Conditioning of a Carbonate Reservoir Model Using Seismic data in an Offshore Upper Mishrif Field in the Middle East

The studied reservoir, a large mature offshore Middle Eastern field, producing from Upper Mishrif limestone series, shows strong heterogeneities (vertical and lateral), due to the original properties of the sedimentary facies and the effect of several subsequent phases of diagenesis and fracturing. This results in complex dynamic field behaviors, which can be difficult to predict including strong saturation variation in the long transition zone, localized dual porosity/permeability flow, early water breakthrough, strong water-cut and uneven/non-uniform pressure support. In order to better predict & optimize future production and/or support additional developments, a robust reservoir model integrating all available data (geological, geophysical and dynamic) is clearly needed. This paper focuses on two key steps in geo-model construction: facies and porosity modeling and more specifically on the use of seismic data for reservoir characterization while taking care of over-utilization. Aims are to improve model reliability and prediction capabilities. Facies and porosity modeling was constrained by numerous data from cored exploration vertical wells and horizontal development wells. Initially reservoir facies represented by eight “electro-facies” are distributed according to the available well data and proportion maps derived from regional knowledge. Then porosity is populated within each facies with dedicated distribution and honoring well data. In parallel, a 3D pseudo-porosity cube was generated by applying the strong linear correlation observed between acoustic impedance and well porosity to an acoustic impedance dataset resulting from a post-stack inversion. It serves as an additional large-scale constraint for modeling: the vertical resolution of the pseudo-porosity cube is roughly 15 meters, while facies and porosity changes occur vertically within 3-8 meter thick layers as described from the well data, it extends over most of the field. It is integrated (1) as a direct soft trend for porosity and (2) as qualitative check of the final geo-model. Once porosity is modeled; it is converted into pseudo-impedance and a synthetic seismic dataset for comparison with original data. In addition the pseudo-impedance was compared to the acoustic impedance cube resampled at reservoir model scale. Both allow significant seismic-scale inconsistencies to be easily detected and understood for further iterative correction. Once areas and origin of mismatch between seismic data and the geological model were identified, reservoir model porosity, and consequently facies, had to be locally modified to ensure consistency with geophysical information. The modified reservoir model porosity was then used to regenerate a synthetic seismic data and compared to the original seismic data in order to check if the modifications were sufficient to obtain optimum coherency between all data. This kind of feedback loop is an efficient iterative process, a time-consuming but necessary step for relevant geo-model quality control. The consistency between all available geological and geophysical information gives high confidence in the result, and ignores seismic noise. It ensures the robustness of the geological model and consequently ensures a sound basis for dynamic modeling.