Errors in the estimation of soil water properties and their propagation through a hydrological model

Abstract. The Agricultural Catchments Research Unit model (ACRU) includes a decision support system (DSS) for estimating the water content of soil at field capacity (θ _fc ) and wilting point (θ _wp ) when these characteristics are not directly measurable. Three methods of estimation are proposed: (a) based on silt and clay content and bulk density, (b) based on clay content only, and (c) based on soil series. These three pedotransfer functions are compared with respect to both the estimation of θ _fc and θ _wp and the propagation of errors when the actual evapotranspiration of a wheat crop (E) is predicted over the growing season by the ACRU model.
The standard error of estimation was between 0.066 and 0.082 m³/m³ for θ _fc , between 0.056 and 0.069 m³/m³ for θ _wp and between 29.9 and 34.8 mm of water for E. The method based on silt and clay contents and bulk density predicted θ _fc and θ _wp for non-swelling soils most precisely. The method based on soil series was better than other methods for swelling soils. It also performed better for estimating available water capacity and consequently for predicting E from a conceptual soil water model. The propagated error of estimating θ _fc and θ _wp using the DSS reached 15–18% of the simulated E. The error in the prediction of E can reach 26–30% when spatial variation in soil properties is also estimated.