Lockdown measures to limit the spread of Covid-19 in France in spring 2020 sharply reduced activities and lowered air-pollution concentrations. This study sought to determine the short- and long-term impacts on mortality in metropolitan France resulting from the temporary decrease in the population's exposure to air pollution. The reduction in exposure to air pollution was estimated by calculating the difference between modeled exposure of the population during the strict lockdown and the gradual lifting, and the simulated exposure that would have been observed in the absence of lockdown. A quantitative health impact assessment was used to estimate both the short-term impact of PM10 and NO2 reductions, and the long-term impact of PM2.5 and NO2 reductions on mortality. Reduced activities during the lockdown lowered NO2 and PM concentrations, resulting in about 2,300 deaths postponed for PM2.5 and nearly 1,200 for NO2, mainly due to avoided long-term effects. This study shows that, even in an unprecedented context that is certainly neither realistic nor desirable to improve air quality in the long run, public interventions appear to have a significant impact on health through reductions in air-pollution levels. In a long-term perspective, the study also reminds us that the total burden of air pollution on health remains a significant risk factor in France. Efforts to reduce ambient air pollution must thus be pursued sustainably for all sources of air pollution with suitably adapted but ambitious policies. Finally, the lockdown restrictions had other consequences, both positive and negative, on the population's health. These consequences highlight the need to conduct more integrated assessments of health impacts that include the multisectoral consequences of interventions, particularly in terms of population compliance with mitigating restrictions, behavior and mental health and, more broadly, climate change.
Short- and long-term forecasting of ambient air pollution levels using wavelet-based non-linear autoregressive artificial neural networks with exogenous inputs
