7/12/2023 0 Comments Climate lockdowns uk![]() ![]() To better estimate the impact of restrictions on air quality, we have computed the average monthly percentage difference between predicted and observed NO 2 levels. Source: Frontier analysis of AURN and NOAA data using the ‘rmweather’ r package Interestingly, we see the gap narrowing in autumn, thus suggesting that the second lockdown, which was in place during November 2020, had a minor impact on air quality compared to the first lockdown.įigure 3 : Average daily NO2 levels in London - predicted vs observed concentrations Over the summer, projected concentrations remain significantly lower than observed NO 2 levels. We then observe a divergence between predicted and actual NO 2 levels starting from late March. ![]() This shows that our model is able to anticipate NO 2 quite precisely under business-as-usual conditions. Predicted concentrations closely track observed concentrations in January, February and early March. London’s NO 2 levels in 2020 were much lower than projectedįigure 3 compares average expected and observed NO 2 concentrations in London over the course of 2020. ![]() The model predictions should reflect seasonal and meteorological variations in pollution levels, thus allowing us to attribute differences between predicted and observed levels to lockdowns rather than meteorological effects. For each monitoring station, our model predicts pollutant concentration for each hour in 2020 based on observed past concentrations, associated weather conditions and other trends and cyclical effects. The model is trained using several years of London historical data on pollution levels and weather conditions collected by air monitoring stations that are part of the UK Automatic Urban and Rural Network (AURN). We have created our counterfactual using a random forest machine-learning model developed by researchers at the University of York (Grange, 2020) and made available to the public through the ‘ rmweather’ package on the r statistical software. This boils down to creating a credible counterfactual that shows how pollution levels would have evolved absent the lockdown. In order to assess the causal impact of lockdowns on London’s air quality one would ideally need to compare the actual observed pollution levels to those that would be recorded under the same weather conditions if it was business as usual. Machine learning can help isolate the impact of lockdowns on air quality Failing to do so introduces the risk of attributing to policy (lockdowns in this case) changes in air quality that would have naturally occurred given different weather conditions. Source: Frontier analysis of AURN and NOAA data using the ‘openair’ r packageįor these reasons, weather conditions should be taken into account when trying to assess the impact of a policy on air quality. However, several confounding factors are at play, so one should be careful not to interpret the apparent correlation between lockdowns and better air quality as a proof that mobility restrictions indeed caused the observed reduction in pollution levels.įigure 1 : NO 2 daily concentrations on London’s Marylebone Road NO 2 then increased again in autumn and winter but remained lower than it was in the first months of the year. As can be seen from Figure 1, NO 2 levels significantly decreased after the first lockdown was put in place in late March 2020, likely as a result of reduced emissions from road traffic, and stayed low over the summer. Machine learning can help identify key trends in pollutant concentrations while controlling for confounding factors such as weather conditions - as our analysis of London air quality data shows.Ī first glimpse at London Nitrogen Dioxide (NO 2) concentrations suggests that lockdowns had a material effect on air quality. However, assessing the impact of lockdowns on air quality is not a straightforward exercise. Covid-19 restrictions have provided the set for a unique experiment: what happens to air pollution when a large city grinds to a halt? ![]()
0 Comments
Leave a Reply. |