Hydrological projections are usually based on hydrological models (HMs) driven by climatic simulations from Global Climate Models (GCMs). Therefore, uncertainties of hydrological projections can be propagated from HMs and GCMs. Meanwhile, downscaling techniques that solve the spatial resolution differences between GCMs and HMs can introduce future uncertainties to projections in streamflow. The first part of the study projects future changes in floods and water availability across China based on 40 combinations of eight HMs driven by five GCMs, and evaluates uncertainties associated with HMs and GCMs. The results show that increasing floods but decreasing water availability is detected in south China under Representative Concentration Pathway 8.5 scenario. The insignificant changes in precipitation combined with increased evapotranspiration may reduce water availability. Uncertainties from GCMs/HMs predominate the uncertainties in the wet/dry areas in eastern/northwestern China. Model agreements are higher in simulated floods than water availability because increasing precipitation extremes are more consistent among different GCM outputs compared to mean precipitation. The control experiments in the second part of the study show that the increase in inter-daily variability of precipitation and temperature, which is an important statistical feature of more extreme climate but usually not well considered in statistical downscaling, can lead to considerable impacts on hydrological simulations.