Numerical Zonation of Log Suites and Logfacies Recognition by Multivariate Clustering

Abstract

ABSTRACT Zonation of well logs and the correlation of zones between wells are routine fundamental tasks in subsurface geological analysis. Objective numerical multivariate clustering is an effective procedure to simultaneously comprehend a suite of well logs and partition it into geologically meaningful zones and for identifying zone-specific logfacies. Discrimination of zones and logfacies from log suites was established by a hierarchical clustering algorithm that defines clusters so that their within-cluster dispersion is minimal. Logfacies are defined by an unconstrained analysis of unzoned data, whereas the special requirements of the stratigraphic context of the zonation are observed by an adjacency constraint which prohibits merging individual depth levels or lower order clusters if their members are not vertically contiguous. Two test cases, one involving a succession composed of several alternating lithologies and the other entailing a uniform carbonate sequence with only subtle changes, demonstrate the applicability and efficacy of the method. In both cases, the numerical zonation matched the manual lithostratigraphic division extremely well. The geological validity of the numerically defined zones was also borne out by density-neutron crossplots. The unconstrained analysis successfully identified logfacies which are zone-specific, indicating that logfacies can serve as important leads for the recognition of reverse faults in boreholes and for interwell correlations.