Date of Award

Spring 5-1-2008

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Geography, Geology, and Anthropology

First Advisor

Qihao Weng

Second Advisor

Paul Mausel

Third Advisor

Brian Ceh

Abstract

Increasing population and economic growth has resulted in rapid urban expansion in the past decades. Timely and accurately mapping urban land-use and land-cover (LULC) and associated population dist1ibution are often required for many applications such as urban management and planning, and environmental monitoring and assessment. Although many techniques have been developed, urban LULC classification is still a challenge based on remote sensing data. This research explored the integration of Landsat ETM+ image and census data for improving urban LULC classification accuracy, focusing on the distinction of different densities of residential use. The housing data were examined for use at three different stages of image classification, that is, at preclassification for selection of training sample plots, during the classification as an extra channel, and at post-classification by sorting the classified image. The results indicated that the use of housing data was effective in improving overall urban LULC classification accuracy, especially useful for separating residential classes by post-classification sorting. Much research has been conducted for population estimation with remote sensing data over the past decades. Although many techniques have been used, it is still difficult to select a proper approach to estimate population and to achieve high accuracy. The perfonnance of population estimation models vaiies with methods used, details of population data set (block, block group, tract, or derived pixel level), and di ffcrent settings of study areas. This research compared different methods for population estimation based on the same ETM+ image and census data in order to reach some general conclusions. The results indicated that the use of residential classes can provide better estimation perfonnance than use of other variables. The major advantage [or using residential classes was that it can be easily transfel1'ed to other data sets for population estimation, because other sources of remotely sensed data can also be used to extract residential land classes.

Download

Share

COinS