An investigation of the relationship between repose time and cumulative moment magnitude preceding volcanic explosions

• Main Author: Tan, Darren Pei Kun
• Other Authors: Benoit Taisne
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2020
• Subjects: Science::Geology::Volcanoes and earthquakes
• Online Access: https://hdl.handle.net/10356/141792
• Institution: Nanyang Technological University

Analysing earthquakes induced by volcanic activity is critical for eruption forecasting as they shed light on the movement of magma underneath the volcano’s edifice. In previous work by Thelen et al. (2010), the relationship between repose times of volcanic eruptions and cumulative moment magnitude (CMM) of precursor seismic swarms was studied across 5 stratovolcanoes. These two parameters demonstrated a potentially linear trend in log-log space, and it was proposed that the continual growth of volcanic plugs over longer repose times creates increasingly competent conduit features which impede magma ascent. Overcoming the obstructions to magma ascent involve greater seismic energy release, and hence a higher CMM. We revisit this study with improved explosion chronologies across 10 volcanoes, each exhibiting long dormancy (≥15 years) prior to reactivation, and attempted linear regression on the log-log plots. The resultant coefficient of determination (R2 value) for all data points is 0.62, while the R2 values for datasets of individual volcanoes, where applicable, range from 0.54 at Redoubt volcano in Alaska to 0.84 at Bezymianny volcano in Russia. These results illustrate that given the robust documentation of explosive activity, CMM and explosion repose time show a substantial linear fit on a log-log scale. Such a plot enables volcano-specific explosion behaviour to be characterised by their precursor seismicity, and it can potentially be used to predict volcanic explosions that follow seismic swarms. This utility, demonstrated in this thesis for 5 volcanoes, yields a combined confusion matrix with 68.3% accuracy and a 55.2% F1-Score. Challenges to this study include setting sensible and accurate constraints on the earthquake spatial filters and temporal windows prior to each volcanic explosion. These carefully selected numbers, as well as the compiled explosion chronologies, are at best supported by existing literature and Global Volcanism Program bulletin reports, which can be qualitative at times. Nonetheless, sensitivity tests are conducted for these filter parameters, and we demonstrate that our results are not vulnerable to the chosen values, even if poorly constrained. Future work involves the incorporation of more volcanoes by mining global databases (such as WOVOdat) and refining the volcanic explosion prediction tool with additional explosion samples.