Causes of reduction in airborne NO2 and CO emissions in Europe during the COVID-19 crisis

Road traffic reduction during the spring 2020 lockdown caused significant NO₂ decrease.

The most presumed causes of decrease in observed air pollution are reduction in road traffic and people staying at home due to the imposed travel bans. In this analysis, changes of NO₂ and CO emissions are compared against the three variables:

• The average population in Europe;
• Road density in Europe;
• Vessel routes density in the Mediterranean Sea and the black sea.

This research was supervised by Prof. Dr. Bodo Bookhagen and Dr. Taylor Smith.

Main air pollutants and their sources in Europe

EEA’s Air Quality 2019 report illustrates the total emissions of pollutants in the EU-28, indexed as a percentage of their value in the reference year 2000. This report further gives an overview of each sector’s contribution to total emissions for all chosen pollutants in the EU-28, for 2017.

Trends in EU-28 emissions, 2000-2017 (as a % of 2000 levels): SO_x, NO_x, NH₃, PM₁₀, PM_2.5, NMVOCs, CO, CH₄ and BC. Also for comparison EU-28 gross domestic product (GDP) is shown (GDP, expressed in chain linked volumes as a % of 2000 level. This shows that there were fewer emissions for each unit of GDP produced per year consecutively. Figure from EEA’s Air Quality 2019 report .

Contribution to EU-28 emissions from the main source sectors in 2017 of SO_x, NO_x, primary PM₁₀, primary PM_2.5, NH₃, NMVOCs, CO, BC and CH₄. The road transport sector was the most significant contributor to total NO_x (nitrogen oxides is a generic term for the mono-nitrogen oxides NO and NO₂) emissions and the second largest contributor for NO₂ emissions was the energy production and distribution sector. The highest and 50% contributor to total CO emissions was the commercial, institutional and households sector and the second largest contributor to total CO emissions was the transport sector. Figure from EEA’s Air Quality 2019 report

The contributions from different emission source sectors to air pollutant concentrations and air pollution impacts depend not only on the amount of pollutants emitted, but also on the proximity to the source, emission/dispersion conditions and other factors such as topography. Emission sectors with low emission heights, such as traffic and household emissions, generally make larger contributions to surface concentrations and health impacts in urban areas than emissions from high stacks.

Processing candidate variables

Population Density

The estimated average number of persons for a square kilometer in 2020 in Europe [NASA SEDAC at the Center for International Earth Science Information Network].

OpenStreetMap Europe data was used to extract the road network of Europe. First, the “.pbf” file was split into sub-parts using OSM converter, with the option; using a border-box to limit the geographical region. Then the linestring geometries of the split files were filtered and the “.pbf” file was converted to a GeoJSON using GDAL.

#UK is the name of a single .pbf file that was split before and 'lines' at the end instructs GDAL to extract only the line geometries from the file.
ogr2ogr -f GeoJSON UK.json UK.pbf lines

Then these datasets were merged in Python, in a way that the final dataset doesn’t have duplicated roads because of the clipping boundaries considered. Each distinctive road is identified by a unique ‘Osm id’.

Road network of Europe extracted from OSM data; the highways tagged as a motorway, trunk, primary, secondary, tertiary, unclassified, and residential in OSM [Processed dataset].

Next, the road densities and the vessel densities of geolocation points used to create the pollutant gas maps and population maps were calculated. The number of roads passing through a 7 km radius from each of these points was calculated using Rtree spatial joins and spatial index, in python.

#Defining radius as 7 km (7km ≈ 0.06306°).
Radius = 0.063063

#Adding this buffer to all geolocation points (grided points)
grid_df['geometry'] = grid_df.geometry.buffer(Radius)

#Importing Geopands as gpd
import geopandas as gpd

#Taking the spatial joins of the two data frames created above (grid_df has point geometries as the geometry and roads_df has Linestring geometries as the geometry.)
grided_roads = gpd.sjoin(grid_df, roads_df, how='inner', op='intersects')

#Taking the sum of roads crossing the boundary of each buffered point.
roads_num_df = grided_roads.groupby('Point_ID').size().reset_index()

Road density of Europe; the number of roads within a 7 km radius from each considered geolocation point. Each distinctive road is identified by a unique ‘Osm id’.

Vessel routes density in the Mediterranean Sea and the Black Sea; the number of vessel routes within 7 km radius from each considered geolocation point. Each distinctive route is identified by a unique ‘Osm id’.

Processing air pollution variables

In this analysis, the time between 1^st of March 2020 to 31^st of April 2020 is the time where strict confinement policies were applied across Europe, and thus is considered as the ‘lock-down period’.

The following three variables are used to indicate the changes in NO₂ and CO emissions. Data was preprocessed in Google Earth Engine (GEE) Code editor and further analysis was done with python:

• Average NO₂ and CO emissions during Covid-19 in Europe.
• The slope of the time series change of NO₂ and CO emissions during the lock-down period.
• Percentage change of NO₂ and CO emissions during the lock-down period compared to the average values of the previous year in the same period.

