Estimating the Impact of Daily Weather on the Temporal Pattern of COVID-19 Outbreak in India

The COVID-19 pandemic has spread obstreperously in India. The increase in daily confirmed cases accelerated significantly from similar to 5 additional new cases (ANC)/day during early March up to similar to 249 ANC/day during early June. An abrupt change in this temporal pattern was noticed during mid-April, from which can be inferred a much reduced impact of the nationwide lockdown in India. Daily maximum (T-Max), minimum (T-Min), mean (T-Mean) and dew point temperature (T-Dew), wind speed (WS), relative humidity, and diurnal range in temperature and relative humidity during March 01 to June 04, 2020 over 9 major affected cities are analyzed to look into the impact of daily weather on COVID-19 infections on that day and 7, 10, 12, 14, 16 days before those cases were detected (i.e., on the likely transmission days). Spearman's correlation exhibits significantly lower association with WS,T-Max,T-Min,T-Mean,T-Dew, but is comparatively better with a lag of 14 days. Support Vector regression successfully estimated the count of confirmed cases (R-2 > 0.8) at a lag of 12-16 days, thus reflecting a probable incubation period of 14 +/- 02 days in India. Approximately 75% of total cases were registered whenT(Max),T-Mean,T-Min,T-Dew, and WS at 12-16 days previously were varying within the range of 33.6-41.3 degrees C, 29.8-36.5 degrees C, 24.8-30.4 degrees C, 18.7-23.6 degrees C, and 4.2-5.75 m/s, respectively. Thus, we conclude that coronavirus transmission is not well correlated (linearly) with any individual weather parameter; rather, transmission is susceptible to a certain weather pattern. Hence multivariate non-linear approach must be employed instead.