In all researchers’ effort to quantify the portion of emission from the transportation sector, models of three resolution levels have been established: macroscopic, mesoscopic and microscopic. The finer the resolution of a model is, the higher accuracy the prediction is. While mesoscopic models have less demanding requirements on data input and can guarantee an acceptable accuracy, there are few studies focused on mesoscopic models and the few existing ones have poor model formulations. In this study, we utilized micro traffic simulation software and MOVES to create data basis. With the data, regression models and neural network models are established for the estimation of traffic emission. In particular, neural network models with long short-term memory architecture were employed to capture the information in data sequence.

CICTP