Seismic array analysis of ocean induced microseisms

Microseisms are weak continuous oscillations of the solid Earth. These oscillations form the background signal on all seismograms, hence the alternative name 'ambient noise'. In the period range 0.5 - 20 seconds, these vibrations are generated in the oceans and their presence can be attributed to two generation mechanisms. Primary microseisms are directly correlated to the period of the ocean waves, and are generated close to coastal areas with a shallow and sloping sea bottom topography. Secondary microseisms show frequencies of double that of their ocean wave sources and are generated in the deeper oceans along storm paths and in coastal areas. An improved understanding of the ambient noise wavefield therefore advances seismological techniques with wide potential applicability, e.g. to ocean wave studies in areas with limited direct observation. A popular technique to study the weak ambient noise wavefield is plane wave beamforming, which utilizes all sensors in a seismic array simultaneously to improve the detection of coherent seismic energy. Beamforming allows estimatation of the directional and slowness dependent seismic wave propagation present in the wavefield. The accuracy of such analysis is governed by the size and configuration of the seismic array and the beamforming technique. To date, the most commonly used plane wave beamforming technique is frequency wavenumber analysis, which was invented over half a century ago. Given that the number of seismic arrays suitable for the analysis of microseisms is limited, the resolution of the frequency wavenumber analysis has been restricted to the detection of the strongest sources only. Improving the beamforming technique has the potential to improve the estimation of weaker signals in the wavefield and allow the application of beamforming to a wider range of arrays. This thesis is divided into two parts, the improvement of beamforming techniques for the array analysis of ocean induced microseisms and the application of these new implementations to seismic arrays in Australia. In the first part, multiple beamforming algorithms and preprocessing schemes are evaluated to design an optimal beamforming framework for the analysis of ocean induced microseisms. The emphasis is on the accurate detection of multiple simultaneous seismic waves, the overall improvement of the beamforming power spectrum resolution and the extension from narrow to broadband (in terms of observed frequency range). This task is achieved by using the minimum variance distortionless response (Capon) beamformer supported by diagonal loading in an 'incoherently averaged' signal approach, hence the name, IAS Capon. The performance is evaluated using data from multiple seismic arrays and is found to be superior to the existing standard techniques. Further, the CLEAN algorithm originally developed in radio astronomy and the Richardson-Lucy deconvolution from visible light astronomy are explored, to further enhance the resolution capabilities of plane wave beamformers in seismology. Deconvolution is used to remove the imprint of the array response function, which is a result of the geometrical configuration of the array stations. It reduces the bias of weaker sources in the power spectrum and also reduces artefact/ghost sources. The best performance is achieved with a modified implementation of the CLEAN algorithm and demonstrated with multiple seismic arrays. CLEAN, as implemented here, iteratively removes energy from the strongest source in the power spectrum. As the parameters of the removed energy are known, it can be placed into a separate 'clean' power spectrum without any array response contributions. Hence, the algorithm is able to decompose the wavefield into it most fundamental signal contributions. The technique is successfully extended to three components and is capable to accurately decompose the three dimensional microseism wavefield into its fundamental energy contributions. In the second part, the research presented in this thesis makes use of the newly implemented algorithms to analyse secondary microseisms in the southern hemisphere and western Pacific, which have received relatively little attention with regard to generation locations and seasonal patterns of the microseism wavefield. Over two decades of ambient noise data from the Warramunga array in Northern Territory, Australia, are evaluated with the IAS Capon approach. This is carried out on an hourly basis for multiple microseismic sources over a broad frequency range. P wave generation regions are found to be frequency dependent and display seasonal patterns. For surface waves, a transition from Rayleigh to Lg waves is observed. A correlation with an ocean hindcast strengthens confidence in the accurate retrieval of weaker sources with IAS Capon. Finally, the CLEAN approach is used to decompose the three dimensional wavefield with the Pilbara array located in Western Australia, for a full calendar year (2013). The analysis allows the separation of multiple types of surface waves and a mean power representation is given. Special emphasis is placed on Love waves, as their generation mechanism in the secondary microseism range has not yet been established. The results shed new light on the generation regions of Love waves and suggest conversion from Rayleigh waves at sedimentary basin boundaries.