Impact of climate change on SARS-CoV-2 epidemic in China

buffer at the boundary of China’s administrative regions, and then extract the weather data via the mask. (2) Conversion of hourly data into daily data. For the TP, the daily data were summed by the hour; for other factors, the daily data were averaged by the hour; and the data processed resulted 606 views. Meanwhile, the units of the data were converted: the original units of SP was Pa, which was converted to kPa after dividing by 1,000; the original units of H was kg�kg-1, which was converted to percentage after multiplying by 100; the original units of UV was J�m- 2, which was converted to J�m-2�h-1 after obtaining the average by the hour, and was converted to kJ�m-2�hour-1 after dividing by 1,000. (3) Integrating population density and weather factors. SARS-CoV-2 is closely related to population distribution and can be used to obtain weather factor data under the effect of human activities. The resolution of the weather data was 0.25˚ × 0.25˚, which was much lower than that of the population data (30’’ × 30’’). The weather data were resampled to the popula- tion resolution to avoid missing values at the boundary caused by multiplying the weather and population data. The LandScan population distribution was used as the weight, and the weather data were calculated to obtain the weighted weather data. The weather data featured a lower resolution, which eliminated the effect of mixed pixels and included the characteristics of human activities, as shown in Eq (2). Wit ¼ X g2iog Wgt ð2Þ

where ω is the population distribution weight, W the weather variable, i the provincial admin- istrative region, g the grid in administrative region i, t the specified time, and gt grid g at time t. Finally, ArcMap’s zonal statistics were used to calculate the weighted weather data for each provincial administrative region. A comparison of the effects before and after data weighting is shown in S2 Fig. The SP, TP, UV, and WS were highly consistent before and after processing, whereas T and H were relatively sensitive to the population density, particularly H, indicating that human activities imposed a greater effect on these variables. (4) Elimination of cross-sectional differences (decentralization). The cross-sectional data refer to data obtained from different study areas at a specific time and correspond to a one-dimensional dataset composed of other spatial objects simultaneously. They allow a phe- nomenon at a particular time to be investigated and highlight the differences between spatial objects. The salient feature of cross-sectional data is their high dispersion. Additionally. cross- sectional data highlight individual differences and typically show irregular instead of actual Table 1. Data information. Name Original Units Units Short Name Abbreviation Param ID Number of data (scenes) Format Dataset 2-meter temperature K ˚C 2t T 167 2,424 NetCDF ERA5-Land hourly data from 1950 to present Total precipitation m mm tp TP 228 2,424 NetCDF Surface pressure Pa kPa sp SP 134 2,424 NetCDF 10-meter u-component of wind

m�s-1 m�s-1 u10 WS 165 2,424 NetCDF 10-meter v-component of wind m�s-1 m�s-1 v10 166 2,424 NetCDF Downward UV radiation at the surface J�m-2 kJ�m- 2�hour-1 uvb UV 57 2,424 NetCDF ERA5 hourly data on pressure levels from 1959 to present Specific humidity kg�kg-1 % q H 133 2,424 NetCDF Note: The selected 6 weather factors are included. https://doi.org/10.1371/journal.pone.0285179.t001 PLOS ONE Impact of climate change on SARS-CoV-2 epidemic in China PLOS ONE | https://doi.org/10.1371/journal.pone.0285179 July 27, 2023 5 / 21