Joint inversion of seismic AVA and CSEM data needs rock-physics models to link seismic attributes to electrical properties. Ideally, we can develop physical-based models (e.g., Gassmann’s equation and Archie’s law) from nearby borehole logs to connect them through reservoir parameters (e.g., porosity and water saturation). However, in the exploration stage, this could be very difficult because information available are typically insufficient for choosing suitable rock-physics models and for obtaining reliable estimates of their associated parameters. The unsuitable rock-physics models and the inaccuracy of their subsequent estimates of model parameters may result in very misleading inversion results. However, it is fairly easy to derive statistical-based relationship (e.g., covariance matrix and regression equations) among seismic and electrical attributes from distant borehole logs. The statistical-based relations may improve estimates of both seismic and electrical properties even with uncertainty. We test such a hypothesis in this study by developing a Bayesian model for joint inversion of seismic AVA and CSEM data. We estimate seismic P- and S-wave velocity, density, electrical resistivity, and lithotypes (shale, brine sand, or oil sand) as functions of depth by conditioning to seismic AVA and CSEM data. The spatial dependences of those quantities are carried out by lithotypes through Markov random fields. We demonstrate the effectiveness of the developed model using a 2D synthetic model and derived datasets that has been developed to closely mimic field conditions. We use borehole logs at two different locations (one with and the other without reservoir) to obtain statistical rock physics relations; we invert seismic AVA and CSEM data at a third location with reservoir using the derived rock-physics models. Comparison of the inversion results shows the developed Bayesian model is effective and combination of seismic AVA and CSEM data improves estimates of geophysical attributes. The developed model can be extended to incorporate estimates of reservoir parameters (e.g., porosity and fluid saturation).
Joint inversion of seismic AVA and CSEM data has been demonstrated to be beneficial for reservoir parameter estimation (Hoversten et al., 2006; Chen et al., 2007) because they provide complementary information on reservoir parameters. This estimation requires rock-physics models to connect seismic attributes to electrical conductivity, which are often derived from suitable nearby borehole logs in practice. First, an appropriate family of rock-physics models is chosen and then those parameters associated with the rock-physics models are estimated by fitting them to the selected borehole logs. However, in the exploration stage, the derivation of detailed rock-physics models could be very difficult due to the lack of nearby wells. However, we can often find logging data at more distant locations where parameter covariance- or regression-based relations can be derived sufficiently. The statistical-based relationship, with associated uncertainty, may improve estimates of seismic and electrical properties (thus reservoir parameters) even though they are less precise than physical-based rock-physics models.
In this study, we develop a Bayesian method based on statistical rock-physics models derived from borehole logs to jointly invert seismic AVA and CSEM data. We apply the developed stochastic model to a synthetic dataset generated from a 2D synthetic model that closely mimics field conditions.