A reliable simulation of the manganese cycle in water supply system is essential for timely and effective manganese treatment strategies. A data-driven model estimating manganese levels in the treated water was developed through a threshold-based approach using raw water and supernatant return manganese levels. This data-driven model and a three-dimensional manganese cycle model for the source reservoir were integrated to simulate the overall manganese variations from the lake to the treated water. A decision support system based on the developed data-driven model was established to allow operators to react to situations that could result in elevated levels of manganese in the treated water. Scenario analyses were completed to examine the manganese variations in the reservoir, raw water and treated water due to different precipitations and wind conditions and to allow the identification of optimal strategies for manganese management from the reservoir to the water supply system.