Determination of urban development in Istanbul using the spectral unimixed Landsat images

Urbanization has increased in the world since 1950s [21]. In Turkey, urbanization gained momentum after 1980s. Immigration from rural areas to cities due to the limited agriculture economy, and insufficient job and education opportunities played a fundamental role in this increase. This migration was primarily to Istanbul and cities like Ankara, Izmir, Bursa, Adana and Antalya where industrialization and/or tourism is at the forefront. This increase in urbanization causes problems such as uncontrolled growth of the cities, unplanned infrastructure and misusage of natural resources. In monitoring these changes in the urban environment, land cover and land use (LCLU) maps which are obtained through standard digital image processing methods where ‘maximum likelihood’ and ‘isodata’ classification techniques are employed, are used as well. However, inaccurate classification results may come out especially on the heterogeneous lands where the satellite data with low spatial resolution is used in these thematic maps obtained. The reason for this is representation of each pixel by only one class value. On the other hand, each pixel is represented by the endmember fractions that constitute itself with the spectral mixture analysis (SMA) method. With this analysis based on linear mixed model, the earth is represented more sensitively in substrate (S), green vegetation (V) and dark surfaces (D) spectral endmember (spectra that are chosen to represent pure surface materials in a spectral image) projections. In this study, different LCLU changes that occur in Istanbul are determined by applying SMA method to LANDSAT images of the city from 1984, 2000 and 2011; and urban growth is monitored and analyzed.