Increased concern for the impact of urban runoff on receiving water quality makes acute the need to develop runoff quality control models for the various stages of urban runoff control planning. Analytical probabilistic models, which compare favourably with the STORM simulation model, have been derived to predict the long-term pollution control performance of storage-treatment systems from rainfall statistics. With knowledge of the cost functions of storage-treatment systems, the developed cost optimization precedure can be used to determine the least-cost combination of storage and treatment systems which can achieve various levels of quality control. This information is useful for specification of design performance and subsequent design level analyses of control systems.