Preprints
https://doi.org/10.5194/amt-2021-177
https://doi.org/10.5194/amt-2021-177

  20 Aug 2021

20 Aug 2021

Review status: a revised version of this preprint is currently under review for the journal AMT.

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

Douglas Finch1,2, Paul Palmer1,2, and Tianran Zhang3,a Douglas Finch et al.
  • 1National Centre for Earth Observation, University of Edinburgh, UK
  • 2School of GeoSciences, University of Edinburgh, UK
  • 3National Centre for Earth Observation, Kings College London, UK
  • anow at: Satellite Vu, UK

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 × 28-pixel images of TROPOMI data (corresponding to 266 × 133 km2) and find that the model can identify plumes with a success rate of 90 %. Over our study period, we find over 310,000 individual NO2 plumes of which 9 % 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 % of total CO2 emissions, with the remaining 8 % mostly due to a large number of small sources < 0.2 gC/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.

Douglas Finch et al.

Status: final response (author comments only)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse
  • RC1: 'Comment on amt-2021-177', Anonymous Referee #2, 07 Sep 2021
    • AC1: 'Reply on RC1', Douglas Finch, 01 Nov 2021
  • RC2: 'Comment on amt-2021-177', Anonymous Referee #1, 08 Sep 2021
    • AC2: 'Reply on RC2', Douglas Finch, 01 Nov 2021

Douglas Finch et al.

Douglas Finch et al.

Viewed

Total article views: 878 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
624 238 16 878 6 6
  • HTML: 624
  • PDF: 238
  • XML: 16
  • Total: 878
  • BibTeX: 6
  • EndNote: 6
Views and downloads (calculated since 20 Aug 2021)
Cumulative views and downloads (calculated since 20 Aug 2021)

Viewed (geographical distribution)

Total article views: 792 (including HTML, PDF, and XML) Thereof 792 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
1
 
 
 
 
Latest update: 26 Nov 2021
Download
Short summary
We developed a machine learning model to detect plumes of nitrogen dioxide satellite observations over two years. We find over 310,000 plumes, mainly over cities, industrial regions and areas of oil and gas production. Our model performs well in comparison to other datasets and in some cases, finds emissions that are not included in other datasets. This method could be used to help locate and measure emission hotspots across the globe and help inform climate policies.