1. Susan L. Beck
George Zandt
Kevin M. Ward
Jonathan R. Delph

2. Work Flow for Ambient Noise Tomography Analysis
After Bensen et al. 2007

3. Green’s Function (Earth’s response)
A. Raw Data
What we need to obtain Rayleigh-wave phase velocities from Ambient Noise Tomography:
Continuously recorded seismic waveform data
Contemporaneously operating stations
Correct (or common) instrument responses
Source Function
Instrument Response
Green’s Function (Earth’s response)
Wavelet

4. B. Pre-Processing Data Temporal normalization
After trend, mean, and instrument responses removed, we normalize amplitudes in the waveforms.
Down-weights large amplitude signals that can dominate waveform
Leads to poor cross-correlations
Good for removing earthquake signals
Raw Data
After temporal normalization (running absolute mean method)

5. B. Pre-Processing Data Spectral Whitening/Normalization
Removes spectral power biases from microseismic energy and other consistent noise sources
Before spectral whitening
After spectral whitening

6. C. Cross-Correlations and Stacking Time Matters
Cross-correlation signal-to-noise ratio improves with increasing data
(small improvements after ~1 year)

7. C. Cross-Correlations and Stacking Moveout
Clear moveout between station is observed
These waveforms can be analyzed via Frequency-Time Analysis to solve for fundamental-mode Rayleigh wave phase velocities between stations

8. C. Cross-Correlations and Stacking Symmetric Component
In theory, positive and negative time lags should be the same.
Reasons they aren’t:
Local multipathing/scattering
Direction and magnitude of energy input
Stack positive and negative time lags to get “average” waveform (Green’s Function) and increase signal-to-noise ratio

9. D. Frequency-Time Analysis
Uses a Gaussian filter on a seismogram and calculates the “power” around a frequency (period) of interest to obtain velocity
Interstation Dispersion Curves!

10. E. Tomographic Inversion
Will be discussed later.