Particulate matter on New Years Eve

Charting air pollution at the dawn of 2019

5 min readJan 31, 2019

When the clock strikes twelve the air fills with screams and bangs of firework. The light of the moon and stars fade away by the colourful flashes and sparkles of thermally excited sodium, calcium, barium, strontium, and copper that fill the dark sky. People on the street and in houses toast to a bright future, big dreams, and long life, while watching the celestial show of pyrotechnic explosions. While sipping my champagne I started pondering about the soot, smoke, and fumes left behind by the joyful chemical reactions.

Particulate matter

The Earths atmosphere contains many different pollutants such as ozone, nitrogen dioxide, sulphur dioxide and particulate matter. Particulate matter¹ (PM) are microscopic solid or liquid particles suspended in the air (fine dust). Depending on size, 70–95 % of fine dust is estimated to be of anthropogenic origin². PM concentrations are commonly used as proxy indicator for air pollution, which I will also do in this blog.

Particulate matter is divided in different categories. Particles that have a diameter of 10 microns or less are called PM10, and are small enough to penetrate into the deep regions of the lungs. Smaller particles with a maximum diameter of 2.5 microns (PM2.5) can even cross the lung barrier and end up in the blood stream. No wonder that chronic exposures to PM are associated with elevated risks of developing cardiovascular and respiratory diseases, and increased mortality. According to the WHO it affects more people than any other pollutant³.

Dutch sensor readings

In the past couple of years quite a few people have bothered building their own particulate matter sensor⁴, and putting the data online. The website https://opensensemap.org/ shows that there is an increasing amount of available sensors, mostly in Europe. I have used this website’s API to download the data between December 31st and January 1st. The data contains almost 1.5 million measurements of PM10 from 1182 active outside sensors worldwide. Retaining only the countries with more that ten active PM10 sensors results in the following table.

Table with the number of outdoor PM10 sensors active during New Years Eve per country.

Germany contains by far the most sensors, while The Netherlands still hosts 51 outdoor PM10 sensors. What are the Dutch PM10 concentrations when the clock strikes twelve?

The Dutch PM10 concentration median filtered on consecutive epochs of 15 minutes. This graph also shows a small peak around 19 h and a very large peak rising sharply at midnight. The shaded area indicates the data between the 10th and 90th percentile. There is quite a large variance in PM10 concentrations.

The graph shows that there is quite some variance between the sensors, but a midnight peak can very clearly be discerned. The peak rises far above the WHO recommended 24 hour mean limit⁵, but luckily fades away in about four hours or so. This graph is quite similar to the figures from the Dutch National Institute for Public Health and the Environment (RIVM)⁶, or the visual created from luftdaten.info⁷.

The European context

Since I also have data from other countries, it is pretty straightforward to plot that data as well and ask ourselves the question: Is the midnight peak typically Dutch?

The PM10 concentration median filtered on consecutive epochs of 15 minutes for countries with more than ten active outdoor PM10 sensors. This graph shows a large peaks for Germany and The Netherlands, while other countries show smaller peaks.

The graph shows interesting behaviour. Germany and The Netherlands have sharp peaks at midnight, while Bulgaria has a very broad peak. Other countries show small peaks (Austria and Italy) or no peaks at all (France).

It is also interesting to overlay the PM10 concentrations on a geographical map. Therefore I need to do two things:

Calculate the average PM10 concentrations over one hour after midnight, keeping in mind that the local time may differ between countries;
Assigning sensor readings to particular regions based on the sensor locations and the regional boundaries. In this case I have used the geocode standard NUTS level 1 from the European Union⁸.

The plot below shows the resulting PM10 concentration averaged over one hour after midnight.

The average PM10 concentration in Europe one hour after midnight. The concentration in some regions can be based on very few sensors only. Note the high values in southern Germany and The Netherlands. In particular the region Saarland has interesting high values. Only sensors in countries listed in the table above are taken into account.

The figure shows that the highest concentrations PM10 are found in southern Germany, part of Austria and The Netherlands. Quite a few other regions in Europe seem to have way lower concentrations.

Conclusions

The figures above clearly show sharp peaks starting at midnight. It is known that the cheap fine dust sensors may deviate from professional equipment readings, in particular in highly humid conditions⁹. However, there is no doubt in my mind that these high PM10 concentrations have been caused by the soot and smoke of the fireworks. People with respiratory diseases suffer from high concentrations of air pollution, also when the exposure is short. The effects in healthy individuals are not known, but assumed to be negative¹⁰.

There is however quite some variance between the sensors as becomes clear from the wide range between the 10th and 90th percentile in the first figure. This suggests that there is locational variance in the PM10 concentration: some sensors show high readings other low readings. This could quite likely be caused by local variation in the amount of lit fireworks. It could be that there is less fireworks in rural areas, or the firework tradition is location dependent¹¹. Plotting the concentrations on a map indeed shows regional variation.

One would say that almost 1200 sensors with together 1.5 Million data points is a lot; some would even use the term ‘big data’. And indeed, about eight hundred sensors will probably reliably reflect the German regional variation. But fifteen sensors are clearly not enough to draw accurate conclusions for a country like France.

When I look up from my screen, the sound of pyrotechnical explosions have faded away, and the firework fumes have disappeared in thin air. The PM10 concentrations are indeed very high at many different locations, far above the daily limit. But not for too long; at sunrise the smoke is gone, PM10 concentrations are mostly back to normal and all what is left is memories and a transient cough of the chemical spectacle.

Happy 2019!

Acknowledgements

I would like to acknowledge Roeland Roeterdink and Salomon Tetelepta for their suggestions for improvement.