‘Silent’ Dome Emplacement into a Wet Volcano: Observations from an Effusive Eruption at White Island (Whakaari), New Zealand in Late 2012

The 2012–2016 White Island (Whakaari) eruption sequence encompassed six small explosive events that included one steam driven and five explosive phreato-magmatic eruptions. More enigmatic, a dome was observed at the back of the vent and crater lake in November 2012. Its emplacement date could not be easily determined due to persistent steam from the evaporating crater lake and because of the very low levels of discrete volcanic earthquakes associated with its growth. During this period, seismicity also included persistent tremor with dominant frequencies in the 2–5 Hz range. Detailed assessment of the tremor reveals a very slow evolution of the spectral peaks from low to higher frequencies. These gliding spectral lines evolved over a three-month time period beginning in late September 2012 and persisting until early January 2013, when the tremor stabilised. As part of the dome emplacement episode, the crater lake progressively dried, leaving isolated pools which then promoted persistent mud/sulphur eruption activity starting in mid-January 2013. We interpret the emplacement of the dome as a non-explosive process where the hot, mostly degassed, magma intruded slowly through the hydrothermal system in late September 2012 and cooled in a relatively quiet state. The tremor evolution might reflect the slow contraction of subsurface resonant cavities, which increased the pitch of the peak resonant frequency through time. Alternatively, spectral evolution might reflect a ‘comb function’ due to clockwork beating of the slowly cooling dome, although direct evidence of clockwork beats is not seen in the waveform data. Finally, it might represent frothing of the hydrothermal system ahead of the slowly propagating magma.


Introduction
White Island (Whakaari in Te Reo Maori) is a frequently active and hazardous composite cone volcano with most of its relief lying below sea level within the south-eastern Bay of Plenty ( Figure 1A). While White Island produces only small eruptions by global standards, the tragic 9 December 2019 event illustrates the possible societal impacts that relate to small eruptions at volcanoes that are frequented by tourists. Such eruptive activity was also present during the 2012-2016 period ( Figure 1) which proceeded from the onset of unrest in mid-2011 [1,2], produced several well documented explosive The hydrothermal system at White Island results from the interaction between magmatic heat and juvenile fluids with water from meteoric and oceanic sources that percolates into the crust [16]. Heat and fluids from the magma drive convective circulation that establish fluid phase transitions below the surface and a hydrostatically controlled phase equilibrium with its enclosing hydrothermal brine. In particular, at the magma-hydrothermal system interface, a single-phase gas enveloped by a two-phase fluid composed of liquid and gas bubbles dominates, whereas a single-phase liquid will eventuate at shallow levels and within the crater lake. These conditions are generally stable but can evolve in different ways depending on the position of magma within the system and externally modulating effects like rainfall ( Figure 2). Eruptions at White Island are driven by persistently active magmatic degassing that is evident over the historical record [7]. This degassing requires long-term magmatic injection into shallow portions of the volcanic edifice, which may be accomplished via convective overturn within a conduit [8][9][10] or persistent injection of small batches of magma [11,12].
A small effusive dome was also emplaced on the back wall margin of the active lake filled eruption vent system, and hence this is an excellent example of magmatic propagation through a 'wet' volcano hydrothermal system. The dome effusion had significant consequences for the evolution of the hydrothermal system, promoting the drying of the lake, and a switch to persistent and well documented mud/sulphur eruption activity [13][14][15][16]. Dome forming eruptive activity is a frequent occurrence in volcanic systems with viscous magmas and have drawn significant scientific interest due to their persistent medium-to long-term hazards. These hazards are greatest when the domes are perched at elevation on unstable slopes of a volcanic edifice such as at volcanoes like Soufriere Hills, Montserrat [17]; Redoubt, Alaska [18]; or Merapi, Indonesia [19]. Domes may also form within positions of lower relief (e.g., Mt. St. Helens [20][21][22]), and in such cases, the eruptions can be emplaced through well-developed hydrothermal systems.
