Fig. 1: Accuracy of the predicted risk of COVID-19 symptom onset by the extended and original Weight Kernel Density Estimation (WKDE) models, under the scenario of Wuhan lockdown.

The predicted onset risk is a standardised value between 0 and 1, indicating risk relative to the highest predicted risk among all locations on the date for which the risk of symptom onset is predicted, hereafter termed “the prediction date” (see Methods section for further detail). The prediction accuracy is defined as the percentage of the confirmed cases in the areas in which the predicted onset risk was >0.8²² on the prediction date. The time interval denotes the period between the base date and the date of prediction. The horizontal line in the box denotes the median, while the lower and upper edges of the box represent the respective first and third quartiles. The lines emanating from the box upwards and downwards represent the respective maximum and minimum values.