Abstract:

Population is an important social and ecological environmental factor, and mastering population information is beneficial to resource allocation and environmental management. This paper takes Nanjing's street-level administrative region as the basic object unit, and uses NPP/VⅡ RS night light data, POI data in big data and IBI of LANDSAT 8 satellite OLI image from the three scales of city level, suburb level and urbanization level, population size and population density. The results show that at the municipal level, the effect of population density simulation is better than that of the total population, POI data simulation is the best, with certainty coefficient of 0.87, followed by building land index of 0.81 and night light number of 0.77. At the suburban level, the fitting effect of the population density in the suburbs is better than that in the urban areas. In terms of urbanization degree, the power function relationship between the street population density of moderate urbanization and NPP/VⅡRS is the best, with the certainty coefficient of 0.99, while the logarithmic relationship between the street population density of low and high urbanization and POI is better, with the certainty coefficients of 0.65 and 0.44, respectively. The multivariate linear regression model of the total population and the data of three factors in the medium and low urbanization streets has the best effect, with the certainty coefficients of 0.91 and 0.78, respectively. It shows that the population simulation effect of moderately urbanized streets is the best. This study expands the breadth and depth of urban population estimation and can provide a reference for related work.

Key words: NPP/VIIRS image, IBI index, POI data, degree of urbanization, population modelling, street level units, Nanjing