Discriminating Between Spatial and Temporal Variations in Seismic Anisotropy at Active Volcanoes

This thesis addresses the measurement and interpretation of seismic anisotropy around active volcanoes via shear wave splitting analysis. An overpressured magma reservoir will exert a stress on the surrounding country rock that may or may not be manifest as observable strain. Shear wave splitting analysis can be a useful indicator of stress in the crust and hence, the pressure induced by magma movement. Changes in shear wave splitting have already been observed at Mt. Ruapehu following eruptions in 1995/1996 and are inferred to be caused by changes in local stress in response to magma pressure. One of the main problems with the interpretation of temporal changes in shear wave splitting is the possibility of spatial variations being sampled along differing raypaths and being interpreted as temporal changes. Using a dense observational network and an automated shear wave splitting analysis, we examine local earthquakes occurring in 2008 within 100 km of Mt. Ruapehu. We note a strong azimuthal dependence of the fast direction of anisotropy (phi) and so introduce a spatial averaging technique and a two-dimensional tomography of recorded delay times (dt), to observe the spatial variation in more detail. Using this new method of mapping shear wave splitting parameters, we have created a benchmark of spatial variations in shear wave anisotropy around Mt. Ruapehu, against which future temporal changes may be measured. The observed anisotropy is used to define regions in which phi agrees with stress estimations from focal mechanism inversions, suggesting stress-induced anisotropy, and those in which phi aligns with structural features such as fault strikes, suggesting structural anisotropy. Data from past deployments of three-component seismometers have been analysed in the same way as those recorded during the 2008 experiment and the results compared. We identify a stable region of strong anisotropy, interpreted to be caused by schistose mineral alignment, and a transient region of strong anisotropy centred on the volcano during the major magmatic eruption of 1995. We also introduce a method of analysing temporal variations in seismic anisotropy at active volcanoes by using tight clusters of earthquakes and highly correlated multiplets. At Mt. Ruapehu, changes in shear wave splitting parameters associated with the 2006 and 2007 phreatic eruptions are detected using a cluster of earthquakes to the west of the volcano. Similar analyses using another cluster and multiplets from the stable region of strong anisotropy do not reveal temporal changes, although examination of the waveform codas of the repeating earthquakes reveals systematic changes that we interpret as being caused by seismic scatterers associated with the 2006 and 2007 eruptions. These scatterers appear to contaminate the shear wave coda and so inhibit the detection of any subtle changes in shear wave splitting parameters. Finally, we apply some of these methods to data from the 2008 eruption of Okmok volcano, Alaska. Shear wave splitting analysis at Okmok reveals a change in anisotropy associated with the 2008 eruption. This change however, is attributed to a change in dominant hypocentre location. Multiplet analysis at Okmok volcano reveals a similar scatterer contamination of the shear wave arrival. This spurious phase is interpreted to be an S to P conversion from interaction with the magma reservoir