| dc.description.abstract | Although considered one of the more productive oil and gas reservoirs in the United States, the Pennsylvanian-age Granite Wash reservoir remain poorly understood. Amongst a myriad of issues that hinder development of hydrocarbon reserves are unusually low porosity and permeability estimates, varying grain sizes, mineralogy, cementation and the presence of micro-fractures. These heterogeneities not only influence the reservoir performance but have also make the targets difficult to image seismically.
To address this later issue, I apply state-of-the-art seismic processing and data conditioning techniques to a 3D seismic volume acquired over the study area. Due to surface conditions overlying this complex play, the seismic data are highly contaminated by coherent and random noise, such as ground-roll, reverberation and air-blast events, resulting in a seismic processing challenge. To improve seismic interpretation, I reprocessed the raw field gathers through coherent noise suppression, prestack Kirchhoff migration, and other sophisticated data conditioning techniques such as spectral-balancing and structured-oriented filtering to improve the quality of the re-processed data.
To understand the reservoir geomorphology and lithological heterogeneity, using seismic geometric, textural attributes and inversion volumes, I construct what I believe to be the first seismic facies analysis of the Desmoinesian-Cherokee wash of Wheeler and Hemphill counties, Texas. An unsupervised latent space Generative Topographic Mapping (GTM) technique provides images of rock-facies types and reservoir quality using facies predictions from a previous well-based study in the same area as ground truth. These facies map provide images of specific alluvial fan depositional environments and reservoir facies from seismic data as well as identifying productive chaotic facies using these attributes. | en_US |
