The complex, static displacement of a very long period seismic signal observed at Soufrière Hills volcano, Montserrat, WI

Highlights

• Restitution of VLP ground displacement by including the spectral components that go beyond the natural period of the seismometer

• Joint interpretation of velocity and displacement seismograms

• Extending the range of intermediate band (30 and 60 s) seismometers into broader band sensors

• Moment tensor inversion of VLP signals

Abstract

In this study we demonstrate how very-long period (VLP) volcanic seismic signals can be processed in order to obtain essential and detailed information about the seismo-volcanic source process. As an example we use the VLP signal observed on 23 March 2012 during an outgassing event at Soufrière Hills volcano, Montserrat, acquired by instruments with different natural periods. The aim of this study is to highlight the importance of retrieving the correct source time function by a complete restitution process. When ground displacement cannot be retrieved through the restitution process due to very narrow band-pass limited instrument response, we compare synthetic and observed waveforms in the velocity domain and determine the best model by generating a synthetic velocity seismogram using the band-limited seismometer characteristics. Furthermore, we show how this approach of forward modelling can reveal much more detail of the source process, since small changes in displacement are enhanced in the velocity seismogram. Using these restituted and modelled displacements we perform a moment tensor inversion combined with a grid search locating the source at 600 m depth below sea level and estimating the source volume change to be in the range of 0.6−1.1 × 10³ m³.

Introduction

Volcanogenic seismic signals cover a broad frequency range and fall into three main categories, and their interpretation and modelling are at the core of any attempt to forecast volcanic eruptions. Volcano-tectonic (VT) earthquakes, generated by the brittle failure of the rocks around the fault plane due to the stress changes of magmatic emplacement or due to pressure changes as a result of water-magma interaction in hydrothermal systems (Neuberg, 2020) have the same characteristics as tectonic earthquakes: clear P- and S-wave onsets and frequency content of 1–20 Hz. Low-frequency (LF) earthquakes have successfully been used in forecasting volcanic eruptions (e.g. Chouet, 1996). They have a spectral range between 0.2 and 10 Hz, with end members of the continuum being Long-Period (LP) events and hybrid events, which are similar to LP events but have additional high frequency onset (Neuberg, 2020). Their source processes differ significantly from the ones for generation of VTs. LF earthquakes originate at the boundary between magmatic fluid and solid surrounding rock (e.g. Chouet, 1988; Neuberg et al., 2000) or can be caused by slow, quasi-brittle low stress-drop failure driven by short-lived upper-edifice deformation (Bean et al., 2014). The deployment and widespread use of broadband seismic networks in the 1990s made studies of very-long period (VLP) signals possible (Kawakatsu et al., 1992; Neuberg et al., 1994) and we focus in our study on this category. VLP signals, whose periods range from several seconds to several minutes, have been observed on almost every type of volcano around the world (Chouet and Matoza, 2013). When the periods of these signals fall into the far end of the range they are often referred to as an ultra-long period (ULP) seismic signals. The event we are describing in this study falls into that range of ULPs, however we choose to call it a VLP as the source process between these two types of signals does not differ. Their source processes are usually attributed to fluid-rock interactions such as mass movement of volcanic fluids (e.g. Chouet and Dawson, 2011) generating abrupt pressure changes inside the volcanic edifice. As VLPs have been observed prior to caldera collapse (Kumagai et al., 2001; Michon et al., 2009) and prior to phreatic eruptions (Kawakatsu et al., 2000; Jolly et al., 2017) the need to study them is of great importance for understanding the underlying physical processes. Therefore, it is essential to retrieve the exact source time history in addition to amplitude and moment tensor components. The major advantage VLP signals offer is direct insight in the deformation of the source process. This fact was recognised and studied by Legrand et al., 2000, Legrand et al., 2005. In this study we emphasise the importance of taking into account how different seismometers influence the observed signals and what the necessary processing steps are in order to retrieve the maximum amount of information from the observed waveforms. These processing steps go beyond the usual “instrument removal” applied as a routine by seismic processing packages, which considers the frequency range in the pass-band of the instrument only. In contrast, we try to retrieve information cut-out by the instrument and subsequently, use this information in our moment tensor inversion to estimate the location and mechanism of the source. As an example we use a VLP signal observed on 23 March 2012 during a outgassing event at Soufrière Hills volcano (SHV), Montserrat.