--> Constraining Geostatistical Reservoir Models with Seismic Attributes

Datapages, Inc.Print this page

Constraining Geostatistical Reservoir Models with Seismic Attributes

By

Richard L. Chambers1, Jeffrey M. Yarus2

(1) Quantitative Geosciences, Inc, Broken Arrow, OK (2) Quantitative Geosciences, Inc, Houston, TX

 Complex seismic trace analysis appeared with the advent of seismic sequence stratigraphy in the mid 1970s. Vail and his colleagues expected that seismic attributes would eventually quantify their seismic facies parameters. Since then we have seen a proliferation in the number (hundreds) and often times an inappropriate use of seismic attributes for reservoir characterization. Efforts to understand the meaning of the plethora of seismic attributes include the use of linear and non-linear techniques, such as Fourier spectral analysis, Principle Components, Discriminate Function, neural networks, and others. The idea was that perhaps combinations of attributes might make sense when individually the attributes lacked clear geological significance, except that they revealed some sort of pattern. Most of the attributes are highly correlated simply because they are derivatives of one another and there is no guarantee that their correlation with a reservoir property is meaningful. Great care must be taken when choosing seismic attributes, because it is not unusual to find spurious or false correlations that do not reflect any physical basis for the relationship and the probability of finding false correlations increases with the number of seismic attributes considered and is inversely proportional to the number of data control points. It is time to return to “first principles” and establish a clear relationship between a reservoir attribute, be it facies, porosity, or lithology, for example, and a seismic attribute(s).

We illustrate the use of seismic attributes following a four step procedure: 1) Calibration phase, 2.) Choice of the seismic attribute, 3) Cross-validation, and 4) Prediction and Uncertainty Analysis using Collocated Cokriging and Collocated Cosimulation.