Optimization of Modelling Population Density Estimation Based on Impervious Surfaces

Population data are key indicators of policymaking, public health, and land use in urban ecological systems; however, traditional censuses time-consuming, expensive, laborious. This study proposes a method modelling population density estimations based on remote sensing Hefei. Four models with impervious surface (IS), night light (NTL), point interest (POI) as independent variables constructed at the township scale, optimal model was applied to pixels obtain finer distribution. The results show that: (1) (IS) can be effectively extracted by linear spectral mixture analysis (LSMA) method; (2) there is high potential multi-variable estimate density, an adjusted R2 0.832, mean absolute error (MAE) 0.420 from 10-fold cross validation recorded; (3) downscaling predicted scale using stepwise regression achieves more refined provides promising for rapid effective prediction interval years, support planning management.