Geostatistical Analysis of Uphole Syrvey Data in the Penola Trough, South Australia..
Honours Degree, 2001
University of Adelaide
Obtaining a good subsurface image from onshore seismic data requires accurate static corrections. Field statics are often derived from uphole survey data, for example in the Otway and Cooper Basins of South Australia. Typically the near-surface velocity structure is interpreted for each uphole, and this is used to build a deterministic model of the near surface from which static corrections at any required location can be interpolated.
This project uses geostatistics to analyse the spatial variability of uphole data, as an interpolation method, and to investigate the density of upholes required to obtain satisfactory static corrections. The analyses were performed on "depth slices", which are data sets of uphole time for selected fixed depths. The use of depth slices avoids the need for a detailed interpretation of every uphole.
The 16km ´ 12km study area analysed is covered by the Haselgrove 3D seismic survey in the Penola Trough of the Otway Basin. The surface geology consists of lacustrine deposits and shoreline barrier grainstones. One hundred and twenty-nine upholes for the 3D survey and 92 upholes for earlier 2D seismic lines were available for this area.
Semivariograms were computed for depth slices at 2 m intervals. The 3D survey upholes showed anisotropy with the longest range in the direction of 1350. The nugget effect proved to be a significant fraction of the sill, indicating an important degree of randomness in the data. At least some of this randomness may be due to timing errors and vertical interpolation of missing uphole levels. Conditional simulation of the 22 m depth slice provided a visual indication of the high level of randomness implied by the interpreted semivariogram model.
Kriging of the depth slices using the 3D survey upholes revealed a region of large time values (i.e. low velocities) in the SE corner of the study area corresponding to the outcropping of the barrier shoreline deposits. Kriging does not work well across the sharp boundary between the barrier and lacustrine units. Cross validation of the upholes produced a cluster of large errors, from -3 ms to 3 ms, in this region while rest of the study area had a random distribution of cross validation errors. This is due to the smoothing effect of kriging, and to the likely non-stationarity of the semivariogram across the boundary. The kriged depth slices also displayed horizontally restricted anomalies with amplitudes of about 1 msec centered on many of the upholes. These are believed to be due to the relatively large nugget effect in the semivariograms. Because of the randomness in the depth slices, kriging probably does not provide significant improvement over other interpolation techniques such as linear interpolation between the upholes for this data.
A series of uphole programs were simulated, in which the number of upholes was successively diminished by 10% of the original number. The deleted upholes were evenly scattered over the study area. Static corrections were computed for each program, starting with recomputation of the semivariograms. Using the static corrections provided by Origin Energy as the "ground truth", it appears that the number of upholes shot for the Haselgrove 3D survey could have been reduced by up to 40% (58 upholes) with little effect on the accuracy of the resulting static corrections. This conclusion may be a useful guide for reducing the cost of future 3D surveys in the area.
The methodology used in this project is not area-specific, and it may prove useful in other areas where a dense uphole database is available, such as the Cooper Basin.