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

Abstract

Abstract. We use a convolutional neural network (CNN) to identify plumes of nitrogen dioxide (NO2), a tracer of incomplete 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 28\u2009×\u200928-pixel images of TROPOMI data (corresponding to 266\u2009×\u2009133\u2009km2) and find that the model can identify plumes with a success rate of 90\u2009%. Over our study period, we find over 310,000 individual NO2 plumes of which 9\u2009% 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 92\u2009% of total CO2 emissions, with the remaining 8\u2009% mostly due to a large number of small sources <\u20090.2\u2009gC/m2/day for which our NO2 plume model is less sensitive. We argue the underlying CNN approach could form the basis of a Bayesian framework to estimate anthropogenic combustion emissions.\n