Article ID: | iaor19951017 |
Country: | United Kingdom |
Volume: | 43 |
Issue: | 2 |
Start Page Number: | 385 |
End Page Number: | 398 |
Publication Date: | Apr 1994 |
Journal: | Applied Statistics |
Authors: | Walden A.T. |
Keywords: | petroleum, statistics: multivariate |
The positioning of development wells for efficient production from an oil-field depends critically on being able to find accurately the edges of the reservoir. This problem is addressed by using as the basic data over 1000 short time series collected by seismic reflection profiling. All the time series extend over the ‘reservoir interval’, i.e. the section of rock potentially containing hydro-carbons. Each time series is first reduced to three physically meaningful variables. These variables capture information on the sand/shale ratio in the reservoir interval at the location of the time series. Spatial changes in this ratio will be expressed in the three variables extracted from the set of time series, the locations of which also vary spatially. The cloud of three-dimensional points is then subdivided by using the graphical tools of spinning and projection pursuit, and the edge of the oil-field is hence delineated by mapping these clusters back into their geographical co-ordinates.