Download files
Abstract
Urban green space plays an important functional role in the urban environment by providing environmental, social and economic values to the growing cities and regions. Estimating urban green space patterns and changes with socioeconomic characteristics has become increasingly important for ecologically oriented city planning and environmentally sustainable urban development. The specific aim of this thesis is to investigate the spatial relationship between urban green space distribution and human socioeconomic status in Canberra, Australia to better understand human-nature interactions.
The study first examined and compared three remote sensing classification techniques to select the best method to quantify urban green space patterns using Landsat TM autumn images from 1996 and 2006. Object-based image classification achieved the highest total accuracy of 82.42% among other traditional pixel-based classifications, and was therefore selected as the most appropriate method for this study. Urban green space patterns were extracted and changes over time were calculated.
Global linear regression and geographically weighted regression (GWR) were then used, along with Australian Bureau of Statistics census data, to investigate the spatial relationship between vegetation density (as independent variable) and seven selected socioeconomic variables (as dependent variables) in urban residential areas. The results showed that the GWR significantly outperforms the global regression in capturing variable variance, as well as providing more insight into the local variability of the relationships within the study area. The results also suggested that the spatial variability of urban green space patterns in Canberra was most strongly and consistently related to housing density, education level and immigration status for both 1996 and 2006. The GWR analysis results were mapped using standard residuals and local R² values to assist in visualizing the spatially heterogeneous processes. The changing of the relationships over the ten years was discussed, based on the output maps. Finally, a detailed look at the vegetation density and median household income in various conditions provided valuable information on the human-dominated environmental inequity issues in an urbanizing world. This research is useful in demonstrating the location-specific social and environmental conditions that may assist city planners to better allocate environmental resources for the less advantaged neighbourhoods and for sustainable urban development policy making.