Precipitation is the most important driving factor for rainfall-runoff models. Typically, only point measurements made with rain gauges are available. Interpolation is required to estimate precipitation in other parts of the catchment. The current version of SWAT uses a nearest neighbourhood approach. However, many other interpolation methods are available (inverse distance weighting, inverse square distance weighting, kriging with a time-stable variogram, kriging with daily variograms). The first objective of this work was to compare the interpolation accuracy of the 5 abovementioned interpolation methods through cross-validation. Daily precipitation time series from a network of eighteen stations in the Lahn-Dill region (Germany) were used in this study. The cross-validation exercise showed that all interpolation schemes had a similar accuracy, except for the nearest neighbourhood interpolation that had a lower accuracy. The second objective was to analyse the impact of different interpolation schemes on the quality of automatically calibrated SWAT simulations both with respect to different components of the water cycle and the optimised model parameters. The 5 interpolation methods were used to generate precipitation fields for the Aar catchment, a 134 km2 catchment within the Lahn-Dill region. The interpolation methods resulted in different water balances with the strongest relative increase in the deep aquifer recharge. Against our expectations, the automatic calibration was only partly able to compensate for the differences in the precipitation fields, which was attributed to the dominance of the lateral flow component in the Aar catchment.