4-D Transdimensional Tomography of Iceland Using Ambient Noise

Located at the east of Greenland and immediately south of Arctic Circle, Iceland is the largest volcanic island in the world and represents a unique region of particular interest to geosciences. Various seismological imaging techniques have been deployed to shed light on composition and thickness of the Icelandic crust with serious geodynamic repercussions (for a recent review, see Foulger (2010)). Due to an abundance of active volcanoes, Iceland can be considered a natural laboratory for studying volcanic earthquakes with anomalous seismic radiation (e.g. Tkalcic et al., 2009; Fichtner and Tkalcic, 2010). Temporal changes in the velocity field due to volcanic processes effect seismic waveforms and are important to consider in the context of seismic sources, whose understanding relies on complete understanding of Earth structure. Apart from reflection and refraction studies and teleseismic signals, ambient noise tomography has been recently utilised to image shallow subsurface of Iceland (Gudmundson et al., 2007). The confluence of North Atlantic and Arctic oceans delivers a strong and relatively evenly distributed noise field, therefore making Iceland an ideal place for an ambient noise study. We initially attempt to confirm previous results of Gudmundson et al. (2007) using conventional surface wave tomography derived from Rayleigh wave group velocity dispersion, with fast marching method as a method of choice for forward modelling (Rawlinson and Sambridge, 2005). We perform cross-correlation over several three-month time intervals of ambient noise obtained from the HOTSPOT experiment (Foulger et al., 2001) distributed across Iceland and we discuss seasonal variation observed in cross-correlograms. To extend conventional methods of imaging, trans-dimensional and hierarchical Bayesian sampling methods are used to produce a multidimensional posterior probability distribution of seismic velocity field. We use a trans-dimensional Bayesian inverse method, as it has an excellent property that it treats the number of model parameters (e.g. the number of basis functions) and the noise in the data as an unknown in the problem (Bodin and Sambridge, 2009; Bodin et al., 2012). This approach is advantageous over general optimisation framework where we need to assume the knowledge of noise in the data, define the smoothness and damping, etc. The level of data noise is crucial because it effectively quantifies the usable information present in the data (a very noisy dataset does not contain much retrievable information) and here it naturally controls the quantity of information that consequently should be present in the model (i.e. the number of model parameters). This work is further extended to image the crust for different time intervals and study ongoing dynamics of the Icelandic crust. References: Bodin and Sambridge (2009), Geophys. J. Int., 178, 1411-1436. Bodin et al. (2012), J. Geophys. Res., 117, B02301. Fichtner and Tkalcic (2010), Earth Planet. Sci. Lett., 297, 607-615. Foulger et al. (2001), Geophys. J. Int., 146, 504-530. Foulger (2010), Plates Vs Plumes: A Geological Controversy, 328. Gudmundson et al. (2007), Geophys. Res. Lett., 34, L14314. Rawlinson and Sambridge (2005), Explor. Geophys., 36,341-350. Tkalcic et al. (2009), Bull. Seismol. Soc. Am., 99, 3077-3085.