PERS_1-14_Flipping - page 91

PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING January 2014 91 PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING January 2014 91
Abstract
Impervious surface estimation has been widely conducted
with medium-high resolution optical imagery. Challenges
however, remain in aspects of spectral similarity of differ-
ent objects, mixed pixel, and shadows of tall buildings or
large tree crowns. In order to reduce these uncertainties, this
study explores the synergistic use of optical and radar remote
sensing datasets in highly populated urban areas. A com-
parative analysis is performed between the RADARSAT-2 full
POL SAR image and the SPOT5 optical image. For both datasets,
the C5.0 decision tree algorithm is applied to select features
extracted to build a decision tree for urban impervious surface
classification. It is found in this study that optical and POL SAR
images possess different merits in delineating urban land
surfaces. The POLSAR data is helpful for delineating bright
impervious surfaces (e.g., buildings) and bare soils, which is
often a difficult task for optical data. However, the confusion
between dark impervious surfaces and bare soils becomes a
severe problem due to similar surface scattering mechanisms.
The intrinsic characteristics of radar scattering (e.g., layover
and shadow effects) also result in high uncertainties in dense
urban areas. These problems are leveraged when both optical
and decomposed POL SAR features are considered. The results
from this study indicate that the synergistic use of optical and
POL SAR data could be an efficient approach to improving the
estimation of urban impervious surfaces.
Introduction
Impervious surfaces are commonly defined as the entirety of
surfaces through which water cannot pass, including build-
ings, paved roads, parking lots, sidewalks, and other urban
infrastructure (Arnold and Gibbons, 1996; Esch et al., 2009).
The impervious surface area often serves as a key indicator in
assessing the impacts of urbanization on environmental and
ecological conditions on terrestrial lands (Hu and Weng, 2009;
Huadong Guo, Zhongchang Sun, and Xinwu Li are with
the Laboratory of Digital Earth Science, Institute of Remote
Sensing and Digital Earth, Chinese Academy of Sciences, No.9
Dengzhuang South Road, Haidian District, Beijing, China
100094 (zcsun@ceode.ac.cn).
Huaining Yang is with the Laboratory of Digital Earth Science,
Institute of Remote Sensing and Digital Earth, Chinese
Academy of Sciences, No.9 Dengzhuang South Road, Haidian
District, Beijing, China 100094; the University of Chinese
Academy of Sciences, No. 19A Yuquan Road, Shijingshan
District, Beijing, China 100049; and the National Earthquake
Response Support Service, No.1 Yuquan West Street,
Shijingshan District, Beijing, China 100049.
Cuizhen Wang is with the Department of Geography,
University of South Carolina, 1600 Hampton Street,
Columbia, SC 29208.
Photogrammetric Engineering & Remote Sensing
Vol. 80, No. 1, January 2014, pp. 91–102.
0099-1112/14/8001–91/$3.00/0
© 2014 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.80.1.91
Weng, 2011). Facing rapid urbanization all over the world,
the above-mentioned issues have triggered a surge of inter-
est in impervious surface studies. From the recent literature,
impervious surface area has been used to investigate urban
change detection (Yang et al., 2003a; Powell et al., 2008; Xian
and Homer, 2010; Lu et al., 2011; Gao et al., 2012), urban heat
island effects (Yuan and Bauer, 2007; Xiao et al., 2007; Weng
and Lu, 2008; Zhang et al., 2009; Xu, 2010), land-use planning
and land-cover mapping (Reilly et al., 2004; Lu and Weng,
2006), water quality assessment (Brabec et al., 2002; Conway,
2007), stream health (Weber and Bannerman, 2004), and rain-
fall runoff (Pappas et al., 2008; Sun et al., in press).
Due to their relatively low cost and suitability for large-
area mapping, satellite images have been widely applied
for impervious surface estimation. Various satellite images
are used in estimating impervious surfaces, including: (a)
medium resolution optical images such as Landsat TM/ETM +
(Wu and Murray, 2003; Yang et al., 2003b; Wu and Yuan,
2007; Weng and Hu, 2008; Esch et al., 2009; Zhang and
Guindon, 2009; Lu et al., 2011), Terra ASTER (Hu and Weng,
2009; Hu and Weng, 2011a) and the Advanced Land Imager
(ALI) and Hyperion data (Weng et al., 2008); (b) high spatial
resolution images such as Ikonos (Goetz et al., 2003; Lu and
Weng, 2009; Hu and Weng, 2011b), QuickBird (Zhou and
Wang, 2008) and SPOT5 (Jiang et al., 2009); and (c) synergistic
use of medium and high-resolution images (Yang et al., 2003b;
Jiang et al.; 2009). Various image processing approaches have
been applied to extract impervious surfaces from these satel-
lite images. These approaches include: (a) spectral mixture
analysis (SMA) (Wu and Murray, 2003; Lu and Weng, 2006;
Weng et al., 2008; Yang et al., 2010); (b) the classification and
regression tree (CART ) algorithm (Yang et al., 2003b; Jiang
et al., 2009); (c) artificial neural networks (ANNs) (Lee and
Lathrop, 2006; Weng and Hu, 2008; Hu and Weng, 2009; Sun
et al., 2011); and (d) multi-process classification models (Luo
and Mountrakis, 2010; Mountrakis and Luo, 2011; Luo and
Mountrakis, 2012).
Most of the above-mentioned studies are based on opti-
cal remote sensing. In relatively dense urban areas, however,
it is often difficult to accurately extract impervious surfaces
from optical imagery. Some limitations include: (a) because of
the heterogeneity in urban landscapes and the complexity of
urban impervious surface materials, the mixed-pixel problem
has been recognized as a constraint to the accuracy of imper-
vious surface estimation (Weng and Hu, 2008); (b) Some natu-
ral materials have spectral properties similar to impervious
surfaces; for example, low albedo objects such as water bodies
Synergistic Use of Optical and PolSAR Imagery
for Urban Impervious Surface Estimation
Huadong Guo, Huaining Yang, Zhongchang Sun, Xinwu Li, and Cuizhen Wang
ASPRS MEMBERS AND SUBSCRIBERS, CLICK HERE TO READ THE ENTIRE ARTICLE