Prediction of Greenhouse Gas Emissions in Downtown Toronto Using Deep Sequence Learning

Mitigating the substantial undesirable impact of transportation systems on the environment is paramount. Thus, predicting Greenhouse Gas (GHG) emissions is one of the profound topics, especially with the emergence of intelligent transportation systems (ITS). We developed a deep learning framework to predict link-level GHG emission rate (ER) (in CO_2eq gram/second) based on the most representative predictors, such as speed, density, and GHG ER of previous time steps. In particular, various specifications of the long short-term memory (LSTM) networks with explanatory variables were examined and were compared with clustering and the autoregressive integrated moving average (ARIMA) model with explanatory variables. The downtown Toronto road network was used as the study area, and highly detailed data were synthesized using a calibrated traffic microsimulation and MOVES. It was found that LSTM specification with speed, density, GHG ER, and in-links speed from three previous minutes performed the best while adopting two hidden layers, and when the hyper-parameters were systematically tuned. Adopting a 30-second updating interval slightly improved the correlation between true (simulated) and predicted GHG ERs (from predictive models), but contributed negatively to the prediction accuracy as reflected in the increased root mean square error (RMSE) value. Efficiently predicting GHG emissions at a higher frequency with lower data requirements will pave the way for various applications, e.g. anticipatory eco-routing in large-scale road networks to alleviate the adverse impact of global warming.