The heteroscedasticity of watershed hydrological series caused by climate change and high intensity human activities has destroyed the precondition of Box-Jekins ARMA model. In view of this, the main research contents and approaches were proposed based on the domestic and foreign research progress and current major scientific problems. The observed series of finance, economics and signal processing are similar to the heteroscedasticity fluctuation characteristics of hydrological non-stationary series in the changing environment of watershed, and many successful cases have been obtained in these areas. So, it may be feasible to apply the heteroskedasticity principles and methods of these fields to the stochastic modeling of hydrological non-stationary series. Also, it is a practical modeling approach compared with the distributed hydrological models. Some ideas proposed in this paper are expected to shed some light on the accurate description of the changing law of hydrological elements in the changing environment, and to provide a basis for the planning, design and scientific management of water-related projects, as well as drought risk regulation and control.