Intro

Welcome to the documentation and development site for Zephyr: an open-source seismic full-waveform inversion toolkit. Zephyr is designed to offer a convenient, easy-to-use set of tools and components that enable seismic FWI. Most importantly, Zephyr is free and open-source, so you can use it for your own projects and/or contribute to its development. Zephyr continues to gain features and improve daily.

Poster: AGU 2015

Download

To download the source code for the Zephyr project, please visit the project page on GitHub.

Branches

master
The current main branch of development for the Zephyr project.
ani_testing
Ongoing work testing anisotropic modelling code
distributor
Ongoing work on distributed computing schedulers for composite problems.
disctests
Ongoing work to write tests for the discretizations.
inversion
Ongoing work to implement a full-featured waveform inversion workflow
oldzephyr_simpeg-integration
Development from 2014 and early 2015, which is still being laboriously merged into master. This includes some of the more advanced work on distributed computing, which is gradually being included in zephyr.middleware.distributors.

Dependencies

Zephyr is written in Python, and tested using the CPython implementation on Mac OS and Linux (at present the authors know of no reason that it wouldn't work on Windows, but this is untested).

Zephyr makes extensive use of NumPy and SciPy in order to support numerical operations. At present (and by design), zephyr.backend depends only on NumPy, SciPy and the Python standard library.

The frontend and middleware layers of Zephyr depend on IPython for some of the parallel computing functionality.

The frontend and middleware layers of Zephyr depend on SimPEG, which is a flexible toolkit for solving geophysical inversion problems, written in Python. Visit the SimPEG GitHub page to download the project, or install with pip.

The frontend command-line interface depends on Click.

Zephyr aims to be compatible with MUMPS, and a corresponding Python-based wrapper called pymatsolver, but this functionality is experimental.

Plotting and analysis tools make use of matplotlib.