In the past decades, little abatement efforts have been implemented on China's non-point source water pollution, and studies aiming at non-point sources regulation were also rare. Watershed abatement trading between point and non-point sources may serve as a cost-effective way to deal with the problem. The inherent uncertainty of non-point emissions, however, could affect the feasibility and outcome of point-non-point effluent trading. The purpose of this paper is to model the watershed point-non-point abatement trading incorporating the uncertainty of non-point source emissions, and to examine its impacts on trading equilibrium and trading ratio. The uncertainties of non-point emissions were taken into consideration by setting an acceptable probability by which the watershed emission constraints were achieved. Using the watershed optimization model, the optimal abatement allocation and trading ratio were explicitly illustrated. It was found that they were affected significantly by the variances of non-point emissions, the reliability requirement assigned to the non-point abatement, and the marginal abatement costs of point and non-point sources. Since the variances of non-point emissions may increase or decrease at the abatement level, the impacts of these factors were discussed in different circumstances. Based on the illumination of the trading model, future directions and implications of point-non-point trading in China were discussed.