Jérôme Barré, Hervé Petetin, Augustin Colette, Marc Guevara, Vincent-Henri Peuch, Laurence Rouil, Richard Engelen, Antje Inness, Johannes Flemming, Carlos Pérez García-Pando, Dene Bowdalo, Frederik Meleux, Camilla Geels, Jesper H. Christensen, Michael Gauss, Anna Benedictow, Svetlana Tsyro, Elmar Friese, Joanna Strużewska, Jacek W. Kaminski, John Douros, Renske Timmermans, Lennart Robertson, Mario Adani, Oriol Jorba, Mathieu Joly, Rostislav Kouznetsov
This study provides a comprehensive assessment of NO2 changes across the main European urban areas induced by COVID-19 lockdowns using satellite retrievals from the Tropospheric Monitoring Instrument (TROPOMI) onboard the Sentinel-5p satellite, surface site measurements, and simulations from the Copernicus Atmosphere Monitoring Service (CAMS) regional ensemble of air quality models. Some recent TROPOMI-based estimates of changes in atmospheric NO2 concentrations have neglected the influence of weather variability between the reference and lockdown periods. Here we provide weather-normalized estimates based on a machine learning method (gradient boosting) along with an assessment of the biases that can be expected from methods that omit the influence of weather. We also compare the weather-normalized satellite-estimated NO2 column changes with weather-normalized surface NO2 concentration changes and the CAMS regional ensemble, composed of 11 models, using recently published estimates of emission reductions induced by the lockdown. All estimates show similar NO2 reductions. Locations where the lockdown measures were stricter show stronger reductions, and, conversely, locations where softer measures were implemented show milder reductions in NO2 pollution levels. Average reduction estimates based on either satellite observations (−23 %), surface stations (−43 %), or models (−32 %) are presented, showing the importance of vertical sampling but also the horizontal representativeness. Surface station estimates are significantly changed when sampled to the TROPOMI overpasses (−37 %), pointing out the importance of the variability in time of such estimates. Observation-based machine learning estimates show a stronger temporal variability than model-based estimates.