Assessing the influence of climate on wintertime SARS-CoV-2 outbreaks

Abstract

High susceptibility has limited the role of climate in the SARS-CoV-2 pandemic to date. However, understanding a possible future effect of climate, as susceptibility declines and the northern-hemisphere winter approaches, is an important open question. Here we use an epidemiological model, constrained by observations, to assess the sensitivity of future SARS-CoV-2 disease trajectories to local climate conditions. We find this sensitivity depends on both the susceptibility of the population and the efficacy of non-pharmaceutical interventions (NPIs) in reducing transmission. Assuming high susceptibility, more stringent NPIs may be required to minimize outbreak risk in the winter months. Our results suggest that the strength of NPIs remain the greatest determinant of future pre-vaccination outbreak size. While we find a small role for meteorological forecasts in projecting outbreak severity, reducing uncertainty in epidemiological parameters will likely have a more substantial impact on generating accurate predictions.

Fig. 1. Wintertime outbreaks in New York City.

Estimated and projected R₀ values (top plot) assuming a 35% and b 15% reduction in R₀ due to NPIs. Corresponding time series show the simulated outbreaks in the climate (blue) or constant (black/dashed) scenarios, with middle row plots assuming a 10% reporting rate and bottom row plots assuming a 3% reporting rate. Corresponding susceptible time series are shown in orange (susceptibles = S/population = N). Case data from New York City are shown in gray. Surface plots (top) show the peak wintertime proportion infected (infected = I/population = N) in the scenarios with c the constant R₀ and d the climate-driven R₀. e shows the difference between the climate and constant R₀ scenario. The timing of peak incidence in years from July is shown for the f constant and g climate scenarios. The difference between climate and constant scenario is shown in h. Points in c–h show the scenarios is a, b. Dashed line shows estimated susceptibility in New York based on ref. .

Fig. 2. Climate sensitivity of outbreaks across global locations.

a The climate effect on R₀ assuming a 35% reduction due to NPIs shown for August and December. b The effect of climate, changing susceptibility, and NPIs on peak proportion infected (infected = I/population = N), post July 2020, for nine global locations.

Fig. 3. Climate variability and wintertime cases in New York.

a Climate-driven R₀ and corresponding infected time series (infected = I/population = N) based on the last 10 years of specific humidity data for New York, assuming a 35% reduction due to NPIs. b The effect of changing susceptibility and NPIs on the coefficient of variation of peak incidence for simulations using specific humidity data from 2008 to 2018. Dashed line shows estimated susceptibility in New York based on ref. .

Fig. 4. Contribution to uncertainty in New York wintertime 20/21 peak size.

The relative importance of NPI efficacy [0–35%], immunity length (10–60 weeks), reporting (1–100%), climate sensitivity of the virus [−32.5 to −227.5], and interannual weather variability [10 years] in determining wintertime peak size. Immunity length and reporting rate collectively determine susceptibility, S.