Household transportation lifecycle greenhouse gas emission prediction

Abstract

This investigation develops a model to predict household transportation life-cycle greenhouse gas (GHG) emissions and identifies the strongest determinants of these emissions. The impact of many variables on household transportation GHG emissions is examined. Ten machine learning methods are used for modeling and prediction. Shapley additive explanation is then applied to detect the relative influence of variables on household GHG emissions. Partial dependency plots are also employed to capture the direction of influence of top variables on household GHG emissions. Further analyses suggest that considering tail-pipe emissions rather than life-cycle emissions leads to underestimating the GHG emissions by roughly 20%. Replacing all gasoline vehicles with electric vehicles would reduce GHG emissions in Montreal by 57%. Then, the modal shifts required to meet the Government of Canada’s goals for transportation GHGs are determined. Finally, a scenario analysis is applied, and a number of GHG emission scenarios are presented.