Search

Under certain conditions, ocean surface gravity waves (SGW) interact with the seafloor underneath to trigger relatively faint but measurable seismic waves known as ocean microseisms. Cyclonic storms (e.g. hurricanes, typhoons) wandering over the ocean are major (non-stationary) sources of the former, thus opening the possibility of tracking and studying cyclones by means of their corresponding microseims. For this purpose, we identified storm-related microseisms hidden in the ambient seismic wavefield via array processing. Polarization beamforming, a robust and well-known technique is implemented. The analyses hinge on surface waves (Love and Rayleigh) which, in contrast to P-waves, are stronger but only constrain direction of arrival (without source remoteness). We use a few land-based virtual seismic arrays surrounding the North Atlantic to investigate the signatures of major hurricanes in the microseismic band (0.05-0.16 Hz), in a joint attempt to continuously triangulate their tracks. Our findings show that storm microseisms are intermittently excited with modulated amplitude at localized oceanic regions, particularly over the shallow continental shelves and slopes, having maximum amplitudes virtually independent of storm category. In most cases no detection was possible over deep oceanic regions, nor at distant arrays. Additionally, the rear quadrants and trailing swells of the cyclone provide the optimum SGW spectrum for the generation of microseisms, often shifted more than 500 km off the "eye". As a result of the aforementioned and added to the strong attenuation of storm microseisms, the inversion of tracks or physical properties of storms using a few far-field arrays is discontinuous in most cases, being reliable only if benchmark atmospheric and/or oceanic data is available for comparison. Even if challenging due to the complexity of the coupled phenomena responsible for microseisms, the inversion of site properties, such as bathymetric parameters (e.g. depth, seabed geomorphology), near- bottom geology or SGW spectrum might be possible if storms are treated as natural sources in time-lapse ambient noise investigations. This will likely require near-field (land and underwater) observations using optimal arrays or dense, widespread sensor networks. Improved detection and understanding of ocean microseisms carries a great potential to contribute to mechanically coupled atmosphere-ocean-earth models., Universität Hamburg, poster

The thesis concentrates on monitoring seismic speeds, especially in active fault zones. By correlating signals generated by background noise, one can estimate the Green's function of a medium. When continuously following these functions, wave speed changes in the medium can be detected. Monitoring methods are applied to data from an active fault zone in Parkfield, California, where two wave speed drops, which coincide with regional seismic events, are detected. The largest corresponds to an event close to the stations. Both speed drops are followed by a gradual postseismic relaxation. In order to understand the reliability of the measurements, we perform laboratory experiments. One interesting result of these experiments shows that an exact reconstruction of the Green's function is not necessary for monitoring, which opens up many possibilities of applications to seismology. Armed with this knowledge, the Parkfield data is analysed again. By improving the temporal resolution to 1 day, we show that the observed speed drop is coseismic with the Parkfield event. We establish that the speed fluctuations are not simply correlated to variations in noise source distribution obtained by beamforming. Finally, the developed methods are applied to an event in Japan. Since the array is spatially much larger than the one used at Parkfield, this data is analyzed to study the dependence between station-event distance and the measured seismic speed drop. STAR, La thèse se concentre sur le suivi temporel des vitesses sismiques, notamment dans des zones de faille actives. En corrélant les signaux générés par le bruit ambiant, il est possible d'estimer la fonction de Green du milieu. Par le suivi continu de ces fonctions, des changements de vitesse dans le milieu peuvent être détectés. Les méthodes de suivi temporel sont appliquées aux données provenant d'une zone de faille active à Parkfield, Californie, ce qui permet de détecter deux chutes de vitesse. Ces dernières coïncident avec des évènements sismiques régionaux, la plus importante concernant un évènement proche des stations. Les deux chutes de vitesse sont suivies d'une récupération postsismique progressive. Pour mieux comprendre la fiabilité des mesures on a effectué des expériences en laboratoire. Un résultat intéressant de ces expériences montre que la reconstruction exacte de la fonction de Green n'est pas nécessaire pour le suivi temporel, ce qui ouvre la voie à de nombreuses possibilités d'applications en sismologie. Grâce à cette connaissance, la série de données de Parkfield a été ré-analysée. En améliorant la résolution temporelle à 1 journée, on montre que la chute de vitesse observée est cosismique avec le séisme de Parkfield. On a établi que les fluctuations de vitesse ne sont pas simplement corrélées aux variations de la distribution de sources du bruit obtenue par formation de voies. Enfin, les méthodes développées sont appliquées à un séisme au Japon. Le réseau étant de taille beaucoup plus grande que celui utilisé pour l'étude de Parkfield, ces données sont analysées pour étudier la dépendance entre la distance stations-séisme et la chute de vitesse mesurée.

Geophysical Journal International, 214 (3), ISSN:0956-540X, ISSN:1365-246X