Extraction of Urban Impervious Surface Using
Two-Season WorldView-2 Images: A Comparison
Cai Cai, Peijun Li, and Huiran Jin
Abstract
Although multispectral images acquired during the summer
season have been used extensively in impervious surface extrac-
tion with relatively high accuracy, the area of impervious surface
extracted is generally underestimated. In this study, a quantitative
comparison of urban impervious surface extraction was conduct-
ed using WorldView-2 images of the summer and winter seasons
over two urban areas in a temperate region of Northern China. A
hierarchical object-based classification method was adopted to
extract urban impervious surfaces. The results showed that the
impervious surface extraction from the winter image achieved an
accuracy comparable with that from the summer image. However,
the area of impervious surface extracted from the winter image
was much greater than that from the summer image, which was
mainly attributed to seasonal variations of deciduous trees. There-
fore, winter images are recommended for impervious surface
mapping in temperate regions using very high resolution images.
Introduction
Impervious surface is generally recognized as an anthropo-
genic feature through which water cannot infiltrate into the
soil. This includes roads, rooftops, and other features in close
contact with human activities and habitation (Blair, 1996; Slo-
necker
et al.,
2001; Weng, 2012). It is also a key environmen-
tal indicator that has been used in urban related studies, such
as urban land use analysis (Madhavan
et al.,
2001; Phinn
et
al.,
2002; Lu and Weng, 2006), residential population estima-
tion (Lu
et al.,
2006; Wu and Murray, 2007), and urban heat
island analysis (Weng
et al.,
2004). The impervious surface is
also very useful information for the evaluation of the impact
of urbanization on surface runoff, water quality, air quality,
biodiversity, and microclimate (Arnold and Gibbons, 1996;
White and Greer, 2006; Luo
et al.,
2009; Lee
et al.,
2010).
Thus, timely and accurate information on impervious surface
is crucially important for many urban and environment re-
lated applications (Sung and Li, 2011; Weng, 2012).
Remote sensing is a major data source for the extraction
of impervious surface from global to local scales. Medium
spatial resolution images such as Landsat Thematic Mapper
(
TM
) , Enhanced Thematic Mapper Plus (
ETM+
) images, and
ASTER
images have been widely used in impervious surface
extraction (Wu and Murray, 2003; Lu and Weng, 2006; Weng
and Hu, 2008; Weng
et al.,
2009; Zhang
et al.,
2009). Because
of the mixed pixel problem caused by the limited spatial
resolution and the heterogeneity of urban landscapes (Small,
2003), spectral mixture analysis has been commonly used in
the extraction of impervious surface. The results extracted
from these medium resolution images usually have limited
accuracy (Wu and Murray, 2003; Lu and Weng, 2006).
In recent years, very high resolution (
VHR
) images, where
mixed pixels are significantly reduced, such as those from
Ikonos, QuickBird, Geoeye-1, and WorldView-2 satellites,
have been used in the extraction of impervious surface (e.g.,
Mohapatra and Wu, 2009; Lu
et al.,
2011, Li
et al.,
2011).
Different methods have been developed to extract impervious
surface using the
VHR
images, such as the sub-pixel analysis
(Lu and Weng, 2009; Small, 2003; Wu, 2009), the pixel-based
method (Lu
et al.,
2011; Xu, 2013), the object-based method
(Li
et al.,
2011; Hu and Weng, 2011) and the pixel- and object-
based hybrid method (Zhang
et al.,
2013). Many studies have
demonstrated that the object-based method outperforms the
pixel-based method (Yuan and Bauer, 2006; Myint
et al.,
2011; Lu
et al.,
2011). The object-based method has therefore
been considered to be effective for impervious surface map-
ping from
VHR
images (Li
et al.,
2011; Hu and Weng, 2011).
However, there are other problems in impervious surface
extraction using
VHR
images. These include the spectral
confusion between impervious surface and other land cover
classes. This is caused by the limited spectral resolution of
VHR
images and the high spectral variation of impervious
surface; shadows cast by tall objects, such as, buildings and
tree crowns, and their spectral confusion with dark impervi-
ous surface and water (Dare, 2005; Zhou
et al.,
2009; Lu
et al.,
2011). In some studies, the shaded area (shadow) was classi-
fied as a separate class and untreated (Yuan and Bauer, 2006;
Sugg
et al.,
2014), which led to an underestimation of imper-
vious surface in the shaded area. In certain studies the shaded
area was manually edited and corrected (Myint
et al.,
2011;
Hu and Weng, 2011; Lu and Weng, 2009; Lu
et al.,
2011),
which is time-consuming and costly, and also dependent on
available reference data. Further classification of the shaded
area is an alternative method that has been used in some stud-
ies (Zhang
et al.,
2012; Li
et al.,
2011; Zhou
et al.,
2009).
Most previous studies have used single-temporal multi-
spectral images (both medium resolution and
VHR
images)
in mapping impervious surface, in particular, multispectral
images of the summer season, because of its distinct spectral
profile compared to vegetation (Lu
et al.,
2011; Myint
et al.,
2011; Zhang
et al.,
2012). The studies conducted in tem-
perate regions showed that summer images achieved more
accurate classification results than those images acquired in
other seasons (e.g., autumn or winter images) (Wu and Yuan,
2007; Weng and Hu, 2008; Hu and Weng, 2009; Hu and Weng,
2010). The studies attributed the lower mapping accuracy
of winter images to the lack of foliage in temperate regions,
which reduced spectral contrast between the impervious sur-
face and vegetated cover. In addition, bare soil underneath the
tree canopy, which is detectable only in winter, is also spec-
trally confused with the impervious surface (Wu and Murray,
2003; Lu and Weng, 2004; Weng
et al.,
2009). Thus, summer
images were recommended for urban impervious surface
Cai Cai and Peijun Li are with the Institute of Remote
Sensing and GIS, School of Earth and Space Sciences, Peking
University, Beijing 100871, P R China (
).
Huiran Jin is with the Department of Geographical Sciences,
University of Maryland, College Park, MD 20742.
Photogrammetric Engineering & Remote Sensing
Vol. 82, No. 5, May 2016, pp. 335–349.
0099-1112/16/335–349
© 2016 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.82.5.335
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
May 2016
335
