Authors | Hao Wu and David Carslaw |
Compilation date | 01 May 2020 |
Customer | The Scottish Government |
Approved by | Stuart Sneddon |
Copyright | Ricardo Energy & Environment |
EULA | http://ee.ricardo.com/cms/eula/ |
Contract reference | Report reference |
This note has been produced for The Scottish Government and examines the impact of lockdown measures on evolving ambient air quality data. This analysis focuses on Scottish data and uses proven modelling techniques to discount the influence of weather on ambient pollutant concentrations. Details of this methodology can be found in this blog. This analysis focuses on NOx and NO2. At roadside locations NOx concentrations will be closely linked to primary emissions and should show the direct impact of reduced local traffic on air pollution. NO2 will be from a mixture of primary emissions and secondary chemical reactions but should again be closely linked to local traffic reduction. Other pollutants such as PM and O3 are more complicated due to their secondary atmospheric formation, requiring more extensive detailed analysis which is not covered here.
N.B. Both measured and modelled data reported here are provisional pending full QA/QC processes.
First let’s look at the daily concentrations of NOx and NO2 since March 2020 and get a general sense of the trend of pollutant concentrations. The pink and green shadings indicate the periods when social distancing was advised and when lockdown was enforced.
A perennial problem when comparing air pollution data before and after certain interventions is the effect of the weather. To counteract the effect of weather, we use a model to simulate NOx and NO2 concentrations from 2018-01-01 to 2020-02-29 using wind speed/direction, temperature, hour of the day, weekday and Julian day as predictors. This model is then used to predict NOx and NO2 concentrations from March 2020, which can be seen as the normal NOx and NO2 concentrations expected if no intervention had taken place.
The method for simulating the normal concentration is our current best estimate. However as things progress this method will evolve and get refined.
The differences between simulated and measured NOx and NO2 concentrations are shown below. The coloured vertical segments represent the difference between modelled and measured NOx and NO2. Purple lines mean the model overestimated the concentrations, whereas orange means the model underestimated. Before 23rd March, when lockdown was enforced, there was roughly the same amount of overestimation and underestimation by the model, suggesting that the measured NOx and NO2 was comparable to the usual levels at this time of the year and under normal business activities. The modelled (i.e. ‘business as usual’) NOx and NO2 concentrations are predominately higher than the measured ones from 23rd March, which suggests that reduced emissions from traffic and industry are being seen in the measurements.
Another way to look at how much change has taken place is to plot the cummulative sum (cusum) of the modelled and measured difference. By accumulating these differences, the effect is to amplify the changes and also provide an indication of when changes may have occurred. The cusum lines stay around zero for most sites before 16th March. However, all of the sites show a consistent decrease after the lockdown started.
To put the magnitude of the decrease into perspective, the mean concentrations of NOx and NO2 after lockdown for measured and modelled (BAU, Business as Usual) concentrations are shown below. The percent increase/decrease is the calculated as the of differences between measured and BAU concentrations divided by the BAU concentrations. In general the roadside sites show larger relative decrease compared to background sites.
For further information, specific questions or bespoke analysis, please use the contact below.
Name | Stuart Sneddon |
Address | Ricardo Energy & Environment, 18 Blythswood Square, Glasgow, G2 4BG, UK |
Telephone | +44 1235 753015 |
stuart.sneddon@ricardo.com |