Detecting fractional land-cover change in arid and semiarid urban landscapes with multitemporal Landsat Thematic mapper imagery

Pixel-based approaches are commonly used for urban land-cover classification and change detection, but the results are often inaccurate in arid and semiarid urban landscapes due to the mixed-pixel problem and similar spectral signatures between impervious surface areas (ISAs) and bare soils. This research proposes a subpixel-based approach to examine land-cover change in Urumqi and Phoenix urban landscapes using multitemporal Landsat Thematic Mapper (TM) imagery. Linear spectral mixture analysis (SMA) was used to unmix TM multispectral imagery into four fractions -high-albedo object, low-albedo object, green vegetation (GV), and soil. ISA was determined from the sum of high-albedo and low-albedo fraction images after removal of non-ISA in both fraction images. The ISA, vegetation abundance, and soil images at different dates were used to examine their change over time. The results indicate that this subpixel-based approach can successfully detect small changes of urban land covers in medium spatial resolution images which pixel-based approaches cannot.