Bayesian Ambient-Noise Inversion
Invited paper
University of Victoria
Tuesday 2 june, 2015, 15:40 - 16:00
0.8 Rome (118)
Abstract:
While ambient noise in the environment has generally been considered
detrimental to traditional controlled-source ocean acoustic or seismic
inversions for geophysical profiles, in recent years it has been recognized
that noise itself can provide an extremely useful passive source for
geoacoustic/seismic remote sensing. Potential advantages to ambient-noise
inversion over active acoustic and direct (coring) methods include simpler
instrumentation and deployment procedures, the ability to carry out
unobtrusive surveys or long-term/large-area measurements with low cost and
power requirements, and reduced environmental impact (e.g., minimal disruption
to marine mammals). This talk presents several approaches to estimate seabed
and terrestrial geophysical profiles based on ambient-noise inversion. These
approaches include estimating seabed reflection-coefficient data from
beamformed ambient noise due to wind-blown sea-surface waves, estimating
seabed Scholte-wave dispersion data from ambient-noise measurements at
seafloor hydrophones, and estimating Rayleigh-wave dispersion data on land
using seismometer arrays at both small scales (10s of metres penetration depth
for site assessment of earthquake hazards) and large scales (10s of kilometres
depth for geologic/tectonic crustal mapping). In all cases nonlinear Bayesian
(probabilistic) inference methods are applied in inversion to quantify the
information content of the ambient-noise data to resolve geophysical profile
structure and provide rigorous uncertainty estimates. An important component
of this involves model selection methods to determine appropriate geophysical
model parameterizations consistent with the resolving power of the data, with
both the Bayesian information criterion and trans-dimensional sampling
considered for various ambient-noise data sets.