Comparison of Post stack Seismic Inversion Methods:A case study from Blackfoot Field, Canada

Seismic Inversion methods have been routinely used for estimating attributes like P-impedance, Simpedance, density, ratio of P-wave and S-wave velocity and elastic impedances from seismic and well log data. These attributes help to understand lithology and fluid contents in the subsurface. The objective of this research is to use several standard seismic post-stack inversion methods for reservoir characterization and compare their results. Model-based (MBI), Colored (CI), Sparse-spike (SSI), and Band-limited (BLI) inversions are applied to the post-stack seismic data from the Blackfoot field, Alberta, Canada. In each case the data is inverted into P-impedance and density volume. The final stacked section shows high-resolution images within the time-depth ranges of 300 to 1300ms. All inversions show mutually consistent results with low-impedances within the target hydrocarbon sand within the channel. All post-stack inversion methods produces accurate and reliable results and unequivocally confirm the presence of reservoir in the channel area at 1060-1065ms time. Model based inversion methods shows higher correlation coefficient (0.99) and least RMS Error (778 m/s*g/cc) and hence better for Blackfoot seismic data. Geostatistical methods-Probabilistic neural network is also employed to estimate the petrophysical (porosity) variations within the sand channel (reservoir) of the Blackfoot field. The Correlation coefficients between the predicted and measured porosities following that the probabilistic neural network show that Sparse spike Inversion when used as an external attribute, is more accurate and produces high-resolution image compared to that estimated with the use of model based and Colored Inversion as an attribute. From the estimates, the predicted logs show correlation of 0.81, 0.84 and 0.86 using neural network algorithm and MBI, CI and LPSSI as external attribute, respectively. Of the four post stack seismic inversion methods used here the probabilistic neural-network method with sparse spike inversion provides a higher correlation coefficient than that estimated for other methods.