Skip to content

Abstract and other details

Jump to bottom
Glenn Thompson edited this page Nov 4, 2021 · 2 revisions
  • Title: Automatic Reclassification of Volcano-Seismic Signals from Soufrière Hills Volcano, Montserrat, 1996-2008
  • Authors: Glenn THOMPSON, Alexis FALCIN, Jean-Philippe METAXIAN, Francois Beuducel
  • Poster ID: V25D-0146
  • Session: V25D - Volcano Seismology and Acoustics: Recent Advances in Understanding Volcanic Processes V Poster
  • When: Tuesday, 14 December 2021, 17:00 - 19:00
  • Where: Convention Center - Poster Hall, D-F
  • URL: https://agu.confex.com/agu/fm21/meetingapp.cgi/Paper/888700

Abstract

Seismic activity during the eruption of Soufrière Hills volcano comprised various transient signals, which were classified visually by the Montserrat Volcano Observatory (MVO), considering waveforms recorded at several stations. For 217,290 transients detected on the MVO digital seismic network between 1996/10/21 and 2008/10/16, five main classes have been identified: rockfall (ROC: 58%), hybrid (HYB: 19%), long-period (LPE: 11%), lp-rockfall (LP-ROC: 5.8%), and volcano-tectonic (VT: 3.1%). Temporal trends in the rate and energy release of these different transients (in addition to swarms and tremor) were key to short-term forecasting of eruptive activity. However, visual classification is highly subjective and non-repeatable, and the inconsistency of the catalog is a barrier to research. In a pilot study, we automatically removed waveforms with dropouts, and manually verified transient classifications until we had approximately 100 transients of each class (total 522). We found ~21% of these transients were incorrectly classified at MVO. Our re-labelled dataset was then used as a starting point for supervised learning, using code from [http://github.com/malfante/AAA]. This code was used by Malfante et al. (2008) to classify 109,609 transients at Ubinas volcano with a 93.5% accuracy. They transformed each waveform into a set of 102 features: 34 features for each of three domains (time, spectral, cepstral). We added 6 frequency features of our own, including band ratios, peak frequency, median frequency, bandwidth, and frequency change. The resulting 108-point vectors of features were then used for modeling. The dataset is randomly divided 50 times into training and testing datasets, to produce a robust model. One model is produced per channel. We use the Random Forest Classifier algorithm from the scikit-learn library. For each waveform, a probability is computed for each class.

Initial results are promising. Separate models for 3 channels yield accuracies of 76-80%. If the LP-ROC class is omitted (following Langer et al, 2006), accuracy rises to 82-85%. If only VT and LP classes are considered, accuracy is 96-99%. We intend to expand our labelled dataset to 1000 events, add new features, build models for each channel, and reclassify the catalog of 217,290 transients by a weighted average of probabilities.

Plain language summary

Many different types of small volcanic earthquakes were recorded during an eruption on the Caribbean island of Montserrat that lasted from July 1995 to February 2010. Sometimes hundreds of earthquakes were detected each day. Analyzing so many earthquakes and correctly classify them was a challenging but important task, as trends in the rates, sizes and locations of different earthquake types helped forecast what might happen next. So as MVO Seismologist in 2001, I reached out to Italian colleagues for help in automatically classifying the earthquakes. This led two papers led by Horst Langer in 2003 and 2006 with a classification accuracy of 80%. A real-time system was never realized, but reclassifying the entire Montserrat catalog of >200,000 earthquakes has remained a goal of mine, as we suspect a classification error rate of 20-30%. Objective reclassification will give us a better understanding of the eruption. In 2018, Marielle Malfante led a paper describing software she had developed and applied to more than 100,000 earthquakes from Ubinas volcano in Peru, achieving 93.5% accuracy. I corresponded with Marielle and Jean-Philippe Metaxian, the seismologist on the Ubinas project. This led to the present collaboration, and initial results suggest accuracies of 75-85% are possible.

Session abstract

Hazardous phenomena at volcanoes includes ash plumes, gas emissions, explosions, pyroclastic density currents, lava flows, lahars, and mass wasting. All of these processes produce seismic and infrasound signals that can provide key real-time information for assessing hazardous surface activity. We can also gain insights into the activity state of volcanoes by identifying and tracking the movement of subsurface magma and hydrothermal fluids using seismicity and seismic imaging techniques. Recent advances in hardware technology and data analysis have promoted more precise characterization and quantification of the physical mechanisms leading to and producing hazardous volcanic phenomena. Nevertheless, volcano seismology and acoustics remains a rapidly developing area of research. We welcome submissions that present new seismic and acoustic observations, interpretations, models, instrumentation, or techniques that improve our understanding of volcanic processes and assist future monitoring efforts.

Clone this wiki locally