Bayesian probabilistic forecasting for ship emissions

Abstract This paper proposes a Bayesian forecasting algorithm to extrapolate ship movements and accordingly ship emissions based on probabilities extracted from current ship movements, sailing configurations and ship particulars. Total amount of ship emission on a given sea field may be estimated by using the Automatic Identification System (AIS) records. However, emission predictions cannot be generated since future ship movements are not available. Therefore, predictive exercises usually lie on extrapolations of mass amount of emissions independent from ship movements and any transformations of ship fuel and engine characteristics. In this circumstance, a Bayesian ship traffic generator is developed to simulate long-term predictions of ship movements based on cumulative ship traffic. The empirical study reflects the case of the Port of Singapore and the forecasting horizon for years of 2020 and 2025, with 2018 being the baseline year for comparison analysis. By utilizing the proposed algorithm, policy makers can visualize and simulate the impact of various regulations and implementation on emission control, fuel standards or technical changes in ship design or engine simultaneously. In other words, the proposed simulation testbed also enables prescriptive analytics of emission factors by adjusting technical features of ships.