A fast and accurate water quality pollutant risk assessment and early warning system, which has great practical significance for decision making in accident management, is urgently needed for water protection and management. Based on a fast mobile early warning system named MEWSUB, this paper modified its framework to make it generate data more automatically and accurately. By adapting manning formula and particle swarm optimization (PSO) for parameters optimization, the accuracy of water quantity and water quality simulation results has been improved. The modified system was successfully applied in an antimony tailings dam leakage accident that happened in China. The coefficient of determination (R2) of the prediction result was higher than 0.9 and relative error (ree) was less than 0.1, which indicated that the accuracy of MEWSUB was high enough for realistic water quality pollutant risk early warning.