Abstract: In this paper the kernel density estimation model based on kernel density estimation theories is established for time series of single variable.It belongs to a class of data-driven approach and avoids the form of probability distribution (normal or pⅢ)and the form of dependence (linear or nonlinear).The model has clear concept and singe structure.The model is applied to the stochastic generation of daily discharge time series at single station.The results indicate that the suggested model is suitable for stochastic simulation of hydrology time series.