The hydrothermal system at White Island results from the interaction between magmatic heat and juvenile fluids with water from meteoric and oceanic sources that percolates into the crust [16]. Heat and fluids from the magma drive convective circulation that establish fluid phase transitions below the surface and a hydrostatically controlled phase equilibrium with its enclosing hydrothermal brine. In particular, at the magma-hydrothermal system interface, a single-phase gas enveloped by a two-phase fluid composed of liquid and gas bubbles dominates, whereas a single-phase liquid will eventuate at shallow levels and within the crater lake. These conditions are generally stable but can evolve in different ways depending on the position of magma within the system and externally modulating effects like rainfall ( Figure 2). Acidic brine fluids underlie meteoric water/condensate lenses (both depicted green) and encapsulate the magmatic/fumarolic vent system (blue and grey). Magma ascends along the main conduit (designated by the red line in B), preceded by increased heat and gas flow which evaporates the twophase vapour-liquid region to single-phase vapor along its way (grey). Three possible resonant systems are portrayed in the inset diagram (C). The bubble filled cavity (1). A clockwork stick-slip rupture mechanism (2), or the frothed expanding boiling front (3 and its inset) may produce harmonic signals discussed. The arrows in B show the expanded two-phase system that results from the propagating magma injection. See text.
From a seismological perspective, the shallow hydrothermal system may produce the full range of seismic observations seen at White Island and its global analogues like Kawah Ijen volcano, Indonesia [4,23]. Examples include discrete long-period (LP) [24], volcano-tectonic (VT) [2,25], very long period (VLP) seismicity [2], and persistent volcanic tremor [1]. Tremor is an important feature of volcanic seismicity [26]. It is characterised by a continuous, banded, or spasmodic signal, which is detectable only when exceeding the background seismic energy level [26]. Several data reduction methods [27] have been proposed to characterize its short-and long-term time evolution, that can lead to the identification of significant transitions between volcanic processes [28], sometimes triggered by external or internal events such as tectonic or volcano-tectonic (VT) seismic events [29]. Volcanic tremor can also occur precursory to an eruption [30−33]. However, the origins of tremor may vary at different volcanoes, and are often poorly understood; moreover, many different seismic sources can act at the same time and combine to produce the signal of interest [27].
Harmonic tremor, consisting of a fundamental frequency and evenly spaced overtones, is a common occurrence at volcanic systems worldwide [34,35], and is important both to determine possible source processes and as a monitoring tool. Changing spectral patterns in harmonic tremor (e.g., migration of the spectral lines, i.e., so-called gliding) is increasingly recognised at volcanoes [36−38] as an important short-term precursor that may occur several minutes prior to explosive Acidic brine fluids underlie meteoric water/condensate lenses (both depicted green) and encapsulate the magmatic/fumarolic vent system (blue and grey). Magma ascends along the main conduit (designated by the red line in (B), preceded by increased heat and gas flow which evaporates the two-phase vapour-liquid region to single-phase vapor along its way (grey). Three possible resonant systems are portrayed in the inset diagram (C). The bubble filled cavity (1). A clockwork stick-slip rupture mechanism (2), or the frothed expanding boiling front (3 and its inset) may produce harmonic signals discussed. The arrows in B show the expanded two-phase system that results from the propagating magma injection. See text.
From a seismological perspective, the shallow hydrothermal system may produce the full range of seismic observations seen at White Island and its global analogues like Kawah Ijen volcano, Indonesia [4,23]. Examples include discrete long-period (LP) [24], volcano-tectonic (VT) [2,25], very long period (VLP) seismicity [2], and persistent volcanic tremor [1]. Tremor is an important feature of volcanic seismicity [26]. It is characterised by a continuous, banded, or spasmodic signal, which is detectable only when exceeding the background seismic energy level [26]. Several data reduction methods [27] have been proposed to characterize its short-and long-term time evolution, that can lead to the identification of significant transitions between volcanic processes [28], sometimes triggered by external or internal events such as tectonic or volcano-tectonic (VT) seismic events [29]. Volcanic tremor can also occur precursory to an eruption [30][31][32][33]. However, the origins of tremor may vary at different volcanoes, and are often poorly understood; moreover, many different seismic sources can act at the same time and combine to produce the signal of interest [27].
Harmonic tremor, consisting of a fundamental frequency and evenly spaced overtones, is a common occurrence at volcanic systems worldwide [34,35], and is important both to determine possible source processes and as a monitoring tool. Changing spectral patterns in harmonic tremor (e.g., migration of the spectral lines, i.e., so-called gliding) is increasingly recognised at volcanoes [36][37][38] as an important short-term precursor that may occur several minutes prior to explosive activity. Systematic evolution of spectra over longer time periods are less common; however, an assessment of such slowly evolving systems may also be important from a process and hazard standpoint [39].
