Assessing the dispersion of hydraulic conductivity values estimated using the Dupuit, Thiem, and Boulton methods on repeated pumping tests in an unconfined aquifer

Hydraulic conductivity (K) is rarely directly measured but can be estimated commonly through the Dupuit, Thiem, and Boulton methods. But which method can give more reproducible values? In this study, dispersion is assessed by examining the range of tolerance needed around the average value to get half of the values of a dataset available for a given borehole. A total of 403 values from 14 boreholes situated at a 1,150-m(2) test site in southwestern France were acquired from 2016 to 2020 using pumping tests of 2 h or 5 days. Estimated values of K ranged from 5.0 x 10(-6) to 9.6 x 10(-5) m/s. The Thiem method, usually providing a value between two boreholes, was modified to give an estimation for one borehole. This method presented the smallest dispersion: half of the values are within the interval of +11% of the average value, followed by the Boulton (+12%) and then Dupuit (+23%) methods. The Thiem and Dupuit methods can be used after a 2 h pumping in pseudo-steady-state condition; in contrast, the Boulton method requires long-term pumping. The Thiem method appeared dependent on the borehole quality, especially when considering the effect of clogging during short-term pumping. Finally, using the Dupuit method, average estimated K values from 2 h or 5-day pumping tests are close (3.1 x 10(-5) and 2.0 x 10(-5) m/s, respectively). The Dupuit method was a good compromise between variability of the estimation and sensibility to site conditions. Also, measurements made with automated probes presented significant variability meaning that human error is not the sole factor of variability.