Reassessing the MADE direct-push hydraulic conductivity data using an improved calibration procedure

Abstract:

In earlier work, we presented and analyzed high-resolution hydraulic conductivity (K) profiles derived from direct-push profiling at the MADE site. The K profiles represent transformations of direct-push injection log (DPIL) profiles, acquired at a vertical sample interval of 1.5 cm, with co-located direct-push permeameter (DPP) tests at selected depths providing the calibration data. The calibration procedure used in our earlier work assumed a linear relationship between lnK and ln(Q/P), where Q/P represents the ratio of the DPIL injection rate and the injection-induced pressure response, with the coefficients of that linear relationship being adjusted to optimize the match between the observed DPP test pressure responses and those simulated using the calibrated K profile. However, although we expect the true ln(Q/P) – lnK relationship to be smooth and monotonic, it is not necessarily linear. In fact, we know that the assumption of linearity has at least one significant flaw. The DPIL injection rates are not high enough to induce measureable pressure responses in the highest K zones, leading to an upper limit on accurately measureable Q/P values. Thus, the true relationship should show ln(Q/P) values leveling out to a roughly constant value above a certain lnK threshold. We are reanalyzing these data using an improved calibration technique that incorporates a smoothing spline representation of the ln(Q/P) versus lnK relationship. This spline function is iteratively adjusted during the course of the inversion process, as the unknown lnK values are adjusted to produce a good match to the observed DPP test pressure responses, allowing for the development of a calibration relationship without requiring specification of its exact form prior to performing the inversion. The resulting ln(Q/P) versus lnK relationship reflects our physical expectations and the calibrated profiles appear to more accurately reflect the high-K zones. We will present a statistical summary of the revised DP K data, comparing it with the results based on the linear calibration procedure and those based on the earlier flowmeter data obtained at the site.