A building volume adjusted nighttime light index for characterizing the relationship between urban population and nighttime light intensity.

The brightness of nighttime lights (NTL) has been proven to be strongly related to population density and thus has been widely used for population estimation from national to county levels. However, the limited availability of fine-grained census data makes an accurate assessment of the pixel-level relationship between NTL intensity and population a challenge. Consequently, pixel-level population estimation bias based on NTL intensity has been rarely investigated. Using fine-grained census data, we quantitively evaluated the correlation between urban population and NTL intensity over the core areas in Shanghai city, China. We also proposed a simple index called building volume adjusted nighttime light index (BVANI) for better characterizition of the relationship between urban population and NTL intensity. Our results found that pixel-level NTL intensity has a minimal correlation with population and the exclusive use of NTL intensity will not improve our ability to model population. While the assessment of BVANI shows that BVANI has an inverse relationship with pupation and the relationship between BVANI and population follows a power-law distribution. The relationship strength with population can be significantly improved by using BVANI with a correlation coefficient of 0.60. We believe that BVANI can be used as an important modeling factor for mapping fine-scale urban populations. • NTL intensity has a minimal correlation with population at pixel level. • An index called building volume adjusted nighttime light index (BVANI) was proposed. • The relationship between BVANI and population follows a power-law distribution. • BVANI reduces pixel-level population estimation biases significantly. [ABSTRACT FROM AUTHOR]