COVID-19 and its environmental conditions

A friend of mine recently quipped that everyone seems to act like probability can be assigned two values 0 or 1. The same sort of logic seems to apply when it comes to talking about the environmental parameters which might affect the progress of COVID-19, such as temperature, humidity, and density. Many people seem to strenuously want to deny there is any plausible evidence that COVID-19 might exhibit seasonality. There is a fair amount of correlational work which suggests that there is an environmental factor shaping the spread and depth of COVID-19. And, we know three out of the four previous coronaviruses exhibit seasonality.

Well, I noticed this note on medRxiv today, Stability of SARS-CoV-2 in different environmental conditions. It’s a very short write-up of their experimental results. I don’t really know much about virology so I can’t evaluate it well, but you can see the figure above. As you increase the temperature the virus titer seems to drop much faster. At a very high temperature of 70 Celsius, they basically can’t detect anything after 1 minute.

Here is one of the better correlational analyses, using some sophisticated techniques, Causal empirical estimates suggest COVID-19 transmission rates are highly seasonal:

Nearly every country is now combating the 2019 novel coronavirus (COVID-19). It has been hypothesized that if COVID-19 exhibits seasonality, changing temperatures in the coming months will shift transmission patterns around the world. Such projections, however, require an estimate of the relationship between COVID-19 and temperature at a global scale, and one that isolates the role of temperature from confounding factors, such as public health capacity. This paper provides the first plausibly causal estimates of the relationship between COVID-19 transmission and local temperature using a global sample comprising of 166,686 confirmed new COVID-19 cases from 134 countries from January 22, 2020 to March 15, 2020. We find robust statistical evidence that a 1◦C increase in local temperature reduces transmission by 13% [-21%, -4%, 95%CI]. In contrast, we do not find that specific humidity or precipitation influence transmission. Our statistical approach separates effects of climate variation on COVID-19 transmission from other potentially correlated factors, such as differences in public health responses across countries and heterogeneous population densities. Using constructions of expected seasonal temperatures, we project that changing temperatures between March 2020 and July 2020 will cause COVID-19 transmission to fall by 43% on average for Northern Hemisphere countries and to rise by 71% on average for Southern Hemisphere countries. However, these patterns reverse as the boreal winter approaches, with seasonal temperatures in January 2021 increasing average COVID-19 transmission by 59% relative to March 2020 in northern countries and lowering transmission by 2% in southern countries. These findings suggest that Southern Hemisphere countries should expect greater transmission in the coming months. Moreover, Northern Hemisphere countries face a crucial window of opportunity: if contagion-containing policy interventions can dramatically reduce COVID-19 cases with the aid of the approaching warmer months, it may be possible to avoid a second wave of COVID-19 next winter.

To be clear. Does this mean weather/climate determine whether COVID-19 will spread or not? No. Rather, I think that weather/climate has some effect on the margin on the R0. I am not sure of the exact reason, but if the virus degrades much faster in hot climates, that could be one explanation of why spreading is more limited. It also does not seem to be the case that tropical countries are going to avoid mass healthcare crises. Rather, as these countries formulate policies to decrease R0, it may not be as long of a haul.

I believe that many are worried that if there is some relationship between temperature and COVID-19, people will think they are safe in a particular climate. The way to deal with this is not to ratchet up skepticism to an inordinately high extent. Rather, it is to be more clear and careful in how one presents the data.

Similarly, I think density has some impact. But, South Korea, Japan, and Taiwan show that density does not seal one’s fate.


16 thoughts on “COVID-19 and its environmental conditions

  1. I continue to be bemused by the low Japanese numbers. I am suspicious that they result from very low rates of testing. But if that is the case, I don’t know where they are hiding the bodies.

    I don’t believe social distancing. I have been on commuter trains in Tokyo, and I have opened less tightly packed cans of sardines. I have read several anecdotal accounts of how dissatisfied Japanese people are with their government’s ‘lax’ approach, but I am not seeing any consequences of that; quite the reverse.

    It’s a real conundrum.

  2. I think the whole debate about the “weather effect” being politically manipulated. Trump said it first, and he was largely right, so you can’t support his position by publicly approving.

    The other issue is the same as with masks (no panic, no sold out, keeping it for the health sector), those in charge in policy and mass media might not want people to “act on their own”, to not take measures seriously just because they rely on masks or the weather, even though every thinking individual should know by now that masks and the weather help, period!
    Anyone who denies these facts is lying or has no idea. Its just that both masks and the weather are no absolute protection from infection, they just reduce the likelihood of getting infected.

    There is some kind of intention with holding the information of the weather effect back, the plot thickens if they don’t communicate it soon. Either it is about manipulating the behaviour of the people or about being “the great winner” with “the right measures” afterwards, while playing the weather effect down.

