Deriving urban sealed surface density map using linear spectral unmixing analysis

Medium resolution earth observation satellites system such as Landsat, due to its extensive historic archives, provides wealth of information useful for urban monitoring purposes. One of the main problems in usage of medium resolution satellite images in urban monitoring is high occurrence of mix-pixel within urban area. Mix-pixels which compose of multiple land classes hamper the urban study at a more detailed intra-urban level. In this study, we use linear mixture modeling to calculate the sub-pixel surface components. Three sub-pixel surface components identified are vegetation, soil and sealed surfaces. The resulting sealed surface proportion result is then used to produce urban density map. Our study demonstrates that intraurban characterizing using linear spectral unmixing can be used for urban density study.