Abstract: Seismic Constraints for a Stochastic Reservoir Model, Block I Eocene C6, Lake Maracaibo, Venezuela

Azpiritxaga, I.; A. Correa; E. Hernandez; C. Coll; A.Galli; C.Ravenne; C. Castillo; E. Gonzalez; G. Gedler; C. Vasquez - PDVSA

Proper reservoir characterization is essential for optimum hydrocarbon recovery. Reservoir architecture and rock quality are usually based on well and seismic information. Well information from logs, cores, side-wall samples, and cuttings provide a good vertical characterization of the reservoir, whereas seismic information extends this knowledge in lateral dimension. Geostatistical tools provide the link between the well and seismic data.

Our first approach was to perform a calibration between seismic and well stratigraphy. Attribute calculation was done for each horizon; geostatistical techniques were applied to build a quantitative relationship between rock quality and attributes, taking into account the correlation coefficient and spatial coherence. This relation is then incorporated into a Gaussian Stochastic Model in order to characterize the reservoir.

This methodology was applied in Block I, Eocene-C6 reservoir. Two genetic units were defined in C6, limited by three maximum flooding surfaces (MFS C6, C6i, and C7) that were sedimentologically and seismically identified. These two units called Upper (C6-C6i) and Lower (C6i-C7) were deposited in a tidal dominated delta in distal and proximal environments respectively. Each environment has a specific vertical and lateral sedimentary facies distribution.

Seventeen attributes distributed in 9 amplitude, 4 frequency, 2 phase, 1 impedance and 1 thickness from time were calculated for each Unit. Six attributes in the Upper Unit and seven in the Lower Unit had found to be non-linearly correlated. The attributes were also divided into their spatial components resulting in 18 attributes for the Upper Unit and 22 for the Lower Unit. To establish a quantitative relationship between seismic attributes and rock quality properties, a cross validation using Kriging with external drift was applied, using the seismic data as the external variable and well data as the hard variable. Pseudo-property (% sand, f, K, Vsh, and RQI) maps were generated for each unit resulting from the combination of attributes. Each unit was correlated with different attributes stemming from vertical and lateral rock quality variation.

The attributes that correlated best with the sand percentage in the Upper Unit were component 2 of the average absolute value of amplitude (aaa2) and the component 3 of the thickness (thc3). The best attributes for the Lower Unit were the thickness (thc) and the ratio of positive to negative sample (rpn). These relations were introduced into a non-stationary stochastic model and an improvement in the definition and distribution of sand bodies was obtained comparing the models with and without seismic constraints. Based on cross validation techniques, a 20% improvement was obtained. This study indicates that seismic constraints are essential to establish a stochastic reservoir model, which help to decrease reservoir uncertainty to predict new drilling areas.

AAPG Search and Discovery Article #90933©1998 ABGP/AAPG International Conference and Exhibition, Rio de Janeiro, Brazil