초록 열기/닫기 버튼
This paper explores a field-based population estimation using a dasymetric mapping technique that incorporates land cover data as a means to redistribute the original census population value into a surface grid. The three methods reviewed by Eicher and Brewer (2001) are tested in Athens-Clarke County, Georgia. USA. Using the three (Binary, three-class, limiting variable) methods and the conventional choropleth method, I estimate total populations of 54 census block groups to quantify how well those models reflect real population distribution. Bivariate regression analysis is used to show how estimation errors vary across cases. All three dasymetric methods perform significantly better than conventional choropleth method. In terms of RMS error and mean coefficient of variation, limiting variable method performs best followed by binary method and three-class method. The correlation coefficients for dasymetric methods are high, ranging from 0.916 to 0.94. Also, simple form of error distribution maps is used to visualize how estimation errors are spatially distributed for each estimation model.
