An overview of the annual hydroclimatology of the United States is provided. Time series of monthly streamflow, temperature, and precipitation are developed for 1337 watersheds in the United States. This unique data set is then used to evaluate several approaches for estimating the long-term water balance and the interannual variability of streamflow. Traditional relationships which predict either actual evapotranspiration or the interannual variability of streamflow from an aridity index $phi = {rm overline {PE}}/{bar P}$ are shown to perform poorly for basins with low soil moisture storage capacity. A water balance model is used to formulate new relationships for predicting actual evapotranspiration and the interannual variability of streamflow. These relationships depend on both the aridity index $phi = {rm overline {PE}}/{bar P}$ and a new soil moisture storage index. A physically based approach for estimating the soil moisture storage index is introduced which requires monthly time series of precipitation, potential evapotranspiration, and an estimate of maximum soil moisture holding capacity. The net results are improved expressions for the long-term water balance and the interannual variability of streamflow which do not require either calibration or streamflow data.