With conventional methods, the non-linearity and local characteristics of water quality indicator time series are hardly addressed. Aiming at solving this problem, a new water quality time series identification and analysis method combining the merits of STL time series decomposition algorithm and Mann-Kendall trend test algorithm is proposed. In order to isolate the trend terms of the indicators, this method regressively decomposes the time series data of water quality indicators using STL time series decomposition algorithm, then uses Mann-Kendall trend test algorithm to identify and analyze the variation trends and characteristics of the terms. The data source of eight water quality indicator time series from 2014 to 2018 of 12 monitoring stations in Minjiang River Basin was analyzed using this method. The results show that the overall water quality of the Minjiang River Basin is good and is improving steadily. The water quality of the upper reaches of Minjiang River is generally better than that of the lower reaches; however the organic matter pollution is reversed. In the lower reaches of Minjiang River, the concentration of NH3—N and TP has decreased significantly, but the DO value is lower than that in the upper reaches and has become the dominant factor affecting the water quality.