Hydrological simulations can be improved by improving the quality of precipitation data. In this study, three precipitation data products are derived by fusing North-American Land Data Assimilation System version 2 (NLDAS-2) precipitation data and NEXRAD MPE precipitation data using a multiscale precipitation data fusion algorithm. Improvements made by the derived precipitation data are assessed using the Noah land surface model (LSM) in terms of the simulations of streamflow, soil moisture contents and evapotranspiration over 14 watersheds in the Ohio River Basin for a period from 2003 to 2005. To have a comparable and consistent assessment among the different precipitation data associated with different spatial scales, a multiple objective optimization scheme is developed to calibrate the Noah LSM parameters with multiple precipitation inputs. Results of this study show that precipitation data fusion is a statistically effective approach to improve not only the streamflow simulations but also the spatial patterns of soil moisture and evapotranspiration. Among the 14 watersheds investigated, eight watersheds show essential improvements in their streamflow simulations, and only two watersheds show somewhat deteriorations. Our results also show that even for watersheds whose improvements on the streamflow simulations are modest, but the influences of the fused precipitation on the spatial patterns of the simulated soil moisture contents and evapotranspiration are significant when compared with those simulated with either NLDAS-2 or NEXRAD MPE precipitation data alone.