Calculation of water budgets is critical to developing long-term management and remediation plans for legacy mine waste facilities. The effects of several modeling decisions commonly made by practitioners constructing numerical hydrologic models can significantly alter estimates of percolation and resultant management strategies. We investigated the effects of variations in several common modeling decisions for an unlined legacy tailings emplacement in Lincoln County, Nevada using predictions from HYDRUS 1D and 2D simulations. Modeling decisions investigated include choice of hydraulic property model, hydraulic properties, numerical mesh discretization, boundary condition representation, dimensionality, and geometry. Hydrologic parameterization was informed by measurement of field-saturated hydraulic conductivity, in-situ density, field moisture content, bulk (dry) density, and water retention curve by hanging column, pressure plate, and chilled mirror hygrometer. Numerical hydrologic models resulted in percolation estimates from 0 to 24 mm/year through the tailings. Choice of hydraulic properties and temporal resolution of boundary condition data contributed the most to variations in percolation estimates. For comparison, percolation was predicted with a simple water balance model (HELP), resulting in a percolation rate of 51 mm/year. Such a wide range of percolation rate predictions demonstrates the potential for significant variability in model output within a range of defensible modeling decisions, assumptions, and parameterization methods. To appropriately bound this uncertainty, several model assumptions and iterations should be considered.