Deconvolving Social and Climatological Influences on the Propagation Rates of COVID-19

Abstract:

While the seasonal cycle of SARS-CoV-2 is unknown, it is critical to estimate, quantify, and predict the social and climatological influences on the propagation rates of COVID-19. Our work attempts to model the COVID-19 pandemic spread in the continental United States from the beginning of March until June. We have gathered and processed a variety of datasets such as the European Centre for Medium-Range Weather Forecast (ECMWF) reanalyses to derive variables that may contribute to modeling the disease dynamics. To isolate the effects of the climatological variables from other social factors, we estimate a cross-sectional model as a starting point to learn the structural determinants of epidemic spread at the county level. The climatological variables are currently cross-sectional and relatively crude, but yield a model that predicts 80% of the variation in coronavirus growth. The model result shows that humidity and shortwave radiation affect the time-average COVID-19 case growth rate, while controlling for the population density, urbanized population, demographics, daily number of air travelers entering each county, the average vehicle miles traveled within the county, road network connectivity, containment policies, and state fixed effects. Given co-evolution of humidity during the warming seasonal period over which the analysis was performed, in subsequent work, we will determine whether such relationships are causal, or merely correlational within demographic, social, and climatological gradients. With comparable standardized coefficients, a one standard deviation change in population density, in-flow air passengers, and humidity produce more of a change in case growth rate than a one standard deviation change in other covariates. With granular weekly time series data, we were able to conduct a special Granger causality test, which allows variable lags of influence between causes and effects. Work has been done so far indicates that the COVID-19 outbreaks in the New York metropolitan area, Bay Area, Southern California, and Southern Florida are mainly caused by the number of air travelers entering these regions. Further investigation is required to assess the relative strengths of social and climatological influences on the COVID-19 spread for other regions in different periods of the epidemic.