Date of Award

2008

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Geography, Geology, and Anthropology

Abstract

Timely and regular information on urban environmental quality (UEQ) is essential for urban planning. UEQ studies using the technologies of remote sensing and geographic information system (GIS) have not been extensive. There are a few attempts to focus on the properties of spatial context and scale-dependency in UEQ. This research provided a multiple spatial and temporal scale analysis of UEQ in Indianapolis, U.S.A, based on the synthetic indicators of physical components extracted from remotely sensed data and socioeconomic components derived from census data. The geospatial algorithms applied in this study included fractals, texture, spatial autocorrelation indexes, and spatial metrics of Shannon's diversity and Contagion Metric. The effectiveness of these algorithms for characterizing urban landscapes at multiple scales was first evaluated. It was found the triangular prism method was the most robust in estimating fractal dimension (FD) values for all types of image data in the digital number format. The red band and the near infrared band were the most spatially complicated among all spectral hands of the Landsat and ASTER images and the IKONOS data, respectively. Among the five transformed images derived from the original spectral bands, the Normalized Difference Vegetation Index (NDVI) contained the most information. The most fragmented land use and land cover (LULC) type in the study area was grassland. FDs of the Landsat derived LULC maps revealed the urban landscape of the study area had experienced great change during 1975-1991 and 1991-200. The cropland and pasture continued to loss over the time. In general, the results of fractal analysis, spatial autocorrelation analysis, and the spatial indices were consistent. The utility of the selected geospatial methods in mapping urban LULC was also examined. The analysis was carried out on the use of different window sizes in generating texture bands for image classification by using an ETM+ image. Evaluation of the quality of texture bands was conducted by correlation analysis, statistics of coefficient of variance, skewness, kurtosis, and local variance, fractal analysis, Moran's I index, separability analysis, accuracy assessment indexes, and pairwise significant Z test of kappa coefficients. It is found the textural data, especially those extracted from fractals, were useful in classifying medium and low density residential areas. Finally, the potential of the integration of remote sensing and census data in assessing the temporal change of UEQ was explored within a GIS framework. Physical environmental variables such as LULC data, NDVI and other transformed remote sensing variables, and land surface temperature were derived from two Landsat images. Socioeconomic variables including population density, house characteristics, income, andEducation level were extracted from US census 1990 and 2000 block group (BG) data. Correlation analysis and factor analysis were performed after the two types of variables were integrated at the BG level. For each year, four factors were identified and interpreted as greenness, crowdedness, general physical environmental condition, and economic status. By assigning different weights to each factor, two synthetic UEQ indexes were generated for the two years. A comparison of the two synthetic indexes revealed the significant temporal changes of UEQ pattern.

Download

Share

COinS