Modeling fault activation due to fluid production: Bayesian update by seismic data
C. Zoccarato, M. Ferronato, A. Franceschini, C. Janna, P. Teatini
Dept. of Civile, Environmental and Architectural Engineering,
University of Padova, Padova, Italy
ABSTRACT
The reactivation of faults and the generation of fractures can be caused by stress changes due to
injection and/or production offluids into and/or from the subsurface. The simulation of these processes,
which could be associated with (micro-)seismicity,is affected by a high uncertainty. The aim of this
work is at developing a mathematical framework to quantify and possiblyreduce the prior modeling
uncertainties by assimilation of seismic data. The mechanics of fault (re-)activation is simulatedby a
Finite Element (FE) numerical model where the discontinuous displacements between the fault surfaces are
suitably considered using appropriate Interface Elements (IEs). The study is carried out by using a
stochastic approach, with a globalsensitivity analysis (gSA) based on Sobol' indices initially performed
to estimate the influence of the input parameters onthe model solution. Then, a Markov Chain Monte Carlo
(MCMC) sampling technique based on the generalized PolynomialChaos expansion (gPC) surrogate solution is
used to update the prior information conditioned on seismic observations. Themethodology is tested on a
3D synthetic test case. The uncertain input is the natural stress regime and the Mohr-Coulomb parameters
characterizing the fault activation criterion. A good reduction of the prior uncertainty is obtained,
showing that the assimilation of seismic data can have a promising potential for improving the subsurface characterization.
