GEOGRAPHICAL RESEARCH ›› 2009, Vol. 28 ›› Issue (1): 19-26.doi: 10.11821/yj2009010004

• Geo-information Science • Previous Articles     Next Articles

Study on the relationship between residential area from multi-source remote sensing images and multi-level population data

YANG Cun-jian1,2, BAI Zhong1, JIA Yue-jiang1, CHEN Xi1, DENG Li-li1   

  1. 1. Research Center of Remote Sensing and GIS Applications, Education Ministry Key Lab of Southwest Land Resources Evaluatiion and Monitor, Sichuan Normal University, Chengdu 610068, China;
    2. The Faculty of Geography and Resources Sciences, Sichan Normal University, Chengdu 610054, China
  • Received:2008-05-12 Revised:2008-09-05 Online:2009-01-25 Published:2009-01-25
  • Supported by:

    四川省青年基金项目(08ZQ026-047);国家自然科学基金项目(40771144);四川省教育厅重大培育项目(07ZZ029)

Abstract:

The relationship between the residential area extracted from multi-source remote sensing images and the population data at the three levels of city and prefecture, county and village in Sichuan province is discussed in this paper. This includes the following steps. Firstly, the rural and urban residential areas are extracted from Landsat TM images in Sichuan province, and the rural residential area and its building land are extracted from Quickbird Images in Juntun town, Xindu district, Chengdu City. Secondly, the residential areas for each unit of the three levels are obtained by overlaying and statistical analysis. Thirdly, the correlation relationship between the total residential areas and the total population, urban and town residential areas and the non-farm population, and the rural residential area and the farm population are analyzed respectively for city and county levels. The non-farm population strongly relates to the urban and town residential areas for city level with the correlation coefficient of 0.962 and county level with the correlation coefficient of 0.791.The non-farm population estimation models based on the urban and town residential areas are formulated respectively for the city and county levels by using regression analysis, whose judgment coefficients are respectively 0.926 and 0.625. Finally, the correlation relationship between the rural population, rural residential area and its building land are analyzed at the village level, and their correlation coefficients are respectively 0.806 and 0.825. The farm population estimation models based on the rural residential area and its building land are formulated by using regression analysis, whose judgment coefficients are respectively 0.65 and 0.68. It is shown that Landsat TM images are suitable for the estimation of the non-farm populations on a large scale, and Quickbird images are suitable for the estimation of the farm population on a small scale.

Key words: population, residential area, correlation analysis, remote sensing