Abstract
Volcanoes can produce a range of eruptive behavior even during a single eruption, changing quickly from effusive to explosive style, and the other way around. The changes in eruption phases (e.g. phreatic explosion, magmatic explosion, lava extrusion, etc.) can lead to different volcanic hazards and require timely assessment for the implementation of mitigation measures. Here we explore how to correlate a given eruption phase with changes in the monitoring data using statistical analysis and conditional probabilities. We calculate the success of detection of an eruption phase using a threshold of monitoring data, which includes the uncertainty on the eruption phase dates with a Monte Carlo simulation. We apply the method to dome forming eruptions of Mt. Merapi (Indonesia) and evaluate their time occurrence using an exceptionally long monitoring time series (from 1993 to 2012, over nineteen years) of Multiphase (Hybrid) Seismic Energy. We identify the seismic energy threshold that is associated with the lava extrusion phase with an accuracy of 90 ±2%, precision of 73 ± 2%, specificity of 96 ± 1%, and sensitivity of 56 ± 1%. We further test our method with the recent 2018 eruption (not used in the thresholds calculations) and we identify the lava extrusion with a precision of 67%, specificity of 70%, and sensitivity of 92%. We also seismically detected the 2018′s onset of the lava extrusion phase 14 days earlier than the visual observation. Given the link between dome-collapse pyroclastic flows and growth episodes of the lava dome at Merapi, our analysis also allows us to establish that 83% of the most energetic pyroclastic flows occur within the first 3 months after the onset of lava extrusion phase. Our method can be applicable to a range of time series of monitoring data (seismic, deformation, gas) and to other volcanoes that have a significant number of past events.
Keywords
Dome formation, Eruptive phases detection, Merapi, Multiphase seismic energy, Probabilistic analysis