Well log stacking: a new technique to improve laterial prediction and sequence stratigraphic interpretation with wireline data
David G. Quirk, Andrew E. Stocks and Fresenay Misker
Journal name: First Break
Issue: Vol 17, No 5, May 1999 pp. 131 - 143
Info: Article, PDF ( 632.21Kb )
The correlation of wireline (well log) data forms an essential part of the interpretation of geological strata in sedimentary basins. The technique usually involves matching by eye the log curves produced from digitally sampled data in one well with those from one or more adjacent wells. Correlation is used primarily to access the lateral extent of rock bodies in the subsurface. However, it is a highly subjective process and is hampered by the difficulty in visually discriminating between non-geological noise, local geological trends and regional geological trends when comparing a number of logs. Variations in the thickness of an interval or differences in the data acquisition between wells compound these problems. Correlation is also constrained by the number and spacing of the wells in question. Thus it is hard to determine what is happening between widely separate wells whilst it is difficult to interpret more than a few log curves at one time due to their complexity. The solution is typically to concentrate on only a few obvious features such as a minimum or maximum peak or the diagnostic response of a specific rock type and to make some simple assumptions as to the lateral extent of other features. Even where additional geological data such as biostratigraphic or core information are available, it is often impossible to be sure that a log pattern in one well is not coincidentally similar to that in another well and regional geological trends will often be disguised by higher amplitude, shorter wavelength local variations. As a consequence of the problems mentioned above, the methodology used in correlating well logs is typically rather arbitrary and is unlikely to use all the data to full advantage. Hence it is difficult to objectively evaluate the validity or confidence level in correlations when assessing, for example, whether reservoirs are likely to be connected across a field or what the chance are of finding a known reservoir in a new well. The subject of this paper is a new automated tool which aims to improve on traditional manual correlation techniques by incorporating all equivalent data from every relevant well into a single log curve. The process suppresses the effect that local variations have on the log signal and enhances the regional response.