Multi Attribute Approach for Quantifying Competing Time Lapse Effects and Implications for Similarity Indicators in Data Assimilation

Abstract

AbstractThis study aims to reduce uncertainties related to non-uniqueness in the interpretation of competing 4D effects and their impact to the interpretations and data assimilation into reservoir models. The study is based on synthetic and observed 4D seismic data, tracers, laboratory measurements, production and geophysical well log data. The methodology involves: (1) building and calibrating a petro-elastic model; (2) forward modeling of each separate physical effect such as saturation, pressure, salinity and noise; (3) combining the effects according to simulated production scenarios; and (4) correlating the modeled with observed 4D seismic data. We generate synthetic logs and seismic traces to quantify the time-lapse observations and to analyze how the combination of effects may affect the seismic character.This work demonstrates a major impact of competing 4D effects in the number and types of true possible interpretations. Yet, estimating their influence on the magnitude and polarity of 4D signal is still achievable. We show application examples from a Brazilian turbidite reservoir, in a setting where a combination of 4D effects occur simultaneously and the confidence level on the 4D interpretations is high due to good quality and frequent seismic data from a permanent reservoir monitoring system.In the example presented, a scenario where only water saturation effects occurs (e.g. aquifer water invading oil zone), 12 to 16% P-impedance (IP) increase (hardening effect) is expected. However, according to salinity measurements and tracers, injected sea water with lower salinity than the formation water decreases the hardening effect towards 8 to 10% in IP change. Having identified the impact of salinity changes in the 4D effects, an approach to include such changes in the petro-elastic model is proposed to generate 4D attributes that reproduce this interaction of effects. The outcome is a better match between modeled and observed 4D attributes, as compared to modeled attributes from a conventional petro-elastic model that considers salinity as constant. Additionally, combining these effects with gas going out from solution due to depletion below bubble point, the IP decreases significantly, reversing the polarity of 4D signal, and a 10% increase in gas saturation produces a softening effect (15% IP decrease).Competing 4D seismic effects are often mentioned but rarely quantified, and their resultant non-uniqueness impacts on similarity indicators used in data assimilation processes. As these similarity indicators are generally based on comparisons between observed and modeled seismic, the methodology presented results in an improved confidence on the workflows. Additionally, the methodology proposed is straightforward and adequate to reduce uncertainty related to 4D seismic interpretations.