Development of a machine learning-based sketch planning model for predicting mobile emissions

Abstract

Metropolitan Planning Organizations (MPOs) are required to measure emissions impacts of transportation plans and programs utilizing an emissions estimator such as MOtor Vehicle Emission Simulator (MOVES) or EMission FACtor (EMFAC) models demanding intensive data and time consumption to run models for scenario planning or sensitivity tests. Over time, transportation planning practitioners have developed and applied various sketch planning models in scenario planning and sensitivity tests. Still, many sketch planning models require extensive data collection/preparation and are overly complicated. This paper discusses an approach for a simple sketch planning method based on a Random Forest (RF) algorithm with a machine learning technique that practitioners may use to predict emissions with significantly short time and effort in data preparation and model execution. In this study, the algorithm is trained to simulate emissions estimates based on multi-year Constrained Long Range Plan (CLRP) data in the National Capital region for eight case studies. The study results showed that the proposed model predicted mobile source nitrogen oxides, volatile organic compounds and greenhouse gas emissions within +/- 10 percent of accuracy against MOVES counterparts with fewer resources in a shorter amount of time. The authors think that the model could be a sketch planning model evaluating emissions impacts of variables such as land use, trip purpose, network enhancement or mobility.

Keywords

• Air quality
• Travel demand model
• MPO
• Machine-Learning
• MOVES
• Random forest