It is surmised that movement of magma below or within a hydrothermal system should produce significant changes in seismic observations [40][41][42] (Figure 2). Hence, the extrusion of magma through a well-established hydrothermal system and onto the surface provides an excellent opportunity to re-examine remote monitoring data for hazards implications. The dome forming eruption at White Island is interesting in this context because its discovery was truly enigmatic. It was first documented by GeoNet on 11 December 2012 but was observed by White Island tour operators possibly a fortnight earlier (Volcanic Alert Bulletin WI-2012/16; https://www.geonet.org.nz). Surprisingly, its emplacement occurred without dramatic seismicity revealed on the permanent White Island seismic station ( Figure 1A). Likewise, an evaluation of the White Island web camera system did not reveal the appearance of the small dome, mostly due to persistent obscuring steam plumes for the period from the initial onset of eruption activity on 5 August 2012 [1]. It also appeared to have been quite effusive, based on the smooth lumpy texture of the surface ( Figure 1C). The dome, with an extent of 20-30 m in diameter and a height of 10-15 m, has a volume on the order of 1−7 × 10 3 m 3 . Below, we retrospectively document the pre-eruption seismicity from the onset of eruptive activity from August to the observation of the dome in December 2012.

Seismic Data and Results
White Island seismic monitoring was composed of a single broadband seismic sensor at site WIZ ( Figure 1A), which is comprised of a Guralp 3ESP seismometer and a Quanterra Q330 digitiser, sampling at 100 Hz and telemetering data in real time to the GeoNet data center. The archived seismic data were processed in two ways: (1) calculation of the real-time seismic amplitude measurement value (RSAM) by taking the time-series data from the vertical component sensor, correcting it to velocity, and computing the root mean square (RMS) amplitude within one minute long, non-overlapping windows ( Figure 3A); and (2) computing the spectra via the fast-Fourier transform (FFT) for each one-minute window ( Figure 3B). For the latter, we visualised the data by picking the peak amplitude from the spectra and monitoring that over the period of interest (before the first eruption to the end of the dome forming phase). To aid in interpreting the seismic observations, example waveforms ( Figure 4A) and spectra ( Figure 4B) are plotted from the first minute of each day for the period 1 July to 31 December 2012 (Figures 3 and 4 are presented in UTC). Although we realize that this procedure potentially produces an aliasing effect, the individual sample waveforms and spectra match closely a denser spectral analysis and confirm the observations in Figure 3. Hence, we regard the observations in Figures 3 The one-minute RSAM observations ( Figure 3A) show the first eruption on 5 August 2012 (RSAM~2800 nm/s), as well as a persistent high amplitude signal in early September (~2100 nm/s), followed by sustained moderate level tremor (500-1000 nm/s) through the time of the dome observation in late November 2012. The spectral analysis ( Figure 3B) shows that the peak frequency of tremor is generally focused within the~1.0-4.0 Hz band, with specific and persistent peaks and notable evolutionary patterns. We note with particular interest the spectral changes associated with high amplitude tremor periods, including (1) a shift from broader spectrum tremor to 2.5-3.5 Hz in mid-June 2012; (2) the onset of a migration of the spectral peaks (termed gliding spectral lines in the literature [36]) in mid-July; (3) the onset of broader spectra tremor with slowly gliding spectral lines in early September; and (4) establishment of stable (non-gliding) persistent spectral peaks in late December (see red dots in  To assess possible harmonic patterns, we computed spectrograms using two-minute-long windows. Each time window is detrended with a mean and a linear function before tapering using a Hanning window (10% on each side). The frequency resolution is 0.0083 Hz. For each column of the spectrogram the resulting FFT amplitudes are color-coded and shown in Figure 5. After smoothing the results with a median of five days, we picked the central frequencies of continuous spectral lines: We then assess if the data were consistent with a harmonic tremor source. In fact, although peaks often appear to be equally spaced in frequency, the dominant frequency is rarely seen. We therefore developed an algorithm to automatically compute this possibly buried dominant frequency. To aid in interpreting the seismic observations, example waveforms ( Figure 4A) and spectra ( Figure 4B) are plotted from the first minute of each day for the period 1 July to 31 December 2012 (Figures 3 and 4 are presented in UTC). Although we realize that this procedure potentially produces an aliasing effect, the individual sample waveforms and spectra match closely a denser spectral analysis and confirm the observations in Figure 3. Hence, we regard the observations in Figures 3 and 4 as robust, illustrating both the longer-term features of the tremor and the slowly evolving migration from lower to higher spectral frequencies.
