Accounting for Bias and Boundary Condition Effects on Measurements of Saturated Core Hydraulic Conductivity

Abstract

Most hydrologic studies require knowledge of saturated soil hydraulic conductivity, Ks This parameter is often measured using saturated soil cores and a constant applied hydraulic head device. A standard approach to reduce uncertainty in the result is to conduct replicate tests at a single hydraulic head gradient. Low‐permeability soils are often tested at large hydraulic head gradients to decrease measurement time. Our objective was to test the common assumptions implicit in calculating Ks from constant‐head laboratory tests, i.e., the theoretical linear relationship between head gradient and flux density exists in experimental data, and the relationship passes through the origin. In this study, we used linear regression analysis to test these assumptions and determined Ks from a broad range of head gradients for: 4.5‐cm (diameter) by 10‐cm (height) intact cores of sandy loam soil; 10‐cm (diameter) by 10‐cm (height) intact cores of clay loam soil; and repacked sand columns of various sizes. We found nonlinear relationships between hydraulic head gradient (i) and flux density (q) for tests conducted on intact cores of both soils, especially for head gradients greater than unity. When we calculated Ks by linear regression of data from intact cores, we found average values approximately one‐third greater than the “standard” method of averaging several replicate tests at a single hydraulic head. The difference between the regression and standard analyses was attributed to experimental bias, which is removed by the linear regression. Although no consistent i or q “thresholds” were identified to predict the onset of nonlinearity in i vs. q data, the intact core results imply that i < 1 and q < 5 × 10−3 cm s−1 may be advisable.