Vertical hydraulic conductivity and layered heterogeneity: from measurements to models

Abstract

Published laboratory and field estimates of vertical hydraulic conductivity (K_v) and the anisotropy ratio (K_h/K_v, where K_h is horizontal hydraulic conductivity) have been compared with equivalent groundwater model values, mainly for sedimentary strata. The results show that model K_v values tend to be higher and K_h/K_v values tend to be lower than field estimates for vertical length scales ≥10 m, particularly for coarse-grained unlithified sediments. This difference is attributed to the widespread use of the assumption of K_h/K_v = 1 and particularly K_h/K_v = 10 in groundwater models, regardless of the length scale or strata type represented. The origin of the K_h/K_v = 10 assumption is obscure and not founded on rigorous data analysis. K_h/K_v, and by inference K_v, is frequently an unimportant parameter in model construction and calibration. On balance, this model artefact is attributed to the common reliance on summary head calibration statistics that hide the inadequacies of the MODFLOW paradigm (K_xx = K_yy = K_h, K_zz = K_v) when used for large-scale hydrostratigraphic units with uniform parameterization, fixed K_h/K_v, and 1 ≤ K_h/K_v ≤ 10. However, thin, high-permeability aquifer models or well-defined aquifer/aquitard models are examples where such simplifying assumptions for K_h/K_v are workable, given groundwater quantity objectives. More realistic model values of K_v and K_h/K_v at length scales greater than field estimates could be obtained by independent calibration of K_v and K_h, use of larger-scale field estimates of K_v and multi-level piezometer/observation wells, and calibration to vertical head gradients separate from summary head calibration statistics.