Seismic precursors to the Whakaari 2019 phreatic eruption are transferable to other eruptions and volcanoes

Volcanic eruptions that occur without warning can be deadly in touristic and populated areas. Even with real-time geophysical monitoring, forecasting sudden eruptions is difficult, because their precursors are hard to recognize and can vary between volcanoes. Here, we describe a general seismic precursor signal for gas-driven eruptions, identified through correlation analysis of 18 well-recorded eruptions in New Zealand, Alaska, and Kamchatka. The precursor manifests in the displacement seismic amplitude ratio between medium (4.5–8 Hz) and high (8–16 Hz) frequency tremor bands, exhibiting a characteristic rise in the days prior to eruptions. We interpret this as formation of a hydrothermal seal that enables rapid pressurization of shallow groundwater. Applying this model to the 2019 eruption at Whakaari (New Zealand), we describe pressurization of the system in the week before the eruption, and cascading seal failure in the 16 h prior to the explosion. Real-time monitoring for this precursor may improve short-term eruption warning systems at certain volcanoes.

To calculate the DSAR, procced as follow: (1) Integrate the bandpass filtered MF and HF data with time.
(2) Take the absolute value and compute averages on 10-minute intervals.
(3) Compute the ratio between integrated MF and HF.
For computing the DSAR median feature proposed as a precursor in this paper follow: (1) Every 10 minutes in the DSAR data stream, take a 48 hours window (looking backwards).
(2) For each window, compute the median.
(3) Construct the feature time series with point every 10 minutes corresponding the medians computed from the 48 hours windows.
The second precursor corresponds to the DSAR rate variance (named for the tsfresh feature that measures the change quantiles variance between percentiles .4 and .6) and is calculated: (1) Every 10 minutes in the DSAR data stream, take a 48 hours window (looking backwards).
(3) For data points between the percentiles (in each window), calculate the difference between subsequent pairs (obtaining a vector of differences). (4) Compute the variance in the differences vector.  Table S1. Basic information on the six volcanoes included in this study indicating the country, the station and network used and its distance to the crater, the number of eruptions recorded, the type and year of eruptions, the length of the seismic record analysed, and the % of continuous data on the record.

Eruption
Time ( Table S2. Catalogue of eruptions included in this study, indicating its time, volcano explosivity index (VEI), eruption type and duration, and column height. See Figure S1. Ruapehu 2009 July 13 wasn't reported as an eruption but as phreatic activity by Geonet (Scott 2013; https://www.gns.cri.nz/static/pubs/2013/SR%202013-045.pdf). By inspection of the available data (seismic, lake temperature, lake level, and rainfall data; see Figure S10), we consider that it might be a missed small eruption given the sharp increase in the lake level (without rain) and a significant peak in the RSAM (MF and HF data too) (given the presence of the lake that could obscure small events, missing small eruptions in Ruapehu is reasonable). By closer inspection, we found that a series of high amplitudes nDSAR median cycles occurred before the event ( Figure S2). Based on these attributes, we decided to incorporate it into the analysis. However, we acknowledge that this is not recognised as an eruption event by GeoNet.

Phase
(1) Magmahydrothermal system interaction Decrease in DSAR median, due to a rapid decrease in MF. Peak in every frequency during decrease of DSAR median (~16 hours before eruption). Inverse RSAM shows a peak and a linear decrease (~16 hours before eruption).
Strong peak in every frequency of the data at the eruption time. Then the signals decrease for 1 day, followed by an increase activity for 4 days (between Dec 10 and Dec 13/14). The system remains with lower activity after that ( Figure S10).

Hypothesis
Interaction between magma and the hydrothermal system causes the harmonic tremor.
Episodic pulses of magmatic degassing that cause sharp fluctuations in the in data (MF more sensible and HF due to several surface activity).

Seal consolidation (low permeability increase; less fluids or gases moving up).
System is sealed and accumulates pressure beneath it. This causes a decrease in HF (less surface activity) and in MF (less gas rising to the surface). System becomes 'quiet'.
Seal breaks due to a perturbation on a critically pressurized system. Eruption is trigger by a gas flux pulse that cause a cascading material failure that leads to an explosive eruption. Table S4. Description of the proposed phases leading to the Whakaari 2019 eruption. Figure S1. Subplots show the pre-eruptive one-month feature time-series for the 18 eruptions for the feature DSAR rate variance (DSAR change quantiles variance (0.6-0.4)). Figure S2. Subplots show the pre-eruptive one-month feature time-series for the 18 eruptions for the feature DSAR median.           Distribution of K-S test statistics from repeat testing of nDSAR median and nDSAR rate variance archetypes prior to eruptions (denoted "in eruption"; blue & red) and randomly selected from the non-eruptive record (denoted "out eruption"; cyan & yellow). Increasing differentiability is indicated by a distribution that clusters closer to zero.