Paul Crutzen Publication award 

COVID-19 was the major event of 2021, and presented a unique opportunity to understand the atmospheric effects of widespread reductions in air pollutant emissions. However, the effects on atmospheric chemistry were short-lasting and therefore very challenging to separate reliably from the large variability in meteorology. In a very effective application of machine learning, Keller et al. achieved this separation for thousands observation sites in 46 countries. The results reveal reductions in nitrogen dioxide concentrations of as much as 60%, but much more complex responses of ozone that depend on season, environment, and non-linear chemical processes. The results provide a unique dataset to test our understanding of atmospheric chemistry and to more-reliably predict future reductions in ozone pollution as nitrogen dioxide emissions decline.