Bristol’s Air Quality in the COVID-19 Lockdown

This brief report summarises the air quality impact of the lockdown policy on air quality measurements in Bristol.

Introduction

This report summarises changes to ambient air quality, represented by concentrations of regulated pollutants, that have occurred during the COVID-19 lockdown period. Bristol City Council continued to monitor air quality throughout lockdown as this work was classed by the government as an essential activity.

Traffic levels declined rapidly from slightly before lockdown leading to a clear reduction in emissions of key pollutants. This is apparent for roadside and background sites.

The changes in air quality characterised by comparisons of raw data between two periods cannot be solely attributed to the lockdown measures, because weather and other variables strongly influence ambient air quality. In order to account for the influence of weather and other covariates, a statistical modelling approach has been adopted which can remove the effect of the weather and identify the changes in concentrations which would arise if meteorological conditions and temporal effects are held constant. This approach can be used to ascribe the changes in air quality to lockdown measures with more certainty than simply by comparing raw data between two periods.

Summary of changes

The comparison of raw data between 2019 and 2020 was for the period 2019-03-25 to 2019-07-01 and 2020-03-24 to 2020-06-30.

Analysis of air quality data from Bristol City Council’s continuous air monitoring network comparing the lockdown period in 2020 to the same period in 2019 shows a significant change in nitrogen dioxide (NO₂), a traffic pollutant, of around -43%. Reductions in NOx (oxides of nitrogen), which can be considered a surrogate for direct exhaust emissions, fell even further with a mean change of -54%. These changes cannot be attributed solely to the lockdown measures because of the effect of weather and the small effect of changes in the vehicle fleet between the two comparison periods.

When the effect of weather is taken into account the mean reduction in NO₂ at all sites was -42%.

Measures of particulate matter (PM) - PM₁₀ and PM_2.5 also fell but the reduction was less. This is because the local contribution to ambient PM is a small part of the total. There are significant regional and background components present which are unaffected by the lockdown measures. For PM₁₀ the roadside increment in 2018 was 30% when comparing a background and roadside site in the central city.

Ozone (O₃) rose when compared to the baseline period. This is expected because as NOx declines, less ozone is chemically reduced in the photochemical reaction between these two species and hence concentrations of ozone may rise.

Data from the NO₂ diffusion tube network, which gives greater spatial coverage than the continuous network has been analysed. The changes in measured nitrogen dioxide for the months of April and May are shown.

All of Bristol’s air quality data are available through our open data portal.

Traffic

Traffic flows started to decline when the first measures were announced on the 16th March. By the time full lockdown was operating (from 24th March), traffic flows had declined by more than 50% compared to normal levels. Data are from BCC’s Urban Traffic Control (UTC) system and represent smoothed daily mean speeds and total flows at routes or counters close to the continuous air quality monitoring sites.

It can be seen that traffic flows began to climb slowly again after the first week in April.

Fig 1. Daily traffic flows and speeds: 2020

Google and Apple are providing access to their mobility data during the COVID-19 lockdown. The chart below shows smoothed trends in mobility by sector for Bristol from Google’s dataset. The baseline for comparison is the median value, for the corresponding day of the week, during the 5-week period Jan 3 – Feb 6, 2020. The initial reduction in mobility for all sectors except residential is apparent. In this context the residential category includes deliveries to homes. There does seem to be a slight increase after the first week in April of non residential sector mobility and from mid - May there was a steep increase in mobility associated with use of parks.

Fig 2. Traffic flows by sector: 2020

Summary of changes by pollutant

The table below shows the percentage changes in each pollutant aggregated by site between the two lockdown periods. Bristol City Council’s data are ratified according to processes detailed with our annual status reports. Data from the national network sites (Bristol St. Pauls and Bristol Temple Way) are not fully ratified at the time of writing.

Table 1. Average change in pollutant concentrations in ugm-3

The table below shows the mean percentage reductions by site and pollutant. Not all sites measure all pollutants.

Table 2. Average change in pollutant concentrations by site and pollutant (ugm-3)

Fig 3. Locations of current continuous monitoring sites

Comparison with other cities

The changes in Bristol can be shown in context with other core cities by using data from the national AURN network. The plot below shows the mean values of NO₂ for the lockdown period in each of the previous four years. Bristol’s national roadside site at Temple Way was not operating in 2017.

It is not possible to infer much meaning from the relative differences in changes between cities as there are a range of site - specific factors that could influence this, but the data are presented to show the national context.

Fig 4. National comparison: nitrogen dioxide

Changes in air quality at Bristol’s continuous monitoring sites

Bristol’s city centre site at Colston Avenue is the most polluted roadside site on the network and so is an interesting example of the reductions during the lockdown. The chart below shows weekly mean concentrations of NOx, NO₂ and PM₁₀ for the baseline and lockdown period. Operational issues between 9th April and 15th April mean that data is incomplete for this period.

Fig 5. Chart of Colston Avenue site by pollutant

Other sites on the network show similar patterns. For the Wells Road site, the difference in concentrations between the two periods was not as pronounced as for Colston Avenue.

