ABSTRACT
CONTENTS 2.1 Introduction ................................................................................................. 21 2.2 Study Area and Data.................................................................................. 24 2.3 Methodology................................................................................................ 26
2.3.1 Artificial Neural Networks ........................................................... 26 2.3.2 Activation-Level Maps................................................................... 28 2.3.3 Neural Network Structure ............................................................ 28 2.3.4 Training and Testing Sample Selection....................................... 29
2.4 Results........................................................................................................... 30 2.4.1 Output Transformation.................................................................. 31 2.4.2 Result Verification .......................................................................... 31
2.5 Conclusion ................................................................................................... 34 References ............................................................................................................. 35
Over the last century, especially since World War II, the process of urbanization has increased and intensified all over the world. In addition to the internal growth and restructuring of cities, emigration and demand for more living space (Clarke and Gaydos, 1998) play a vital role in the spread of urbanized areas. It is likely that this trend will continue in the twenty-first century, eventually leading to the formation of Gigalopolises or supercities containing hundreds of millions of people (Clarke and Gaydos, 1998). According to a United Nations Population Division report, by the year 2015, the number of cities that have an urban population > 5 million will go up to 58 from 39 in 2000 (UN, 2001). One of the problems associated with
rapid urbanization is the concurrent increase (Xian and Crane, 2005) in impervious surfaces, which include roads, sidewalks, parking lots, and various rooftops, through which water cannot infiltrate into soil (Arnold and Gibbons, 1996). A study (Elvidge et al., 2004) partly funded by NASA’s land-cover land-use change program found that the aggregate impervious surface area of continental United States was slightly less than the total area of the state of Ohio. Although the percentage of impervious surface to the total landmass of the United States is just over 1%, it is a matter of great concern as impervious surface limits soil infiltration, threatens water quality through increased flow of polluted runoff, contributes to flooding, and creates heat islands (USEPA, 2003). Therefore, accurate estimation of impervious surface is crucial for sustainable urban development and planning (Arnold and Gibbons, 1996; Flanagan and Civco, 2001; Goetz et al., 2003; Wu and Murray, 2003; Dougherty et al., 2004; Xian and Crane, 2005; Yang, 2006).
