The problem of capricious log response is one that has persistently troubled formation evaluation experts since the Schlumberger brothers ran their first log in Pechelbronn, France. Since the advent of 6FF40 induction logs in the 1950’s, subtle differences have been noted between laterolog and induction response. As field resistivity measurements have evolved to array induction and array laterolog tools, resultant resistivity variability has increased. (Gianzero, 1999) This paper examines how the resistivity discrepancies between laterolog and induction response in an electrically anisotropic rock can greatly affect calculated water saturations (Sw), and ultimately oil in place. Further, several possible solutions are posited to resolve the riddle of resistivity.
The root cause of the differences between the two measurement techniques is in how each tool measures the vertical resistivity (Rv) and horizontal resistivity (Rh) in addition to dielectric effects. In isotropic formations, the difference between Rv and Rh is miniscule. However most organic shales and many laminated low porosity formations are anisotropic. (Klein et al., 1997) In anisotropic formations, the ratio of Rv/Rh is not constant over the possible range of resistivities. This ratio has been observed to be as high as 5 at less than 1 ohmm of Rh, and approaches unity at infinite resistivity.
Due to the high Rv/Rh ratio, at low resistivities, differences between laterolog response (Rh + fraction of Rv) and induction response (Rh) has a dramatic impact on resultant water saturation values. Laterolog array measurements exhibit a systematically higher resistivity than array induction measurements in the same formation. Variances in Sw as high as 30% has been observed. Since most North American unconventional fields have a mix of historical laterolog and induction data from different eras, it is imperative to address this apparent contradiction in values.
Further confounding the issue, the mud salinity required to run both tools at peak performance is nearly mutually exclusive. This complicates efforts to resolve the conundrum because the tools cannot be run simultaneously. The closest measurements on the same rock come from sidetracked wells where one has a laterolog and the other induction. The next best possible measurement is the tri-axial resistivity which can be used to model the Rv and Rh. The issue with tri-axial tools, is that the laterolog apparent resistivity does not conform to either end member of the Rv or Rh.
Since the detailed field measurements have been lost to time and only the measured resistivities are available in public LAS data sets, several practical solutions have been devised by the authors to untangle this mess. First, sets of proximal wells (<1000 ft apart) with either tool were depth-shifted and oriented for analysis. Wells with tri-axial resistivity modeled Rv and Rh supplemented the data set. Once the data was collected, the authors utilized simple x-y regression, multilinear regression, artificial neural net, and random forest regression to predict true Rh. The results of each predictor algorithm is discussed, as the optimal solution is situationally dependent.