Seismic velocity changes at White Island volcano, New Zealand, using ten years of ambient noise interferometry

<p>Seismic velocity changes at volcanoes carry information about stresses present within hydrothermal and magmatic systems. In this thesis, temporal velocity changes are measured at White Island volcano using ambient noise interferometry between 2007–2017. This period contains multiple well-documented eruptions starting in 2012, following an inactive period that extends back over a decade. Three primary objectives are identified: (1) investigate what seismic velocity changes can tell us about dynamic changes beneath the volcano, (2) investigate non-volcanic sources and their possible influence on interpretations, and (3) consider the potential for real-time monitoring using ambient-noise. These objectives extend beyond White Island volcano, with implications for ambient noise monitoring of volcanoes globally. Two different approaches are used to measure velocity changes at White Island. The first involves cross-correlating noise recorded by pairs of seismic stations. Velocity changes are sought by averaging changes recorded across ten station-pairs that consist of an onshore station and a station on the volcano. The second approach involves cross-correlating the different components of individual seismic stations. This represents a less traditional approach to monitoring volcanoes, but is well-suited to White Island which has one permanent station active throughout eruptive activity. Single seismic stations located onshore are also processed to investigate background regional changes. Two periods of long-term velocity increases are detected at the volcano. The first occurs during a highly active period in 2012–2013 and the second occurs in the months preceding an explosive eruption in April 2016. Comparison with velocities recorded by onshore stations suggest a meteorological source for these changes is unlikely. Velocity increases are therefore interpreted to reflect cracks closing under increased pressures beneath the volcano. Similarly, a rapid decline in the velocity within 2–3 months of the April 2016 eruption is interpreted to reflect depressurization of the system. In addition to volcanic sources, we also find clear evidence of non-volcanic processes influencing velocity changes at the volcano. Two clear co-seismic velocity decreases of approximately 0.05–0.1% are associated with a Mw 5.2 earthquake in 2008 — within 10 km of the volcano — and the Mw 7.1 East Cape earthquake in 2016. The East Cape earthquake — located 200 km away from the volcano — produces significant velocity decreases over a large region, as detected by stations onshore and on White Island. This likely reflects dynamic stress changes as a result of passing seismic waves, with an eruption two weeks later interpreted here to have been triggered by this event. Finally, we identify similarities between annual variations recorded by onshore stations and changes at the volcano, suggesting an environmental influence. Velocity changes at White Island therefore represent a complex interaction of volcanic and non-volcanic processes, highlighting the need for improved understanding of external sources of change to accurately detect short-term eruptive precursors.</p>

Seismic velocity changes at White Island volcano, New Zealand, using ten years of ambient noise interferometry

<p>Seismic velocity changes at volcanoes carry information about stresses present within hydrothermal and magmatic systems. In this thesis, temporal velocity changes are measured at White Island volcano using ambient noise interferometry between 2007–2017. This period contains multiple well-documented eruptions starting in 2012, following an inactive period that extends back over a decade. Three primary objectives are identified: (1) investigate what seismic velocity changes can tell us about dynamic changes beneath the volcano, (2) investigate non-volcanic sources and their possible influence on interpretations, and (3) consider the potential for real-time monitoring using ambient-noise. These objectives extend beyond White Island volcano, with implications for ambient noise monitoring of volcanoes globally. Two different approaches are used to measure velocity changes at White Island. The first involves cross-correlating noise recorded by pairs of seismic stations. Velocity changes are sought by averaging changes recorded across ten station-pairs that consist of an onshore station and a station on the volcano. The second approach involves cross-correlating the different components of individual seismic stations. This represents a less traditional approach to monitoring volcanoes, but is well-suited to White Island which has one permanent station active throughout eruptive activity. Single seismic stations located onshore are also processed to investigate background regional changes. Two periods of long-term velocity increases are detected at the volcano. The first occurs during a highly active period in 2012–2013 and the second occurs in the months preceding an explosive eruption in April 2016. Comparison with velocities recorded by onshore stations suggest a meteorological source for these changes is unlikely. Velocity increases are therefore interpreted to reflect cracks closing under increased pressures beneath the volcano. Similarly, a rapid decline in the velocity within 2–3 months of the April 2016 eruption is interpreted to reflect depressurization of the system. In addition to volcanic sources, we also find clear evidence of non-volcanic processes influencing velocity changes at the volcano. Two clear co-seismic velocity decreases of approximately 0.05–0.1% are associated with a Mw 5.2 earthquake in 2008 — within 10 km of the volcano — and the Mw 7.1 East Cape earthquake in 2016. The East Cape earthquake — located 200 km away from the volcano — produces significant velocity decreases over a large region, as detected by stations onshore and on White Island. This likely reflects dynamic stress changes as a result of passing seismic waves, with an eruption two weeks later interpreted here to have been triggered by this event. Finally, we identify similarities between annual variations recorded by onshore stations and changes at the volcano, suggesting an environmental influence. Velocity changes at White Island therefore represent a complex interaction of volcanic and non-volcanic processes, highlighting the need for improved understanding of external sources of change to accurately detect short-term eruptive precursors.</p>

