Aims and Background: As part of a project on optimization of vineyard irrigation, we discuss the validity of two modeling approaches for soil-water transfers in terms of accuracy and parameterization cost. We compared an empirical ‘bucket-like’ model to an innovative 1D model (M2), based on a dual porosity nonrigid aggregated representation of the soil-water medium. Methods: The two models are coupled to the same canopy growth and radiation absorption models. Root water uptake is described as a bilinear change in relative transpiration rate as a function of the fraction of extractible soil water. The accuracy of the two models was evaluated in vineyards varying in soil type, rooting depth and irrigation procedure in the Languedoc region (France), by comparing soil moisture and predawn leaf potential, a reference indicator of soil water deficit. Results: A sensitivity analysis shows that, provided an accurate estimate of rooting depth, both approaches give relevant and similar results. M2 is robust to most of the hard-to-obtain parameters as root repartition versus depth. However, the two models calculated very different water losses by deep drainage out of the root zone. Conclusion: Considering the robustness of M2 to parameters uncertainty, we privilege it as a decision-support tool for irrigation, and as a ground for further modeling developments.