Data assimilation of surface displacements to improve
geomechanical parameters of gas
storage reservoirs
C. Zoccarato, M. Ferronato, G. Gambolati, P. Teatini
Dept. of Civile, Environmental and Architectural Engineering,
University of Padova, Padova, Italy
D. Bau'
Dept. Civil and Structural Engineering, University of Sheffield, Sheffield, UK
A. Alzraiee
Sandia National Laboratories, Carlsbad, New Mexico, USA
ABSTRACT
Although the beginning of reservoir geomechanics dates back to the late 1960s, only recently
stochastical geomechanical modelling has been introduced into the general framework of reservoir
operational planning. In this study, the ensemble smoother (ES) algorithm, i.e., an ensemble-based data
assimilation method, is employed to reduce the uncertainty of the constitutive parameters characterizing
the geomechanical model of an underground gas storage (UGS) field situated in the upper Adriatic
sedimentary basin (Italy), the Lombardia UGS. The model is based on a nonlinear transversely isotropic
stress-strain constitutive law and is solved by 3-D finite elements. The Lombardia UGS experiences seasonal
pore pressure change caused by fluid extraction/injection leading to land settlement/upheaval.
The available observations consist of vertical and horizontal time-lapse displacements accurately measured
by persistent scatterer interferometry (PSI) on RADARSAT scenes acquired between 2003 and 2008.
The positive outcome of preliminary tests on simplified cases has supported the use of the ES to jointly
assimilate vertical and horizontal displacements. The ES approach is shown to effectively reduce the spread
of the uncertain parameters, i.e., the Poisson's ratio, the ratio between the horizontal and vertical Young
and shear moduli, and the ratio between the virgin loading (I cycle) and unloading/reloading (II cycle)
compressibility. The outcomes of the numerical simulations point out that the updated parameters depend
on the assimilated measurement locations as well as the error associated to the PSI measurements. The
parameter estimation may be improved by taking into account possible model and/or observation biases
along with the use of an assimilation approach, e.g., the Iterative ensemble smoother, more appropriate for
nonlinear problems.
