Abstract:

In order to bridge the gap between aggregated census data and geocoded data，different dasymetric mapping techniques were developed to disaggregate census data.The satellitemeasured DMSP/OLS nighttime light data was widely used for regional level mapping of socioeconomic activities due to its high temporal resolution，free availability and wide swath.However，due to the coarse resolution，data saturation and overglow effects of DMSP/OLS data，any application need to take into account the limitations of using this data source.Firstly，although the DMSP/OLS sensor has a nominal resolution of 1 km，this has been resampled from the 27 km native resolution of the sensor.The coarse resolution of the nighttime lights data lower the accuracy of dasymetric mapping.Secondly，the overglow effect due to surface reflection and scattering and refraction in the atmosphere results in the overestimation of lighted areas.Thirdly，the low radiometric resolution of 6 bits (i.e.the digital number value ranges from 0 and 63) results in data saturation over brightly light builtup areas.Vegetation indexes like Normalized Difference Vegetation Index (NDVI) are negatively correlated with the impervious surfaces and can be used for estimation of builtup areas.The incorporation of vegetation index (NDVI) can reduce the errors occurring in estimating human settlements from the DMSP/OLS nighttime light imagery due to data saturation and other factors.In addition，elevation is an important variable in population distribution modeling because most human settlements occur on lower elevation in China.This paper provides an approach for rapid and accurate estimation of population on a per pixelbasis using a integration of two coarse spatial resolution remote sensing data namely DMSP/OLS and SPOT NDVI，as well as the DEM data.The DMSP/OLS stable nighttime light image for year 2010 was firstly combined with SPOT NDVI data to develop a Human Settlement Index (HSI) image.Due to the complementary characteristics between DMSP/OLS data and NDVI，the resultant HSI image conveys more information than both the individual datasets.Then the DEM was involved in the HSI.The model for population density estimation was developed based on the significant linear correlation between the population and HSI with elevation effect correction.The HSI image for year 2010 was then used for modeling the population density of Zhejiang Province at a resolution of 1 km×1 km.The results showed that the error of population estimates was reduced by approximately 5 percent due to the elevation effect correction of HSI.8861% of total population in Zhejiang distributed in areas with an elevation lower than 500 m.The average population density of Zhejiang Province was 515 people/km2 and the mean relative error was 183%.The present paper provides a integrated approach for rapid and accurate estimation of population at regional scale using coarse spatial resolution images.The DMSP/OLS nighttime light imagery，SPOT NDVI and DEM data for any region can be freely downloaded.Hence，the approach developed in this paper can be used to map the population distribution in other regions of China