Annual meteorological data are routinely used in models of contaminant migration through the vadose zone. In arid and semi-arid regions, models based on such data yield negligibly small net infiltration rates that are insufficient to cause groundwater contamination. We conduct a series of flow and transport simulations, in which daily data from a weather station serve as input, to demonstrate that precipitation patterns typical of (semi-)arid regions make the reliance on annual data questionable. We demonstrate that the accuracy of temporally averaged predictions is influenced by the degree of nonlinearity of the Richards equation describing flow in partially saturated porous media. Additional errors are introduced when one ignores topographical and/or urban features that tend to focus and increase local infiltration rates.