Forecasting air pollutant emission intensity for maritime shipping based on machine learning and statistical time-series models

Abstract As the International Maritime Organization (IMO) continues to tighten its emission regulations to control shipping-related pollution, forecasting ship pollutant emission intensity has become an essential prerequisite for evaluating the effectiveness of mitigation measures and ensuring industry compliance. This study aims to identify an optimal model for maritime emission forecasting by comparing the performance of multiple predictive models. Moreover, in conjunction with IMO’s emission-reduction targets, this work aims to evaluate the effectiveness of current reduction policies. Using daily time-series data, the performances of the seasonal autoregressive integrated moving average with exogenous factors (SARIMAX) model, the Prophet model, the long short-term memory (LSTM) network, and the Holt–Winters model in forecasting ship emission data are compared. The findings indicate the following: (1) The LSTM model exhibits broad applicability in forecasting shipping emissions; (2) shipping-related CO₂ emissions are expected to reach 549 million tons by 2030; and (3) under the dual regulatory framework of the Energy Efficiency Design Index and Carbon Intensity Indicator, the shipping industry is expected to successfully pass its first temporal checkpoint.