Air quality during the corona crisis
Did you notice fresher air after the first measures were enforced to prevent the spread of the new coronavirus (SARS-CoV-2)? It started in the Netherlands from the beginning of March 2020. The roads became devoid of cars and many events were cancelled. The skies seemed to clear up and outdoor exercise felt healthier. It felt like an interesting period that formed a significant anomaly in our modern history. These experiences are mainly anecdotal and maybe you can relate. But does the data also reflect these changing times? Did the air we breath truly became more palatable?
The graphs below may demonstrate what we already suspect. But before we dive into the results, let us break down the first graph as it contains a lot of information. Each solid line represents the smoothed moving average of the measurements by each corresponding variable as labelled on the right side of this line. The filled area under the line indicates the value of each variable where the base of this area rests on the timeline. I have left out the y-axis for each variable as we will purely look at the trend. Now how do we determine the trend? This is done by linearly fitting several sine and cosine waves with different amplitudes and frequencies onto the measurements. These waves will recognise a seasonal trend and continue this over the whole time series. This is called a harmonic trend which is indicated by a dashed line. When the measurements of a variable deviate from this trend, you will see a discrepancy between the solid and dashed line.
Let us start at the top variable, which is the temperature measured by a weather station located at Schiphol. The harmonic trend (dashed line) is barely visible as it stays neatly on the smoothed measurements (solid line). This suggests that there are not really any deviations from the trend, following the expected seasonal curve. The same applies more or less for the other weather variables, which makes sense as weather on a short term is less affected by sudden human changes. However, the simple harmonic trend did not account for all deviations over the past 6 years as most weather patterns are subject to numerous unaccountable fluctuations. Near the very end there is an interesting break in the trend for the wind gusts and the horizontal view, more on this later.
Below the weather variables is the road traffic count, which is the amount of driving cars/trucks/etc measured by several sensors throughout Amsterdam. A very clear deviation from the trend is visible in the year 2020, where the road traffic drops well below the trend based on the years before.
Following the weather and road traffic, the measured nitrogen dioxide (NO2) concentrations by sensors in and around Amsterdam are displayed. The final variable is the NO2 as measured by the Sentinel-5 Precursor satellite (capturing data from mid 2018). You may have already seen it, but there seems to be a clear deviation from the seasonal trend in the year 2020! It is still somewhat hard to see, therefore, in the graph below the seasonality is removed by subtracting the smoothed measurements from the harmonic trend. Now it purely shows the difference between these lines. The filled area above and below the difference (solid line) indicates the distance from the smoothed measurements to the harmonic trend. The trend deviations become even more evident now. Where 2020 shows a clear dip in road traffic and NO2 concentrations, for both ground sensors as for the satellite data.
Could this dip in 2020 mean the corona crisis measures had a positive impact on our air quality? Possibly! But let not get too far ahead of ourselves. In 2017 another major trend deviation can be seen in the opposite direction. Namely at the ground NO2 concentrations and some weather variables (except for the missing satellite data). In an earlier blog post, it was already established that wind gusts (i.e. wind speed) and the horizontal view (i.e. how clear is the horizon?) has an inverse correlation with air quality. So faster winds mean reduced NO2 concentrations. We can see this in the trend deviations in 2017. The road traffic of course did not care, and continued its usual business.
Now fast forward to 2020. A very consistent drop in NO2 concentrations can be seen over all air quality measurement sensors similar to the drop in road traffic. Simultaneously, the wind speed seems to increase and the horizon becomes more visible (meaning less particles in the air). Is this, together with the improved air quality, a result of the corona crisis and its economic and societal changes? Or is it just a coincidence and did the weather instead of the corona crisis affect the air quality? I am yet undecided, but if you have any thoughts, please let me know!
In the next blog post we will bring the work done so far together by scientifically testing the prediction accuracy at more locations and time intervals as previously done, as well as quantifying the trend analysis from this blog post.