Specific spectral peaks become well established in late August 2012, which then perceptibly migrate towards higher frequencies over a three-month period. On~29 August 2012, the specific peaks of interest are at 2.1 Hz ( Figure 4B). These peaks shift to a stable frequency of 2.3 Hz by 30 November 2012 ( Figure 4B), about the same period that the dome must have been emplaced based on the tremor and the first observation of the dome. A subsidiary peak is also observed, which weakly emerges at 2.9 Hz after the ashing episode and persists after the first dome observation at a modestly higher frequency of~3.2 Hz. There are also stationary spectral peaks observed at other specific frequencies (Figures 3 and 4B) as part of a possible harmonic pattern which began in early September and evolved until the dome was observed.
To assess possible harmonic patterns, we computed spectrograms using two-minute-long windows. Each time window is detrended with a mean and a linear function before tapering using a Hanning window (10% on each side). The frequency resolution is 0.0083 Hz. For each column of the spectrogram the resulting FFT amplitudes are color-coded and shown in Figure 5. After smoothing the results with a median of five days, we picked the central frequencies of continuous spectral lines: 1.4, 1.6, 1.8, 2.0, 2.2, 2.3, 2.7, 2.9, 3.2, 3.5, 4.9 Hz. We note that 2.2 Hz and 2.3 Hz seem to merge in September.
Geosciences 2019, 9 FOR PEER REVIEW 7 window (e.g., 4 h or 12 h) by counting for each frequency how many times a peak is detected. Finally, we search for our dominant frequency in the 0.3−1.3 Hz range, and cumulate histogram values not only for that candidate dominant frequency but also for its potential harmonics. The cumulated value is then normalized (i.e. divided by the number of harmonics) because lower frequencies may yield higher numbers of harmonics and higher cumulated values.

Figure 5.
Spectrogram computed using two-minute-long windows after removing the mean, the trend and applying a cosine taper in each window. Yellow colours correspond to high amplitudes, whereas blue colours correspond to low amplitudes.
For each column, we determine the potential dominant frequency, and we superimpose it, together with all its harmonics, on the spectrogram to assess its coherency with time. The results ( Figure 6 for time windows of 4 h, and in Figure 7 for time windows of 12 h) are consistent with a harmonic oscillation over a narrow frequency band, hence confirming that the tremor could originate from a harmonic process. We must acknowledge that the observation is based on the sole single station observation and hence may include other influences such as path and structural effects in this Figure 5. Spectrogram computed using two-minute-long windows after removing the mean, the trend and applying a cosine taper in each window. Yellow colours correspond to high amplitudes, whereas blue colours correspond to low amplitudes.
We then assess if the data were consistent with a harmonic tremor source. In fact, although peaks often appear to be equally spaced in frequency, the dominant frequency is rarely seen. We therefore developed an algorithm to automatically compute this possibly buried dominant frequency.
For each column of the spectrogram (i.e., 2 min), the algorithm determines all the peaks by looking for local maxima in the column. An histogram of these peaks is then built over a wider time window (e.g., 4 h or 12 h) by counting for each frequency how many times a peak is detected. Finally, we search for our dominant frequency in the 0.3-1.3 Hz range, and cumulate histogram values not only for that candidate dominant frequency but also for its potential harmonics. The cumulated value is then normalized (i.e. divided by the number of harmonics) because lower frequencies may yield higher numbers of harmonics and higher cumulated values.
For each column, we determine the potential dominant frequency, and we superimpose it, together with all its harmonics, on the spectrogram to assess its coherency with time. The results ( Figure 6 for time windows of 4 h, and in Figure 7 for time windows of 12 h) are consistent with a harmonic oscillation over a narrow frequency band, hence confirming that the tremor could originate from a harmonic process. We must acknowledge that the observation is based on the sole single station observation and hence may include other influences such as path and structural effects in this complex volcanic edifice.

