yasur_ml

This repository contains the code accompanying the paper "Waveform features strongly control subcrater classification performance for a large, labeled volcano infrasound dataset" by Liam Toney, David Fee, Alex Witsil, and Robin S. Matoza.

Installing

A conda environment specification file, environment.yml, is provided. You can create a conda environment from this file by executing

conda env create

from the repository root.

This code requires the UAF Geophysics Tools package rtm for generation of the labeled catalog. This package and its dependencies are installed when the above command is executed.

You must define two environment variables to use the code:

• YASUR_WORKING_DIR — the path to this repository
• YASUR_FIGURE_DIR — the directory where figure files should be saved

Workflow overview

1. download_3E.py — download the data
2. build_catalog.py — run rtm to create a CSV catalog
3. label_catalog.py — associate entries in catalog to a subcrater
4. extract_features.py — extract features from waveforms and store in Feather file
5. Apply tools in svm/

Citation

If you use the tools contained in this repository, please cite our paper:

Toney, L., Fee, D., Witsil, A., & Matoza, R. S. (2022). Waveform features strongly control subcrater classification performance for a large, labeled volcano infrasound dataset. The Seismic Record, 2(3), 167–175. https://doi.org/10.1785/0320220019

Acknowledgements

This work was supported by the Nuclear Arms Control Technology (NACT) program at the Defense Threat Reduction Agency (DTRA). Cleared for release.