

Indices and Tables