Dense Array Studies of Volcano-Tectonic and Long-Period Earthquakes Beneath Mount St. Helens

A 904 single-component 10-Hz geophone array deployed within 15 km of Mount St. Helens (MSH) in 2014 recorded continuously for two-weeks. Automated reverse-time imaging (RTI) was used to generate a catalog of 212 earthquakes. Among these, two distinct types of upper crustal (<8 km) earthquakes were classified. Volcano-tectonic (VT) and long-period (LP) earthquakes were identified using analysis of array spectrograms, envelope functions, and velocity waveforms. To remove analyst subjectivity, quantitative classification criteria were developed based on the ratio of power in high and low frequency bands and coda duration. Prior to the 2014 experiment, upper crustal LP earthquakes had only been reported at MSH during volcanic activity. Subarray beamforming was used to distinguish between LP earthquakes and surface generated LP signals, such as rockfall. This method confirmed 16 LP signals with horizontal velocities exceeding that of upper crustal P-wave velocities, which requires a subsurface hypocenter. LP and VT locations overlap in a cluster slightly east of the summit crater from 0-5 km below sea level. LP displacement spectra are similar to simple theoretical predictions for shear failure except that they have lower corner frequencies than VT earthquakes of similar magnitude. The results indicate a distinct non-resonant source for LP earthquakes which are located in the same source volume as some VT earthquakes (within hypocenter uncertainty of 1 km or less). To further investigate MSH microseismicity mechanisms, a 142 three-component (3-C) 5 Hz geophone array will record continuously for one month at MSH in Fall 2017 providing a unique dataset for a volcano earthquake source study. This array will help determine if LP occurrence in 2014 was transient or if it is still ongoing. Unlike the 2014 array, approximately 50 geophones will be deployed in the MSH summit crater directly over the majority of seismicity. RTI will be used to detect and locate earthquakes by back-projecting 3-C data with a local 3-D P and S velocity model. Earthquakes will be classified using the previously stated techniques, and we will seek to use the dense array of 3-C waveforms to invert for focal mechanisms and, ideally, moment tensor sources down to M0.