Semisupervised fuzzy clustering for facies analysis using extended elastic impedance seismic attributes

Abstract

A novel method in seismic facies analysis is proposed to resolve the discrepancy between well and seismic facies concepts. Extended elastic impedance (EEI) attributes are ideally suited for facies analysis because they are representative of different elastic parameters of rocks. A semisupervised method aims to accept or reject seismic clustering based on the well facies interpretation. First, specific EEI logs/attributes are calculated after a feasibility study and EEI analysis of the well data set are used for facies analysis. Using EEI logs as input attributes prevents data deficiency as a result of upscaling of well-log data to the seismic scale in the learning process. Then, the seismic EEI attributes are calculated from prestack seismic data and used for seismic facies analysis considering a fuzzy clustering algorithm with parameters estimated from the well facies analysis stage. The fuzzy clustering methods using membership degrees in their algorithms are valuable tools to reduce uncertainties. To evaluate the method’s performance, 3D seismic data of an oil sand reservoir were selected as a case study. The comparison with previous crossplot methods in facies modeling confirmed the advantages of the presented results in seismic facies analysis. The seismic facies sections are sufficiently interpretable according to the geologic setting and correlate more with the well facies codes.