A data-driven assessment of early travel restrictions related to the spreading of the novel COVID-19 within mainland China

Alberto Aleta; Qitong Hu; Jiachen Ye; Peng Ji; Yamir Moreno

Population flow. We obtained migration data from Baidu Quanxi, an open platform based on Baidu Location Based Services (LBS) that provides information about the population flow within China [23] . The dataset comprises two types of data: the Baidu migration index and the Baidu migration ratio. The former is a number proportional to the number of individuals leaving each region. To obtain the proportionality constant, we averaged the amount of individuals leaving Wuhan between Jan. 1st and 10th and compared it to the value of 502,013 estimated by Wu et al. [15] . The second part of the dataset contains the fraction of individuals going from region i to region j, p ij . As such, it is possible to estimate the number of individuals going from region i to j by multiplying the total outflow of the region by the corresponding p ij . The outflow of the individuals is available for the years 2019 and 2020, while the p ij values only for 2020. Nevertheless, as it can be seen in Fig.1 in the main text, it represents a good proxy of the situation in 2019. Indeed, even though some populations were closed after Jan. 23rd, 2020, using these values for 2019 correctly describes the return of the population that had left for the Spring Festival. To gauge the effect of the chosen parameterization of the model, we have repeated the analysis with several values reported in the literature [7, 24] . As expected, if the incubation period is larger or the basic reproduction number is smaller, the overall number of cases decreases. Conversely, when using large values of R 0 or longer generation times, the amount of infected individuals increases. In any case, the mobility model is quite independent from the dynamics of the disease, since the correlation is almost always the same in all cases considered.