Value of Information: Comparing Surface-Wave Dispersion Curves Estimated from Conventional Seismometers Versus Distributed Acoustic Sensing

Abstract:

The efficacy of geophysical data to estimate key subsurface parameters is difficult to quantify given the complexity of both the signal and the earth, among many other factors. To address this, we utilize a metric from decision analysis known as the value of information (VOI). We analyze the uncertainty of surface-wave dispersion curves derived from travel times recorded by two types of seismological sensors: 3-component seismometers and Distributed Acoustic Sensing (DAS), a technique for measuring longitudinal strain in fiber-optic cables. Both data types were recorded at the Garner Valley test site in California. A 45 kN shear-shaker source produced a swept-frequency input from a few Hz to 10 Hz and back over 60 seconds. The geophone and DAS traces were filtered to remove harmonics from the source, traffic and other external noise. Source-Synchronous Filtering (Lord et al., AGU 2015) was applied to obtain waveforms. To measure the travel time from the source to the sensor, multiple zero-crossings are picked for frequencies between 4 and 20 Hz. These picks are used to estimate phase velocities as a function of frequency for both data types by plotting the travel times versus the distance of the sensor from the source. The slope of the best-fitting line provides an estimate of the phase velocity at a given frequency. To assess its uncertainty, we use a nonparametric bootstrap procedure [Efron & Tibshirani, 1986]. The resulting distributions of phase velocities describe the precision of the estimates for each data type. We then plot the phase velocities as a function of their frequency to generate many dispersion curves. These dispersion curves are inverted to estimate the posterior distribution of shear wave velocity with depth for each type of data. We hypothesize that this information is used to make a decision (e.g. how to develop a geothermal field), and therefore the VOI technique can be applied. This approach provides a metric for evaluating the VOI of each of the two data types. This research is part of the larger PoroTomo project http://geoscience.wisc.edu/feigl/porotomo.