Seismic facies

Abstract

The method presented herein describes the seismic facies as representations of curves and vertical matrixes of the lithotypes proportions. The seismic facies are greatly interested in capturing the spatial distributions (3D) of regionalized variables, as for example, lithotypes, sedimentary facies groups and/or porosity and/or other properties of the reservoirs and integrate them into the 3D geological modeling (Johann, 1997). Thus when interpreted as curves or vertical matrixes of proportions, seismic facies allow us to build a very important tool for structural analysis of regionalized variables. The matrixes have an important application in geostatistical modeling. In addition, this approach provides results about the depth and scale of the wells profiles, that is, seismic data is integrated to the characterization of reservoirs in depth maps and in high resolution maps. The link between the different necessary technical phases involved in the classification of the segments of seismic traces is described herein in groups of predefined traces of two approaches: i) not-supervised and ii) supervised by the geological knowledge available on the studied reservoir. The multivariate statistical methods used to obtain the maps of the seismic facies units are interesting tools to be used to provide a lithostratigraphic and petrophysical understanding of a petroleum reservoir. In the case studied these seismic facies units are interpreted as representative of the depositional system as a part of the Namorado Turbiditic System, Namorado Field, Campos Basin. Within the scope of PRAVAP 19 (Programa Estratégico de Recuperação Avançada de Petróleo – Strategic Program of Advanced Petroleum Recovery) some research work on algorithms is  underway to select new optimized attributes to apply seismic facies. One example is the extraction of attributes based on the wavelet transformation and on the time-frequency analysis methodology. PRAVAP is also carrying out research work on an optimized application of Kohonen type neural networks developed by Matos et al. 2003.