Preliminary evidence for seasonality of Covid-19 due to ultraviolet radiation

Much attention has been paid by news reports, publications, press releases, preprints and white papers concerning containment measures for the Covid-19 pandemic, caused by the highly infectious and contagious virus SARS-CoV-2 in the human-to-human transmission chain (Anderson et al., 2020). From the onset of the epidemic, the naming of the new disease was not reminiscent of the previous, related disease SARS (Holmes, 2003), perhaps generating a certain degree of incertitude. The term Covid-19 has obfuscated its similarity and kinship to SARS for the public, with an initial tepid response in the Western world, as the disease started spreading globally. If Covid-19 had not been initially contrasted to influenza, responses might have been stronger and more efficient (Vetter et al., 2020). This is all history. Now, there is an acute need for swift, simple measures that can limit global contagion (Hunter, 2020), and ways to understand the epidemic response to climate factors.

There have been multiple reports and, admittedly, a hope that Covid-19 might exhibit a seasonal pattern, that will slow down the spread of the epidemic (Bukhari & Jameel, 2020). In fact, the initial confusion with flu-like symptoms might have contributed to this propitious hypothesis, obscuring the dangers for the epidemic in areas such as sub-Saharan Africa (Araújo & Naimi, 2020). A host of studies have examined seasonal fluctuations and regional climate parameters, including temperature (Auler et al., 2020; Shi et al., 2020), humidity (Luo et al., 2020), a combination of both (Mecenas et al., 2020; Şahin, 2020; Sajadi et al., 2020; Wu et al., 2020), along with wind speed (as an indicator for long-range airborne spread) (Chen et al., 2020), precipitation levels (Oliveiros et al., 2020), some in conjunction to population density (Liu et al., 2020; Pedrosa, 2020). Yet, only minor correlations with temperature, even controlling for a lag of detected cases, (Ma et al., 2020; Shi et al., 2020) or precipitation (Sobral et al., 2020) have been observed. Press releases reporting these studies state that seasonality can be a significant factor with summer days approaching. However, at the time of initial writing and review in parallel with the appearance of these reports, Northern and Southern hemispheres were at the March (‘Spring’ for the North) Equinox or just after it, confounding real-time analysis and derivation of solid conclusions that connect the epidemic with seasonal fluctuations to a certain extent (Bannister-Tyrrell et al., 2020), with little or no evidence found for local temperature effects, e.g. Spain (Briz-Redón & Serrano-Aroca, 2020) or Japan (Ujiie et al., 2020). Previous findings on the diminished presence of SARS-CoV on surfaces at higher temperatures, attenuating transmission, have been pointed out (Bannister-Tyrrell et al., 2020). Additional evidence for potential seasonality of Covid-19 arises from four other endemic coronaviruses, despite a significant uncertainty in epidemiological parameters, with the risk of a false impression of containment over the summer period (Neher et al., 2020).

Sunlight, as the surrogate and key element of seasonality, emits at infrared (IR), visible and ultraviolet (UV) wavelengths. While IR is responsible for heat transmission with low-energy electromagnetic (EM) waves, UV is responsible for high-energy EM waves – even at low temperatures, e.g. at high altitude or high latitudes¹. Therefore, any assessment for the seasonality of Covid-19 should consider UV emission (typically UV-A or -B for sunlight), also hinting at the targeted deployment of artificial UV-C rays (naturally absorbed by the atmosphere and the ozone layer)². In addition, it should be noted that extensive UV exposure can occur even in overcast skies, as this EM wavelength can penetrate water-droplet clouds that diminish UV radiation by as little as 15% (Calbó et al., 2005). Finally, there is also a correlation between vitamin D and UV, which implicates UV in human physiology (Asyary & Veruswati, 2020; Engelsen, 2010; Juzeniene et al., 2010; Tamerius et al., 2011). Statements concerning the absence of UV effects on Covid-19 seasonality over a short period of time and in an area with multiple confounding factors cannot be conclusive, as previously shown (Yao et al., 2020). More comprehensive studies now appearing suggest otherwise, suggesting a role for UV (Carleton et al., 2020) (Box 1). Consequently, there should be more focus on the effects of UV radiation, as a parameter that might slow virus transmission in open spaces under natural light and solar UV exposure as summer days are longer, in the period examined so far (Chiyomaru & Takemoto, 2020)³. Artificial sources of UV (-C) light for built environments that can be installed to eliminate the spread of infectious viral particles in public spaces can be deployed. Known examples include the use of UV to eradicate SARS-CoV-2 from banknotes, buses and hospitals. Indeed, studies on the related SARS-CoV report virion inactivation using 254nm UV light, heat or chemicals (Darnell et al., 2004).

