A rule-based classifier using Classification and Regression Tree (CART) approach for urban landscape dynamics

An increasing rate of urbanization has led to haphazard growth, increased infrastructure costs, deterioration of living conditions and degradation of the environment. In order to assist urban planners in more effective urban planning and management strategies, there is a great demand for developing accurate and detailed spatial information, understanding urban land cover changes, and the causes of these changes. This research studies the urban landscape dynamics for the city of Columbia, Missouri USA using multi-temporal and multi-date (1984, 92, and 2000) Landsat TM and ETM satellite imageries. The classification algorithms used to classify these images include traditional classifier (i.e. maximum likelihood) and rule-based classifier.