Feasibility study of spectral pattern recognition reveals distinct classes of volcanic tremor

Systematic investigations of the similarities and differences among volcanic tremor at a range of volcano types may hold crucial information about the plausibility of inferred source mechanisms, which, in turn, may be important for eruption forecasting. However, such studies are rare, in part because of an intrinsic difficulty with identifying tremor signals within very long time series of volcano seismic data. Accordingly, we develop an efficient tremor detection algorithm and identify over 12,000h of volcanic tremor on 24 stations at KÄ«lauea, Okmok, Pavlof, and Redoubt volcanoes. We estimate spectral content over 5-minute tremor windows, and apply a novel combination of Principal Component Analysis (PCA) and hierarchical clustering to identify patterns in the tremor spectra. Analyzing several stations from a given volcano together reveals different styles of tremor within individual volcanic settings. In addition to identifying tremor properties common to all stations in a given network, we find localized tremor signals including those related to processes such as lahars or dike intrusions that are only observed on some of the stations within a network. Subsequent application of our analysis to a combination of stations from the different volcanoes reveals that at least three main tremor classes can be detected across all settings. Whereas a regime with a ridge of high power distributed over 1-2Hz and a gradual decay of spectral power towards higher frequencies is observed dominantly at three volcanoes (KÄ«lauea, Okmok, Redoubt) with magma reservoirs centered at less than 5km below sea level (b.s.l.), a spectrum with a steeper slope and a narrower peak at 1-2Hz is observed only in association with open vents (KÄ«lauea and Pavlof). A third regime with a peak at approximately 3Hz is confined to two stratovolcanoes (Pavlof and Redoubt). These observations suggest generic relationships between the spectral character of the observed signals and volcano characteristics such as magma viscosity, storage depths, and the physical properties of volcanic edifices. Similarities among the spectral patterns detected at stations 4km and 8-10km distance from the centers of volcanic activity, respectively, indicate that path effects do not strongly influence spectral shapes at distances of a few kilometers from the inferred source of the signals. Our preliminary work shows that a global comparison of tremor is feasible. Our results suggest that further work on data from a larger sample and diverse range of volcano types will reveal additional classes of tremor signals and plausibly identify fingerprints of source processes that are specific to volcano type, but independent of volcano location.