Development of high-resolution population database from ASTER-derived urban area map

This paper presents a methodology of estimating population at 15-m resolution from ASTER-based urban extent map and existing population grid data. We proposed two methods of estimating population: dasymetric mapping and regression model. Experiment for Tsuchiura City, Japan, and Pasig City, Philippines, showed visually good distribution of the estimated population at 15-m resolution. The data represented population in a better way by representation corresponding to urban structure at 15-m resolution. The result indicated that there were spaces to be improved by introducing new data explaining human activities and iterative machine learning between the two methods proposed.