Hydraulic conductivity variation within and between layers of a high floodplain profile

- We investigated the hydraulic conductivity (K) of a floodplain in the Weihe River.
- K values from individual sediment layers showed normal distribution.
- K values from three or more sediment layers became log-normal distribution.
- The probability models of K values depend on the layout of the sampling locations.

SummaryWe investigated the hydraulic conductivity of a high floodplain of the Weihe River, the largest tributary of the Yellow River in China. We took sediment cores from five layers of the sediments in the high floodplain and performed on-site permeameter tests to determine the horizontal hydraulic conductivity (Kh) of these undisturbed cores. Although the sediments within each layer were relatively homogeneous, the Kh values exhibited variation among the testing locations within each layer. The Kh values from any two layers were statistically different except for the Kh values from layers 1 and 4, which were statistically from the same population. We used the Jarque-Bera and Lilliefors tests to determine the normality of these Kh values. The Kh values from individual layers (layers 1-3) showed normal distribution. When the Kh values from the three layers were combined as a single data set, both statistical tests failed to confirm the normality; instead, the two test methods suggested log-normal distribution. When the Kh values from the five layers were combined as one data set, they also showed a log-normal distribution. These results suggest that hydraulic conductivity values from an individual layer were likely in normal distribution, although their mean and standard deviation varied for individual layers. We then generated random Kh values of normal distribution based on the mean and standard deviation of the field measurements in the five layers, as well as using modified means and standard deviations. When these normally distributed Kh values were combined into one data set for normality analysis, they became log-normal distribution. Clearly, the probability models of hydraulic conductivity depended on how the layout of the measurement locations was placed on porous media consisting of multiple layers of sediments. Hydraulic conductivity from a single layer cannot be simply assumed to be log-normal. The log-normal-distribution model is not always appropriate for characterization of a hydraulic conductivity field for a relatively homogeneous porous medium.