Published 1994 as part of Computer Applications 3
Copyright © 1994 The American Association of Petroleum Geologists.  All Rights Reserved.
 

The geostatistical method can be applied to quantitatively relate well and seismic data, assess the quality of the resulting map, and estimate the probability of success directly from the available data. To illustrate the steps of the geostatistical method, we present a case study in which the thickness (in feet) of an oil-bearing sand was estimated using seismic amplitude from a three-dimensional seismic survey as a guide.

The geostatistical method is a four-step procedure that calls on several statistical tools. The first step is to quantify the spatial continuity of the well data using variogram analysis. The second step is to find and quantify a relationship between the well and seismic data. The third step is to use what has been learned to grid the well data using the seismic as a guide via kriging with external drift. The last step is to assess the accuracy of the map just made. Traditionally, a geoscientist creates a map that is assumed to be correct until additional information becomes available. Only rarely is an estimate made of the map's accuracy. A geostatistician creates an expected value or average map and has a quantitative estimate of its accuracy. Conditional simulation is a geostatistical tool that yields a quantitative measure of the error in a map.

In the case study, the geostatistical method was used to estimate the expected value and error in a sand isochore. Cumulative probability distributions and risk maps are two tools used to display the range of error expected in the sand thickness map. These tools, along with economic thresholds, were used to directly derive the probability of success from the data available.

Two wells drilled after geostatistical analysis confirmed that using the geostatistical method is more accurate than traditional, nongeostatistical methods.

The application of geostatistics presented in the second part of this paper has been masked to protect proprietary information. Although the data have been distorted, the application of geostatistics has in no way been affected.