LLMG, 
y spectral 
nds. n: 
nsing of 
resolution 
| imagery 
ICtions on 
ation for 
ges Using 
'eoscience 
ctral and 
] Method. 
ensing, 4T: 
ultisensor 
ions, and 
al, 2(2y 
or image 
ods and 
Sensing, 
| of high- 
s. IEEE 
1. 
)9. Super 
S Remote 
0-100. 
rithm for 
1 Images. 
Sensing, 
B., 2010. 
ic quality 
"rnational 
formation, 
K., 2009. 
Computer 
    
  
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B7, 2012 
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia 
MONITORING OF GLACIAL CHANGE IN THE HEAD OF THE YANGTZE RIVER 
FROM 1997 TO 2007 USING INSAR TECHNIQUE 
Hong’an Wu**, Yonghong Zhang”, Jixian Zhang”, Zhong Lu ^ Weifan Zhong * 
a Chinese Academy of Surveying and Mapping, Beijing 100830, China — 
(wuha, yhzhang, zhangjx)@casm.ac.cn, zhongwf59(2)126.com 
? U. S. Geological Survey Center for EROS, Sioux Falls, SD 57198, USA - lu@usgs.gov 
Commission VII, WG VII/6 
KEY WORDS: SAR interferometry, Glacier monitoring, Phase texture analysis, Gray level co-occurrence matrix (GLCM) 
ABSTRACT: 
Accurate monitoring of the glacier changes is essential to evaluate the environmental-ecological health in the scenario of global 
change. Conventional method for glacial monitoring is optical remote sensing. However, affected by cloud and snow cover, it is hard 
to monitor glacier by optical images. With the fast development of InSAR technique, interferometric coherence has been utilized for 
extracting glacial information. However, it is difficult to distinguish glacial area from non-glacial area when their coherence is 
similar, especially for short wavelength radar, such as X-band and C-band. In this case, interferometric phase can play an important 
role to identify glacier. In this paper, phase texture analysis method is proposed to extract glacier. 8 texture features were analyzed 
based on gray level co-occurrence matrix (GLCM), including mean, variance, homogeneity, contrast, dissimilarity, entropy, second 
moment, and correlation. Among them, variance, contrast and dissimilarity can distinguish glacier from non-glacier clearly most, so 
they are chosen for RGB combination. Then the RGB combination image is classified into several land covers by maximum 
likelihood classification (MLC). After post-classification processing, glacial area can be extracted accurately. With this proposed 
method, two ERS-2 SAR single look complex (SLC) images acquired in 1997 and two ENVISAT ASAR SLC images acquired in 
2007 are used to extract glacial area in 1997 and 2007 over Geladandong area, the head of the Yangtze River. The extracted areas are 
validated by Landsat TM data, which indicate that the proposed method can obtain accurate glacial area. The results also 
demonstrate during the 10 years, glacial area over Geladandong decreased fast, with a reduction of 22.97km’, 
1. INTRODUCTION 
Glacier change especially mountain glacier change is one of the 
best natural indicators to global climate change (IPCC, 2001). It 
not only plays an important role in climatology, but also has a 
potential influence in economy around the world (Lu, 2002). 
However, since 1950s, mountain glaciers have recessed fast as 
the global warming, which is obvious in the Tibet Plateau (Li, 
1998; Su, 1999; Wang, 2001). As the solid reservoir, glaciers 
are the most important water resource in the arid and semi-arid 
area, in western China. Several Glacier Lake Outburst Flood 
(GLOF) events took place in this region in the last century 
causing significant damage to infrastructure and livelihoods 
(Richardson, 2000). Thus monitoring glacier change accurately 
is essential to evaluate the recession velocity and glacier melt 
water runoff. 
For mountain glaciers, most of them are developed in sparsely 
populated region with harsh natural environment, so 
conventional method of monitoring glacier can not be 
implemented in these areas. With the rapid development of 
satellite remote sensing technology, high-resolution multi- 
spectral remotely sensed data provide advanced technical 
support to glacier monitoring. Using geographic information 
system, optical remote sensing has been widely used in glacier 
dynamics monitoring (Dwyer, 1995; Haeberli, 2000; Kargel, 
2005). However, because of cloud cover and snow cover effects, 
multi-spectral data is often difficult to accurately monitor 
glacier changes. Due to the ability of penetrating cloud cover, 
  
* Corresponding author. 
all-weather and high resolution, synthetic aperture radar (SAR) 
remote sensing has a more wide application for ice and snow 
research (Lu,2010; Zhou, 2010), which has been an important 
supplement for optical remote sensing to overcome its inherent 
difficulties. 
Existing method for extracting glacier information by SAR is 
mainly based on interferometric coherence, which selects 
appropriate threshold of coherence by comparing the glacial 
surface and non-glacial surface, because the glacial surface is 
normally de-coherent seriously. By now this method has been 
applied successfully (Zhou, 2010; Li, 2001, Li, 2008). However, 
when the coherence contrast is not obvious, this method cannot 
extract glacier information exactly. In this case, interferometric 
phase can play an important role. This paper propose a new 
method to extract the glacier area in mountain regions by using 
texture analysis of interferometric phase based on gray level co- 
occurrence matrix (GLCM) for the monitoring of glacier change 
in Geladandong area, the head of the Yangtze River, from 1997 
to 2007. 
2. METHOD 
2. Different feature of interferometric phase textures in 
glacial and non-glacial region 
For C-band, because of serious decorrelation, interferometric 
phase in glacial region performs noisy texture with no obvious