Infiltration Prediction Based on In-Situ Measurments of Soil-Water Properties

Abstract

In watershed simulation we need a physically based infiltration equation which can accommodate variation in antecedent soil water content and also spatial variability of infiltration-related physical properties. The method of determining equation parameters should be appropriate for routine field use, and the parameters should be sufficiently sensitive to represent significant variations in infiltration associated with soil differences in a watershed. Three different infiltration equations were employed to predict infiltration in well-drained Typic Torrox soils on the island of Oahu, Hawaii. Simple equations for calculating hydraulic conductivity, K (8), and diffusivity, D (8), were derived, and the parameters in the derived equations were determined from field measurements of steady infiltration and redistribution. Subsequently the parameters Sand A in Philip's 2-term equation were calculated from K (8) and D (8). In order to adequately characterize the hydrologic properties of the surface soil to which the infiltration process is especially sensitive, the calculated sorptivity-antecedent water content relation, S(80), was adjusted by an in-situ measured S, which was obtained by the method of Talsma. For the Green-Ampt equation, the field-saturated hydraulic conductivity, Ks, was measured directly in the field. The wetting front potential, Hf, in the equation was calculated from a derived simple algebraic equation determined from field-saturated porosity and redistribution measurements. Assessment of the Philip and Green-Ampt equations by comparison with measured infiltration at seven experimental sites with 14 infiltration measurements showed that results from the Philip equation had an average percentage error of 17% in predicting cumulative infiltration; the Green-Ampt equation was good only for predicting infiltration in relatively dry soil. The Talsma-Parlange equation, which requires S (80) and Ks but does not require K (8) and D (8), appeared especially promising for routine field use. The spatial variability of the field-measured S was best described by a log-normal distribution as indicated by the KolmogorovSmirnov test. In order to simplify obtaining S (80), a linear relation between Sand 8 was assumed and approximated from the geometric mean of field-measured sorptivity and S = 0 at saturation. This linear approximation was tested on seven soil locations with 26 infiltration measurements using the Talsma-Parlange equation to predict infiltration. The results showed an average percentage error of 23% in predicting cumulative infiltration. While infiltration predictions based on K (8) and D (8) obtained from field redistribution data were more accurate than the simplified Talsma-Parlange prediction, the latter is well adapted for extensive field use in watersheds.