This paper presents the implementation of an integrated reservoir modeling approach that tightly connects and integrates different reservoir modeling disciplines. The approach allows the propagation of subsurface reservoir uncertainties across the various modelling domains, from seismic interpretation though to dynamic reservoir simulation and surface facilities modeling. The results achieved by this approach is a geologically consistent ensemble of runs (100 runs or more) that matches observed data and capable of predicting the future field performance under various development scenarios with a higher degree of reliability. Using an ensemble of runs that first honor the geological facies, as initially defined in the prior probability field, and subsequently updated by reservoir production behavior in the post probability field, enables the user to predict future field performance by allowing hydrocarbon recovery probabilities to be calculated, where the impact of subsurface reservoir uncertainties on prediction results and the possible risk in each development decision is estimated. The field studied is located Offshore in Abu Dhabi and has four main, two secondary, and few minor stacked carbonate reservoirs. Available evidence indicates that the field's reservoirs are not in communication and this study focuses on the main and secondary units. The conceptual geological model proposes that all the zones are conformable, with no truncation or pinching, forming a layer cake depositional model in which reservoirs range in thickness, from few feet in thin zones, to tens of feet in thick zones. Over time the field has been affected by different tectonic stress regimes, resulting in complex strike slip faulting, extending vertically across all reservoirs. Quantification of static and dynamic uncertainties together in the applied methodology is improving the team integration and common understanding for the field structural and geomodelling uncertainties and its impact on the dynamic model behavior for different reservoirs. The methodology has been tested where new drilled well data is included at different times and it showed an easy and quick update for the entire workflow from seismic to simulation. The model was calibrated to historical observed data using different geological uncertainties (structural, facies, geomodelling and dynamic uncertainties) which helped to achieve a geologically consistent and reliable models.

To establish relationships between seismic derived acoustic impedance and LWD porosity measurements from several horizontal wells to be implemented into property modeling. This workflow is a sequential process that integrates property relationships from seismic scale to log scale using log data from a dozen of vertical wells and validate results at field scale with log data from about 50 horizontal wells. Overall process functions at grid-block scale in a 100x100mx1ft cell size following the four main phases. The first phase, involves exploratory data analysis and quality check. This is followed by a second phase of model building to concatenate all the required modeling steps. Third phase of model optimization explores the effect of all the parameters and data links defined in the process. Finally fourth phase involves validation to assess residual errors from the resulting porosity distributions and quantifying predictability of the model itself. A comprehensive and robust set of properties is generated by performing a recursive and convergent process of property modeling using lateral coverage from seismic inversion products and vertical resolution near well log scale. Independent analysis of different scales of porosity measurements are reconciled in this systematic approach by defining average distributions and descriptive statistics of reservoir properties at field scale. Variable data types, sample sizes and data resolution evolves across four different phases that integrates a holistic understanding of datasets in different dimensions. Quantitative analysis of seismic data ultimate correlates to a dense dataset from long horizontal wells. Final predictability of the model reaches a high confidence level (about 80% accuracy) when testing the predicted properties vs real measurements in about 50 horizontal wells. Multiple realizations of properties distribution matching all the available data is final output that provides a better understanding of reservoir property. This workflow allows total utilization of log data from horizontal wells into property distribution with no impact on overall statistics. No complex de-clustering operations are required as all the descriptive statistics are defined from vertical wells calibrated to core and seismic data. This methodology maximizes the value of LWD formation evaluation logs in property distribution, by combining the resolution of the logs along long horizontal wells with the strong lateral coverage of seismic inversion cubes.