Counterfactual time series analysis for the air pollution during the outbreak of COVID-19 in Wuhan

TL;DR

This paper uses counterfactual time series models to analyze the impact of COVID-19 quarantine measures on air pollution in Wuhan, demonstrating significant pollution reduction due to these interventions.

Contribution

It introduces a comparative analysis of SARIMA, LSTM, and XGBOOST models for estimating air quality changes during COVID-19 in Wuhan.

Findings

COVID-19 measures significantly reduced air pollution in Wuhan.

LSTM outperformed SARIMA and XGBOOST in model accuracy.

Results support policy decisions for environmental management during crises.

Abstract

Environmental issues are becoming one of the main topics of concern for society, and the quality of air is closely linked to people's lives. Previous studies have examined the effects of abrupt interventions on changes in air pollution. For example, researchers used an interrupted time series design to quantify the impact of the 1990 Dublin coal ban; and a regression discontinuity to determine the arbitrary spatial impact of the Huaihe River policy in China. An important feature of each of these studies is that they investigated abrupt and localized changes over relatively short time spans (the Dublin coal ban) and spatial scales (the Huaihe policy). Due to the abrupt nature of these interventions, defining a hypothetical experiment in these studies is straightforward. In response to the novel coronavirus outbreak, China implemented 'the largest quarantine in human history' in Wuhan on January 23, 2020. Similar measures were implemented in other Chinese cities. Since then, the movement of people and associated production and consumption activities have been significantly reduced. This provides us with an unprecedented opportunity to estimate the changes in air pollution brought about by this sudden "silent" move. We speculate that the initiative will lead to a significant reduction in regional air pollution. Thus, we performed counterfactual time series analysis on Wuhan air quality data from 2017-2022 based on three models, SARIMA, LSTM and XGBOOST, and compared the excellence of different models. Finally, we conclude that 'silent' measures will significantly reduce air pollution. Using this conclusion to further investigate the extent of air pollution reduction will help the country to better designate environmental policies.