The Study For Scaling With Socioeconomic Data

In the filed of Geographical Information Science, spatial scale is the premise of Geo-data process and analysis and the foundation of spatial analysis and decision. Therefore, It is the most significant problem of Geographical Information Science, which affects almost every aspects of GIS application. Among the studies of spatial scale, scaling with socioeconomic data is one of choke points in geonomy and socioeconomic researches, such as diagnosis of environmental health risks, evaluation of natural disaster loss, and comparison of on-site sampling investigations, ect. Based on the theories and methods of spatial analysis, the scaling of socioeconomic data is studied. The thesis focuses on both the development of theory and practices.Firstly, the relevant researches and literatures on scaling are reviewed. And then by taking "spatial analysis" as a clew, the thesis inspects the key concepts in "sciences of scale" and compares the methods to solve the problems of scaling in different fields. Based on these researches, the thesis finds that the choosing of suitable methods in scaling with Geo-data is a challenge in the future.Secondly, according to the achievement of theory and method research, the thesis further takes Heshun, a county in Shanxi Province as an example and successfully applies five methods - point in polygon, areal weighting, Distance-decay model, kernel estimation and grid interpolation method based on genetic algorithms(GA) and genetic programming(GP) - to transform population density data from census to drainage areas. In order to solving the difficulties of traditional scaling methods in influence factors choice and model creation, the thesis explores a novel approach to scaling with population data by integrating GIS ,GA and GP, namely grid interpolation method based on GA and GP. The results of experiments prove that it is a new and feasible method to solve the population distribution scaling problem.Thirdly, the thesis compares and analyses uncertainties of all the applied methods for choosing the most suitable methods and providing experience and reference to future scale studies. The results indicate that as far as the data of Heshun is concerned, the accuracy of interpolation method based on GA and GP is highest, kernel estimation takes second place, point in polygon is lowest. Noticeably, during uncertainty analysis the thesis creatively uses integral best values calculated by integrating those 5 methods in stead of the default actual ones to measure the errors. Meanwhile, it also computes the reliability of integral best values. That method will exploit a new methodology for the uncertainty study of scaling methods.Finally, the thesis concludes by made a summary about the whole works, and also puts forwards some thoughts in the future.