Stratigraphic analysis of logs
Some log analysis techniques are available, which can be used to improve the stratigraphic resolution of wells or sections.
The different between the two methods is that the first is mainly used to asses the amount of time represented by the sediments (log units) and the second is used to find 'natural' breaks between units. Therefore they are best used in conjunction, i.e. first run stratigraphic clustering over a log to decide where unit boundaries should be located and then run a spectral analysis per unit to estimate the duration of that unit.
Cyclicity analysis involves running a statistical analysis over a section of log to detect regularities in the pattern, interpreted as cyclic sedimentary events. The results are typically displayed in a graph with variance on the vertical axis and the harmonics (wavelenghts) on the horizontal axis. This graph therefore shows the variance contained in each harmonic, high values (peaks) usually indicating frequent occurrence of a wavelength in the measured log portion.
Since the thickness of the measured log portion is known, the thickness related to each harmonic (wavelength) is known. By comparing the thickness ratios of wavelenght frequency peaks, we can compare these ratios with those established by Milankovic for the earth's orbital parameters. If these ratios are similar, we infer that the measured frequencies are caused by these orbital parameters and the duration of the measured log portion can be calculated by multiplying the harmonic with the Milankovic cycle duration.
The method can therefore be used to check an existing stratigraphic subdivision by measuring the durations of the proposed units and comparing these to those already proposed.
If no subdivision exists, Cyclicity Analysis can be used to generate one. This is achieved by repeatedly running the analysis over a fixed number of data points (a data window), each time moving the window one data point down. The resulting output shows the location of the peaks of each analysis, thereby delineating log portions with stable sediment accumulation rates (where peak locations do not shift significantly) and log portions where the absence of peaks indicates that rapid changes of sediment accumulations rates have produced an absence of peaks. Iteration of this process with different data window sizes results in a number of these patterns which, when put next to each other, improve the confidence in the reality of the breaks.
The proposed subdivision should be checked against other proposed subdivisions and can be verified by measuring durations over the proposed units.
The limitation of the method is the number of data points over which one can use the statistical methods and is therefore mainly related to the resolution of the measured log and the thicknesses of proposed units. In practice, using less than 254 data points may result in spurious results when the Fast Fourier analysis is used (e.g. 254 * 0.15 m = 38.1 m, 0.15 m is the usual data spacing for an electric log). The Maximum Entropy analysis allows for less data points (approximately 100). See ten Kate (1992) Tables 1 and 2 (p.218 and p.220) for a summary of advantages and disadvantages of the methods.
Stratigraphic Clustering uses a full log suite over an interval and differs from the usual Log Clustering method in the additional stratigraphic constraint added to the algorithm. It results in a proposed subdivision of the measured interval based on the cluster dispersion cut-off value and leaves it to the interpreter to choose the desired cluster cut-off value.
This method is a valuable help in subdividing a section without any prior knowledge and is typically used as the proposed subdivision for Cyclicity Analysis if no other stratigraphic data, such as biostratigraphic data, is available.
The method is limited in that it takes into account only those phenomena which can be measured by electric logs, thus ignoring grain size, grain angularity, burrowing etc.
The presentation of the results shows one of the used logs (usually the Gamma Ray log), with the proposed subdivisions (based on all logs and taking a 25, 50 or 75% dispersion cut-off) and the cluster pattern. Other subdivisions can be obtained manually using different cut-off values in the cluster pattern.
Gill, D., Shomrony, A. & Fligelman, H., 1993. Numerical zonation of log suites and logfacies recognition by multivariate clustering. AAPG Bulletin Vol.77 No.10.
ten Kate, W.G.H.Z. & Sprenger, A., 1992. Rhythmicity in deep water sediments, documentation and interpretation by pattern and spectral analysis. Ph.D. Thesis, Free University of Amsterdam. Drukkerij Elinkwijk, Utrecht, The Netherlands. ISBN 90-9005486-3.
Send mail to email@example.com with questions or comments about this web site.