Estimating the impact of COVID-19 on the PM2.5 levels in China with a satellite-driven machine learning model

Abstract:

Ground monitoring showed a significant reduction of PM_2.5 levels during the COVID-19 pandemic in major cities worldwide. China implemented an aggressive locked down procedure immediately after the outbreak in January 2020. As China emerges from the impact of COVID-19 on national economic and industrial activities, it became the site of a large-scale natural experiment to evaluate the impact of COVID-19 on regional air quality. However, ground measurements of PM_2.5 concentrations do not offer comprehensive spatial coverage especially in suburban and rural regions in China. In this study, we developed a machine learning method with satellite-derived aerosol optical depth, meteorological fields and land use parameters as major predictor variables to provide spatiotemporally resolved daily PM_2.5 estimates. Our study period consists of a reference period (November 1, 2018 – April 30, 2019) and an epidemic period (November 1, 2019 – April 30, 2020), with a total length of 363 days. Each period was then divided into period 1 (November and December), period 2 (January and February) and period 3 (March and April). The reference period model obtained a 10-fold cross-validated R² (RMSE) of 0.79 (17.55 ug/m³) and the epidemic period model obtained a 10-fold cross validated R²(RMSE) of 0.83 (13.48 ug/m³) for daily PM_2.5 predictions. Using trained PM_2.5 models and predictor variables, we estimated daily PM_2.5 at a resolution of 5 km × 5 km across China during the study period. Our prediction results showed that PM_2.5 levels were lowered by 4.8 ug/m³ during the epidemic period comparing to the reference period and PM_2.5levels during period 2 decreased most by 18%. The Southeast region was affected most by the COVID-19 outbreak with PM_2.5 levels during period 2 decreased by 31%.