There are many uncertain factors in the multi-source water irrigation system, along with the changes in the system environment and the effects of uncertainty, leading to dynamic characteristics of the water distribution process.Based on the water distribution characteristics of irrigation system, interval-parameter multi-stage stochastic programming model was constructed and improved to consider effect of sensitive index of various stages and water irrigation on crop production, which introduced water sensitivity index weights and made a case study over rice at different growth stages in Heping irrigation area.The study area of this paper had two water sources: the surface water and the ground water, which also could be divided into three different projects by the water intake mode: water diversion project, water lifting project and well irrigation project.At the same time, Liuhe reservoir was taken as a water external source, where all the water system constituted a complex multi-water source supply system.In this study, four stages of rice growth were selected as the research period, i.e.tillering stage, jointing stage, heading stage and milk stage.The water sensitive index weight coefficients in each growth stage were 0.37, 0.46, 0.11 and 0.06 respectively.Inflow level of different growing stage was random variables and closely related to hydrological factors such as rainfall and runoff, hence the probability density function was introduced to represent uncertainty, and discrete interval was used to show other hydrologic and economic uncertainty.Multi-stage stochastic programming model could allocate water between different phases and different growing stages under a series of inflow level.Because of the uncertainty of inflow water, a four-period(five-stage) scenario tree and improved interval-parameter multi-stage stochastic programming model were used to carry out dynamic distribution of water in multiple stages of growth.Research results showed that in the context of different inflow level, managers could determine the water sensitive index of crop growth in different stages in accordance with the weight coefficient in each growth stage.The greater weight coefficient of the water sensitive index was, the greater the impact of water shortage had on the output, hence it belonged to the key water requirement stage; the smaller weight coefficient of the water sensitive index was, the smaller the impact of water shortage had on the output, hence it belonged to the non-key water requirement stage. When inflow water was excessive, excess water could be allocated to the next stage; when inflow water was insufficient, managers needed to seek a trade-off between reducing irrigation water and increasing transfer water.Tillering stage(t=1) and elongation stage(t=2) belonged to the key water requirement stage, managers wanted to fully satisfy the water needs of crops to ensure production, hence, in these two growth stages, no water deficiency existed, water deficiency value was 0.In different inflow level, when the irrigation water could not meet the demand for water, water supply project would transfer water from external water sources to meet the distribution targets in various growth stage; heading stage(t=3) and milk stage(t=4) belonged to non-key water requirement stage, so water deficiency had little effects on crop yield, and water supply project would make water deficit choice.When inflow level was not determined, it would take a lot of risk to meet the high water demand, and also increase the punishment of water deficiency; as for the low water demand, it took less risks, meanwhile, the punishment and water use benefit was also lower, which means water supply, risk and economic benefits are interwoven with each other.Through the analysis of optimal allocation of water supply project, the relationship between external water and irrigation water supply quantity, this paper realized the dynamic distribution of multi-water source irrigation system.This model can ensure the crop yield when the irrigation water resources are rationally configured in the growing period of the crop, and effectively avoid the waste of water resources, and improve the efficiency of irrigation water, which is of great significance for sustainable use of water resources.