For us scientists, it is tempting to throw our expertise into the fight against COVID-19. For spatial ecologists, this often means: predicting the risk of spread of the virus based on species distribution models, trying to identify its climatic niche – and thus where it is likely to show up as the season develops. After all, we have heaps of papers proving the validity of our modelling approaches for a whole range of species (and many diseases do indeed have a biogeography). This has already resulted in a myriad of preprints linking the spread of COVID-19 to climate, as summarized here.
Unfortunately, our ecological modelling techniques might not be sufficient in this case, for a variety of reasons. Many of these reasons are neatly summarized in the fantastic work from Chipperfield et al. , including the facts that 1) the distribution data for the virus is uncertain at best and highly biased at worst, due to undertesting, and 2) that the virus is far from its (climatic) equilibrium. These models might thus not only be unhelpful, but also perhaps counter-productive (think about the hopes that warm spring weather will easily get rid of the virus).
There is one problem with these models that I would like to stress a bit more, based on my own experiences as a microclimate ecologists: the climate we have at hand to model the distribution of the virus is just plain wrong. The coronavirus does not have much of a link with longterm averages in free-air temperatures measured by weather stations, and will thus not behave accordingly. We have been hammering on this nail for years now for a variety of organisms, like tundra plants (Lembrechts et al. 2019) and soil microbes (Lembrechts et al. 2020).
The problems might be even more acute for a virus than for these organisms, and they boil down to this: the spatiotemporal scale at which the organism operates, is so different from the climate data we have at hand, that there are likely several degrees of mismatch between the climate we use to model, and the climate as experienced by the organism.
It will likely not help to switch from long-term averages to current weather data, although that does take away the error caused by the fact that spring might have been much warmer this year than the average in many parts of the northern hemisphere. It will also not suffice to take into account urban heat island effects, even though the virus is spreading fastest in cities – which are indeed warmer than the average climate predicted by weather stations. The main issue lies in the fact that the coronavirus spends so little of its ‘lifetime’ in free-air: a lot of its time is spend either in human bodies, or indoors (where most of the transmissions take place) and on objects (gloves and mond masks, to name a few). None of these ‘habitats’ has temperatures only remotely near to what a weather station would give. Only for free-air transmission, this might be the case.
And this is an important note: humans are actively trying to make microclimates in their habitats – the indoor world – as comparable as possible across the whole world. We are warming our houses there where it’s cold, and cooling them where climate is warming. The effective result is a globally homogenised climate that is much more similar between human habitats across the globe than between these habitats and the weather stations closed to them.
So let that be a take-home message: modelling and predicting the climatic niche of any organism is tremenduously tricky, as long as we do not have the actual climate it experiences. This means that for so many organisms, we are far from ready to accurately predict their future distributions. And the coronavirus is just one example.