Chapter 3 - Fractal Inversion

Abstract

Inversion of seismic data plays a vital role in reservoir characterization. High-resolution inversion methods provide models with higher resolution than those built from the conventional stacked seismic traces (e.g., Rowbotham et al., 2003; Sen, 2006). Such models are good for reservoir characterization and model building. Edited well logs provide the most accurate and the best vertical resolution of geophysical information of a subsurface reservoir. Though vertical resolution of well logs is very good, horizontal resolution is poor because of sparse availability and the small extent of lateral depths of investigation. One approach to obtain a high-resolution 3D description of a reservoir is to use geostatistical interpolation, extrapolation, or simulation (Hass and Dubrule, 1994) using available well-log data at a few sparse locations. Accuracy of such models is dictated by the number and spatial distribution of the wells over a reservoir. Therefore, seismic data being the most continuous information available (although at a lower vertical resolution), a stochastic inversion of seismic data that integrates seismic data and well-log data can add great value in reservoir characterization (Francis, 2006a) as it combines seismic data with well logs and uses vertical resolution from well logs and good horizontal resolution from seismic data into the estimated model. The fusion of seismic and well-log data is possible using seismic inversion, which converts seismic data information into petrophysical properties such as acoustic impedance and shear impedance (e.g., Dimri, 1992; Russell and Hampson, 2006; Srivastava and Sen, 2010; Vedanti and Sen, 2009). Merging seismic data directly with well-log data is difficult because they have a different range of scale/frequency of measurement compared to well logs and also their recording is basically in a different domain, viz., seismic data in the time domain, whereas logs are recorded with depth.