Modeling urban growth in the Atlanta, Georgia metropolitan area using remote sensing and GIS
MetadataShow full item record
This study examines the dynamics of human-induced land use/cover changes, especially urban growth, in Atlanta, Georgia metropolitan area. Land use/cover changes of Atlanta between 1987 and 1997 were observed from LANDSAT TM satellite images and linked with the biophysical and socio-economic data using image processing, spatial analysis and modeling integrated in the framework of a geographic information system(GIS). Normalized Vegetation Difference Index (NDVI) differencing and temporal logic were employed to improve the accuracy of land use/cover change detection. A Markov chain model was used to conduct a ‘what-if’ analysis to predict the quantity distributions and spatial patterns of the future land use/cover based on the 1987-1997 land use/cover transition probabilities. In addition, a logistic regression model was used to identify the factors governing the process of urban growth and to predict the most probable sites of the growth. This study found that from 1987 to 1997 the most intense land use/cover change in Atlanta was deforestation for urban development. During the ten years, high-density urban area increased 12.94% from 876.78km2 to 990.27 km2; low-density urban area increased 42.57% from 2468.62 km2 to 3519.56 km2; and forest decreased 10.62% from 9217.04 km2 to 8238.28 km2. The Markov chain simulation revealed that urban use will continue to grow at the expense of forest. The proportion of urban area was 20.93% in 1987, and it will be 35.45% in 2020. Forest will decrease from 57.67% in 1987 to 43.64% in 2020. Urban growth will occur in the rural and urban-fringe areas, taking on a fragmented pattern. It was found from the logistic regression model that urban growth tends to occur around existing urban areas and close to major roads (road influenced growth), while some new urban clusters located at a distance from the existing urban areas can also form(diffusive growth). Whereas the logistic regression model can incorporate human dimensions, the Markov chain model is more temporally dynamic. Recommendations were made for further exploration of a more realistic land use/cover change model that can deal with space, time, and people simultaneously.