ABSTRACT
CONTENTS 6.1 Introduction ................................................................................................. 94 6.2 Data Used..................................................................................................... 95 6.3 Methodology................................................................................................ 95
6.3.1 Linear Spectral Mixture Analysis................................................. 95 6.3.2 Endmember Selection..................................................................... 96 6.3.3 Impervious Surface Estimation .................................................... 99 6.3.4 Evaluation of Impervious Surface Images................................ 101
6.4 Analysis of Results ................................................................................... 103 6.4.1 Results of Spectral Mixture Analysis ........................................ 103
6.4.1.1 Four Endmembers of High Albedo, Low Albedo, Vegetation, and Soil .............................. 103
6.4.1.2 Three Endmembers of High Albedo, Low Albedo, and Soil.................................................... 104
6.4.1.3 Three Endmembers of High Albedo, Low Albedo, and Vegetation ....................................... 105
6.4.1.4 Three Endmembers of High Albedo, Vegetation, and Soil....................................................... 107
6.4.1.5 Three Endmembers of Low Albedo, Vegetation, and Soil....................................................... 109
6.4.2 Results of Impervious Surface Estimation................................ 113 6.5 Discussion .................................................................................................. 114 6.6 Conclusions................................................................................................ 116 Acknowledgments ............................................................................................. 117 References ........................................................................................................... 117
Many techniques have been developed to estimate and map impervious surfaces using remotely sensed data. Spectral mixture analysis (SMA), as a subpixel information extraction algorithm, is gaining interest in the remote sensing community in recent years. The linear SMA model assumes that the spectrum measured by a sensor is a linear combination of the spectra of all components within the pixel (Adams et al., 1986). Because of its effectiveness in handling the spectral mixture problem associated with mediumresolution (10-100 m) satellite imagery (such as Landsat TM=ETMþ, and Terra’s ASTER images), linear SMA has been widely used in the estimation of impervious surfaces (Ward et al., 2000; Madhavan et al., 2001; Phinn et al., 2002; Wu and Murray, 2003; Lu and Weng, 2006a,b). Three distinct methods based on the SMA model have been developed for estimation of impervious surface. These include (1) extraction of impervious surface as one of the endmembers in the standard SMA model (Phinn et al., 2002), (2) estimation by the addition of high-albedo and low-albedo fraction images, both as the SMA endmembers (Wu and Murray, 2003), and (3) the combination of several impervious surface endmembers from a multiple endmember SMA model (Rashed et al., 2003). However, these SMA-based methods have a common problem, that is, impervious surface is often overestimated in the areas with a small amount of impervious surface, but is underestimated in the areas with a large amount of impervious surface.
