Taking the mega city Shanghai as the research object,we fuse multi-source data such as nighttime lighting of LJ-01,points of interest,DEM,sub-spatial land use,rivers,etc.,to establish a multi-source feature database,and based on the GridSearchCV score evaluation,the XGBoost model was constructed to realize the 100 m×100 m population spatialization in Shanghai,and the accu-racy is compared with the Worldpop population dataset. The results show that the population distribution in Shanghai is characterized by multi-center distribution. LJ-01 nighttime lighting and POI data play an important auxiliary role in population spatialization. Al-though land use data is not of high importance to population spatialization,the spatial division of land use is equally important. It can reflect the differences in spatial functions of different land uses,and the accuracy of the results of this study (R2=0.98) is higher than that of the Worldpop population dataset (R2=0.78),indicating that the XGBoost model has high reliability and can provide ref-erence for other large urban population spatialization research.
关键词
人口空间化/珞珈一号夜间灯光/兴趣点/分空间土地利用/XGBoost模型
Key words
spatialization of population/nighttime lighting of LJ-01/points of interest/sub-spatial land use/XGBoost model
维普
万方数据