Estimating macropore parameters for HYDRUS using a meta-model

Abstract

The vertical change in the number of macropores causes a variation of the relative macroporosity (w(f)) and the effective aggregate width (d(ag)) over the soil profile. Both parameters are used in HYDRUS to represent this variation, increasing the number of parameters and making automated calibration challenging. The working hypothesis is that we can improve an analytical estimation of w(f) and d(ag) developed in previous research by inverse estimation with a meta-model for HYDRUS 2D/3D, using disk infiltrometer data of infiltration at zero pressure head. We generate a meta-model that describes the vertical heterogeneity of the macropore number with a general function using four parameters: the relative macroporosity at the soil surface (w(fs)), the effective macropore radius (r(m)), the maximum depth of macropores (z(max)) and the shape parameter of the w(f) curve (m). The meta-model computes the variation of w(f) and d(ag) over depth, thus reducing the parameters for automated calibration with HYDRUS. We theoretically described how to directly obtain the meta-model parameters with disk infiltrometer data, providing an example for field conditions. A complete parametrization of matrix and macropore parameters for HYDRUS 2D/3D was generated from these data and previous studies, which were updated by automated calibration. Only w(fs) was calibrated, increasing by about similar to 3.5 times the initial measurement. We tied several macropore parameters to w(fs) during calibration by their physical or mathematical relations. This methodology can be utilized to estimate HYDRUS parameters for risk assessment or detailed plot studies. Highlights A meta-model to estimate macropore parameters for HYDRUS 2D/3D is presented. The meta-model reduces the number of macropore parameters for HYDRUS 2D/3D. Initial estimations of meta-model parameters are obtained by disk infiltrometer. The parameters were updated through calibration with HYDRUS 2D/3D. Preferential flow is predicted by a two-dimensional model.