Our study has some limitations. Firstly, the SEIR model was set up based on a number of necessary assumptions. For example, we assumed that no super-spreaders exist in the model, but there is currently no supportive evidence. Secondly, the accuracy of the estimation model depends largely on the accuracy of the parameters it used, such as incubation period. With more precise parameters obtained as the epidemic progresses, our estimation model will also be more precise. Our estimates of the reproductive number from 3.1 to 0.5 are based on previous studies and experience from SARS control. However, this measure may change substantially over the course of this epidemic and as additional data arrives. Besides, using a fixed Rt value in each phase may incur potential bias because Rt is essentially a dynamic parameter. Thirdly, these estimated data may not be sustained if unforeseeable factors occurred. For example, if some infections were caused by multiple exposures to animals, these estimates will be exposed to a big uncertainty. Fourthly, the epidemic trend shows great difference between Wuhan and Hubei Province and regions in mainland China outside Hubei Province according to the NHC reported data. It is thus inappropriate to generalize the estimations in Wuhan to regions in mainland China outside Hubei Province. The dynamics model for the other locations in mainland China remains to be developed and specific parameters need to be redefined. Lastly, we do not provide model fit information in the current study. SEIR model is a prediction model forecasting the number of infections in the future. The data corresponding to actual situation in the future cannot be determined and this makes model fitting almost impossible during the outbreak. We would carry out model fitting according to the real data in pace with more information and knowledge about the characteristics of COVID-19 and the epidemics in the future.