The intervalparameter twostage robust stochastic programming model (ITRM) was put forward for supporting optimization and management of water resources in Mudanjiang Irrigation District, Sanjiang Plain, which was in such a situation, with multiwater source, multicrop and multirobustness coefficient. The ITRM incorporated intervalparameter programming (IPP) and twostage stochastic programming (TSP) with a robust optimization framework to deal with uncertainties. Compared with the traditional TSP, ITRM can overcome the deficiency of uncontrollable risk in system. The robustness coefficient was introduced to represent the system risk, which was effectively reflected in the model. Let the robust coefficient be 0, 0.4, 1, 2，3 and 5, respectively. With the increase of robust coefficient, the feasibility of the model and the stability of the system were enhanced. It was a sufficient and effective balance between system cost and risk when the robustness coefficient was 3. A more stable allocation of water resources was generated and the system cost was between 1.90327×109 RMB and 2.63475×109 RMB. Based on this model, a series of scenarios under different levels of preallocation water was done and different degrees of water surplus and shortage were obtained correspondingly. All these results were valuable for watersaving irrigation agriculture and improving water use efficiency and economic value, and thus it provided a technical support for decision maker in reasonable allocation of agricultural water resources.