| Spatialized population grid data could overcome the shortcomings in statistical demographic data, such as low spatial resolution, poor time comparability, inconvenient visualization and spatial analysis. It also can precisely delineate spatial pattern of population distribution, and facilitate integration with varied natural and cultural elements of data for further analysis and application. In order to meet the demands of various applications, enlarge the applied domain of spatialized population data, and improve the accuracy of spatial population data, it is necessary to establish a systematic spatial population database with various grid-scales by substantial research on spatial population modeling approach at multi-scale.AnHui province is selected as the study area, this study mainly focus on the spatialization modeling of county-level population census data in 2010 at different grid scale by using multi-source data fusion analysis, remote sensing and GIS technology. To improve the model precision of remotely sensed land cover data for stimulation of spatial population distribution, a processing of population secondary partitioning is carried out. Then, these subareas are used to build population models, respectively. Combining with land use data and nighttime light (DMSP/OLS), the study presents a method, reclassification of urban residential areas, for improving the spatial distributing of census data. Based on the population regionalization, multi-factor linear regression (MFLR) model and geographically weighted regression (GWR) model are employed to integrate the reclassified urban residential land-use data with the rural residential land-use data. This study establishes different population spatial data sets at 1km,5 km and 10 km gird scale. Finally, study results are investigated by comparing grid population and census data in county-level and rural town-level, with statistical analysis and spatial analysis. The results are listed as follows:(1) The 1-km grid scale test results show that the average relative error of population regionalization modeling is 24.77%, with a decrease of 16.28% compared with that of the whole province unified modeling. This study indicates that, with the processing of secondary population partition, it is helpful for regional unified modeling in the population subareas with similar characteristics in demographic distribution. The accuracy of population regionalization modeling is much better than that in the whole province.(2) By using nighttime light (DMSP/OLS) as an information source indicating the urbanization level, the urban residential land-use data is reclassified. This processing is helpful for improving the shortcomings of traditional statistical approaches in revealing the variations in population distribution within a single land-use type.(3) Based on the reclassification of urban residential land-use data, different spatialized population data sets at lkm,5km and 10km gird scale are established by MFLR model and GWR model. By comparison of two models’precision at each scale, this study shows that the accuracy increases with the grid scale by using the MFLR model, and the highest accuracy is acquired by the 10-km grid data sets. While for the GWR model, the accuracy decreases as the increasing grid scale, and the highest model accuracy is acquired at 1-km scale. Overall, the GWR model has a higher accuracy than the MFLR model when taking into account the geographic location and local modeling.As a whole, this study presents a new method for improving the precision of population spatialization model by using nighttime light to reclassify the urban residential land-use. And it also reveals the relationship between the spatialized population grid data and the grid scales by researching the transformation of grid size. This study could provide a scientific basis for the production and application of population spatial data, and also provide a reference of spatialization for other types of statistical data in the future. |
PDF Full Text Request