It is known that UV light (i.e. UV-C) affects nucleic acids in a detrimental fashion for organisms, including viruses and bacteria, with ssRNA viruses being more sensitive than other viral types by 2- to 3-fold (Tseng & Li, 2007). Germicidal UV is standard practice for cell culture and sterilization protocols (Li et al., 2005; Rae et al., 2008). Studies on the H1N1 influenza virus report inactivation with low doses of 222nm UV light, suggesting potential low dose-rate UV-C radiation solutions for reducing the spread of viral infections across indoor public spaces (Welch et al., 2018). Interestingly, a recent study on the effect of meteorological factors on influenza virus in Northern Europe has found some correlation between UV and spread of the epidemics during 2010–2018 (Ianevski et al., 2019), thus pinpointing one of the elements of seasonality, albeit for a different infectious disease. It is not known how resilient SARS-CoV-2 can be under strong, short-wave UV light, yet application of technologies, such as appropriate LED or UV-arrays, no matter how expensive, could in principle be deployed in crowded spots, such as mass transport systems, with a certain urgency.

There is limited information for non-human hosts of coronaviruses in the wild. It is remarkable, however, that a survey of populations for eight species of seabirds in the Southwestern Indian Ocean – including Réunion and Madagascar – did not detect any coronavirus presence in 338 samples, of which only a handful (39) were collected during the winter (Lebarbenchon et al., 2013). This serendipitous observation might suggest that the summer sun keeps these populations in the wild healthy and protects them from viral infections, however mild. More research into that direction will definitely be worthwhile, as the analogy of this finding cannot be readily translated for public health: neither the role of UV radiation nor seasonality can be inferred from these remarkable observations (C. Lebarbenchon, personal communication, University of Réunion). How wildlife responds to coronavirus infections with regard to seasons and climate change is not entirely understood at present.

Regarding seasonality, our guide should be coronavirus epidemiology in general (Neher et al., 2020; Nickbakhsh et al., 2020), and the SARS/MERS outbreaks and their containment in particular – maybe more so than the flu. While models accept, simulate and interpret seasonality (Carleton & Meng, 2020; Kissler et al., 2020; Li et al., 2020), they do not usually refer to UV radiation in an explicit manner (rate, i.e. strong sunshine, or duration, i.e. day length). In the case of SARS and MERS, these epidemics did not spread widely, due to containment and mitigation strategies, including limited quarantine, which is not comparable to the scale we are experiencing today globally (Nickbakhsh et al., 2020). For instance, seasonal fluctuation for MERS was primarily limited at latitudes as low as 20⁰ – in Saudi Arabia with high temperatures – although most infections occurred within hospitals (Zumla et al., 2015), and over a short period, at 40⁰ latitude, e.g. in South Korea (Kim et al., 2017). We need to know how SARS-CoV-2 responds to UV and what its viability is on surfaces (van Doremalen et al., 2020), under simulated sunlight, which is currently unknown. Any chance of seasonal variation will provide valuable time in the North and ring an alarm in the South, as modelling the variation of UV radiation and temperature suggests a decline during summertime (Merow & Urban, 2020), with this study being the first following our initial hypothesis connecting UV radiation with Covid-19 growth rates (Karapiperis et al., 2020).

