Roles of meteorological conditions in COVID-19 transmission on a worldwide scale

Biqing Chen; Hao Liang; Xiaomin Yuan; Yingying Hu; Miao Xu; Yating Zhao; Binfen Zhang; Fang Tian; Xuejun Zhu

Considering that there is a latency stage from the day one get infected to the day being confirmed, a time delay of the day COVID-19 was confirmed from the day weather data were collected needs to be taken into consideration. As it is reported that the latency period for COVID-19 is 3~7 days on average and 14 days at most, four time points delay of virus infection were taken into consideration, that is, weather data were collected on the day, three days before, seven days before, 14 days before collecting the epidemiological data. At first, each meteorological variable was plotted against the confirmed new case counts for the Wuhan dataset, with four time delays display on one plot. Only one city Wuhan was chosen for illustrating the time delay effect because it is the original city where SARS-CoV2 was first uncovered, there could not be any imported cases for Wuhan, which might obscure the correlation between weather and virus transmission. A Loess regression interpolation approach was adopted to visually identify the relationship between meteorological variables and confirmed new case counts. After choosing the appropriate time delay, data from the discovery dataset were fitted into generalized linear model or non-linear model (basically polynomial and inverse models) according to the indentified relationship by Loess regression. Each of the four meteorological variables was fitted into models solely, and then two or three variables were combined together to fit complex models.