Large eddy simulations coupled with two uncertainty quantification (UQ) methods, latin-hypercube sampling and polynomial chaos expansion, were carried out to quantify the effects of model parameters and spray initial and boundary conditions on spray development. Evaporating, non-reacting n-dodecane and iso-octane experimental spray data under typical diesel and gasoline engine-like conditions was used to compare penetration lengths and probability contours. Five spray initial and boundary conditions were used for single-hole n-dodecane sprays. The Morris one-step-at-a-time method was used to identify parameters with the highest impacts for multi-hole iso-octane sprays. The resulting four most important parameters, including two model parameters and two spray boundary conditions, were chosen for further study. Penetration lengths and corresponding standard deviations derived from both UQ methods were found to be quite similar with experiments. Probability contours were very close to the uncertainty of experiments, though the probability bands were slightly wider in the simulations. Correlations between input variables and responding penetration lengths at selected times were analyzed.