Impervious surfaces are mainly anthropogenic features such as paved roads, rooftops, driveways, sidewalks, and parking lots that are covered by impenetrable materials. With the urban expansion, vegetation and soils are replaced by impervious surfaces, which become a major ecological and environmental concern. This is because the increase of impervious surfaces generally leads to the decrease in vegetation, wetlands and agricultural lands, and consequently, to a series of environmental problems, such as the decease of groundwater recharge, the increase of surface runoffs and flood frequency and urban heat islands. The percent cover of impervious surfaces, as well as its spatial pattern, has been widely used as an indicator to quantify the urbanization level and urban environmental quality, and is essential to understand the interactions between human and the environment. Therefore, accurate mapping and estimating impervious surfaces is crucial for environmental and resources management. In this study, we compared and evaluated two methods: the Normalized Difference Vegetation Index (NDVI) based binary approach and the Linear Spectral Unmixing (LSU) method. These two approaches have been frequently used in mapping impervious surfaces. With the NDVI based binary approach, impervious surfaces are extracted based on information on vegetation distribution that can be well represented by NDVI,. Then vegetation fractional coverage was first estimated from a scaled NDVI, and then impervious surfaces were estimated as by subtracting the vegetation fraction from 1. This approach had the merit of simplicity. However, large errors may occur in impervious surface estimation. The LSU approach is based on the vegetation-impervious surface-soil (V-I-S) model proposed by Ridd in 1995, a novel conceptual model for remote sensing analysis of urban landscapes. The VIS model indicated that land cover in urban environment is a linear combination of these three components, that is, vegetation, impervious, and soils. The LSU approach has been widely used for remote sensing of impervious surfaces. This method provides a suitable technique to detect and map urban materials, and to address the mixed pixel problem in medium spatial resolution imagery. Taking the Beijing-Tianjin-Tangshan urban agglomeration as a case study, this research compared these two approaches on estimating impervious surfaces. The study area included Beijing City, Tianjin City, Tangshan City and Sanhe City, including a region with more than 40 000 square kilometers. Landsat 5 TM image data acquired in 2010 was used for mapping and estimating the impervious surfaces. A layer of impervious surfaces derived from ALOS images with spatial resolution of 2.5 m was used as a reference to evaluate the accuracies of the two methods. The results showed that the NDVI based binary approach had a root-mean-square error (RMSE) of 40.2%. The LSU approach was much better for impervious surfaces estimation than the NDVI based method, resulting in a RMSE of 20.0%. The residuals of the LSU approach ranged from -0.4 to 0.4. This accuracy was comparable to those from previous studies that were mostly conducted at a smaller geographical area, generally several thousands square kilometers. Our research expanded the knowledge of existing studies by proving that the LSU approach could be applied to a large study area for mapping impervious surfaces with acceptable accuracy.