Using multiple seismic attributes improves the accurate estimation of reservoir thickness over using any single seismic attribute.

To use 3-D seismic data effectively for reservoir development, one must relate seismic attributes to one or more reservoir properties. It is common practice to relate a single seismic attribute (for example, seismic amplitude, instantaneous frequency, phase, time or thickness) to the reservoir property most critical to reservoir performance (for example, porosity, porosity feet or sand thickness). One can improve the accuracy of the estimation of a reservoir property using several seismic attributes to estimate changes in the reservoir property over the reservoir rather than one seismic attribute. The key to successfully using multiple seismic attributes to estimate a reservoir property is that each seismic attribute must show a strong correlation to that property.
A reason multiattribute reservoir property estimation has not been fully utilized is because the techniques and algorithms necessary to implement the process typically are not available in a single set of software and therefore do not work together easily. However, a recent project used a single software system for all of the calculations and analysis. The software provides an integrated set of tools for petrophysical, geological and geophysical interpretation of seismic data.
A suggested methodology for the inclusion of seismic attributes starts with construction of a good regional geological model and proceeds to the calculations of reservoir properties at the wellbore locations based on petrophysical analysis of available logs. Seismic horizons then can be interpreted and used as the foundation for the calculation of various attributes. These attributes should be individually checked for significant correlation with reservoir properties. Attributes with good statistical correlations can be combined using a simple multiple linear regression technique to increase overall accuracy. The combined attributes can be mapped to extend the interpretation of reservoir properties beyond the wellbores.
This methodology was used with data from the Sooner field in Weld County, Colo. The field, discovered in 1985, is part of the D Sandstone play in northeast Colorado's Denver Basin. The D Sandstone is a member of the Dakota group and is encased between marine shales, the Garaeros and the Huntsman. At the end of the Huntsman period sea levels began receding, and shallow marine sandstones (D Sandstone) were deposited over the region. As the seas lowered further, a shoreline formed toward the field's northeast corner, and rivers and streams incised through the older D Sandstone into the underlying Huntsman shale. Following this regressive event, a rise in sea level filled the incised valleys with fluvial and estuarine sandstones and siltstones. Most of the producing fields in the D Sandstone are from stratigraphic traps with little to no correlation to structural closures. Prolific wells usually are associated with areas of thicker valley-fill.
A project database developed for the Sooner field included a high-resolution 3-D seismic survey and digital well logs from 27 key wells. These well logs were used to construct a series of cross sections to correlate several key markers such as the Niobrara, X Bentonite, D Sandstone, Huntsman and J Sandstone. Researchers calculated reservoir properties for the D Sandstone interval from a suite of normalized logs. Volume of shale (VSH) curves were computed from the gamma-ray curves using sand and shale baseline parameters as interpreted from histograms. Density curves were normalized between 0 and 1, and water saturation was calculated using a standard Archie model. Net pay was determined using a porosity cut-off of 0.08, VSH of less than 0.35 and water saturation of less than 0.6.
A normal incident synthetic seismogram, constructed using a sonic curve and a zero-phase wavelet, tied the well data to the seismic data. The synthetic seismogram had a strong positive response at the top of the D Sandstone (black peak on the wiggle-trace display) and generally fit well with the seismic data (Figure 1). The synthetic seismogram also showed that the D Sandstone is too thin to produce distinct events marking its top and base on the seismic data. The sonic curve was altered several times to increase and decrease the thicknesses of the D Sandstone. Synthetic seismograms modeled varying thickness of D Sandstone and verified that the amplitude of the seismic peak marking the top of the D Sandstone increases with increasing D Sandstone thickness. Since the D Sandstone is too thin to be time-resolved, the base of the D Sandstone (top of the Huntsman) is best predicted by the zero crossing below the D Sandstone peak.
Four key markets were chosen on the seismic data: the zero crossing above the D Sandstone, the D Sandstone peak, the zero crossing below the peak and the trough below the peak. These horizons were used to construct two isochrons and served as guide windows for various attribute extractions. In addition to amplitude extractions, attributes were extracted for instantaneous frequency, instantaneous phase and instantaneous amplitude (Figure 2). Individual attributes were cross-plotted against the gross reservoir thickness and net pay at each of the 27 wellbore locations. Seismic attributes with good visual and statistical correlation were used in the construction of two composite horizons for gross and net pay sand using multiple linear regression, which predicts a response based on the linear relationship with one or more variables. To be used in the final gross and net sand estimation, a seismic attribute had to have a high correlation individually with net and gross sand (a correlation coefficient greater than 0.65 in the case of gross sand). A physical reason also had to explain why changes in sand thickness would affect the seismic attribute.
Four attributes demonstrated a strong correlation with the reservoir's gross thickness: instantaneous frequency, average amplitude, isochron of the peak-to-trough and the depth-converted horizon of the Huntsman. These attributes, combined using the multiple linear regression technique, yielded an overall correlation coefficient of 0.85 and a goodness of fit of 0.72 (Figure 3).
The technique also was applied to net pay. Instantaneous phase, isochron of the zero crossings, average amplitude and the ratio of the peak-to-trough amplitude ratio attributes had poor to weak correlation with the well data. The linear regression using these attributes produced an overall correlation coefficient of 0.5 and a goodness of fit of 0.25.
Two wells in the field were not used in the multiple linear regression calculation. Instead they were used to check the ability of the technique to measure sand thickness. For one well, the logs showed gross and net sand thickness of 41.9 ft (12.8 m) and 9.6 ft. (2.9 m), respectively, while the multiple seismic attribute technique estimated a gross sand thickness of 39 ft (11.9 m) and a net sand thickness of 3.5 ft (1.1 m) at the well location. In the second well, the logs showed net and gross sand thickness of 42 ft (12.8 m) and 14 ft (4.3 m), respectively, and the multiple seismic attribute technique estimated gross and net thickness of 43 ft (13.1 m) and 19 ft (5.8 m), respectively. At the first test well location the technique estimated gross sand within 7%; at the second test well location the error was just slightly more than 2%.
Conclusions
The accuracy of predicting a reservoir property at each bin in a 3-D seismic survey from seismic attributes can be improved by combiningthe seismic attributes that show a statistically significant correlation to the reservoir property. A critical step in this process is the calculation of reservoir properties at each wellbore location. The seismic attributes used in multiple seismic attribute analysis should be simple to derive and have a good visual and statistical correlation to the well data.
References
Sippel, M.A., Cammon, T.J., "Advanced secondary recovery Project for the Sooner 'D' sand unit Weld County, Colorado," US Department of Energy, DOE/BC/14954-14, 1996.
Wittick, T.R., "Using 3-D seismic data to find new reserves in Quitman Field," The Leading Edge, pp. 450-456, April 1998.

Comments