Automatic Reclassification of Volcano Seismic Signals from Soufrière Hills Volcano, Montserrat, 1996 to 2008

• Abstract and other details

Motivation

Eruption. Seismic monitoring best technique. Continuous data problems: need to identify anomalous signals from noise. Andesitic eruptions dominated by transient seismic event signals. 6 classes used by MVO (state them). MVO team seismic overwhelmed by events: often hundreds (sometimes thousands) per day, impossible to classify consistently and objectively. Needed to objectively reclassify catalog. Goal was to detect, classify, locate and quantify every event in real-time. GT part of MVO team seismic in 1996 and 2000-2004. Approached Langer in 2001, led to papers in 2003 and 2006.

Current project workflow

1. MVO Seisan event database to Pickle/CSV files
2. Preparing a set of labelled events from Pickle/CSV files
3. Applying Malfante's AAA code base

Miscellaneous codes

Desired project workflow

(Major refactoring would be necessary)

1. MVO Seisan event database to PyMSEC format
2. Preparing a set of labelled events from PyMSEC format
3. Applying Malfante's AAA code base

Poster outline including results

Data formats & directory structures

• Seisan event database
• PyMSEC format - Python Meta Seismic Event Catalog format