Discussion
Whilst the migration in spectral peaks is very small compared to other well-known examples like Soufriere Hills [36] and Redoubt [38], the spectral gliding observed for White Island occurs over

Discussion
Whilst the migration in spectral peaks is very small compared to other well-known examples like Soufriere Hills [36] and Redoubt [38], the spectral gliding observed for White Island occurs over a much longer time period of ~3-4 months. This timescale is more comparable to the one observed at

Discussion
Whilst the migration in spectral peaks is very small compared to other well-known examples like Soufriere Hills [36] and Redoubt [38], the spectral gliding observed for White Island occurs over a much longer time period of~3-4 months. This timescale is more comparable to the one observed at Villarrica volcano [43], although in that case the frequency changes were more abrupt. In all of these cases, the spectral changes can be linked to eruptive activity at the volcano. The gliding frequencies at White Island are unusual in two respects. First, they persisted over several months, starting with the onset of low-level ash eruptions in early September 2012, and the observation of a dome over 2.5 months later in late November. Second, while the gliding spectral lines observed prior to large eruptive events (e.g., Soufriere Hills [36]; Redoubt [38]) seem to have persistent monotonically increasing frequencies (concave upward trending spectral peaks) for fundamental frequencies and overtones, the observations here include increasing spectral peak-lines with a reduced rate of change over time (i.e., a concave downward increase in frequency). This observation might suggest a different mechanism for the two processes ( Figures 3B and 5).
Harmonic features in seismic data are often regarded as part of three potential processes, including fluid-elastic [39,40,[44][45][46][47][48][49][50][51][52][53][54], frictional-induced [38,41], and permeable flow-controlled resonance [42,46]. In the first source process (see Figure 2C label (1)), fluid-elastic resonance involves the vibrations of cavities as a result of the motion of fluids in response to short-term perturbations (e.g., possibly the superposition of hybrid earthquakes thought to occur as magma proceeds through the glass transition at shallow depth [55]. In such a case, the different modes of vibration vary with the impedance contrast (i.e., with the properties of the fluid filling the cavity) and geometry [35,[56][57][58]. In the second source process ( Figure 2C label (2)), a frictional mechanism involves the generation of harmonic tremor by the superposition of highly regular repetition of mostly identical stick-slip earthquakes [41]. In such a case, the dominant frequencies vary with the elastic properties of the medium, stressing rates, and geometry, whereas overtones are the result of a Dirac comb effect. In the third source process (see Figure 2C label (3)), a permeable flow-controlled mechanism involves the spontaneous vibration of cavities while gas escapes toward the surface and is the result of the transient porous flow of magmatic/hydrothermal gases through the permeable medium that caps the cavity. In such a case, the pressure inside the cavity is governed by the equation of a non-linear oscillator, which reduces to a linear first-order harmonic oscillator for highly-fractured and thin ( 100 m) caps [42]; hence, the specific frequencies depend on the permeability and porosity of the cap, gas properties, supply, and geometry of the cavity. In particular, a harmonic tremor may emerge through a Dirac comb effect when gas feeds the cavity regularly (e.g., through bubble clouds or foam collapse, [59][60][61]. Alternatively, a harmonic tremor may arise as a non-linear effect [46] or for thick ( 100 m) caps as result of the high-order terms controlling the pressure oscillations of the cavities [42].
At White Island, the observation of the dome in late November 2012 and the lack of strong seismicity associated with its appearance is enigmatic and can provide some constraints regarding the governing tremor mechanism. Persistent steam obscured the back of the crater lake and this might imply that the dome had been emplaced much earlier, possibly around the 2 September ash venting period shown in Figure 4. If this did mark the dome emplacement, then the slow cooling of the dome and conduit may have contributed to the spectral gliding within the tremor. For example, the slow progressive cooling and degassing might provide both the required seismic trigger mechanism and the higher impedance contrast cavity to produce harmonic tremor through fluid-elastic resonance. If a magmatic root remained beneath the dome, such a feature could hold exsolved gases within a bubble-rich root structure; Neuberg and O'Gorman [58] have shown how such a conduit could produce resonant tremor. If this conduit became progressively further degassed or if progressive top down solidification occurred, this could produce a smaller conduit with time. The slowly evolving changes in tremor might then reflect variations of the resonant root structure.
As an alternative, the slowly evolving spectral features could be regarded as a feature of the emplacement process if the dome were emplaced shortly before its observation in late November.
The 2-5 Hz tremor source process at White Island, thought to be dominantly within the hydrothermal system, is a long-term feature of the shallow volcanic system [1]. In this case, the modulation of the shallow seismicity might reflect a longer-term intrusion process at depth. If the sub-surface magma began its interaction with the shallow hydrothermal system about 2.5 months before the dome's observation, then the ascent rate of intrusion would have been very slow. We regard this alternative as less likely due to the apparent fluidity of the dome features, however it might be relevant to the spectral observations occurring prior to the 5 August explosive eruption (see Figure 3B).
