Data-driven assessment of air quality and health benefits from future shipping emission controls in coastal China

The lack of toolkits for assessing the shipping-related atmospheric impacts limits China's ability to formulate effective shipping emissions control policies to address coastal air pollution and mitigate related health burdens. Here, we developed a deep learning model, DeepShip, to efficiently predict shipping-related PM_2.5 concentrations and further coupled it with a multi-task learning and generative-adversarial training strategy to enhance the sensitivity of the data-driven model to variations in small emission sources. Based on DeepShip, we comprehensively analyzed the response of shipping-related PM_2.5 to changes in anthropogenic emissions based on 210 scenarios involving emission reductions of shipping and land-based sectors. Furthermore, sulfur and nitrogen emission control scenarios that China might implement in the future were established to assess their cost, air quality improvement, and health benefits. We found that shipping-related PM_2.5 shows an almost linear relationship with shipping emissions, while exhibiting a nonlinear relationship with land-based emissions. Considering the cost and environmental-health benefits, future shipping emissions control should prioritize progressively enhancing the NO _<i>x</i> emission standard while coordinating with land-based emission reductions.