In this dissertation I have developed three techniques to evaluate hydrocarbon saturation in reservoirs: the Support Vector Machine (SVM) method, frequency-attenuation method, and spectrum decomposition method.; The presence of a small amount of gas in a shallow reservoir can drastically reduce the P-wave velocity, similar as high gas saturation effect, so it is difficult to distinguish low-saturation gas from economic gas with P-wave velocity. However, in deep water environment, quantitative study (Han and Batzle, 2002) shows that it is possible to discriminate low-saturation gas and economic gas with combination of different attributes and local calibration.; In Chapters 3 and 4, I have developed new quantitative methods using SVM in classification and regression. I have applied the SVM technique to a gas field and its adjacent prospect in the Gulf of Mexico. The result shows a clear discrimination of low saturation gas and economic gas in deep-water, poorly consolidated sand reservoirs.; In Chapter 5, I have applied frequency-attenuation analysis to seismic data of commercial gas and low-saturation gas reservoirs. The attributes of peak frequency and bandwidth suggest that low-saturation gas has a higher attenuation than economic gas does, similar to the mean frequency observed by Li and Han (2005).; Finally, in Chapter 6, I have applied spectral decomposition method to analyze 4D seismic data in an oil reservoir. Time shift in 4D affects the peak frequency for relatively thick reservoirs; the reflectivity contrast between reservoir top and bottom affects the peak frequency at both thin and thick reservoirs. This result will help to monitor water flooding in an oil reservoir. |