Reservoir Facies Prediction from Geostatistical Inverted Seismic Data

Abstract

Abstract The estimation of oil in place in a reservoir is a key element in reservoir management. Based on these estimations, economical decisions can be taken. Therefore, it is necessary to take into account all possible sources of uncertainty in reservoir characterization. To address the uncertainty from seismic inversion, we use pre-stack geostatistical inversion to arrive at a family of related IP and IS volumes. The geostatistical inversions explore the possible elastic subsurface models that are constrained by the seismic and well data together with geostatistical modeling parameters. These IP and IS volumes are transformed to a family of seismic facies volumes with the use of a well data derived cross-plot. On the cross-plot, seismic facies regions are interpreted while each seismic facies region contains points of the well identified geological facies. The seismic facies models obtained from the inversions are used to obtain geological facies models. To arrive at such a model, the uncertainty in the points sampled by the well is used to define thresholds for a truncated Gaussian simulation. In this manner, the uncertainties from the seismic data inversion are propagated to the geological facies model. With the availability of probability functions for the petrophysical parameters and the volumes of the reservoir grid cells, the oil in place can be computed. Repeating the procedure for each seismically inverted data volume, a histogram of the oil in place is obtained in which the uncertainty from the seismic inversion is propagated to a geological facies model which can be utilized in an integrated oil in place calculation. Introduction It is standard practice in seismic reservoir characterization to invert seismic data into seismic impedance. When angle stacks are available, inversion for elastic parameters becomes possible, Pendrel, et al.1 and Ma2. The derived elastic parameters can be interpreted separately or together to optimally discriminate between geological facies for the construction of reservoir models. In contrast to standard inversion techniques, which produce a single ‘optimal’ impedance model at the scale of the seismic resolution, geostatistical inversion generates a family of 3D inversion results at higher resolution. These inverted data volumes are constrained by geostatistical modeling parameters, well data and seismic data inside the available bandwidth, Haas and Dubrule3. When angle stacks are available, pre-stack geostatistical inversion has the potential to explore the inversion uncertainties on both the P- and S-wave impedances. As a result of pre-stack geostatistical inversion, a family of high-resolution elastic models of the subsurface is obtained which need to be interpreted and analyzed. The interpretation of this large amount of data makes it impractical to interpret each volume of elastic parameters individually. Hence, methods are developed to analyze the inverted data statistically, Barens, et al.4. To further utilize the uncertainties obtained from the inversion, they need to be brought into the Oil In Place (OIP) calculations. Here, we present a method to include the inversion uncertainties into the computation of OIP. The pre-stack geostatistical inversion method is briefly described. Following the inversion, a description of a method to translate the obtained elastic parameters to geologic facies is discussed. These geologic facies models can serve as an alternative start point for the OIP calculations. When for each of the geological facies present in the model distributions of their relevant petrophysical parameters are known together with the cell volumes, a histogram for the OIP can be computed which includes the uncertainty from the seismic inversions. In an example, we highlight the methodology.