Best-basis analysis of broadband tremor signals

Active volcanoes usually generate highly non-stationary broadband tremor signals. Short-time shock events with a frequency content of several decades are superimposed on a stationary narrow band continuous tremor. Tremor signals of this type can be observed in the near field of many active volcanoes. In this paper we will demonstrate the analysis of such signals using a specific tremor signal of Mt. Stromboli (Sicily). We used the Best-Basis Algorithm (BBA) in order to compute a spectrogram which is adapted to signal properties on highly different scales. It turns out that the BBA can reveal better fitting properties of the tremor in the time-frequency plane compared to standard methods like Short-Time Fourier Transformation (STFT). Moreover, this very effective algorithm can be used for real time monitoring in the time-frequency plane, for data compression or for de-noising of the tremor signals.