Fig 6. NOx and NO2: Wells Road

Changes in nitrogen dioxide at diffusion tube sites

Diffusion tubes are used to monitor ambient nitrogen dioxide. Nitrogen dioxide from the air diffuses into a metal grid in the tube and the tubes are sent for analysis in a laboratory. Diffusion tubes are placed according to a monthly calendar specified by Defra and BCC’s diffusion tubes are exposed according to this calendar. April 2020 was the first complete month where tubes were entirely exposed under the lockdown measures.

The boxplot below shows that there was a significant reduction in concentrations at sites where data was available for all the years analysed when April and May 2020 is compared to those months in other years.

Fig 7. mean concentrations of NO2 in April diffusion tubes by year

Time variation

Analysis of the time variation of NOx shows a significant reduction at all hours of the day when compared to the same period in 2019. Again, while this cannot be be attributed solely to the lockdown measures, it is likely that traffic reductions arising from lockdown play a significant role.

Fig 8. Time variation of NOx at roadside sites in lockdown period by hour

A similar reduction is shown when plotted by day of the week.

Fig 9. Time variation of NOx at roadside sites in lockdown period by day

For NO₂ the reduction is less than for NOx, because NO₂ is not directly proportional to exhaust emissions, but is the result of photochemical reactions with NOx and other gases which happen in a timescale of hours.

Fig 10. Time variation of NO2 at roadside sites in lockdown period by hour

As explained earlier, PM₁₀ reductions are not as significant as NOx or NO₂ because lockdown measures only affect the locally generated component of PM₁₀ and not the regional or secondary components.

Fig 11. Time variation of PM10 in lockdown period by hour

Accounting for Meteorology

Analysis of the wind speed and direction from Bristol Lulsgate show that although wind speeds from the NE quadrant were higher in the 2020 lockdown period than in the baseline period in 2019, it is unlikely that this can explain all of the dramatic reduction shown, as the wind from the dominant direction for this period will be somewhat polluted by other urban upwind emissions. Further de - weathering analysis is required to establish this with more certainty and is summarised in the following section.

Fig 12. Wind roses for lockdown periods: Bristol Lulsgate

De - weathering ambient air quality measurements

The deweather functions of the openair package were used to remove the effect of the weather on concentrations of regulated pollutants measured by Bristol City Council’s and Defra’s monitoring sites in Bristol. The deweather package uses a boosted regression tree approach for modelling air quality data. This technique builds a statistical model of the air quality data and thereby takes account of the many complex interactions between variables as well as non-linear relationships between the variables.

Predictions of daily mean concentrations are derived from the modelled hourly means of data that are aggregated by site type. The accepted classifications of “Urban Background” (distant from busy road) and “Urban Traffic” (close to busy road) were used, as well as a classification of “City Centre” to cover the most polluted site, Colston Avenue, as this site represents the most polluted air in the city. Sites with a data capture less than 90% are not used in the analysis. The table below shows the classes for each of the sites analysed.

Table 3. Sites aggregated for model analysis

The chart shown below is the predicted “de - weathered” concentrations of three pollutants grouped into site types. The removal of the influence of the weather indicates that the reductions in measurements of traffic pollutants are probably not due solely to the weather. Reductions in traffic emissions due to COVID-19 lockdown measures is the likely explanation.

For NO₂ it is noticeable that at the city centre site (Colston Avenue) concentrations started to decline around the 16th March, and continued to decline further in the days immediately following lockdown on the 24th March. The post - lockdown decline was also apparent in the urban traffic and urban background site classes. A small increase in NO₂ at urban traffic and urban background sites was seen around 8th - 12th of April. This could be explained by a regional pollution episode that also increased PM₁₀ concentrations during the same period. Unfortunately operational issues at the city centre site mean that data was unavailable from the 9th April to the 15th April.

For NOx, a similar reduction was seen at the city centre site after the 16th April and a small reduction in concentrations at urban traffic sites is apparent.

PM₁₀ concentrations did decline in the immediate post - lockdown period but then rose steeply during the pollution episode over Easter weekend. The boosted regression tree models used in the de - weathering process take into account wind speed and direction but cannot account for elevated pollutant levels in the incoming air and hence are unable to remove the effect of regional pollution episodes such as the one that occurred at this time.

Fig 15. Deweathered chart by site class

The de - weathering process can also be applied to individual sites. The chart below shows how the de-weathered daily predictions vary from the mean levels for traffic pollutants for 2020 for each site. It is clear that from the date of lockdown, significant reductions have ocurred at all sites for NOx and NO₂ and that these reductions are not primarily driven by weather. It can be seen that deweathered concentrations start to rise from mid June.

Fig 15. Deweathered chart by site: NOx and NO2

The predicted changes in the deweathered NOx and NO₂ concentrations are summarised in the table below. The comparison is between the period 2020-01-01 - 2020-03-23 and 2020-03-24 - 2020-06-30. It is clear that Colston Avenue experienced the largest drop in NO₂ and NOx concentrations.

Table 4. Summary statistics for de - weathered NOx and NO2 concentrations

Plotting the weekly mean concentrations of the traffic pollutants NOx and NO₂ shows the effect of lockdown rules on weekly concentrations by controlling for the effect of the weather. It can be seen from the plot below that concentrations remained low throughout May despite some evidence that traffic levels are climbing since mid April. There is some evidence of rising concentrations towards the end of June, particularly at Colston Avenue.

Fig. 16. Predicted weekly mean NOx and NO2: De - weathered predictions