Methods and Computational Techniques for Investigating and Monitoring Seismic Velocities in the Earth's Crust

<p>This thesis is concerned with scrutinising the source, distribution and detectability of seismic velocity phenomena that may be used as proxies to study conditions in the crust. Specifically, we develop modelling techniques in order to analyse the directional variation of seismic wave speed in the crust and test them at Mt. Asama in Japan and Canterbury, New Zealand. We also implement both active source and noise interferometry to identify velocity variations at Mt. Ruapehu, New Zealand. Observations of temporal variation of anisotropic seismic velocity parameters at Asama volcano in Japan indicate that there is some process (or processes) affecting anisotropy, attributed to closure of microcracks in the rock as it is subjected to volcanic stress in the crust. To test this assertion, a 3D numerical model is created incorporating volcanic stress, ray tracing and estimation of the anisotropy to produce synthetic shear wave splitting results using a dyke stress model. Anisotropy is calculated in two ways; by considering a basic scenario where crack density is uniform and a case where the strength of anisotropy is related to dry crack closure from deviatoric stress. We find that the approach is sensitive to crack density, crack compliance, and the regional stress field. In the case of dry crack closure, modelled stress conditions produce a much smaller degree of anisotropy than indicated by measurements. We propose that the source of anisotropy changes at Asama is tied to more complex processes that may precipitate from stress changes or other volcanic processes, such as the movement of pore fluid. We develop a generalised anisotropy inversion model based on the linearised, iterative least-squares inversion technique of Abt and Fischer [2008]. The model is streamlined for use with results from the MFAST automatic shear wave splitting software [Savage et al., 2010]. The method iteratively solves for the best fitting magnitude and orientation of anisotropy in each element of the model space using numerically calculated partial derivatives. The inversion is applied to the Canterbury plains in the region surrounding the Greendale fault, using shear-wave splitting data from the 2010 Darfield earthquake sequence. Crustal anisotropy is resolved down to a depth of 20 km at a spatial resolution of 5 km, with good resolution near the Greendale fault. We identify a lateral variation in anisotropy strength across the Greendale fault, possibly associated with post-seismic stress changes. We perform active source and noise interferometry at Ruapehu in order to investigate potential seismic velocity changes and assess their use as a possible eruption forecasting method. Six co-located 100 kg ammonium nitrate fuel oil explosives were set off serially at Lake Moawhango, situated approximately 20 km south-east of Mount Ruapehu. Two methods of interferometry, using moving window cross correlation in the time and frequency domains, respectively, were applied to the recorded signal from each explosion pair in order to determine velocity changes from the signal coda waves. We identify possible diurnal velocity variations of ~ 0:7% associated with strain caused by the solid Earth tide. Synthetic testing of velocity variation recoverability was also performed using both methods. Interferometry of noise cross-correlations during the period was also performed using moving window cross correlation in the frequency domain. Analysis of velocity variations in the ZZ, RR and TT component pairs show little coherency. This, combined with results from synthetic testing that show that the frequency domain interferometry technique employed is unstable above velocity variations of 0.1%, indicate that the method may not be suitable for determining velocity variations at Ruapehu.</p>

