Rapid detection and early warnings of water pollution in urban water supply systems are crucial for the safety of drinking water. The tap water of a city was monitored continuously for an extended period using a fluorospectrophotometer. A parallel factors (PARAFAC) algorithm was used to analyze the data and spectra. The fluorescence peaks of protein-like and humic-like substances were both in a stable state. The sudden change and rise of the fluorescence peak of protein-like substances were probably caused by the influx of some fluorescent organic pollutants into the raw water; the variation of the fluorescence peak of humic-like substances was probably related to the climate change. Sudden pollution in city water can be detected via 3D fluorescence reliably and in a timely manner; this was verified by using the leakage of ethylbenzene to simulate water pollution.