APPLICATION OF SEISMIC REFLECTION DATA TO DISCRIMINATE SUBSURFACE LITHOSTRATIGRAPHY

Abstract

The ever increasing demand for energy has necessitat ed the exploration of hydrocarbons in stratigraphic traps. The seismic technique can be effectively used to elucidate subsurface stratigraphy and lithology. tk interpret seismic responses of geologic sections in terms of subsurface strati graphic and lithologic information it is necessary to establish a correlation between lithology and suitable para meters abstracted from the seismic response. The present work deals with (a) simulating mathematical models for sedimentation processes and calculating their response with the above objective and (b) applying the above concept and methodology develop ed on synthetic data to r~al seismograms to infer lithostratigraphic information. Depositional situations may be modeled by using Markov chains. These involve the concept of memory where the nature of successor lithologies are predetermined by preceding lithologies according to certain probabilities. Markov chains with one step memory are therefore applied to model two different depositional conditions of a formation -XVin a sedimentary basin in India4". Accordingly, two Areas X and Y are considered for the purpose of this study. Area X corresponds to a dominantly sandy (sand = 53 percent) part of the basin together with coal (26 percent) and shale (2l percent) constituents. Geophysical well logs of this area have been used tu calculate the probability of upward transition from one lithology to another at a four metre sampling interval for a particular formation. These were used to generate 255 different synthetic stratigraphic sequences which are collectively designated as Model E. Area Y corresponds to a dominantly shaly (shale = 60 percent) part of the same basin, with sand (37 percent) and coal (3 percent) constituents. Another 253 synthetic sequences generated for this area and designated as Model F were syn thesized on the basis of the probabilities of transitions from one lithology to another as calculated from well log data- The five hundred and eight sedimentation sequences thus generated represent sedimentary sequences deposited in changing environments- Seismic response in time and frequency domain for these models have been calculated. Part of the work embodied in the thesis is based jn real field data, courtsey, Oil and Natural Gas Commission, Dehradun, India. The locations and name of basin have been suppressed. The sedimentary basin is referred to as Basin 2, two areas within the basin as Areas X and Y and the hydrocarbon bearing formation as Formation K. -XVIThe models used in this study are composed of homogeneous, isotropic and perfectly elastic layers. The acoustic impedances of these layers were calculated from the velocity and density logs available for the area. The impulse response was calculated and then convolved with a source wavelet to yield conventional looking seismograms- The autocorrelation function, and the power spectrum using maximum entropy methods were computed. Seventeen variables were picked from the auto correlation function (AGF) and are A,/*q, a2/A0' A^/Aq, where A denotes the ACF at the subscripted lag, \in/A0» where A . denotes the minimum value of the ACF, T , T9, T,, mm 1 « j where T denotes time of the subscripted zero crossing in the ACF and T . , the time at which first minima occurs. Nine amm' variables were picked from the power spectrum and are, average power weighted frequency, frequency at which maximum power occurs, frequency at 25th, 50th and 75th percentile values of frequency weighted power, frequencies of 25 , 50 th and 75 percentile of power, and frequency at which logarithm of power decreases to zero. The above mentioned seventeen variables were calcula ted for all the simulated responses of synthetic stratigraphic sequences. Discriminant analysis which was employed showed that a combination of all the variables can maximally seperate, in the variable space, the two different Models E and F. The discriminating seismic attributes characterize the two sedimentation sequences and may aid the inter pretation of field records in terms of subsurface strati graphy. The success achieved in discriminating different depositional situations in computer simulations has led to the test of the method with real seismic data. The forma tion on which the transition matrices were based for simulat ing Models E and F was marked on the seismic sections of Areas X and Y and the 387 seismic traces when subjected to the discriminant analysis allowed to distinguish between litho stratigraphic units of Areas X and Y, thereby endorsing the validity of this approach. Contributions of the seven teen variables towards effective discrimination shows that only seven variables, viz., fQ, A2/AQ, f^ T^^, fM, f^ and A]_/Ao> common to both synthetic and field seismic data, make positive contributions- The variables designated as seismic discriminators of subsurface litho stratigraphy may ultimately help discriminate an oil bearing stratigraphic trap from its barren surroundings in a sedimentary basin. The statistical method presented here has been shown to be a potential tool for the determination of sub surface litho stratigraphy from seismic data. This consti tutes on important additional tool in the exploration for hydrocarbons.