Nitrogen dioxide (NO2) plays a critical role in terms of air quality, human health, ecosystems, and its impact on climate change. While the crucial roles of the vertical structure of NO2 have been acknowledged for some time, there is currently limited knowledge about this aspect in China. The Geostationary Environment Monitoring Spectrometer (GEMS) is the world's first geostationary satellite instrument capable of measuring the hourly columnar amount of NO2. The study presented here introduces the use of mixing height for NO2 in the atmosphere. A thorough examination of spatiotemporal variations in the mixing height of NO2 was conducted using data from both the GEMS and ground-based air quality monitoring networks. A random forest model based on machine learning techniques was utilized to examine how meteorological parameters affect the mixing height of NO2. The results of our study reveal a notable seasonal fluctuation in the mixing height of NO2, with the highest values observed during the summer and the lowest values during the winter. Additionally, there was an increasing diurnal trend from early morning to mid-afternoon. Moreover, the study discovered elevated NO2 mixing heights in the dry regions of northern China. The results also indicated a positive correlation between the mixing height of NO2 and temperature and wind speed, while negative associations were found with relative humidity and air pressure. The machine learning model's predicted NO2 mixing heights were in good agreement with the measurement-based outcomes, as evidenced by a coefficient of determination (R2) value of 0.96 (0.84 for the 10-fold cross-validation). These findings emphasize the noteworthy influence of meteorological variables on the vertical distribution of NO2 in the atmosphere and enhance our comprehension of the three-dimensional variations in NO2.
Keywords: GEMS; Machine learning; Meteorology; Nitrogen dioxide; Vertical distribution.
Copyright © 2023 Elsevier Ltd. All rights reserved.