Data

We considered 54 NUTS3 administrative regions (named départements in French) over 42 weeks (from March 23, 2020 to January 10, 2021). We did not consider all existing regions, but only the 54 for which the IPTCC can be computed. In order to evaluate the consequences of the climatic conditions on the pandemic, we used the weekly numbers of hospital admissions and deaths due to SARS-CoV-2 for the same 54 administrative regions considered. The data are official and publicly available at www.data.gouv.fr. The hospitalizations and deaths due to SARS-CoV-2 were consistently measured over the period and are thus more reliable than the series that report the number of infections, as the latter highly depend on the availability of tests and on the population’s willingness to get tested. Daily data are available since March 19, 2020, but we chose to use the weekly frequency in order to avoid a seasonal effect induced by lower reporting during weekends.

The meteorological data came from the 63 Météo-France stations. These stations are homogeneously distributed over mainland France and they were chosen in order to better represent the diversity of the country’s climate. The dataset provides a daily average for air temperature (measured in °C) and relative humidity (measured in %); these values were calculated as an average of the daily maximum and minimum values of temperature and relative humidity. With them, it is easy to compute the daily absolute humidity for each station, measured in g/m³, using the Clausius-Clapeyron equation⁸.

In order to analyze the potential relation between climate conditions and virus transmission, the IPTCC was created to characterize the potential for virus transmission according to climatic conditions. Following¹⁰, the IPTCC is a function of absolute humidity (AH), relative humidity (RH), and temperature (T), and the formula can be written as follows:

The IPTCC is thus maximal when the temperature reaches 7.5 °C and the relative humidity 75%. The IPTCC is available on a daily basis since January 1, 2020 for all meteorological stations. Then, for each of the 54 regions considered, we built a weekly indicator computed as the average of the daily IPTCC. Visualizations of the IPTCC through 2D and 3D representations are provided in the Supplementary Materials (Figs. ^A1 and ^A2).

As a preliminary step, the evolution of the variables over the period considered can be plotted. The figures for each of the 54 regions are reported in the Supplementary Materials (Fig. ^A3). To summarize them, Fig. 1 reports weekly hospitalizations and deaths (i.e. the total for the 54 regions) and the weekly IPTCC (the average for the 54 regions). The correlation between the three series is quite remarkable (see Table ^A2 for the coefficients). Most notably, the peaks in hospitalizations and deaths correspond to periods during which the IPTCC is high on average. Conversely, throughout spring and summer, hospitalizations, deaths, and the IPTCC are low. We also notice that the IPTCC is more volatile than the epidemiological series. Lastly, the evolution of the IPTCC seems to precede those of hospitalizations and deaths, which is probably due to the time between contamination, the incubation period, and a possible worsening of the disease.

Comparison of the time series of total hospitalizations and deaths and the average IPTCC, March 23, 2020 to January 10, 2021.

To go beyond this first graphical analysis and investigate how climatic conditions affect the spread and severity of SARS-CoV-2, we developed a time series analysis. This was done in two steps. Firstly, we demonstrated a causal relationship between the IPTCC and the hospitalization and deaths, and then we evaluated the responses of the latter to an increase in the IPTCC.