Methods and Computational Techniques for Investigating and Monitoring Seismic Velocities in the Earth's Crust

<p>This thesis is concerned with scrutinising the source, distribution and detectability of seismic velocity phenomena that may be used as proxies to study conditions in the crust. Specifically, we develop modelling techniques in order to analyse the directional variation of seismic wave speed in the crust and test them at Mt. Asama in Japan and Canterbury, New Zealand. We also implement both active source and noise interferometry to identify velocity variations at Mt. Ruapehu, New Zealand. Observations of temporal variation of anisotropic seismic velocity parameters at Asama volcano in Japan indicate that there is some process (or processes) affecting anisotropy, attributed to closure of microcracks in the rock as it is subjected to volcanic stress in the crust. To test this assertion, a 3D numerical model is created incorporating volcanic stress, ray tracing and estimation of the anisotropy to produce synthetic shear wave splitting results using a dyke stress model. Anisotropy is calculated in two ways; by considering a basic scenario where crack density is uniform and a case where the strength of anisotropy is related to dry crack closure from deviatoric stress. We find that the approach is sensitive to crack density, crack compliance, and the regional stress field. In the case of dry crack closure, modelled stress conditions produce a much smaller degree of anisotropy than indicated by measurements. We propose that the source of anisotropy changes at Asama is tied to more complex processes that may precipitate from stress changes or other volcanic processes, such as the movement of pore fluid. We develop a generalised anisotropy inversion model based on the linearised, iterative least-squares inversion technique of Abt and Fischer [2008]. The model is streamlined for use with results from the MFAST automatic shear wave splitting software [Savage et al., 2010]. The method iteratively solves for the best fitting magnitude and orientation of anisotropy in each element of the model space using numerically calculated partial derivatives. The inversion is applied to the Canterbury plains in the region surrounding the Greendale fault, using shear-wave splitting data from the 2010 Darfield earthquake sequence. Crustal anisotropy is resolved down to a depth of 20 km at a spatial resolution of 5 km, with good resolution near the Greendale fault. We identify a lateral variation in anisotropy strength across the Greendale fault, possibly associated with post-seismic stress changes. We perform active source and noise interferometry at Ruapehu in order to investigate potential seismic velocity changes and assess their use as a possible eruption forecasting method. Six co-located 100 kg ammonium nitrate fuel oil explosives were set off serially at Lake Moawhango, situated approximately 20 km south-east of Mount Ruapehu. Two methods of interferometry, using moving window cross correlation in the time and frequency domains, respectively, were applied to the recorded signal from each explosion pair in order to determine velocity changes from the signal coda waves. We identify possible diurnal velocity variations of ~ 0:7% associated with strain caused by the solid Earth tide. Synthetic testing of velocity variation recoverability was also performed using both methods. Interferometry of noise cross-correlations during the period was also performed using moving window cross correlation in the frequency domain. Analysis of velocity variations in the ZZ, RR and TT component pairs show little coherency. This, combined with results from synthetic testing that show that the frequency domain interferometry technique employed is unstable above velocity variations of 0.1%, indicate that the method may not be suitable for determining velocity variations at Ruapehu.</p>

Massive lithospheric delamination in southeastern Tibet facilitating continental extrusion

Significant left-lateral movement along the Ailao Shan-Red River fault accommodated a substantial amount of the late Eocene to early Miocene India-Asia convergence. However, the activation of this critical strike-slip fault remains poorly understood. Here, we show key seismic evidence for the occurrence of massive lithospheric delamination in southeastern Tibet. Our novel observation of reflected body waves (e.g. P410P and P660P) retrieved from ambient noise interferometry sheds new light on the massive foundered lithosphere currently near the bottom of the mantle transition zone beneath southeastern Tibet. By integrating the novel seismic and pre-existing geochemical observations, we highlight a linkage between massive lithospheric delamination shortly after the onset of hard collision and activation of continental extrusion along the Ailao Shan-Red River fault. This information provides critical insights into the early-stage evolution of the India-Asia collision in southeastern Tibet, which has significant implications for continental collision and its intracontinental response.