Heavy metal pollution accidents have the characteristics of high uncertainty, which are difficult to predict and control, thus posing great threats to the ecological security, human health and even social stability. In this study, based on the Bayesian Network, ecological risk was probabilistically evaluated with uncertainty for the accidental Cr6+-Hg2+ mixing pollution at the Dongjiang River downstream watershed, following the risk route “multiple risk sources hazardousness-risk stressors joint exposure-multiple risk receptors vulnerability-ecological risk”, and the relevant risk control and management measures were proposed. The results showed that (1) all electroplating risk sources (67 in total) would cause ecological risk to Dongjiang downstream watershed. The ecological risk level in Huizhou City was the highest compared to that in Shenzhen City and Dongguan City. (2) The ecological risk level was the highest in the tributary (Xizhijiang River), which was superior than that in the mainstream and the river network of the Dongjiang River. (3) Based on the sensitivity analysis, risk node “water body vulnerability” was the critical factor for the ecological risk. The quantitative risk assessment can provide a scientific basis for the ecological risk transfer methodologies, as well as technical support for ecological risk prevention and management at the watershed scale.