Containment measures limit environmental effects on COVID-19 early outbreak dynamics

Abstract

Environmental factors are well known to affect spatio-temporal patterns of infectious disease outbreaks, but whether the rapid spread of COVID-19 across the globe is related to local environmental conditions is highly debated. We assessed the impact of environmental factors (temperature, humidity and air pollution) on the global patterns of COVID-19 early outbreak dynamics during January-May 2020, controlling for several key socio-economic factors and airport connections. We showed that during the earliest phase of the global outbreak (January-March), COVID-19 growth rates were non-linearly related to climate, with fastest spread in regions with a mean temperature of ca. 5 °C, and in the most polluted regions. However, environmental effects faded almost completely when considering later outbreaks, in keeping with the progressive enforcement of containment actions. Accordingly, COVID-19 growth rates consistently decreased with stringent containment actions during both early and late outbreaks. Our findings indicate that environmental drivers may have played a role in explaining the early variation among regions in disease spread. With limited policy interventions, seasonal patterns of disease spread might emerge, with temperate regions of both hemispheres being most at risk of severe outbreaks during colder months. Nevertheless, containment measures play a much stronger role and overwhelm impacts of environmental variation, highlighting the key role for policy interventions in curbing COVID-19 diffusion within a given region. If the disease will become seasonal in the next years, information on environmental drivers of COVID-19 can be integrated with epidemiological models to inform forecasting of future outbreak risks and improve management plans.

Keywords: Absolute humidity; COVID-19, pathogen growth rate; Climate; Global analysis; PM 2.5; Pollution; Population size; Temperature.

Graphical abstract

Fig. 1

COVID-19 growth rate (a) and stringency of containment measures (b) in regions experiencing COVID-19 outbreaks in different periods. The bold lines represent the fit of a generalized additive model, the shaded area its 95% confidence band. The figures report data for regions where outbreaks occurred between February 27 and May 31, 2020, as before that date data were sparse (<50 regions experienced outbreaks between January 22 and February 26).

Fig. 2

Temporal variation of the importance of variables in explaining COVID-19 growth rate. We fitted regression models starting from regions experiencing outbreaks up to February 27, until we included all regions experiencing outbreaks up to May 31, 2020 (n = 586 regions). The partial R² statistic (variance explained by each fixed effect) was taken as a measure of the relative importance of variables. a) Temperature calculated using the 30-days period; b) temperature calculated using the Δ14 days period (see Supplement 1, Fig. S1 for details).

Fig. 3

Temporal variation of the relationships between independent variables and COVID-19 growth rate (standardized coefficients). We fitted regression models starting from regions experiencing outbreaks up to February 27, until we included all regions experiencing outbreaks up to May 31, 2020 (n = 586 regions). The plot includes temperature calculated using the 30-days period; the pattern was identical if a Δ14 days period was used (see Supplement 1, Fig. S1 for details). Shaded areas represent 95% confidence bands. When confidence bands do not cross the horizontal broken line (0 threshold), the effect of a given variable is statistically significant (P < 0.05).

Fig. 4

Variation of COVID-19 growth rate in relation to local mean temperature (30-days period), air pollution (PM 2.5) and stringency of containment measures. Partial regression plots from mixed models of COVID-19 mean daily growth rates fitted for local outbreaks occurring up to March 15 (n = 195 regions), April 15 (n = 529 regions) and May 15 (n = 577 regions) are shown. The shaded areas are 95% confidence bands.