Markov Bayes Simulation for structural uncertainty estimation

Reservoir models are built using disparate datasets each of which may be prone to experimental and interpretational errors and therefore a resulting reservoir model is generally associated with uncertainties. One of the primary sources of uncertainties lies in the structure (or reservoir architecture) estimation from seismic data. Geostatistics can be used to integrate seismic data with well data for the purpose of structural uncertainty estimation. In this paper we present a case study from the Gulf of Mexico, where structural uncertainty associated with a seismic horizon is modeled using Markov-Bayes stochastic simulation. For this simulation, seismic data is used as “soft” or secondary data while well log derived marker depths are used as hard data. Simulation results show uncertainty distributions with smaller variance in the vicinity of the wells. However, in regions away from the wells, the interpreter-picked horizon appears to fall outside the error bounds predicted by our stochastic algorithm. Lack of well control, existence of faults, improper choice of seismic processing parameters (error in time migrated images) and interpreters' bias are some of the plausible causes of this disparity.