document preview

Abstract
This paper presents a specific workflow developed to build a detailed geological model constrained by high-resolution 3D seismic data. The objective of the proposed approach was 1/ to integrate both geological and seismic information in a coherent fine-grid model at early stage of field production, 2/ to account for uncertainties in heterogeneity distribution, and 3/ to prepare the integration of 4D seismic data.

The first and critical step of this approach was the definition of a 3D grid of facies proportions from 3D HR seismic data to constrain the geological modeling process. A novel approach has been developed to account for scale differences between seismic and well data to obtain a detailed geological facies description.

In the second step, geological facies proportions derived from the seismic facies characterization have been optimized to minimize the differences between the seismic impedances of the average stochastic model and the real impedances. In the last step, the truncated Gaussian algorithm has been
used to generate stochastic realizations constrained by well data and geological facies distribution. Multiple realizations have been used to quantify uncertainty on facies spatial distribution.

This approach was applied successfully to the Girassol field, a complex and faulted turbidite field located offshore Angola. The objective was to build an initial geological model for an innovative history matching workflow integrating 4D seismic data (Roggero et al., [1]).

Thanks to the exceptionally high resolution of the seismic data, a 3D matrix of lithofacies proportions has been constructed with a resolution close to that of standard well logs (metric). Together with variograms, these geostatistical parameters have been used to generate equiprobable images of the geological model. Both well data and the seismic constraint are honored. Average sand proportions and volumes computed from ten realizations are in good agreement with the average values computed using the seismic constraint. Variability in the distribution of geological facies has been quantified and can be related to the uncertainty of the seismic constraint. This uncertainty has been reduced in the history matching process by introducing more deterministic
geological information.

Introduction
The petroleum industry has been focusing on giant deep offshore turbidite reservoirs over the past decade (Pettingill and Weimer, [2]). Because of the high costs of field development and production, it has become necessary to monitor the dynamic evolution of the field accurately even at early production stages. Developments in 4D acquisition and processing were the first prerequisites to achieve this challenge (Lefeuvre et al., [3]). The next challenge was to correctly integrate 4D seismic data in history matching, which first requires an accurate characterization of these highly heterogeneous reservoirs before the start of production (Mezghani et al., [4]).

The understanding of deep offshore reservoir architecture relies to a large extent on seismic interpretation combined with knowledge from outcrop analogs (Beydoun et al., [5]; Sikkema and Wojcik, [6]; Eschard et al., [7]; Joseph et al., [8]). Seismic information can be used to constrain initial facies models in various ways (Dubrule, [9]). The integration of seismic data in geostatistical facies models can be direct. In this case, the seismic information is integrated during the estimation/simulation phase, using external drift, coestimation, cokriging, and their variations. Another approach consists of an indirect integration starting with a preliminary estimation of seismic and geological attributes, which is followed by the simulation phase. In this case, 2D maps of seismic facies or quantitative attributes are often used (Doligez et al., [10, 11]; Fournier et al., [12]).

A specific workflow has been developed for the geological modeling of the Girassol field, to integrate quantitatively a 3D constraint of facies proportions based on the interpretation of the high-resolution 3D seismic data. The objective is to prepare the integration of 4D seismic data to update the geological model. This final phase is based on an innovative history matching methodology, as presented by Roggero et al. [1].