As a result of decreasing distance between cities during the rapid urbanization, the water quality of cities located in upper and lower reaches has closer interations. This paper proposes the concept of safe distance between cities to ensure the water quality in downstream cities, which is quantified based on the BP neural network model for water quality. Two adjacent cities along the Yangtze River, Wuhu and Ma'anshan, are chosen as the representative case to evaluate the safety of water quality and quantify the minimum safe distance after city expansion. The results reveal a safe distance of 4.6km between the two cities in 2020, which could ensure the water quality of the control section in upper reaches of Ma'anshan (the downstream city) to meet the class II standard of surface water. However, compared with the year 2010, the water quality of the control section will decline, where the COD concentration is projected to increase by 29.2% and NH₃-N by 23.2%. In order to ensure the water function of the control section, the minimum safe distance between the two cities needs to be 3.2km.