Bayesian facies inversion on a partially dolomitized isolated carbonate platform: A case study from Central Luconia Province, Malaysia

Abstract

We present a case study of geophysical reservoir characterization where we use elastic inversion and probabilistic prediction to predict 9 carbonate lithofacies and the associated porosity distribution. The study focuses on an isolated carbonate platform of middle Miocene age, offshore Sarawak in Malaysia, which has been partly dolomitized — a process that increased porosity and permeability of the prolific gas reservoir. The 9 lithofacies are defined from one reference core and include a range of lithologies and pore types, covering limestone and dolomitized limestone, each with vuggy varieties, as well as sucrosic and crystalline dolomites with intercrystalline porosity, and also argillaceous limestones, and shales. To predict lithofacies and porosity from geophysical data, we adopt a probabilistic algorithm that employs Bayesian theory with an analytical solution for conditional means and covariances of posterior probabilities, assuming a Gaussian mixture model. The inversion is a 2-step process, first solving for elastic model parameters P- and S-wave velocities and density from 2 partial seismic stacks. Subsequently, lithofacies and porosity are predicted from the elastic parameters in the borehole and across a 2-D inline. The final result is a model that consists of the pointwise posterior distributions of facies and porosity at each location where seismic data are available. The facies posterior distribution represents the facies proportions estimated from seismic data, whereas the porosity distribution represents the the probability density function at each location. These distributions provide the most likely model and its associated uncertainty for geological interpretations.

Study on Fine Geological Modeling of Sandstone Reservoir under the Influence of Igneous Rock─A Case of Bohai X Oilfield

“…Xu et al used high-frequency information outside the dominant frequency band of earthquake to clearly reflect the location and lateral distribution of karst caves. Ghon et al used Bayesian method to predict lithofacies and porosity of some dolomitized carbonate platforms by using parameters such as P-wave velocity, S-wave velocity, and density. Third, research on modeling methods for igneous rocks: The distribution and physical properties of volcanic reservoirs are controlled by many factors, including tectonics, paleotopography, paleoclimate, volcanic eruption weather conditions, magmatic evolution, volcanism and postvolcanic eruption burial, diagenesis, and epigenetics. − Therefore, volcanic reservoirs have greater complexity and stronger heterogeneity than clastic reservoirs, and their research difficulty is far greater than that of clastic reservoirs.…”