Detection and attribution of reduced satellite-observed AOD to COVID-19 with machine learning

Abstract:

In this contribution, machine learning is used to attribute observed AOD anomalies in China to COVID-19 related political measures.

Atmospheric aerosols are a key but poorly understood component of the climate system, and as air pollutants also play a critical role for human health. Severe air pollution episodes are a well-known problem in China, a global hot spot of aerosol abundance, and pose a major health risk to the population. Also, climate-system effects of aerosols are particularly pronounced in China and over adjacent seas. Therefore, reaching a quantitative understanding of the potential of short-term air quality measures on the atmospheric aerosol loading in this region is relevant for both policy makers and the climate science community.

Here, the conditions resulting from the policy response to the coronavirus outbreak are used to analyze effects of dramatic reduction in air-pollution sources on atmospheric aerosol loading. A statistical model is designed and applied to estimate daily atmospheric aerosol optical depth (AOD) during the first months of 2020 in China. The model uses information on aerosol climatology, geography and meteorological conditions, and is capable of explaining 70% of the day-to-day aerosol variability. The results are analyzed with a particular focus on large urban regions. In the Shanghai region, AOD is significantly lower than predicted in March 2020, which may be attributed to reduced local emissions. Regional variability is observed, with some parts of China displaying AOD levels above predictions, likely explained by dust transport and wild fire-related aerosol emissions. The findings highlight the use of statistical models in attribution studies, and that strong air-pollution measures, if enforced for a prolonged time, have the potential to substantially reduce local aerosol loadings, with potential implications for human health and climate.