Modeling urban growth and spatial structure in Nanjing, China with GIS and remote sensing

This research focuses on the use of GIS, remote sensing and spatial modeling for studies on urban growth and spatial structure. Previous studies on urban growth modeling have not elaborated the spatial heterogeneity of urban growth pattern, which, however, is well recognized. The census population data is widely used for investigating urban spatial structure, but it has inherent various problems which can lead to biased analysis results. Studies on urban growth and spatial structure of Chinese cities remain limited due to the data availability and methodology development. In this dissertation, I initiate a new analysis framework and a new method to address these critical issues through a case study of Nanjing, China.; The study first set up urban land expansion models for Nanjing in the period of 1988-2000. Landsat imageries are processed and classified to provide land use data in 1988 and 2000. GIS data are used to provide spatial variables inputs for the land use conversion models. A combined land use data sampling is conducted to obtain land use sample points for the proposed models. Classic logistic regression is used to reveal the urban land expansion from a global view. Furthermore, a logistic geographically weighted regression (GWR) model is set up to reveal the local variations of influence of spatial factors on urban land expansion. The study finds that the logistic GWR significantly improved the global logistic regression model and verifies that the influences of explanatory variables of urban growth are spatially varying. An urban growth probability surface is then generated based on the variable and parameter surfaces. This new framework for analyzing urban growth pattern may open a new direction for urban growth modeling.; Second, the dissertation develops a new method, which utilizes detailed urban land parcel and building data to generate population surface of Nanjing in 2000. With this method, populations of small areas at intraurban level can be estimated much more accurately, and moreover, the generation is not constrained to any pre-defined resolution of raster land use data, thus different cell size can be chosen with changing scale and research contexts. The case study finds that despite suburbanization, Nanjing remains a compact city, and population density declines quickly with the increase of distance from the central business district (CBD). Based on the generated population surface, the dissertation also uses exploratory spatial data analysis to investigate spatial associations between non-residential land uses and population density. It finds that population suburbanization in Nanjing has been limited to the inner suburb area where population is densely distributed without substantial commercial and office land development and that commercial activities influence population distribution and suburbanization more significantly than industrial suburbanization.; This dissertation concludes with several suggestions on future research foci.