Annual flow duration curves (AFDCs) are used increasingly because unlike traditional period of record flow duration curves (FDCs), they provide confidence intervals for the median AFDC, they enable one to assign return periods to individual AFDCs, and they offer opportunities for developing a generalized stochastic model of daily streamflow. Previous stochastic models of FDCs and AFDCs were unable to reproduce the variance of AFDCs. We introduce an index flow approach to modeling the relationship between an FDC and AFDCs of daily streamflow series, which is able to reproduce the FDC, as well as the mean, median, and variance of the AFDCs without resorting to assumptions regarding the seasonal or persistence structure of daily streamflow series. Our approach offers additional opportunities for the regionalization of flow duration curves and for the generation of time series of daily streamflows at ungauged sites. Our approach is tested on three river basins in eastern central Italy.