The development of the index was ba

The development of the index was based on the unique spectral response of built-up lands that have higher reflectance in MIR wavelength range than in NIR wavelength range. However, this is not always the case. Some studies showed that the reflectance for certain types of vegetation over the band pass of TM5 increased as leaf water content decreased (Cibula et al., 1992; Gao, 1996). The drier vegetation can even have higher reflectance in MIR wavelength range than in NIR range (Gao, 1996), resulting in positive values in NDBI imagery for these plants. This study also found that the many vegetated areas have positive NDBI values, especially in Fuzhou’s NDBI image where the mean of vegetation is 0.01. Furthermore, in some circumstances, water with high suspended matter concentration (SMC) can also reflect MIR stronger than NIR because the reflectance peak shift to longer wavelength regions as the suspended matter increase. Therefore, the drier vegetation and water with high SMC will have positive NDBI values when computed using Equation 4 and present as noise in a NDBI image. Consequently, the contrast of the NDBI image is not so good as SAVI and MNDWI images (Figure 1), because many pixels of vegetation and water areas having positive NDBI values show medium gray tones and present as noise mixed with built-up features. Wu et al. (2005) employed NDBI to extract urban built-up lands of Xi’an City of China and obtained a low accuracy of 78.7 percent. A similar situation was also encountered in this study (see discussion later). These suggest that the urban built-up land features could not be extracted merely based on a NDBI image. This is why this study combines the NDBI with SAVI and MNDWI to extract urban built-up land features. This combination can remove the vegetation and water noise, and hence improve the extraction accuracy.