Automated Water Classification in the Tibetan
Plateau Using Chinese GF-1 WFV Data
Guoqing Zhang, Guoxiong Zheng, Yang Gao, Yang Xiang, Yanbin Lei, and Junli Li
Abstract
The unique climate and topography of the Tibetan Plateau
produce an abundant distribution of lakes. These lakes are
important indicators of climate change, and changes in
lake area have critical implications for water resources and
ecological conditions. Lake area change can be monitored
using the huge sets of high-resolution remote sensing data
available, but this demands an automatic water classification
system. This study develops an algorithm for automatic water
classification using Chinese
GF
-1 (or Gaofen-1) wide-field-
of-view (
WFV
) satellite data. The original
GF
-1
WFV
data were
automatically preprocessed with radiometric correction and
orthorectification. The single-band threshold and two global-
local segmentation methods were employed to distinguish
water from non-water features. Three methods of determin-
ing the optimal thresholds for normalized difference water
index (
NDWI
) images were compared: Iterative Self Organizing
Data Analysis Technique (
ISODATA
); global-local segmentation
with thresholds specified by stepwise iteration; and the Otsu
method. The water classification from two steps of global-
local segmentations showed better performance than the
single-band threshold and
ISODATA
methods. The
GF
-1
WFV
-
based lake mapping across the entire Tibetan Plateau in 2015
using the global-local segmentations with thresholds from the
Otsu method showed high quality and efficiency in automatic
water classification. This method can be extended to other
satellite datasets, and makes the high-resolution global moni-
toring and mapping of lakes possible.
Introduction
The Tibetan Plateau has an average altitude of approximately
4,000 m
ASL
, making it the highest plateau in the world. The
Tibetan Plateau along with surrounding regions (e.g., the
Pamir-Hindu Kush-Karakoram-Himalayas) comprises 46,000
glaciers and covers an area of 100,000 km
2
(Gardner
et al.
,
2013; Yao
et al.
, 2012). Thus, it is also called the Third Pole
outside the Arctic, Antarctic, and Greenland (Qiu, 2008).
The region is the source of several large Asian rivers which
supply water for more than 1.5 billion people downstream
(Immerzeel
et al.
, 2010). During recent decades, the climate
and water resources of this plateau have undergone signifi-
cant changes (Yang
et al.
, 2011; Zhang
et al.
, 2011, 2013,
and 2014a). Both meteorological station measurements and
high-resolution simulation with the Weather Research and
Forecasting (
WRF
) model and the Global Land Data Assimi-
lation system (
GLDAS
) reveal a warmer and wetter climate
(Gao
et al.
, 2015; Zhang
et al.
, 2017). Meanwhile, the second
Chinese glacier inventory (Guo
et al.
, 2015) has shown that
the glaciated area in the interior of the plateau has shrunk by
9.5 percent between the 1970s and 2004 to 2011. The interior
of the Tibetan Plateau is an endorheic basin whose hydrology
is dominated by large lakes (Zhang
et al.
, 2014b).
The Tibetan Plateau has the greatest coverage of lakes in
China. There are approximately 1,200 lakes larger than 1 km
2
,
i.e., approximately half of the lakes in China, both by num-
ber and area (Ma
et al.
, 2011; Zhang
et al
., 2014b). Lakes in
the Tibetan Plateau have been mapped from the topographic
map produced by aircraft surveys in the 1960 to 80s, i.e., the
first China lake inventory (Wang and Dou, 1998). The second
China lake database was finished in 2005 and 2006 using
data collected with the China-Brazil Earth Resource Satellite
(
CBERS
) (Ma
et al
., 2011). Recently, many studies have focused
on lake area, water level, and volume changes, and their
response to climate change (Crétaux
et al.
, 2016; Zhang
et al
.,
2011, 2013, and 2017).
Remote sensing data provide a useful tool for monitoring
lake changes in the Tibetan Plateau where the station obser-
vation network is sparse and access for field investigations
is difficult because of the remote and rugged environment
(Ma
et al.
, 2011; Zhang
et al
., 2011). The Landsat series of
satellites provides the longest continuous record of remotely
sensed observations, including Landsat Multispectral Scanner
System (
MSS
) (1972 to 1992), Thematic Mapper (
TM
) (1982 to
present), Enhanced Thematic Mapper (
ETM+
) (1999 to pres-
ent), and Landsat-8 Operational Land Imager (
OLI
) (2013−).
Data from other satellites such as
CBERS
(Ma
et al
., 2011) and
instruments such as the Moderate resolution Imaging Spectro-
radiometer (
MODIS
) (Sun
et al.
, 2014) have also been used for
large-scale lake mapping.
The construction of China’s high resolution Earth observa-
tion system will further enrich the store of data in monitoring
the Earth’s resources and environmental changes. The
GF-
1 (or
Gaofen-1, which means high-resolution images in Chinese,
2013 to present) is the first Chinese high-resolution satellite
used for Earth observation. It has the advantages of high spa-
tial and temporal resolution and a wide swath width (Table
1).
GF-
2 (2014 to present),
GF-
3 (2016 to present), and
GF-
4
(2015 to present) have also been successfully launched into
space. Four more
GF
satellites are expected to be launched
over the next few years (Xu
et al.
, 2014).
The first two lake surveys in China (Ma
et al
., 2011; Wang
and Dou, 1998) and the recently released lake survey using
GF-
1 data from Wan
et al.
(2016); all used visual interpretation
for extracting lake boundaries. This method has the advantage
that each lake is checked and edited, but it is labor intensive
work and could involve uncertainties from multiple opera-
tors. An objective, automatic water classification system is
urgently needed for regional lake mapping. Many studies
Guoqing Zhang, Guoxiong Zheng, Yang Gao, Yang Xiang and
Yanbin Lei are with Institute of Tibetan Plateau Research,
Chinese Academy of Sciences, Building 3, Courtyard 16,
Lincui Road, Chaoyang District, Beijing, 100101, China
(
).
Junli Li is with Xinjiang Institute of Ecology and Geography,
Chinese Academy of Sciences, Urumqi 830011, China
Photogrammetric Engineering & Remote Sensing
Vol. 83, No. 7, July 2017, pp. 509–519.
0099-1112/17/509–519
© 2017 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.83.7.509
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
July 2017
509
