SRTM Error Distribution and its Associations
with Landscapes across China
Quan Zhang, Qinke Yang, and Chunmei Wang
Abstract
In this paper the distribution of 3-second elevation error in the
data from the Shuttle Radar Topography Mission (
SRTM
) over
the whole of China and its associations with topographic and
land cover factors were systematically evaluated. The land-
scape features extracted from different datasets at more than
500,000 sites were used to determine the variation pattern in
the errors by the method of single factor analysis. The results
showed that the topographic attributes derived from
SRTM
data
could adequately represent the terrain of China. However,
there were extended and observable areas with abnormalities
in a small proportion of the data. Slope was the dominant
factor affecting elevation error compared with other landscape
features (aspect, vegetation, etc.). The mean errors in glaciers,
deserts and wetlands were -1.05 m, -2.03 m and -2.43 m, and
1.05 m in built-up areas. In general the elevation errors in the
SRTM
data formed a complex pattern of variation across China.
Introduction
Land surface elevation datasets are important foundations
for studying topography, land surface processes, and global
environmental change. With technological advances, remote
sensing platforms (satellites, space shuttles, etc.) are increas-
ingly being used to acquire high-quality surface elevation data
(Nelson
et al.
, 2009). The Shuttle Radar Topography Mission
(
SRTM
) C-band data is a near-global digital elevation dataset
collected and released through collaboration between the Na-
tional Aeronautics and Space Administration (
NASA
) and the
National Geospatial-Intelligence Agency (
NGA
). The
SRTM
data
has been released in two formats: one with a resolution of
1-second and the other with a resolution of 3-seconds (Rabus
et al.
, 2003; Farr
et al.
, 2007). Although the
SRTM
1-second
global dataset is now being released worldwide in phases,
starting September 2014, the 3-second dataset, which has
been released to the public since 2003, has gained a great deal
of attention and has been applied extensively in geoscience
studies worldwide (Zandbergen, 2008; Yang
et al.
, 2011). For
that reason, the 3-second product has been used for the cur-
rent study. Since its initial release, the
SRTM
3-second dataset
has been continuously upgraded by a number of groups of
which one example is the Consultative Group on Internation-
al Agricultural Research (
CGIAR
) version 4 (Jarvis
et al.
, 2008).
However, despite the improvements,
SRTM
still suffers from
certain issues in representing the geomorphology of the land
surface. In some cases, these issues can significantly affect the
accuracy of research outcomes.
The quality of the basic released
SRTM
data is closely relat-
ed to the properties of the remote sensing devices, the interac-
tion between the land surface and radar signals, and the base
data processing. Among these factors, the effect of the device
on the quality of
SRTM
data is usually the result of changes
in the attitude of the space shuttle and signal transmission
anomalies from the sensors. Rodríguez
et al.
(2006) provided
a detailed explanation of the data error characteristics caused
by the properties of platforms and sensors. The land surface
features which interact with radar signals and result in eleva-
tion data errors, include topographic attributes and land-cov-
er. In regard to
SRTM
data, the elevation error in areas with
steep terrain is larger whereas in flat areas the elevation error
is smaller but includes striping noise (Carabajal and Harding,
2006). The pixel values of the
SRTM
data provide effective
height information for an area of the Earth’s surface, but in
many areas the STRM elevation value can be different from
that of the bare surface (Gallant and Read, 2009; Nelson
et al.
,
2009). Currently, most discussions of the effects of surface
objects on the quality of
SRTM
data have focused on increased
effective elevation due to vegetation (Weydahl
et al.
, 2007;
Baugh
et al.
, 2013; Su and Guo, 2014). Man-made objects,
such as urban built-up, can also affect the quality of the
SRTM
data, but only a few studies to date have focused on this effect
(Gamba
et al.
, 2002). In other situations, the
SRTM
radar signal
can penetrate the actual ground surface, including snow-cov-
ered surfaces, ice, deserts, etc., and yield elevation values
that are lower than the actual bare surface. The radar signal
can also penetrate vegetation crowns, so vegetation height
information in
SRTM
data is normally also an underestimate
of the height to the vegetation crowns (Carabajal and Harding,
2006; Kenyi
et al.
, 2009). Because of the impact of the device
properties and land surface features, there were many voids
in the preliminary
SRTM
products. Although many of these
problems have been effectively fixed in later versions through
spatial interpolation and multi-source data fusion (Delaney
et
al.
, 2005; Kuuskivi
et al.
, 2005; Grohman
et al.
, 2006; Reuter
et al.
, 2007), the processing that corrected these problems can
also introduce new errors to the
SRTM
data.
The
SRTM
data incorporating the effects of land-cover height
(vegetation, buildings, etc.) essentially provide estimates for a
Digital Surface Model (
DSM
) (Nelson
et al.
, 2009) whereas the
elevation of the bare ground surface is the objective of a Digi-
tal Elevation Model (
DEM
). Although there are previous studies
focusing on the elimination of land-cover height effects and
particularly of vegetation height effects (Gallant and Read,
2009; Baugh
et al.
, 2013; Su and Guo, 2014), no fully satisfac-
tory and universal methods are yet available for the large scale
conversion of the
SRTM DSM
to a
DEM
. In a number of studies
where a high-accuracy
DEM
at large scale is not available,
SRTM
Quan Zhang is with the College of Urban and Environmental
Sciences, Northwest University, No.1 Xufu Street, Chang’an
District, Xi’an, Shaanxi Province 710127, P.R. China
Qinke Yang (corresponding author) and Chunmei Wang
are with the College of Urban and Environmental Sciences,
Northwest University, No.1 Xufu Street, Chang’an District,
Xi’an, Shaanxi Province 710127, P.R. China
(
).
Photogrammetric Engineering & Remote Sensing
Vol. 82, No. 2, February 2016, pp. 135–148.
0099-1112/16/135–148
© 2015 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.83.2.135
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
January 2016
135
