Abstract: This research took the Dongzhang Reservoir in Fuqing City, Fujian Province as an example, and collected water quality monitoring data (including water temperature, pH, conductivity, turbidity, and dissolved oxygen) and meteorology data (including air temperature, and wind speed) of 295 groups for the bloom period from 2016 to 2017. The BP artificial neural network model was set-up with 80% of the data set for model training, and 20% of the data set for model validation, while the concentration of chlorophyll-a was used as the output parameter. By testing different parameter combinations as input parameters, the results were compared with the measured chlorophyll-a concentrations in order to select the optimal parameter combination. The results showed that the combination of water temperature, dissolved oxygen, conductivity and air temperature were the optimal input parameter sets. Comparing with the measurements, the model performance was satisfied with the coefficient of determination (R2) equals to 0.83, the root mean square error (RMSE) equals to 0.08 μg/L, and the ratio of root mean square error to standard deviation (RSR) equals to 0.43. It was expected that the BP artificial neural network could be used to predict the occurrence of phytoplankton bloom in Dongzhang Reservoir as an early warning system.