Abstract: Census data with the spatial scale of administrative district cannot reveal the spatial differences sufficiently. Meanwhile, this kind of statistical population data cannot be integrated and analyzed with most of other gridded geographic datasets. People in pixels provide an effective way to solve this problem. We established multivariate statistical models for population in 1 KM pixels based on the correlation relationships between population and land use types. The urban population density, traffic conditions, DEM and total amount control were used for model correction. Forty counties with township population data from east, west and middle of China were chose for precision verification. The errors of the spatial population data in these counties are between 4.5% and 13.6% and most of them are less than 10%.

Key words: China, GDP, grid, spatial data, 2005, 2010.