Acta Geographica Sinica ›› 2007, Vol. 62 ›› Issue (10): 1110-1119.doi: 10.11821/xb200710010

Previous Articles    

A Method of Spatialization of Statistical Population

LIAO Yilan1,2, WANG Jinfeng1, MENG Bin3, LI Xinhu4   

  1. 1. Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing 100101, China;
    2. Graduate School of the Chinese Academy of Sciences, Beijing 100039, China;
    3. College of Arts and Science of Beijing Union University, Beijing 100083, China;
    4. Institute of Urban Environment, CAS, Xiamen 361003, China
  • Received:2007-01-22 Revised:2007-07-02 Online:2007-10-25 Published:2010-08-04
  • Supported by:

    National "973" Program, No.2001CB5103; National "863" Program, No.2006AA12Z15; National Natural Science Foundation of China, No.70571076; No.40471111; Knowledge Innovation Program of the CAS, No.KZCX2-YW-3-8


Mapping distribution of population has arisen as an important issue in the fields of geographical and relative researches, due to the necessity of combining with spatial data representing socio-graphic information across various spatial units, such as to evaluate the total numbers of people at environmental health risks or died in natural disasters. However, most existing solutions to this problem focus on selection and quantification of influencing factors and rarely take into account the correlation among selected factors. And much expertise is needed in modeling process to formulate the relationships between influencing factors and population data successfully. It usually not only produces information redundancy but increases the complexity of the problem. This paper explores a novel approach to transform population data from census to grid by integrating genetic programming (GP), Genetic Algorithms (GA) and Geographic Information Systems (GIS). A set of natural and socioeconomic factors which contribute to population distribution are identified and quantified under GIS environment. And then GP and GA are severally applied to build and optimize the population model in the hierarchical form, allowing for the computation of the relevant population data error. The experiment proves that the proposed method performs much better than stepwise regression analysis and adapted gravity model approaches. The GP/GA-based method is the first to introduce such computational intelligence techniques as GP and GA to generate gridded population maps, hence it is a methodological innovation in interpolation of population data.

Key words: spatial interpolation, surface modeling, GP, GA, GIS, Shanxi