Ambient noise tomography with non-uniform noise sources and low aperture networks: case study of deep geothermal reservoirs in northern Alsace, France

We developed and applied a method for ambient noise surface wave tomography that can deal with noise cross-correlation functions governed to first order by a non-uniform distribution of the ambient seismic noise sources. The method inverts the azimuthal distribution of noise sources that are assumed to be far from the network, together with the spatial variations of the phase and group velocities on an optimized irregular grid. Direct modelling of the two-sided noise correlation functions avoids dispersion curve picking on every station pair and minimizes analyst intervention. The method involves station pairs spaced by distances down to a fraction of a wavelength, thereby bringing additional information for tomography. After validating the method on synthetic data, we applied it to a set of long-term continuous waveforms acquired around the geothermal sites at Soultz-sous-Forets and Rittershoffen (Northern Alsace, France). For networks with limited aperture, we show that taking the azimuthal variations of the noise energy into account has significant impact on the surface wave dispersion maps. We obtained regional phase and group velocity models in the 1-7 s period range, which is sensitive to the structures encompassing the geothermal reservoirs. The ambient noise in our dataset originates from two main directions, the northern Atlantic Ocean and the Mediterranean Sea, and is dominated by the first Rayleigh wave overtone in the 2-5 s period range.