richardscottoz / be-a-geoscience-detective

Why choose between Data Science and Geoscience? Do both!

Be a geoscience detective

In this repo I resume work done at the 2018 Geophysics Sprint (organized by Agile Scientific ahead of the Annual SEG conference), which resulted in the error flag demo notebook in the repo in-bruges.

The inspiration came after watching How to be a statistical detective, a Stanford University webinar.

I presented this work as a lightning talk I gave at the Transform 2020 virtual conference organized by Software Underground.

The notebook demonstrates how one can asses the quality of multiple predictions against one another (from Inversions and/or Machine Learning models) and against ground truth (for example from geophysical logs), by:

1. Calculating the difference between prediction and ground truth at each sample
2. Flagging statistically significant differences based on a user-defined distance (in deviation units) from either the mean difference or the median difference

3. Further characterizing the error as percentage within a specific zone, and with a cofindence interval via Boostrapping

Conda setup

To create the conda environment for this tutorial run:

conda env create -f environment.yml

License

The figure used is copyright of the Canadian Society of Exploration Geophysicists.

The rest is open source, and copyright of Matteo Niccoli, 2020:

Text is licensed under a CC BY Creative Commons License.

Code is Licensed under the Apache License, Version 2.0.

Be-a-geoscience-detective

A notebook demonstrating how to quantitatively QC the results of seismic inversion or other geophysical processes from a paper.

To run the notebook interactively with Binder click on the button below: