Parameterizing Haverkamp Model From the Steady-State of Numerically Generated Infiltration: Influence of Algorithms for Steady-State Selection

Abstract

BEST (Beerkan Estimation of Soil Transfer parameters) methods of soil hydraulic characterisation are widely applied for estimating sorptivity, S, and saturated hydraulic conductivity, K_s. Calculating these properties requires choosing the β and γ parameters of the Haverkamp infiltration model. These parameters can be obtained from numerically simulated three-dimensional (3D) infiltration runs reaching steady-state. This investigation tested dependence of the estimated β and γ parameters on the algorithm for steady-state selection using simulated 3D cumulative infiltrations for different soils and initial conditions. Two algorithms used the original simulation outputs and included using (i) a threshold defining steadiness (T-algorithm) and (ii) the last four data points, yielding a reference value of steady-state infiltration rate (R-algorithm). A third algorithm, similar to the R-algorithm, was applied to previously re-sampled infiltration data at fixed time intervals (RR-algorithm). The intercept, b_s, of the straight line fitted to the data describing steady-state on the cumulative infiltration plot depended on the applied algorithm more than the slope of this line. Consequently, β varied with the applied algorithm more than γ. The RR-algorithm, yielding 0.62 ≤ β ≤ 1.99 and 0.74 ≤ γ ≤ 0.98, was preferred since it mediated between advantages and disadvantages of T- and R-algorithms. The influence of the choice of proper values for β and γ on the estimates of S and K_s was evaluated using BEST. Using the default values of β (0.6) and γ (0.75) yielded accurate estimates of S but not of K_s. Soil dependent β and γ values should be used in this case. A check of the reliability of the estimates of b_s can be made by a sequential analysis of the cumulative infiltration data. Future developments include considering sources differing in size and establishing if the suggested β and γ values apply in general to the available BEST algorithms.