Generation of high resolution population distribution map in 2000 and 2010: A case study in the Loess Plateau, China

An adequate knowledge of population distribution in the long term is increasingly being used in both science and policy. In this paper, we proposed a GIS based approach using remotely sensed land use, land cover, night light emissions, and NDVI data to redistribute the aggregated population statistics at township level into a regular 100m * 100m grid in 2000 and 2010 across the core area of Loess Plateau, China. Nighttime light emission data from the DMSP satellites was firstly combined with NDVI to generate a Vegetation Adjusted Nighttime Light Urban Index (VANUI) map with less saturation and more variation within inter-urban area. The land use (or land cover) data was then reclassified and rasterized to provide a 100-m resolution map. Then, VANUI was matched to the land use classes across the research area. The entire township units of the research area were divided into three different zones according to their population density. Stepwise regression method was used to derive the model of relationship between census population counts (at township level) and land use area and night light indicators for each zone. Based on these equations, we redistribute the statistics of every township unit into the 100m * 100m grid. All the relationship models of each zone were seen to be good with a relative high R2 and low SEE and the generated population distribution map is spatially explicit and quantitatively detailed. In summary, the method here is illustrated to be effective to model the population distribution in long term with a high resolution and the population distribution maps in 2000 and 2010 in Loess Plateau is expected to greatly assist related researches in the region.