<p>Risk assessment of severe weather conditions for ocean-going ships provides security guarantees for marine transportation. Due to the diversity of indicators and the uncertainty in marine environment, it is difficult to construct an objective and quantitative risk assessment model. This study proposes a new marine environmental risk assessment model based on the cloud model and projection pursuit. First, through the analysis of marine environmental characteristics and ship accidents, a risk index system and five-level risk classification criteria are proposed. Second, the backward cloud generator is used to generate the cloud parameters of the assessment indicators, and the x-condition cloud generator is applied to build the membership function. Finally, the projection pursuit model is constructed to realize the fusion of membership degrees, and the risk level is obtained by combining the principle of maximum membership degree. The new model is applied to the Milan ship typhoon accident in the South China Sea to evaluate the seasonal risk of each hazard factor and the marine environmental risk of accident nodes. It is found that the risk of each hazard factor in the South China Sea has obvious seasonal characteristics as well as spatial distribution characteristics. The results of the sea area where the accident of the “Milan” ship occurred are extremely high, while the starting point and the chosen docking site are rated as extremely low. This is consistent with the actual situation and confirms the validity and practicality of the model.</p>