HR: 0800h
AN: G11A-0781 [Abstracts]
TI: Bayesian Inference of Lower-Crustal Viscosity Near the Kunlun Fault Based on Geologic, Geomorphic, and
Geodetic Data
AU: * Hilley, G E
EM: hilley@seismo.berkeley.edu
AF: University of California, Department of Earth and Planetary Science, Berkeley, CA 94720-4767
United States
AU: Burgmann, R
EM: burgmann@seismo.berkeley.edu
AF: University of California, Department of Earth and Planetary Science, Berkeley, CA 94720-4767
United States
AU: Zhang, P
EM: peizhen@ies.ac.cn
AF: Institute of Geology, State Key Laboratory of Earthquake Dynamics, Beijing, 00000
China
AU: Molnar, P
EM: molnar@cires.colorado.edu
AF: University of Colorado, Boulder, (4) Dept. of Geological Sciences and Cooperative Institute for Research
in Environmental Sciences, Boulder, CO 80309
United States
AB:
Bayesian methods provide the means of integrating geologic, paleoseismic, seismic, and geodetic data to improve estimates of
fault zone and lower-crustal properties that are important for predicting the long-term deformation of plate boundary zones,
understanding deformation and stress transfer following earthquakes, and estimating future seismic hazard. We developed a
Bayesian methodology that integrates geologic, geomorphic, and geodetic data to provide probabilistic estimates of fault-zone
and lower-crustal properties. Unlike previous studies that have used {\it a priori} information to constrain optimizations,
and bootstrapping to assess uncertainty in model parameters, we explicitly employed Bayes' rule in our analysis and estimate
model parameters using simulation methods.
Our Bayesian methodology provides a means of straightforwardly assessing covariance between and uncertainty within model
parameters, and allows geologic and geomorphic information to be used to quantitatively constrain fault-zone and lower
crustal properties. Because the methodology casts model parameter estimates in a probabilistic framework, inversions
obtained from these methods provide far more information about the behavior of the model parameters inferred from data than
do conventional geophysical inversions. Finally, these methods may be modified to incorporate other a priori information
(e.g., fault zone geometry from microseismicity, and paleo-eathquake timing and recurrence from paleoseismic excavations),
and thus may serve as important tools in seismic hazard assessments that strive to formally incorporate different data types.
We applied this methodology to the Kunlun Fault in northern Tibet, where quantitative estimates of lower-crustal viscosity
are lacking. Here, geologic and geomorphic information constrains the range of permissible long-term slip rates and
coseismically generated offsets. We combined these a priori estimates with GPS velocities using a Bayesian implementation of
an elastic-viscoelastic earthquake cycle model to estimate fault-zone and lower crustal properties in the area. We found
that the non-dimensional relaxation time varies between 0.04 and 3.25 (95% bounds), implying plastosphere viscosity between
$3.8 \times 10^{19}$ Pa s and $6.7 \times 10^{21}$ Pa s. These viscosities are large compared to those required by models of
mid-crustal channel flow, perhaps suggesting that the mid- to lower-crust underneath northern Tibet is significantly more
viscous than previously envisioned.
DE: 7221 Paleoseismology
DE: 8107 Continental neotectonics
DE: 8120 Dynamics of lithosphere and mantle--general
DE: 1243 Space geodetic surveys
SC: Geodesy [G]
MN: 2004 AGU Fall Meeting