Développement et évaluation d'une solution numérique efficace de l'équation de Richards incluant l'extraction racinaire des plantes

[Departement_IRSTEA]RE [TR1_IRSTEA]RIE / PHYLEAU [TR2_IRSTEA]RIE / TRANSFEAUInternational audienceFrom the non iterative numerical method proposed by Ross (2003) for solving the 1D Richards' equation, an unsaturated zone module for large scale hydrological model is developed by the inclusion of a root extraction module and a formulation of interception. Two root water uptake modules, first proposed by Lai and Katul (2000) and Li et al. (2001), were included as the sink term in the Richards' equation. They express root extraction as a linear function of potential transpiration and take into account water stress and a compensation mechanism allowing water to be extracted in wetter layers. The vadose zone module is tested in a systematic way with synthetic data sets covering a wide range of soil characteristics, climate forcing, and vegetation cover. A detailed SVAT model providing an accurate solution of the coupled heat and water transfer in the soil and the surface energy balance is used as a reference. The accuracy of the numerical solution using only the SVAT soil module, and the loss of accuracy when using a potential evapotranspiration instead of solving the energy budget are both investigated. The vadose zone module is very accurate with errors of less than a few percent for cumulative transpiration. Soil evaporation is less accurately simulated as it leads to a systematic underestimation of soil evaporation amounts. The Lai and Katul (2000) module is not adapted for sandy soils, due to a weakness in the compensation term formulation. When using a potential evapotranspiration instead of the surface energy balance, we evidenced a difference in partitioning the energy between the soil and the vegetation. A Beer-Lambert law is not able to take into account the complex interactions at the soil-vegetation-atmopshere interface. However, under field conditions, the accuracy of the vadose zone module is satisfactory provided that a correct crop coefficient could be defined. As a conclusion the numerical method proposed by Ross (2003) coupled with the Li et al. (2001) root extraction module provides an efficient and accurate solution for inclusion as a physically-based infiltration-evapotranspiration module into larger scale watershed models