Regarding the second source process (Figure 2), in order to produce clear overtones in the spectrogram, the repetition interval of seismic events must be very precise. However, we do not observe clear discrete earthquakes in the sequence ( Figure 4A), so the application of a repeating clockwork pattern, and thus frictional-induced resonance, is not easy to invoke for White Island in this period. In any case, it might be possible for interactions of hydrothermal system fluids with the conduit walls to act as a source of persistent micro-seismicity. Whilst the occurrence of a regular clockwork pattern is not apparent in the waveform data ( Figure 4A), it could still be part of the underlying tremor excitation process. This process might also be only weakly periodic but still interact with a nearby resonant cavity [62]. If, on the other hand, the dome was extruded immediately prior to its first observation in November, it is plausible that the tremor observed over the period September to November could be the result of interaction between the ascending magma column and the hydrothermal system through which it is moving. In Figure 2C labeled (3), we illustrate this process with the front of the ascending plug intruding into the overlying two-phase vapor-liquid region extant at the time. This process could explain the earlier June to September spectral features. Temperature gradients above the ascending magma are very high, ranging from an assumed sub-liquidus intrusion temperature of~900 • C to the temperature of the vapour-liquid saturation curve. The overlying conduit, assumed to be a porous fractured medium, is two-phase over its entire length to the surface (see Figure 2C). This conduit could deliver non-condensable gas to an overlying compressible "gas pocket" created beneath a partial mineralogic seal. Such a seal system is argued to be present based on C/S ratios of vent emissions at the time [16].
Gas pockets trapped beneath permeable media or embedded in cracks within the dome may lead to spontaneous vibrations while gas escapes toward the surface [42] but also through thermal instabilities [54]. In these resonant gas pockets, gliding spectral behaviour results from changing lengths of the cavity, with shorter lengths generating progressively higher frequencies over the spectral range of 0.1 Hz to 15 Hz. It is possible that this could result from shortening of the conduit cavity during magma ascent, although this process cannot be definitively confirmed by the existing data. Whereas spectral gliding would be complete when the magma reaches the surface, a tremor would be ongoing during cooling and the ongoing interaction between the plug/dome with the hydrothermal system, as portrayed in Figure 2C.
We regard propagation and emplacement of the dome in early September as more likely due to the strong tremor and the relatively abrupt change of the spectral patterns ( Figure 3). If so, then the perturbation of the hydrothermal system led to high levels of gas and steam discharge, which obscured the dome, and caused the observed spectral gliding. It also may have promoted the slow evaporation of the crater lake system, which was a central feature of the White Island crater vent for more than 10 years before the 2012-2016 eruption episode. Regardless of the timing and mechanism of emplacement and stabilisation, it is remarkable for a dome to be emplaced into an active wet hydrothermal system with such subtle changes of the observed tremor patterns.

Conclusions
We describe the visual and seismic observations related to the onset of eruptive activity at White Island, New Zealand, for the period June to December 2012. The period included an explosive eruption, ash venting, and a dome emplacement episode through a 'wet' volcano hydrothermal system. The period was marked by persistent elevated tremor amplitudes and slowly evolving gliding spectral lines. At the same time, the crater lake system progressively evaporated, which encompassed the full crater floor in June 2012 and progressed to isolated pools by the end of December 2012. Interestingly, this progressive lake loss promoted the next phase of mud/sulphur eruptions which began in mid-January 2013 [13][14][15][16] through the pre-existent Southeast vent system (Figure 2).
We regard the injection of magma into the shallow hydrothermal system as the key driver of the lake loss and shifts in tremor patterns in mid to late 2012. We interpret that the magmatic intrusion into the hydrothermal system began in early September and promoted the range of seismic observations documented here. Future work will be needed to examine the plausibility of the models outlined here with a rigorous analytical and numerical approach.
As a final note, the 9 December 2019 White Island eruption occurred within the timeframe of the development of this research and post-eruption unrest associated with that eruption is ongoing at the time of writing. Hence, this work does not include new insights from this most recent eruptive activity. At the conclusion of the present unrest period, a detailed retrospective assessment of the progression of unrest will be critical to determine similarities and differences for the two time periods. It is hoped that such an analysis will lead to ever-improving outcomes at difficult to monitor volcanoes like White Island. The present work illustrates how persistent shallow magma and well-developed hydrothermal systems may produce observations that may fit many possible conceptual models and produce challenging monitoring conditions. It is a central goal of the volcano science community to document such events and hence to improve monitoring outcomes for similar volcanic systems both in New Zealand and around the globe. This work will help to assess the viability of the various models within the context of the limited available data.