Identification of hydraulically significant fractures from geophysical well logs using cluster analysis

Abstract

Identification of hydraulically significant fractures is critical to groundwater investigations in fractured rock because of the important role fractures play in contaminant transport. In this study, cluster analysis is applied to geophysical well logs to identify hydraulically significant fractures. Cluster analysis is a multivariate technique for the simultaneous analysis of several geophysical well logs over a wide depth range. In this study, 16" normal resistivity, natural gamma, and tube-wave amplitude logs from the Maynardville Limestone on the Oak Ridge Reservation, Tennessee, were examined. The logs show decreases in both 16" normal resistivity and sonic tube-wave amplitude near fractures. The natural gamma log identifies clay-rich zones that exhibit geophysical response similar to fractures and can be used to discriminate these zones from fractures. Second derivatives of the resistivity log and tube-wave amplitude log were calculated prior to cluster analysis in order to enhance anomalous response attributed to fractures and suppress long wavelength responses caused by changes in lithology and water chemistry. Clusters associated with fractures are characterized by a high 16" normal resistivity second derivative, a high second derivative of the tube-wave amplitude, and low natural gamma radiation. The cluster analysis results were calibrated by comparing cluster distribution to the distribution of fractures that had be5h previously identified by core analysis and drilling records in two wells. The calibration was verified by successfully identifying fractures in a third corehole using the characteristics of the fracture clusters from the other wells. The cluster analysis method used in this study is capable of distinguishing wide-aperture fractures or small cavities from other smaller fractures. This method provides a quick and simple way of locating fractures in the vicinity of a borehole, particularly where core or other data are not available, and this technique adds to a growing number of log interpretation techniques for hydrogeologic investigations.