Geophysical modeling & inversion with pyGIMLi

The pyGIMLi tutorial at Transform 2022

Instructors: Thomas Günther¹, Carsten Rücker ², Florian Wagner ³

¹ Leibniz Institute for Applied Geophysics, Hannover, Germany

² Berlin University of Technology, Department of Applied Geophysics, Berlin, Germany

³ RWTH Aachen University, Applied Geophysics and Geothermal Energy, Aachen, Germany

About

pyGIMLi is an open-source library for modeling and inversion in geophysics. This tutorial is particularly suited for new users but adds up to last-years tutorial on Transform 2021 that was covering model building and synthetic modellings on the equation and application levels, and some standard inversion of synthetic and field data, plus how to use an own forward operator. This tutorial will add on to this and go into some more details about the underlying classes but will mainly focus on user-specific inversion:

• DataContainer, Mesh, matrix types, transformations, frameworks
• geostatistical regularization vs. classical smoothness types
• individual treatment of subsurface regions
• incorporation of prior knowledge
• different kinds of joint and coupled inversion
• induced polarization modelling and spectrally constrained inversion
• outlook and overview of other packages in the pyGIMLi ecosystem (BERT, COMET, SAEM, custEM)

Table of contents

• Geophysical modeling & inversion with pyGIMLi II
  • About
  • Table of contents
  • BEFORE THE TUTORIAL
  • Setup instructions
  • Step 1: Prerequisites
  • Step 2: Download material for the tutorial
  • Step 3: Install the tutorial environment
  • Step 4: Start JupyterLab
  • Schedule

BEFORE THE TUTORIAL

Make sure you've done these things before the tutorial on Tuesday:

1. Sign up for the Software Underground Slack
2. Join the channel #t22-tue-pygimli channel. This is where all communication will happen and where we will answer any question about installation and the tutorial
3. Install the pyGIMLi conda environment as described below.

Setup instructions

Quick setup for experienced users

If you are working on Mac or Linux and have worked with conda and have git installed, you can copy & paste these lines separately. For all others, we recommend to carefully read the descriptions of individual steps below.

git clone https://github.com/gimli-org/transform2022
cd transform2022
conda env create
conda activate pg-transform2022
python -c "import pygimli; pygimli.test(show=False, onlydoctests=True)"
jupyter lab

To start the tutorial setup, please follow the next steps:

Step 1: Prerequisites

There are a few things you'll need to follow the tutorial:

1. A working Python installation (Anaconda or Miniconda). For details on how to install Anaconda, we refer to: https://docs.anaconda.com/anaconda/install/
2. A modern web browser that works with JupyterLab or Jupyter Notebook (Internet explorer will not work)
3. Intermediate experience in Python programming (Python, numpy, matplotlib, jupyter)
4. Background on geophysical modeling and inversion

Step 2: Download material for the tutorial

• Download the course material and unzip it a folder of your choice.
• or, alternatively, open a terminal, navigate to a folder of your choice, and execute git clone https://github.com/gimli-org/transform2022.

Step 3: Install the tutorial environment

1. Open a terminal (Linux & Mac) or the Anaconda Powershell Prompt (Windows). Navigate to the folder from step 2 (using the cd command) and type:

conda env create

2. Activate the environment in the terminal by typing:

conda activate pg-transform2022

3. To test if everything works correctly you can do the following:

python -c "import pygimli; pygimli.test(show=False, onlydoctests=True)"

If none of these commands gives an error, then your installation is working fine. If you get any errors, please let us know on Slack at #t22-tue-pygimli.

Step 4: Start JupyterLab

1. Windows users: Make sure you set a default browser that is not Internet Explorer.
2. Activate the conda environment: conda activate pg-transform
3. Start JupyterLab: jupyter lab
4. Jupyter should open in your default web browser. We'll start from here in the tutorial and create a new notebook together.

Schedule

Introduction into pyGIMLi, looking back and forward

Diving into the details of main classes: DataContainer, Meshes and Matrices

Explaining regularization on behalf of a simple model

Region-specific regularization using an aquatic ERT inversion

Inversion with a-priori constraints: from optimizing options to joint inversion

Complex-valued induced polarization modelling and inversion

Outlook and overview on other projects