With the emergence of spatio-temporal big data and the development of data mining technology, scholars have made great progress in estimation of urban population’s socio-economic characteristics on fine spatio-temporal scale, which makes up for the shortcomings of the traditional data with poor timeliness and low spatial precision effectively. However, related studies are restricted in urban informatics, computational social science, geography and other fields, which have not been widely applied to the resource allocation based on spatio-temporal distriution of urban population. This problem restricts the in-depth practice of urban fine management. In order to promote information communication and interdisciplinary dialogue, this paper prospected the research prospects and challenges according to the analysis of data source, estimation methodology, result verification and model evaluation, mature fine scale population characteristic distribution data set. The results found that spatial regression, machine learning and neural network can be used in the fine spatio-temporal scale estimation of urban population characteristics. Among many socio- economic characteristics, research on fine scale population distribution is the most developed, which has transformed from single data source to multiple data sources, from general grid to grid with spatial information, and from low spatio-temporal resolution to high spatio-temporal resolution. The gender, age, race and education of urban population can also be predicted based on social media data, but it is not mature. The application of fine scale urban population characteristics data is rich, which includes disease prevention and control, poverty identification, climate change mitigation, urban risk management, urban development backtracking, urban social spatial structure analysis, etc.
