EGU General Assembly 2020
© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

Characteristics of seismic activity of Villarrica Volcano

Johanna Lehr and Wolfgang Rabbel
Johanna Lehr and Wolfgang Rabbel
  • Christian-Albrechts-Universität zu Kiel, Institut für Geowissenschaften, Geophysik, Kiel, Germany (

Villarrica is one of the most active and dangerous volcanoes in Chile. During the last decade it consisted of a single open vent hosting an active lava lake which produced mild stombolian explosions, persistent tremor and continuous degassing.

We present an analysis of the seismic activity of Villarrica between 2010 and 2012. Periods of increased lava lake activity are characterized by numerous small transient events which exibit a variety of waveforms and spectral characteristics. Statistical analysis of interevent times revealed a periodic occurrence. At comparable volcanic systems (Stromboli, Erebus), such distributions of events indicated unusual periods of activity corresponding to magma injection. Methods of blind signal separation (ICA, PCA) were used to analyse the wavefield. While regional and local tectonic earthquakes can easily be separated, the tremor and transient events from the crater can not.

How to cite: Lehr, J. and Rabbel, W.: Characteristics of seismic activity of Villarrica Volcano, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-15005,, 2020

Comments on the presentation

AC: Author Comment | CC: Community Comment | Report abuse

Presentation version 1 – uploaded on 01 May 2020
  • CC1: Comment on EGU2020-15005, Luca De Siena, 03 May 2020

    Very interesting thanks! The 3 Hz frequency band is well known to produce source-biases in medium-dependent waveform-reconstruction methods. At 6 Hz I would generally expect no influence of the conduit and have been working with attenuation and scattering tomography inferring deep structures from their variations. How long is the 6-Hz frequency signal going to be recorded in your signals? Does it disappear once you are far from the lake/conduit?


    Luca De Siena

    • AC1: Reply to CC1, Johanna Lehr, 04 May 2020

      Thank you for your comment. The frequency content of the signal generally drops below 5 Hz within 1-2km distance from the summit crater. I just did a similar analysis on another set of stations ~2km away from the summit and about 800m lower. There is a very faint peak around 6 Hz in one of the independent components but I am not sure whether that is to be taken seriously. The 3-Hz peak is still present but much less prominent than at the crater stations.

      Do you have a recommendation where to learn more about the 3-Hz band that you mentioned?

      Thank you,

      Johanna Lehr

      • CC2: Reply to AC1, Luca De Siena, 04 May 2020

        All the tomography works done using coda-wave attenuation generally fail in volcanic environments. There is a debate regarding if this is due to resonant sources (e.g., Chouet et al. 1995) or to medium-dependent deformation of the volcanic edifice (Bean et al. 2014, Nature: Geosciences). Obviously, this looks more important for seismicity stronger than the one you use. However, even for volcano-tectonic events, it is nowadays demonstrated that the sensitivity of coda waves is way more surficial than expected in volcanoes. In De Siena et al. 2016 (EPSL) we show how we can map the debris avalanche of the volcano (max 50 m depth) using earthquakes as deep as 18 km. One of my PhDs, Simona Gabrielli, has just published a paper regarding these medium biases (Gabrielli et al. 2020, GJI) where we show this is likely due to the interaction of the wavefield to very local scatterers (as you also mention in the presentation) that transform the body wave into surface waves.

        This is crucial for our "tomographic community" as we tend to  map the 3 Hz frequency as "deep" in volcanoes: (Prudencio et al. 2013a,b, GJI; Del Pezzo et al. 2016; Prudencio et al. 2018, JGR: Solid Earth). Things might change when the level of heterogeneity is not so strong, as in the average faulted crust (Napolitano et al. 2020, Geosciences Frontiers).

        Thanks for your answer and "see you" tomorrow,


  • CC3: Comment on EGU2020-15005, Agnes Dakota Wansing, 04 May 2020

    This sounds interesting. I am not familiar with the seismicity of volcanos, so sorry for a maybe dumb question but you relate the measured fundamental resonance frequency with the length of the pipe, depending on the velocity. Are there other methods you can use to further constrain the velocity? Or are all these velocities from 1000 – 2000 m/s equal likely?

    • AC2: Reply to CC3, Johanna Lehr, 04 May 2020

      At the moment, II treated the velocities equally. Basically, I  use the range to get an idea about what the length of the conduit could be given the fundamental frequency. A better constraint could probably be obtained from the magma composition and especially analysis of the gas content of the magma.

    • CC4: Reply to CC3, Luca De Siena, 04 May 2020

      No dumb question at all! First: there is no easy way to constrain velocity in volcanic edifice, simply because it is very hard to "pick waves" in there. My main concern woul be that the actual value of the velocity could be lower than 1000 m/s. It is demonstrated that Young or Bulk Moduli in volcanic rocks can be as low as 5 GPa (see presentation of Heap et al. in this session). If so, depending on density and assuming shear modulus is not too far from it, you ge velocities way lower than 1000 m/s. If so, this could have effect on the aproximations you use? 

      Ths always gets into the debate source vs medium effect, very difficult to solve...

      All the best


      • AC3: Reply to CC4, Johanna Lehr, 04 May 2020

        The figure shows the conduit lenghts for velocities of 500-2500m/s. For 500m/s the length would be ~50m. Richardson et al., 2014 ( estimated lake depths of 100-200m. Thus 50 m would be very low, meaning the free magma surface of the lake would have been deep inside the conduit. That seems rather unlikely since the volcano was rather active during the period of observations. Hence the magma level was rather high.

        So yeah, if it turned out that the velocity of the magma filling in my simple pipe model would need to be much lower than 1000m/s, this model is likely wrong.

        • CC5: Reply to AC3, Luca De Siena, 04 May 2020

          Thanks Joanna for the time taken to answer. A pity we cannot discuss it in Vienna!