The two videos below show the weekly average NO₂ and CO emissions from 1^st of January to 31^st of July 2020. Areas with no data values are displayed in black colour. Codes used to create the videos can be found here [NO₂, CO].

Weekly average NO₂ emissions of Europe from 1^st of January to 31^st of July 2020, measured in mol/m².

Weekly average CO emissions of Europe from 1^st of January to 31^st of July 2020, measured in mol/m².

Mean NO₂ emission volumes during the lock-down period, during the previous two months and the same maps for 2019 emissions. The emission volumes have dropped significantly compared to the 2019 levels in all of the areas in Europe during the first two months of the year, except in Milan and Turin in Italy. During the lock-down period, NO₂ emissions have dropped in these two cities as well.

</center>

Mean CO emission volumes during the lock-down period, during the previous two months and the same maps for 2019 emissions. The last two maps show the difference in CO mean values of the two periods in 2019 and 2020 respectively. Difference maps show reductions in Europe wide CO emissions in each following month.

The change (trend or slope) of NO₂ emission data during the lock-down period. Negative slopes indicate reductions in NO₂ emission volumes and positive slopes indicate increases in NO₂ emission volumes.

The change (trend or slope) of CO emission data during the lock-down period. Negative slopes indicate reductions in CO emission volumes and positive slopes indicate increases in CO emission volumes.

Percentage change of average NO₂ emissions during the lock-down period compared to the previous year. Percentages less than zero indicate reductions in NO₂ emission volumes and percentages greater than zero indicate increases in NO₂ emission volumes compared to the average values of the previous year in the same period.

Percentage change of average CO emissions during the lock-down period compared to the previous year. Percentages less than zero indicate reductions in CO emission volumes and percentages greater than zero indicate increases in CO emission volumes compared to the average values of the previous year in the same period.

Mean NO₂ and CO emission volumes over the vessel routes in the Mediterranean Sea and the Black Sea during the lock-down period.

The slope of NO₂ and CO emissions over the vessel routes in the Mediterranean Sea and the Black Sea during the lock-down period. Negative slopes of NO₂ emission in the coastal areas in the Mediterranean Sea indicate reductions in NO₂ emission in these areas.

Percentage change average of NO₂ and CO emission volumes over the vessel routes in the Mediterranean Sea and the Black Sea during the lock-down period compared to the previous year's same time period.

Assessing the relationships between the variables

Correlation statistics between emission variables and road density, vessel routes density and population density. Mean CO and NO₂ are measured in mol/m² and road density and vessel density are measured in the number of roads/number of vessel routes within a 7 km radius from each geolocation point. Population density is measured as the number of people living in a 7 km radius from each geolocation point.

Road density has the highest correlation with the pollutant gas variables. All three candidate variables have the highest correlation with the ‘Mean NO₂ volume’ compared to other emission variables.‘Percentage change of NO₂’ has the next highest correlation with all three candidate variables and they are negative because as expected the NO₂ emissions are lower during the lock-down period than in the previous year. In places where there is a high traffic density or a high population density, reduced NO₂ emission volumes are observed.‘Slope of NO₂ volume’ has the next highest correlation with the candidate variables and these correlation statistics are negative. ‘Slope of NO₂ volume’ and the three candidate variables together provide the best explanation for the reduction of NO₂ emissions during the lock-down period because the slope (coefficient of linear time series trend of air pollutants) during the lock-down period is a measurement that shows how rapidly the emission volumes have dropped during the lock-down.

Time series analysis

NO₂ emission time series change in Lisbon-Portugal, Madrid-Spain and Roam-Italy, and RD of these places. NO₂ emissions have dropped significantly during the lock-down period in all these places. Volumes start to drop from mid March and start to increase again in late May.

NO₂ emission time series and the smoothened series change in Barcelona-Spain, Venice-Italy and Istanbul-Turkey, and VD of these places. NO₂ emissions have dropped significantly during the lock-down period in all these places. Similar to the above figure, this drop has also started in mid March but it continues until the end of July with some irregular spikes in between.

CO emission time series change and the smoothened series in Ilford-United Kingdom, Barcelona-Spain and Frederiksberg-Denmark, and PD of these places. CO emission has increased during the lock-down period in all these places. The emission volumes have started decreasing again from early May.

Key takeaways

• Road traffic reduction has been the primary cause of the decrease in NO₂ emissions during the lock-down period in Europe.

• The places that have higher decreases in NO₂ volumes during the confinement period have higher road, vessel, and population densities.

• Road, vessel, and population density data do not provide strong reasoning for the reduction in CO emissions due to several reasons:

  1. CO is a highly dispersing gas, and therefore, it might not be a good approach to compare the geographical location of the sensed CO gas and its sources

  2. People staying at home during lock-down could have created events that have both positive and negative effects on changes in CO emission levels.

• It would add further insights to this analysis if the locations of the industries that contribute to a high level of NO₂ and CO emissions in Europe could be checked against the changes in CO and NO₂ emission levels in these locations during the lock-down period because 50% of the CO emissions and 8% of NO₂ emissions are coming from commercial institutions.