Some 4 billion people are currently affected by lockdown measures connected with the coronavirus pandemic. It is therefore not surprising that monitoring by Europe's Sentinel 5P satellite reveals a strong decrease in air pollution globally. Especially nitrogen dioxide (NO2) levels are a good indicator of air contamination caused by industrial production and road traffic.
But what seems obvious can be deceptive. Long-term DLR analyses show that nitrogen dioxide pollution can be very low at regional levels due to locally prevailing weather conditions, such as the cold snap Europe experienced that transported clean polar air far southward just at the time preventive measures were being instituted.
Only a combination of satellite measurements, in-situ data, and computer models can allow scientifically reliable confirmation of coronavirus effects.
Figure 1: Temporal progression of NO2 pollution over Lombardy from 1 January to 15 April as recorded by the GOME-2 instrument on the MetOP-A satellite. The tropospheric vertical column data shown in µmol/m² give the number of NO2 molecules per unit area. The blue curve shows the development in 2020, the red curve the mean value for the same time interval in the years 2007 to 2019. The bright red area is the variability (standard deviation) during that period.
The graph shows the weekly mean value of nitrogen dioxide pollution over Lombardy for the period from January to April. The blue curve gives the values for the last few weeks, the red curve the mean values for the past twelve years. The red area shows how much the values varied in those years (the standard deviation). Starting in early March, the time of the lockdown in Italy, the blue curve permanently exits the area with long-term comparable measurements. From 8 March on, the Italian government decreed a fast sequence of quarantine measures to slow down the spread of COVID-19. Public events were cancelled, churches and museums closed, and shortly thereafter also businesses and many production facilities.
The measurements shown here are from the GOME-2 sensor on Europe's MetOp-A satellite, which has provided a global overview of pollution for many years. The satellite records not only ground air pollution but the entire air column. With increasing altitude winds can quickly transport or dilute contaminants, or bring them in from distant regions. For that reason, a apparent rapid, unambiguous assessment of measurement data is not possible.
In order to exclude possible weather effects, data from 25 local ground stations in Lombardy were also evaluated. The red curve in the second graph show the mean values for six years during the relevant months. The same trend as before is apparent, but that is still only circumstantial evidence. The influence of weather events like a cold snap or precipitation that washes out air contaminants cannot be excluded.
Figure 2: Measurements of ambient air at 25 ground stations in Lombardy from 15 Feb. to 15 April shown as mean values for 2020 (blue curve) and for the years 2013 to 2019 (red curve). The grey-shaded area is the standard deviation, the pink area the minimal and maximal range of values for these years. (Source ARPA Lombardia)
In order to eliminate the influence of the weather, EOC scientists used a computer model in the last step of the analysis. This model can calculate chemical processes in the atmosphere for various air constituents taking into account existing wind and weather conditions.
The scientists began their analysis with several years' worth of mean pollution emission values to clarify the mean normal situation. This was done to ensure that the model does not incorporate any the effects of coronavirus-related measures when determining pollution emissions, but it does include, hour by hour, the prevailing meteorological conditions. The normally expected NO2 concentrations were thus calculated for each of the 25 ground stations involved. The modelled normal situation was then subtracted from the current ground measurements. Indeed, starting on 8 March the values are clearly in the negative range. The model uninfluenced by coronavirus considerations systematically yields higher pollution levels despite the same weather conditions.
Shutting down road traffic and economic activity has thus led to a reduction of NO2 pollution of about 20µg/m³, which is equivalent to a 45 percent drop.
Figure 3: Difference between NO2 concentrations measured at 25 ground stations in Lombardy between 15 Feb. and 20 April using the simulation of the air quality model POLYPHEMUS/DLR. A reduction of NO2 concentrations is clearly evident from 8 March on.
The various pieces of the puzzle needed for this analysis – satellite measurements, ground station measurements and simulation with a computer model – together produce a clear picture: the lockdown introduced by the Italian government on 8 March results in noticeable improvement of air quality in Lombardy by significantly reducing road traffic and economic activity.
It is clear that the satellite measurements closely match the ground measurements and the model calculations. With satellites it is possible to make global assertions about pollutant concentrations. Although the influence of coronavirus in Lombardy is without doubt scientifically verifiable, the situation in other regions of the world is less obvious. There, current measurements are in some cases within the standard deviation range of past years. Whether this is due to particular local weather conditions or a late institution of lockdown measures cannot be determined without further analysis. Despite the uncertainties described here a global comparison does show definite departures from the long-term mean values.
In contrast to the long-term analysis, the high-resolution comparisons shown here are based on data for the years 2019 and 2020 from the Copernicus Sentinel-5P (Precursor) satellite. For the study, 1.2 trillion individual measurements collected over the course of two months by the TROPOMI sensor were processed with DLR retrieval algorithms and quality tested.
In 2020 the measured tropospheric nitrogen dioxide values (NO2) over Southeast Asia were considerably lower than those for the previous year. The reduction, in some cases over 40 percent, was not limited to large cities. Daily fluctuations caused by the weather were minimized by calculating monthly mean values.
Asia in February 2019 and February 2020:
Concentrations of NO2over Europe between 16 March and 15 April 2020 are also lower by comparison, for example in the Benelux countries, western Germany, northern Italy and southern Great Britain. For hotspots like the Po basin, Madrid, Paris, Milan and Rome reductions in tropospheric NO2 exceeded 40 percent.
Europe, North and South America between 16 March and 15 April in 2019 and in 2020:
The situation in North America between Mid-March and April 2020 resembles that in Europa. Here the reduction in tropospheric NO2 on the East Coast, especially in the region around New York, is about 30 percent.
In South America decreases of 50-60 percent for Buenos Aires and Sao Paolo and of 20-30 percent for Guayaquil and Santiago de Chile have been measured.
India shows for Mumbai and Delhi a decline since the end of March in the order of 40-60 percent.
The NO2 reduction in the troposphere in South Africa is even more than 75 percent in the region around Johannesburg.
Among other clients, the TROPOMI NO2 data generated at DLR are used in the S-VELD project, which is financed by the Federal Ministry of Transport and Digital Infrastructure to analyse the effect of traffic emission on air quality in Germany.
This analysis contains modified Copernicus Sentinel-5 Precursor data (2019-2020) processed by DLR.
The operational GOME-2 data is provided by DLR in the framework of the EUMETSAT AC-SAF project.