Ambient Noise Cross-Correlation Green's Tensor Analysis at The Geysers Geothermal Field, Northern California

We retrieve empirical Green's Functions at The Geysers (TG) in the frequency range from 0.2 to 0.9 Hz for a wide range of interstation distances ranging between 1 to 25 km, corresponding to 0.23 to 5.7 times the wavelength. The Green's functions are estimated from the cross-correlation of ambient noise recorded by a variety of sensors (surface short-period, broadband, accelerometers, and borehole geophones) in and around the reservoir area. The methodology preserves the relative amplitudes of the different components of the 9-component Green's Tensor that allows us to directly compare Noise-derived Green's Functions (NGFs) with normalized displacement waveforms of complete single-force Synthetic Green's Functions (SGFs) computed assuming various 1D and 3D velocity models. In the T-Tangential, R-Radial, Z-Vertical reference frame the TT, RR, RZ, ZR and ZZ components (1st component: force direction, 2nd component: response direction) of NGFs show clear surface waves and even body wave phases for many station pairs and are remarkably similar to SGFs in waveform shape and relative amplitude at all distances. The NGFs are found to be broadly consistent with seismic velocity features inferred from body wave travel-times, and the TT component exhibits body wave geometrical spreading decay at distances < 3.5 km and surface wave decay at distances beyond 3.5 km. We find large anomalous amplitudes in TR, TZ, RT and ZT components of NGFs at small distances (< 4 km) that can be attributed to 10°-30° sensor misalignments at many stations inferred from analysis of long period teleseismic waveforms. After correcting for sensor misalignments, many NGFs for longer paths (> 8 km) across the center of the reservoir area still show significant amplitudes in these components. This region marks an increase in seismic velocities from the NW section to the SE section of TG along with a prominent low VP/VS ratio anomaly, and the 3D velocity model incorporating these features is able to reproduce the complex NGF waveforms well. We show that NGF waveforms present an effective means of evaluating, validating, and refining existing velocity models, which is potentially beneficial for developing velocity models for seismic source analysis because of the lack of source mechanism bias using the complete Green's Tensor.