Predictive analysis of flight traffic at New York JFK Airport on air pollution using machine learning

Abstract

Anthropogenic activities like transportation result in the emission of numerous pollutants like NO2, CO, SO2, PM2.5-10, O3, and Pb, which are identified as major air pollutants by environmental and health agencies. Emissions of those major pollutants result in negative health impacts like cardiovascular and respiratory diseases. Even though emissions from on-road traffic have decreased in recent years due to stricter regulations and technological advancements, less strict regulations on aircraft have resulted in an increase in emissions with the increasing air traffic. This study aims to estimate NO2 emissions from commercial flights at John F. Kennedy Airport (JFK) in New York and their impacts on air quality. The study combines numerical modeling using the AERMOD air dispersion model along with a machine learning model to predict NO2 concentration distributions as a function of space and time. To achieve this goal, departure and arrival flight data of John F. Kennedy Airport (JFK) in New York for the year 2018 is used. After the data is cleaned and prepared for the analysis, AERMOD is used to simulate atmospheric pollutant dispersion. The results of this study indicate that aircraft emissions can lead to significant NO2 concentrations in the vicinity of the airport. The simulated concentrations are then used in the training of a machine learning model. Decision tree-based extreme gradient boosting (XGBoost) is used as a machine learning model. It is shown that training in the emission prediction model has resulted in a well-generalized and well-performing model. Overall, this study demonstrates that machine learning modeling can be an effective tool for estimating pollutant dispersion.