COVID-19: How effective is my lockdown?

(Some of you may have noticed that I struggle with the tone of these blog posts. The intention is to keep them short and pleasurable to read. But of late, the topics I have written about are serious, and it’s a challenge not to trivialize the issue while maintaining a somewhat flippant tone. More ominously, I sometimes realize that I start taking myself too seriously! And then I go back and edit out the pompous prose. But I suspect sometimes that quite a bit of it seeps through into the final post, and I apologize for that. My feeling is that this is going to be one of those.)

COVID-19: How effective is my lockdown?

Vehicular emissions and staying at home

As I write this, stay-at-home orders of some sort are in effect in most large countries around the world. At the draconian end, surprisingly, are a few liberal economies of Europe, where you need curfew passes to step out. At the other end is libertarian Sweden, where people have been trusted to do what they feel is right. And then there is the US, drowned in a cacophony of polarized opinion, with their leader openly fomenting rebellion. Disaster movie makers have got things more right than we’ve ever given them credit for!

Leaders have had to choose between “lives and livelihoods” (as elegantly put by the Indian Prime Minister’s speechwriters). However they have chosen, the fact remains that at no point in history, have such far-reaching decisions been taken, based on so little evidence of their effectiveness. One of the critical unknowns is, what is the right amount of social distancing to flatten the curve while keeping the economy from going aground? Social distancing involves a slew of heterogeneous measures with varying efficacy in reducing person-to-person contact, and does not lend itself to easy grading. It would be nice if we could measure this objectively so that we can relate it to the effectiveness of COVID control in a country.

Grading lockdown severity

One obvious way of doing this is to use the regional mobility reports now published by the good folks at Google (yes! Google again). (https://www.google.com/covid19/mobility/) Google collects anonymized data from users and displays them as percentage change from January and February 2020. Data are reported under several categories such as “Retail and recreation”, “Parks”, etc., and provide a good guide as to how users are behaving in a community. But the relative importance of these individual categories towards social distancing is debatable. Perhaps lounging around in public parks and beaches isn’t so bad (as the Swedes clearly believe: up 84% from before COVID-19!)? In any case, as Google admits, these results depend on “user settings, connectivity, and whether it meets our privacy threshold” (whatever that means!). Moreover, the sample of users from whom these data are collected may not be representative of the entire community.

Vehicular emissions and staying at home

Now I am sure that many of you have thought of this, but weren’t motivated enough to go looking for the data. Though I am for the most part too lazy to do any form of useful organized activity, data dredging for me holds a strange fascination. So the question I asked was if vehicular emissions in a region serve as a good surrogate for compliance with stay at home rules. It seems that you can get data on air pollution for any city in the world from the World Air Quality Index (WAQI) project. The COVID-19 outbreak has brought out the benevolent streak in everyone, and the WAQI team is no different; they have made datasets for air quality available free for anyone to use (https://aqicn.org/data-platform/covid19/verify/4e1efc47-8886-4d50-a27d-002f6f96d925). Now, this is really a lot of data on all the major air pollutants for most of the major cities in the world. So my next task was to decide which one of the emission levels to use as the surrogate; I wasn’t about to do an exploratory analysis for all the pollutants! (I may be a sucker for data, but there are limits!). On reading up a bit, it seemed that it’s a toss-up between carbon monoxide (CO) and nitrogen oxide (NO_x), particularly NO₂ levels. I chose to go with NO₂ just based on one study which showed excellent correlation with black carbon levels, but I could have gone with CO too. (https://www.intechopen.com/books/air-quality-measurement-and-modeling/the-air-quality-influences-of-vehicular-traffic-emissions). Next, to find out if levels really changed following lockdown, I needed to have a suitable control. Since the WAQI folks were kind enough to provide data for previous years as well, I decided to use the previous year’s trends for comparison.

NOx populi

I used Wuhan as the starting point as most people would agree that the lockdown there was possibly the most stringent, and was effective in containing the disease. I chose other cities arbitrarily to reflect different perceived grades of lockdown, with Stockholm representing the lenient extreme.

Not surprisingly, Wuhan had a 58% fall in average NO_x levels during the lockdown phase compared to the same period in 2019, which rebounded with easing of restrictions. (Figure 1) India never ceases to surprise: Mumbai had a similar fall in emission (61%) as Wuhan. Of the European cities, Madrid, where restrictions were most strictly implemented, showed the most impressive reduction in emissions (51%). Paris was close showing a 43% reduction. A pronounced increase in emissions towards the later part of the lockdown in Paris may provide fodder for nudge theorists who apparently advised the British government (https://www.theguardian.com/commentisfree/2020/mar/13/why-is-the-government-relying-on-nudge-theory-to-tackle-coronavirus). (Figure 1)

Figure 1: Comparison of trends in NO₂ emissions between 2019 and 2020

London and New York showed less impressive reductions (35 and 28% respectively). (Figure 2) Most surprisingly, however, Stockholm, where no formal lockdown was imposed, showed a greater reduction (38%) in NO₂ emissions than London, Rome (35%) and New York. (Figure 3) The Google mobility reports for Sweden and the United States show similar reductions for major categories such as “Retail and recreation” and “Transit stations” (though the values for the UK suggest much greater restriction of movement). (https://www.google.com/covid19/mobility/) Perhaps this suggests that the trust reposed by the Swedish government on their people to do the right thing is not entirely misplaced.

Figure 2: Comparison of trends in NO₂ emissions between 2019 and 2020

Figure 3: Comparison of trends in NO₂ emissions between 2019 and 2020

It appears that air pollutants of vehicular origin can be used as a graded measure of the degree to which people abide by stay-at-home rules. This can in turn be used to determine the impact of lockdowns on the control of COVID-19. But there are several obvious caveats. Pollutants in air are affected to a great extent by wind speed, temperature and precipitation. I have not even tried to adjust for these (mainly because I don’t know how to). Emission levels will clearly not be affected by people moving on foot and congregating in large numbers (as has happened in some countries including India). And also, WAQI insists that I say that these data that I downloaded are not fully verified or validated.