Water quality management is subject to some challenges such as inherent uncertainties of the natural system and competing decision objectives.Therefore,as the most used method of optimal decision making for water quality management,mathematical programming is required to represent optimization models with uncertain coefficients and multiple objectives.A reliability-oriented multi-objective decision making approach was proposed for optimization problems with stochastic parameters and multiple decision reliability objectives,based on multi-objective dependent-chance programming (MO-DCP).MO-DCP was solved by the Controlled Elitist Non-dominated Sorting Genetic Algorithm (CE-NSGA-Ⅱ),an advanced multiple objective evolutionary algorithm.To analyze relationship between reliability levels and decision variables,objective-interpretation is developed by means of Back Propagation Neural Network (BPNN).Advantages of the developed decision approach lies in the following points: (1) it provides decision makers with great insight into tradeoffs between reliability levels; (2) a range of alternative solutions can be generated with one single run of the optimization algorithm; (3) subjective pre-determination of reliability levels can be avoided; (4) it provides a shortcut to identify schemes at given reliability levels for decision makers by objective-interpretation.The approach was applied in a case study for optimal nutrient load reduction of the Lake Dianchi Watershed with stochastic information,aiming to minimizing reliability levels of reduction cost and environment capacity.The results demonstrated tradeoff curves between cost and capacity reliability and schemes at different reliability levels.