Seasonal fluctuation can be a factor that will limit Covid-19, and the reason may not be due to high temperatures – as the media, but also a number of epidemiological studies, keep on reporting – but sustained UV radiation (Merow & Urban, 2020), as demonstrated in comparative studies (Lytle & Sagripanti, 2005). One piece of evidence is that, until the March Equinox 2020, Covid-19 was imported into the Southern Hemisphere but did not achieve epidemic status rapidly until recently (Figure 1). It is challenging to compare spread in real time for seasonality (Martinez, 2018), with 90% of the human population residing in the Northern Hemisphere. Using data from Peru, a country with the highest UV levels (Suarez Salas et al., 2017), there is a correlation between elevation and incidence, along population density (Figure 2), as latitude comparisons are not currently feasible. High-elevation countries such as Bhutan and Nepal exhibit low incidence rates for Covid-19, while exceptions due to travel patterns include countries as diverse as Andorra and China, all in the Northern hemisphere.

Figure 1. Pictorial representation of Covid-19 across three sets of countries, over time.

Data were imported from ECDC and refer to the period 31-Dec-2019 to 5-May-2020, © ECDC (2005–2019). Cross-checks with other datasets were performed, e.g. from HealthMap (Xu & Kraemer, 2020) (https://www.healthmap.org/covid-19/). Visualization was facilitated by Charticulator (https://charticulator.com/app/index.html, © 2018 Microsoft Corporation). Dates are shown on the x-axis, per day (vertical green line signifies the March equinox). Countries are classified in three groups on the y-axis (separated by horizontal lines, in orange) and listed alphabetically to avoid clashes over latitudes – top: Northern (73), middle: Southern (10) for contrast to North, bottom (70): equatorial (between 23N and 23S), 153 in total. Daily counts normalized per 100K population (shown in cells), to display spread (light: low values, dark: high values). It can be seen that at most countries in the North, there is some attenuation, except Russia, Turkey, the UK and the USA at present (dark-blue, at the top block). While many other factors contribute to the spread, it remains to be seen whether the pandemic will switch across North and South, indicating a seasonality pattern, with UV as a salient element compared to the more variable parameter of average daily temperature – and despite the fact of a large number of recorded cases. This will crucially depend on mitigation measures, travel restrictions and various other public health strategies around the world.

Figure 2. Covid-19 incidence in Peru compared to elevation as a proxy for UV radiation.

A cursory view of the situation using elevation, as seasonality is currently difficult to track (see main text). We assess incidence in Peru, as the country with the highest UV radiation levels and high elevation contrasts (Suarez Salas et al., 2017). Reported Covid-19 cases in 25 departments (provinces) (from: https://covid19.minsa.gob.pe/sala_situacional.asp), 5-May-2020. Despite the fact that case reports have a resolution per province, there is a trend for high incidence in the low-lying East and coastal plains, while high altitudes with increased UV radiation levels (Suarez Salas et al., 2017) have lower counts. Left panel, generated by ClusVis (Metsalu & Vilo, 2015): Heatmap of PCA-based clusters, using correlation as a metric and complete linkage clustering for elevation, Covid-19 incidence (per thousand population) and population density, the latter used to capture crowding effects. Unit variance scaling applied to columns, scale shown (top right). Annotations: green for high, yellow for low – thresholds: high density >50/km2, high incidence >568pM (world average 13-May-2020), high elevation >200m. Annotations on the left block of the heatmap (also shown on the country map, right panel): 19/25 provinces consistent with high density/incidence and low elevation (6), and low density with low incidence/high elevation (10) or the reverse (3). There are just 6/25 provinces of low density and high incidence/elevation (inconsistent with UV-vs-incidence), four of which being coastal. Two clusters in rows and columns shown, delineating low-density/high-elevation provinces (bottom cluster, row-wise), and connecting Covid-19 incidence/density (right cluster, column-wise). Right panel: With no resolved data for UV per province, we used elevation of capital city as a UV estimate (i.e. altitude – selected from a number of other metrics). Annotation blocks (from left panel, see above) are shown per province; those consistent with UV-vs-incidence have a grey border (19), those inconsistent with UV-vs-incidence have a cyan border (6). The pattern provides, perhaps for the first time, some evidence of correlation between UV and Covid-19 incidence across altitudes and not latitudes – with UV being a strong candidate for an additional driver of seasonality at lower altitudes elsewhere (summer months).

As summer is approaching in the North, with increased UV radiation due to increased day length, there is a chance for a decelerating pace of transmission in the open. Artificial means to irradiate public spaces might be an option.

Data availability