• The most detailed predictive hazard maps being made now are based on soil or rock type.
• S-velocity measurements are rare.
• Our objective is to develop a technique for cheap and easy shallow S-velocity measurments, applicable to a large number of sites.
• It has to work in noisy urban environments.

In July 1997 we tested noise recording by common seismic refraction instruments at the Reno/Tahoe Airport.

• Two crossed 350 m arrays of 24 channels each - many sensors allows better velocity correlation.
• Vertical 8 Hz refraction single phones - most common type of sensor, cheap, setup needs no burial or leveling.
• 500 m from jet touchdown point, 5 m from busy intersection.
• Also recorded surface waves of 3000 kg construction blast 16 km south.
• Kyoto Univ. and Shimizu Corp. colleagues set up arrays of broadband accelerometers in December 1997.
• Nearby Double Spring Flat M6 aftershock records suggested two times the amplitude of rock sites.

Summed power spectrum of all 48 s records from refraction equipment.

• Very surprising that 8 Hz exploration geophone sensors had significant power down to 2 Hz.
• In a cluster test the correlation coefficient between sensors at 2-4 Hz was always above 97%. No phase distortion.

Velocity Spectrum Analysis

1. Low-pass filter records to maximum frequency of interest.
2. Slantstack (p-tau transform) each record from (x,t) to (p,tau).
3. Fourier transform each trace from (p,tau) to (p,f).
4. Compute spectral power at each (p,f).
5. Sum power at -p into transform at +p.
6. Sum the power spectra of several records at each (|p|,f).
7. Find spectral ratio of power at each (|p|,f) against average power at f over all |p|.
8. Identify and pick trends of high ratio that show normal-mode dispersion by sloping to lower velocities at higher frequencies.

Spectral-ratio plots computed as above from the airport noise records. Warmer colors are higher spectral ratios. The vertical axis is slowness increasing linearly down, so velocity increases non-linearly upwards as noted. Frequency increases linearly to the right.

Slopes down to the left are record truncation, aliasing, and slantstack artifacts. Any spatial Fourier analysis would be aliased below the white dotted lines. The slantstack can follow velocity trends, with increasing error, into the aliased area, because more than two traces are used and surface waves arrive in groups and are not continuously harmonic.

• Upper left: curves follow general dispersion trend of 2000 kg blast record down to the right.
• Lower left: stacked sledgehammer record shows no dispersion.
• Upper right: traffic and plane touchdown noise record shows dispersion. Values picked at circles inverted below.
• Lower right: summed noise and blast record shows clear trend.

Energy concentrates near the true velocity due to the non-linearity of the cosine azimuth factor. Assuming noise is coming from all directions, two adjacent |p| traces in these plots cover a third of the azimuthal circle, and a third of the energy.

Noise record dispersion picks and synthetic dispersion at Reno Airport.

• We compute synthetic dispersion using a code of Saito modified by Yuehua Zeng.

Shallow velocity model yielding synthetic dispersion above.

• 350 and 600 m/s velocities and 20 m deep increase constrained well.
• 1500 m/s velocity trades off with 40 m depth.

Comparison of array results from Reno. Plot by T. Iwata of DPRI, Kyoto Univ. of noise dispersion by F-K analysis of 1-km and 100 m accelerometer arrays. Thick dashed line is model above from 24 refraction geophone noise record, from 4-8 Hz.

• Excellent match between different recording and analysis methods.
• The quick refraction geophone arrays can get S-velocity structure to 30-50 m depth.
• Larger arrays needed to get below 50 m.
• No source needed; just record noise.

How sensitive are these velocity spectral analyses? Can they really distinguish a low-velocity region from a normal region? Can the technique be applied to existing seismic survey data sets?

• In March 1998 UNR did a 3.5 km reflection survey using the 8 Hz single phones and 2 kg blasts, to profile a shallow-dipping, basin-bounding normal fault in Dixie Valley, Nevada, 2 hours east of Reno.
• Examine a record from a 48-channel, 720 m array near the fault on bouldery piedmont alluvium, with granite bedrock <350 m deep.
• Also examine a similar record 2.5 km to east of fault, atop >1 km of sediments, with a clay playa at the surface.

Velocity-spectral analyses of the playa and piedmont records.

• Normal-mode dispersion down to right is clear, despite down-to-left aliasing and truncation of body wave artifacts.
• Playa record shows phase velocities do not exceed 350 m/s.
• Piedmont record shows velocities not below 350 m/s, and 2000 m/s velocity of bedrock.

• As long as extended source or receiver groups are not used, existing surveys recording strong surface waves yield dispersions easily.

What if I only have a 12-channel recorder?

• Sections at left test using only 12 of 24 channels of Reno Airport noise record, and 12 of 48 channels of Dixie Valley playa record.
• The 12-channel array is still made as long as possible, 180 m in these cases.

• 12 traces of the traffic noise record show loss of velocity resolution, and more apparent velocity values.
• 12 traces of the explosion record provide better results than 48, since the near-shot traces are more dominated by surface waves.