Global water and energy cycles are two tightly related and critical components of the Earth climate system. In current assimilation systems, model output is only locally constrained by available observations, which are always uncompleted and contaminated by random error and biases. Thus the global water and energy balances are good yet high standard indicators of the performance of an assimilation system. In this study, we use several global observed datasets to evaluate the assimilation results from the validation experiments for the Modern Era Retrospective-analysis for Research and Applications (MERRA) based on NASA GEOS-5 data assimilation system. The observed datasets include CERES ERBE-like TOA radiation fluxes, GPCP precipitation, CMAP precipitation etc. Furthermore, based on comparison with other assimilation systems, deeper understanding about the effect of the interaction between model and analyzed data on the water and energy cycles will be given.