Eruption Type Classification from IR Image Sequences at Erebus Volcano, Antarctica
Abstract
The lava lake atop Mount Erebus, Antarctica, is characterized by a cyclic pattern of features such as temperature, heat flux, lake level, and gas flux punctuated with strombolian eruptions. Classification of these eruptions leads us to recognize valuable insights into the interior mechanisms of the lake and its magma chamber. In the first part of this research, we utilized Support Vector Machines (SMVs), a supervised machine learning method to categorize infrared images of Mount Erebus. We use images acquired every 2 seconds from the crater rim from 2013 to 2016. By visually selecting the training components, we separate images into two categories, "eruption" and "no eruption", then classify several weeks from January through April and November and December 2014. The second part of our research focuses on classifying sequences of 6 images into three categories: "eruption", "post eruption", "no eruption", in the same fashion of a previous work (Dye and Morra, PEPI, 2020). We quantify the improvement obtained by applying Machine Learning to image sequences instead of single images in differentiating between different types of eruptions.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2020
- Bibcode:
- 2020AGUFMV021.0020H
- Keywords:
-
- 8414 Eruption mechanisms and flow emplacement;
- VOLCANOLOGY;
- 8419 Volcano monitoring;
- VOLCANOLOGY;
- 8434 Magma migration and fragmentation;
- VOLCANOLOGY;
- 8439 Physics and chemistry of magma bodies;
- VOLCANOLOGY