지리가중 포아송 회귀모형을 이용한 위성영상 기반 인구추정

This research estimated population counts at the census block group using geographically weighted Poisson regression(GWPR) and land use/cover data derived from satellite imagery in the Atlanta metropolitan area and then evaluated the accuracy of population estimation. GWPR was used to model count data such as population counts, to calculate positive population estimates, and to consider the spatial non-stationarity of local regression coefficients. The accuracy of GWPR was evaluated against ordinary least squares(OLS), spatial lag model(SLM), spatial error model(SEM), geographically weighted regression(GWR), and Poisson regression(PR). Results show that GWPR was a better regression model in the study area, but that GWR was a very close second. Also, the results indicate that spatial non-stationarity may have more impact on the accuracy of population estimation than spatial dependence in the study area. This research suggests that it should be necessary to carefully select a regression model considering statistical distribution, spatial dependence, and spatial non-stationarity for population estimation using satellite imagery.