    However, if people sit close together in kindergarten, in an oversized office or a party club, yes, they still will get infected. A shaded and air conditioned artificial environment won’t have the same effect as naturally aired open space.

  3. It will be over in the US before the frost free planting date:

    “Confirmed COVID-19 cases in the U.S. surpassed 100,000 Friday, doubling in just three days as the pandemic accelerates and the U.S. rolls out broader testing measures.

    “Confirmed U.S. cases passed 50,000 on Tuesday, up from 5,000 last week.

    “At the beginning of the month, there were roughly 100 confirmed cases in the U.S.”

    If the above is correct, and if the doubling period continues to be 3 days, then in about 5 weeks the pandemic will have infected everyone in the US (~330 million).

    The body count will continue to mount for two weeks after that. Mother’s Day (5/10) is the beginning of week 7. The frost free plantig date around 40 North is usually around that time.

  4. U.S. temp is 72 deg F, year around, in offices, homes, and cars.

    I confess that I perceive the mass of my fellow Americans (I travel around the U.S. a lot, mixing with all demographics and classes) to be obese diabetic indoor cats whose summer physical recreation consists of waddling from air-conditioned living spaces into their air-conditioned cars, to go graze from some other bowl of Purina Human Chow, I mean Cracker Barrel.

    So, I believe the stats that there is seasonality in U.S. regular-flu transmission, but I do wonder the mechanism. Is it perhaps down to school year/children vectoring?

  5. A couple of comments.

    Many coronaviruses display seasonality. This is not typically due to temperature. Most virologists and virology textbooks suggest seasonality is due to sunlight hours. More sunlight hours = more UV exposure. This is why viruses are seasonal in Gnome Alaska or Siberia whose summer temperatures are low.

    Most doctors think this is a null clinical point since this is a novel exposure to a immuno naive population. See smallpox through American Indians.

    Google Indonesian Covid and see what is happening in Jakarta. Japan cases increasing as we speak.

    Social distancing + home confinement. Required but not sufficient. For areas with increased population density or high dweller occupancy per bedroom (10 people sharing 3 bedroom apartment) central quarantine likely required.

    See this 5 min video
    “ Central Quarantine May Be Necessary to Control Covid-19”

    Based on slide deck from “Evolving Epidemiology and Impact of Non-pharmaceutical Interventions on the Outbreak of Coronavirus Disease 2019 in Wuhan, China”

    Male morbidly obese (BMI 40 or more) diabetics with questionable blood pressure control are at high risk, but the death of an infant today and multiple teens and twenties whilst at low risk are not invincible.

  6. My wife came across an article, unfortunately only in Japanese, that suggested that BCG vaccine is relevant. Specifically, the particular variant of BCG used in each country has a significant effect on response to Covid-19. This article was linked:

    I also found this article with a map of the different strains of BCG used in each country:

  7. “Is [the seasonal flu pattern] perhaps down to school year/children vectoring?”

    Absolutely not, because similar seasonal patterns are observed throughout the world that are definitely not based on the school year. For example, in Australia the flu season ends quite abruptly in October, and the number of cases is essentially the same in November, December, and January, despite schools not taking their summer break until sometime in December.

    “Most doctors think this is a null clinical point since this is a novel exposure to a immuno naive population.”

    That is a ridiculous claim, but most doctors are terrible at probability and statistics. While a population with no prior exposure and hence no immunity will certainly have a faster spread than one with immunity, other factors in how contagious a disease is still matters tremendously for the pace of spread.

    That’s true even if eventually the entire population gets it. It’s true that in a lot of statistical processes, like percolation, there really are only two end states, one of measure one and one of measure zero, though a surprisingly small change in probability of transmission can switch between the two states. However, even without changing end states, a small change in probability can have a large effect on the time to converge. It’s not like smallpox wiped out the entire continent of American Indians in a single season. Similarly, even if no one is immune to HIV and AIDS, that doesn’t mean that everyone is going to eventually get it, and that changes in behavior or other chances of transmission can’t have a dramatic effect.

  8. “If the above is correct, and if the doubling period continues to be 3 days, then in about 5 weeks the pandemic will have infected everyone in the US (~330 million).”

    First, the confirmed cases you’re using are cases that already were transmitted before any serious shutdowns or distancing occurred, and just have not shown symptoms or been tested until now. While that means that the true rates of sickness are higher than the confirmed cases, it also means that there is excellent reason to believe (based on measurable statistics that are proxies for amount of social contact like traffic, air pollution, and mass transit turnstiles) that the spread in the last week, that will show up in confirmed cases the next two weeks, is slower.

