Seismic Rock Physics Analysis of Organic-rich Shales Using Statistical and Machine Learning Methods

Shale is the most common sedimentary rock, which accounts for approximately 70 percent of the rocks in the crust of the Earth. In a conventional petroleum system, organic-rich shales have been regarded as source rocks generating hydrocarbon resources. Over the past decade, these shale rocks have become both the source rocks and unconventional reservoirs because of the combination of horizontal drilling and hydraulic fracturing. Although drilling and development technologies have been advancing rapidly, there is a lack of exploration involved with many shale resource operations. The current ’drilling on a grid’ strategy causes inefficient hydrocarbon production and unnecessary expense because many shale reservoirs are more heterogeneous and complex than conventional sandstone reservoirs. To identify shale reservoir production ’sweet spots’ from seismic data to help optimize recovery, we need to develop more e↵ective seismic reservoir characterization methods. Also, we have to reduce the impact of greenhouse gas emissions and associated climate change while developing unconventional oil and gas. For this reason, we need a better rock physics model (RPM) to describe shale reservoirs and understand their heterogeneity or anisotropy. However, in contrast to conventional sandstone reservoirs, many important shale reservoir rock physics properties are not well understood. Moreover, the connections between seismic, elastic, and shale rock properties are complex and need more research to improve quantitative inversion and interpretation of seismic data. There are two main geophysical problems for better shale characterization: (1) building a proper seismic RPM of organic-rich shales, and (2) estimating accurate shale-specific reservoir properties from seismic data. To help integrate these multi-disciplinary concepts, this dissertation proposes three topics as part of my PhD research: (1) Improved shale RPM methods to estimate the total organic carbon (TOC) and mineralogical brittleness index (MBI) using statistical and machine learning methods, (2) Seismic amplitude variation with o↵set (AVO) forward modeling and AVO attribute analysis to model an accurate synthetic seismic data and predict important shale properties such as TOC and clay volume, and (3) Enhanced seismic anisotropy characterization methods to estimate shale reservoir properties from anisotropic seismic data. Therefore, successful estimation of shale rock properties from seismic properties may allow us to better characterize the organic-rich shale rocks. Also, these approaches can interpret the e↵ect of variations in porosity, mineralogy, and organic carbon concentration on seismic property changes. In conclusion, these methods potentially improve the shale reservoir characterization and time- lapse monitoring for shale hydrocarbon production. They can also help to provide seismic rock physics frameworks for accurate interpretation and monitoring of the physical property changes, for example, during CO 2 enhanced oil recovery (EOR), carbon capture and storage (CCS), and geothermal energy production.