Evaluation of the effects of the COVID-19 lockdown and the meteorology on surface ozone variability with machine learning techniques.
- 1Istituto Superiore di Sanità, Rome, Italy (roberta.gagliardi@iss.it)
- 2Istituto Nazionale Assicurazione contro gli Infortuni sul Lavoro, Rome, Italy (c.andenna@inail.it)
The Mediterranean Basin, which is considered an hot-spot region in term of air-quality and climate change, frequently experiences surface ozone (O3) exceedances to European legislated values for human health protection. Indeed, its peculiar meteorological conditions, such as frequent clear sky and high solar radiation in summer, as well as the availability of natural and anthropogenic precursors, enhance the formation of photochemical ozone so increasing the risk of O3 pollution events in the area.
O3 is one of the air pollutants of most concern for public health, as it can damage the respiratory and circulatory systems. O3 can also deteriorate the vegetation, ecosystems, and building materials. Moreover, it is also one of the most important greenhouse gases as a so-called short-lived climate factor. Being a secondary pollutant, the variability of O3 is strongly dependent on the precursors, the meteorological parameters and the interactions among them through a series of complex and non-linear functions. Such complexity makes the definition of the O3 control strategies not immediate; this is why the improvement of the extensive networks of continuous O3 measurements as well as of appropriate methods to analyze O3 levels and to predict future changes is required.
The abrupt and unplanned anthropogenic emission reductions determined by the COVID-19 lockdown in 2020 produced an extraordinary real-world opportunity to assess the air quality response to changes in emissions. Taking advantage of this circumstance, the present study aims to characterize the influence of both the precursor emissions and the local meteorological conditions during 2020 on the surface ozone concentrations measured in the Basilicata region, (Southern Italy), which is located in the center of the Mediterranean basin.
To these ends, eXtreme Gradient Boosting (XGBoost) machine learning (ML) models were built to provide both the business-as-usual (BAU) and the meteorological normalized O3 time series. The former allows the estimates of the impact of the lockdown restrictive measures on O3 by means of the comparison with the observed concentrations, the latter accounts for the meteorological variability and other time features for more reliable assessment of O3 trend/changes. Moreover, thanks to the non-black box nature of the developed ML models, the partial dependence of the observed O3 concentrations on each explanatory variables used in the models can be analysed, shedding light on the role of local meteorological processes in the observed O3 variability.
The results obtained made it possible to evaluate the different contribution of emissions and meteorology on the detected changes in the levels of O3 and to provide deeper insights of major drivers of surface ozone concentrations in the studied area. This knowledge could help in defining strategies effective in reducing the negative impacts associated with O3 exposure as well as in optimizing the potential synergies between O3 reduction policies and climate change policies.
How to cite: Gagliardi, R. V. and Andenna, C.: Evaluation of the effects of the COVID-19 lockdown and the meteorology on surface ozone variability with machine learning techniques. , 17th Plinius Conference on Mediterranean Risks, Frascati, Rome, Italy, 18–21 Oct 2022, Plinius17-16, https://doi.org/10.5194/egusphere-plinius17-16, 2022.