    It is thus highly improbable that the doubling period will continue at this rate even though next week. Eventually you also get some effect because of the number of people who have had it and are immune, though that does work much better with a slower inherent rate of transmission.

    To pretend that the doubling rate will stay the same is to make the same mistake as people who thought that we would never get it here at all.

    No pandemic ever stays exponential for that long. The Spanish flu never infected more than one-third the population for that reason.

  9. Google Indonesian Covid and see what is happening in Jakarta. Japan cases increasing as we speak.

    i read about indonesia every day. this does not falsify the idea that environment may have an affect. only a moron thinks that the env. is ALL that matters.

  10. Previously, literally just a few weeks ago, some were hoping tropical weather would more or less prevent mass community spread of the COVID 19 virus. Unfortunately, recent evidence from Philippines, Thailand, Malaysia, and Indonesia (possibly India???) has now invalidated that hope.

    However, one part is still possible: although tropical weather doesn’t prevent mass community spread, it does slow it to some significant extent. Instead of the standard, explosive exponential growth you see in China, Italy, Spain, Iran, etc, tropical nations may see a more gradual, linear increase. And so far, it looks like …..that’s the case?

  11. @AmBer, interesting if UV exposure->vit d (going outside) matters to reduce severity.

    In that case you could see widespread infection plus house quarantine with limited UV exposure and reduction in diet towards less D laden sources (more starch) could spike fatality rates a bit.

  12. I’m willing to buy what AmBer says. Hong Kong is almost on the Tropic of Cancer, only very slightly north, so the sun is virtually directly overhead on the summer solstice.

    Pale skinned people don’t need that much sun exposure to boost Vitamin D sufficiently – 15 minutes on bare arms and face is enough. It’s actually hard to get enough from diet, unless from fortified foods. I’m currently taking a small regular daily supplement because of that.

  13. @john
    “ That is a ridiculous claim, but most doctors are terrible at probability and statistics. While a population with no prior exposure and hence no immunity will certainly have a faster spread than one with immunity, other factors in how contagious a disease is still matters tremendously for the pace of spread.”

    There was a saying at Med school “Statistically significance does not imply a meaningful clinical outcome”. Watching my hospital resources being gobbled by Covid in real time is striking. This virus is really, really “sticky”, using a nontechnical term.

    The combination of an asymptomatic, highly-transmissible host with environmental factors such as high population density or high occupancy dwellings (many people per bedroom) is a toxic mix. These effects seem to overwhelm half-hearted social distancing attempts, or any environmental temp/UV benefit.

    Viral modeling using SIR compartment models is covered in a standard medical statistics class for doctors (not necessarily Biostats, medstats is different). The first part of the curve is exponential. The curve crests then slowly peters out with a long tail as potential hosts dry up and decrease in spatial and temporal distribution. The “flatten the curve” graphic has normal distributions. Helps with the visualization. Anybody who brings this up has a mere minor quibble.

    The two best lines from Razib are “I think that weather/climate has some effect on the margin on the R0……..It also does not seem to be the case that tropical countries are going to avoid mass healthcare crises”. He is on point.

    Jakarta today announced 62 deaths with 627 cases. A top-down approach using rough 1% fatality to case number would suggest a minimum of 6300 cases. Wuhan data indicate unacertained cases represent 60% of actual total so now your over 15,000 cases minimum. A bottom-up approach starts with UW Virology estimates that the first community case represents an underlying case load around 200. If you double every 3 days for a few weeks that is probably a realistic number. 21 days in gives 25,600 cases. Finding when that case happened in Jakarta might be hard, but Pittsfield, MA (top 3 density cases in US) reported first community acquired case March 3rd. I’ll stick by my estimates as ballpark range.

    Which brings me to this current paper. The data regarding Covid cases is minimally accurate. Firming up these numbers will take time. And they didn’t even run any sensitivity analysis wrt the Covid numbers!

    Many academics are trying to stake claims with Covid as a career building move, although it may be based on quicksand. Of course the authors are UChicago economists. The conclusion they make
    “it may be possible to avoid a second wave of COVID-19 next winter”,
    in the words of Dr Fauci is “aspirational”. Seems to ignore everything mankind knows about pandemic and viral behavior.

  14. Northern-Italy-is



  15. That’s good to know for sterilization purposes (if you were worried about take-out, you could put it in an oven pre-heated to 160 degrees Fahrenheit and effectively destroy any Covid-19 on the container in two minutes), but it doesn’t seem it helps us much in day-to-day life. Even at 37 degrees C ambient temperature (98 degrees F), it takes six hours before it makes a significant dent in virus amount.


Comments are closed.