Search

Additional file 1: Figure S1. Seismic velocity (single station cross-component approach) estimated using (a) the stretching and (b) the MWCS approaches between 0.1 and 1 Hz. The pink period in 2012–2013 delineates the magmatic eruption and the last two red lines mark the phreatic eruptions. The black and red lines correspond to the daily mean values across different combinations of components (EN, EZ and ZN) smoothed using a Hodrick–Prescott filter with different values, 500,000 and 100,000,000, respectively. See Fig. 4 for the description of the vertical lines corresponding to key time periods. Key periods are (1) 2011 earthquake swarm (2) dome extrusion (3) April 2016 eruption. Each tick on the x-axis corresponds to 1 January of each year.

Additional file 2: Figure S2. Waveforms (left) and corresponding spectra (right) of the main classes of volcanic earthquakes. The amplitudes are normalized and a log scale if used for the x axes of the spectra.

Additional file 3: Figure S3. Compilation of the different observables during the time period studied (a) lake level height above the outlet (cyan; Christenson et al. 2017) and percentage of lake level (dark blue; see Hamling, (2017; 2020) for a description) (b) fumarole F0 (dark blue) and lake temperatures (cyan, Christenson et al. 2017) (c) SO2 flux measured with a DOAS (light blue lines and smoothed using a Hodrick Prescott filter in brown) and with a Cospec (magenta circles) (d) CO2 flux (red circles and smoothed line) and CO2/SO2 ratios (green circles) (e) RSAM (filtered between 2 and 5 Hz) computed using the seismic data at WIZ (black circles and smoothed version using a Hodrick Prescott filter in cyan). Each tick on the x-axis corresponds to 1 January of each year. Roman numbers correspond to different periods of reference described in the Chronology section.

Additional file 4: Figure S4. Time evolution of earthquake seismicity. Bars correspond to the total number of earthquakes of each type per week. Red bars denote eruptions and the pink section encompasses the unrest period from 2012 to 2014.The very-long period (VLP) families are from Park et al., 2020 (this issue). High frequency (RHF), volcano-tectonic (VT) and long-period earthquake (LP) characteristics are described in the text. Note that earthquake detections are not available for (a), (c) and (d) in 2018. Each tick on the x-axis corresponds to 1 January of each year. Roman numbers correspond to different periods of reference described in the Chronology section.

Additional file 8: Figure S8. Violin plot showing the yearly distribution of the 5-min main frequencies (dashed lines show median and interquartile). A kernel density estimator is used to show the distribution shape of the data; wider sections representing higher probabilities. Violin plots are useful to assess how data are distributed and if a single/multiple sources are activated.

Additional file 9: Figure S9. DSAR results since 2000. See Fig. 4 for the description of the vertical lines corresponding to key time periods. The 1-day DSAR data have been smoothed with a Hendrick Prescott filter, which prevents temporal aliasing (blue line (parameter = 1000) and orange (parameter = 108) lines). The few days of data in 1989 had DSAR around 1.

Additional file 7: Figure S7. Seismic velocity variations (single station cross-component approach) for the different combinations of components (EN, EZ and ZN). The 1-day results have been smoothed with a 30-day rolling mean. See Fig. 4 for the description of the vertical lines corresponding to key time periods.

Additional file 10: Table S1. Analytical results for fumarole F0. SD refers to sample date, ST is sampling temperature (°C), RCO = log(CO/CO2) and RCH = log(CH4/CO2). Concentrations are µmol/mol.

Additional file 6: Figure S6. Correlation between lunar cycles and seismic amplitude for different window lengths. Faded colors correspond to measurements that are below the 5σ confidence level for correlation between the variables. Blue symbols are for positive correlation and red symbols are for negative correlation. See Fig. 4 for the description of the vertical lines corresponding to key time periods.

Additional file 2. A dendrogram based on the absolute value of correlation coefficient. Two families (F1 and F2) belong to a cluster (orange lines) except for one low SNR waveform, and all U events denote black lines.

Additional file 4. Peak-to-peak amplitude and cluster tree of a F1 and b F2 sub-families resulted from threshold of r = 0.85. The seven eruptions, unrest and swarm periods are denoted by reversed triangles (red), shaded areas (green) and horizontal bars (black), respectively. The colours in both top and bottom plots correspond.

Additional file 1. A list of triggered time (UTC: dd-mm-yyyy hh:mm:ss), peak-to-peak (P2P) amplitude (μm/s), event order in the cluster tree, and family index (F1, F2 and U) for VLPs recorded at WIZ and WSRZ stations. This table supports Fig. 2.

Additional file 5. Cumulative amplitude of F1 (blue) and F2 (red). The seven eruptions are denoted by reversed triangles (red). Time map of two VLP families (F1 and F2). Axes indicate the time intervals between each VLP and its preceding (x axis) and subsequent (y axis) events. Swarm and intermittent VLPs are separately clustered.

Additional file 3. Template matching results (gray dots) without semblance and SNR filtering. This catalogue includes two families (F1 and F2) as well as teleseismic data and lower SNR events. The vertical axis is peak-to-peak (P2P) amplitude (μm/s) computed using the vertical component of WIZ station.