APPLICATION OF SEISMOLOGY-BASED VOLCANO MONITORING TECHNIQUES AT WHAKAARI/WHITE ISLAND VOLCANO

Volcanoes present some of the most immediate and unpredictable natural hazards, particularly when they erupt without clear precursors. Phreatic and phreatomagmatic eruptions are especially challenging to forecast because they involve rapid interactions between shallow magma, hydrothermal fluids, and sealed rock layers. Whakaari volcano in New Zealand exemplifies this challenge. It is one of the most active and hazardous volcanoes in the country, with a long history of phreatic eruptions. The December 2019 eruption highlighted the need for better monitoring tools capable of detecting early warning signals, even at isolated volcanoes with sparse seismic networks. This thesis investigates the pre-eruptive changes of Whakaari in 2019 through an integrated seismic approach. By combining shear wave splitting (SWS), which is sensitive to anisotropy and crack alignment, with ambient noise interferometry (ANI), which detects isotropic seismic velocity changes, we characterize temporal and spatial variations in the subsurface. The central aim is to assess whether combining these complementary methods can reveal consistent precursory indicators of unrest and improve understanding of the processes leading to phreatic eruptions. The first part of this research focuses on shear-wave splitting and Vp/Vs ratio analysis using two permanent GeoNet stations, WIZ and WSRZ, which provided continuous monitoring of seismic activity in the Whakaari region. Using a catalog of earthquake locations from the temporary seismic network Back-Arc Rifting in New Zealand (BARNZ), we measured fast polarization directions (φ) and delay times (δt) to track temporal changes in crack alignment and possibly stress field orientation. The analysis also examined the Vp/Vs ratio, which provides insights into the fluid content of cracks (e.g., gas versus liquid) and the nature and evolution of seismic anisotropy. We conducted shear-wave splitting and isotropic velocity change analyses between 2018 and 2020, covering periods of quiescence, unrest, and the eruption itself. In total, 3,145 shear-wave splitting measurements and 2,656 Vp/Vs ratio measurements were examined. Comparing splitting parameters from similar earthquake paths across different times suggests that the observed temporal changes are unlikely to reflect variations in earthquake paths through a spatially heterogeneous medium. Instead, these variations more likely stem from evolving anisotropy, controlled by changes in crack alignment due to stress redistribution or varying fluid content. The results show pronounced variability in anisotropy and Vp/Vs ratio leading up to the 2019 eruption. Fast directions alternated between northeast–southwest and northwest–southeast orientations, consistent with rotations of the stress field and reactivation of pre-existing structural fabrics. More specifically, at WSRZ the fast orientation gradually rotated from N–S in September to E–W by December, leading up to the eruption. Similar changes were observed at WIZ, where the fast direction shifted from E–W/ESE–WNW to E–W/SE–NW. Delay times also exhibited a slowly varying pattern, with slight variations during the May–June 2019 earthquake swarm, but showed little overall change throughout the unrest period, even as the Vp/Vs ratio increased. These temporal variations are interpreted as evidence for dynamic crack formation, fluid pressurization, and temporary sealing within the volcanic edifice. The analysis also revealed that the anisotropy at Whakaari is shallow, likely confined to the upper 1–2 km of the edifice, as evidenced by the rapid spatial changes in splitting parameters between nearby stations. This is consistent with models in which the hydrothermal system and overlying fractured rock play a dominant role in controlling stress-induced anisotropy. Importantly, the temporal variations in SWS parameters and the Vp/Vs ratio correspond closely to periods of elevated seismicity, highlighting the potential of these parameters as a monitoring tool for tracking evolving stress, crack geometry, and fluid content prior to eruptions. The second part of the thesis investigates seismic velocity changes (dv/v) at Whakaari volcano from 2014 to 2020, a period encompassing the 2019 eruption. Using the MSNoise package on continuous data from the WIZ and WSRZ stations, we applied single-station cross-component interferometry in the 0.1–1.0 Hz band. Cross-correlation functions were computed in two-hour windows and stacked into 10- and 30-day averages. Relative velocity changes were then measured in the 20–80 s coda window using the Moving-Window Cross-Spectral (MWCS) method to assess their relationship with volcanic activity. Our analysis revealed clear temporal velocity variations across the three component pairs (EN, EZ, NZ), with two key episodes of dv/v change correlated with volcanic unrest. The first, during the May–June 2019 seismic swarm, showed a sharp 0.5–1% decrease in velocity likely reflecting crack formation and fluid movement associated with intensified activity. The second, during the October–December 2019 unrest cycle, was characterized by a general downward trend but with notable short-term fluctuations in velocity. The unrest began with a magnitude 5.65 earthquake that had a high energy density at the volcano of 12 J/m³. The velocity oscillations suggest alternating phases of crack opening, partial sealing, and reactivation before velocities reached minima immediately prior to the 9 December eruption. Importantly, we also observed both increases and decreases in velocity across different components, indicating heterogeneous responses within the volcanic edifice. This sequence may reflect initial crack formation, brief periods of stabilization or closure, and subsequent reopening or creation of new fractures. Depth sensitivity kernels indicate that the observed changes are most pronounced in the shallow edifice (upper 10 km). During the unrest period, our analysis of RSAM and seismic parameters, together with GeoNet reports of elevated SO₂ gas output and volcanic tremor, showed that activity had risen to its highest recorded levels, providing independent evidence of heightened volcanic unrest. In addition, a concurrent decrease in cross-correlation coefficients (CC) during the unrest period indicates a pronounced decrease in correlation, pointing to contributions from both medium changes and seismic source effects, such as increased tremor. A similar pronounced decrease in correlation was previously observed before the 2016 eruption, suggesting a recurring seismic anomaly that may provide short-term forecasting potential. These findings highlight the potential of single-station ambient noise interferometry as a tool for monitoring volcanic activity, even with limited instrumentation. The observed patterns align with those found at other volcanoes and suggest that seismic velocity changes can serve as precursors to volcanic eruptions, as demonstrated by the 2019 eruption at Whakaari. By integrating seismic velocity measurements with other geophysical observations, this approach offers a pathway to improving early warning capabilities for volcanic hazards by providing insights into conditions preceding eruptions. The third part of the thesis integrates anisotropy and interferometry results to examine how stress-induced crack changes and isotropic velocity variations interact. To achieve this, we rotated horizontal seismic components by 45° to align with the dominant anisotropic fast directions identified in SWS analysis. This allowed for a direct comparison of velocity changes along anisotropy-aligned paths. The integrated analysis revealed a complex, intermittent relationship between crack alignment and bulk seismic velocity changes. These were sometimes consistent with the expected poroelastic response along the fast direction and sometimes strongly decoupled, highlighting the non-uniform nature of the crack–fluid–stress system. During the swarm period, anisotropy showed shifts in fast direction while ANI detected a sharp dv/v decrease, indicating simultaneous stress rotations and crack pressurization. During the unrest period, velocity changes became more variable, with both decreases and increases observed, consistent with a sealed hydrothermal system undergoing cycles of pressurization and partial release. Both methods converged on the interpretation that shallow crack dynamics played a key role in destabilizing the system prior to the eruption. Comparisons show that the faster velocities from cross-correlations do not always coincide with shear-wave splitting fast directions, implying that the two methods may be sampling different regions of the edifice, either laterally or with depth. Beyond Whakaari, this research contributes to a broader understanding of phreatic eruption mechanisms, demonstrating that small changes in crack geometry and fluid content, detectable through seismic anisotropy and interferometry, can precede explosive outcomes. These findings support the development of multi-parameter early warning systems that integrate seismic, geodetic, and geochemical data, and the thesis outlines a framework for future monitoring in which incorporating shear-wave splitting and interferometry analyses into operational observatories could significantly enhance short-term eruption forecasting.