Automated detection of atmospheric NO2 plumes from satellite data: A tool to help infer anthropogenic combustion emissions

We use a convolutional neural network (CNN) to identify plumes of nitrogen dioxide (NO2), a tracer of combustion, from NO2 column data collected by the TROPOspheric Monitoring Instrument (TROPOMI). This approach allows us to exploit efficiently the growing volume of satellite data available to characterize Earth's climate. For the purposes of demonstration, we focus on data collected between July 2018 and June 2020. We train the deep learning model using six thousand 28ĝ×ĝ28 pixel images of TROPOMI data (corresponding to ĝ‰ ĝ266ĝkmĝ×ĝ133ĝkm) and find that the model can identify plumes with a success rate of more than 90ĝ%. Over our study period, we find over 310ĝ000 individual NO2 plumes, of which ĝ‰ ĝ19ĝ% are found over mainland China. We have attempted to remove the influence of open biomass burning using correlative high-resolution thermal infrared data from the Visible Infrared Imaging Radiometer Suite (VIIRS). We relate the remaining NO2 plumes to large urban centres, oil and gas production, and major power plants. We find no correlation between NO2 plumes and the location of natural gas flaring. We also find persistent NO2 plumes from regions where inventories do not currently include emissions. Using an established anthropogenic CO2 emission inventory, we find that our NO2 plume distribution captures 92ĝ% of total CO2 emissions, with the remaining 8ĝ% mostly due to a large number of small sources (<ĝ0.2ĝgĝCĝm-2ĝd-1) to which our NO2 plume model is less sensitive. We argue that the underlying CNN approach could form the basis of a Bayesian framework to estimate anthropogenic combustion emissions.