Accuracy of top soil moisture simulation using a mechanistic model with limited soil characterization

For many applications, soil water flow models require knowledge of soil properties and conditions that is difficult to obtain. In this study, we analyzed the impact of approximations on model prediction of soil water content when only limited soil property data are available. The TEC model of Chanzy and Bruckler (1993), representing soil heat and water flows, was selected since it is based on soil properties that can be estimated with the pedotransfer function, and it is driven by standard climatic data. The impact of the choices made for soil hydraulic properties, initial soil water conditions, and lower boundary conditions for a wide range of soil were studied. In this study, emphasis was placed on estimating moisture in the surface soil layer (0-30 cm) and was limited to wet and moderately dry soil conditions, which can be critical for applications in agriculture to determine trafficable days and in surface hydrology. The results from this study indicate the importance of soil hydraulic properties and moisture profile initialization, which may lead to errors larger than 0.10 m(3) m(-3) on soil moisture in the surface layer. The pedotransfer function proposed by Wosten (1997) led to the best simulated soil moisture in the top 30 cm, but initialization has a strong impact. Thus a "warming up'' period of 70 days is needed to mitigate the initialization error. This duration can be reduced according to the amount of precipitation. We propose starting simulations 48 hours after the last significant rainfall with an initial water potential of -10 kPa. At the bottom of the soil, a gravitational flux at a depth of at least 80 cm is recommended to compute the boundary conditions. These recommendations for implementing a soil water transfer model were evaluated against independent experimental data, and a moisture accuracy (root mean square error) of 0.04 m(3) m(-3) was obtained.