Passive capillary samplers, which sample water from the vadose zone via a hanging water column in a fiberglass wick, have shown potential to provide better estimates of actual soil percolation fluxes than alternative field methods. Unsaturated and saturated flows (water and solutes) are extracted continuously and without external vacuum generator from a non-disturbed soil volume, through a significant area (typically 900cm2). In order to achieve a minimal disturbance of the native flow regime, the wick type (hydraulic conductivity and section), length and number and the contact area have to be adjusted to match as close as possible the expected soil pressure/flow conditions. To estimate the optimal characteristics of the wick, an approach has already been proposed (Knutson and Selker, 1994, Soil Sci Soc. Am. J., 58: 721-729), based on analytical solutions of the Richards equation relying on strong assumptions like an exponential potential-conductivity relationship, one-dimensional and permanent flow, unit gradient soil water profile. Here, the Hydrus 2D code (Simunek et al., 1999, International Ground Water Modeling Center, 227pp) solving the Richards equation for simulating two-dimensional unsaturated flow was used to evaluate the uncertainties in flux estimation by a passive capillary sampler. Two sources of uncertainties were examined. Those associated with the assumptions of the dimensioning of the wick and those associated to experimental uncertainties. The numerical experimentation was conducted on two reference soils submitted to a 11 days actual hyetograph. Results showed that the analytical dimensioning method is relevant. But significant errors on the observed fluxes occur when the soil and wick properties do not exactly match, which is the common case since the range of available characteristics of fiberglass wicks is limited. Also, uncertainties in the hydraulic conductivity properties of the soil, in wick length appear to have an important influence on the representativity of the wick fluxes against the actual soil drainage fluxes. Finally, this work outlines the potential of a 2D numerical simulation approach to evaluate and correct the possible bias of observed drainage fluxes by passive capillary samplers.