Introduction
In recent decades, glaciers on the Tibetan Plateau have retreated rapidly due to climate warming, not only affecting water resources and hydrological processes in the region (Reference Yao, Wang, Liu, Pu, Shen and LuYao and others, 2004, Reference Yao, Pu, Lu, Wang and Yu2007), but also causing the expansion of glacial lakes (Reference YaoYao and others, 2010) and the potential for glacial lake outburst floods (GLOFs) (Reference Richardson and ReynoldsRichardson and Reynolds, 2000; Reference NayarNayar, 2009). To the people living downstream, the GLOFs loom large because of their devastating effect on infrastructure and social resources. The Zhangzhangbo GLOF in 1981 (Reference XuXu, 1988; Reference Bajracharya, Mool and ShresthaBajracharya and others, 2007b), the Dig Tsho GLOF in 1985 (Reference Mool, Bajracharya and JoshiMool and others, 2001b; Reference Bajracharya, Mool and ShresthaBajracharya and others, 2007b) and the Luggye Tso GLOF in 1994 (Reference Mool, Wangda, Bajracharya, Kuzang, Gurung and JoshiMool and others, 2001a; Reference Bajracharya, Mool and ShresthaBajracharya and others, 2007b) are examples of the destructive consequences of GLOFs. However, owing to the relatively remote location of glacial lakes, it was not until recently that researchers began to assess the hazards using scientific methods (e.g. Reference Bajracharya, Shrestha and RajbhandariBajracharya and others, 2007a; Reference Bolch, Buchroithner, Peters, Baessler and BajracharyaBolch and others, 2008; Reference Fujita, Suzuki, Nuimura and SakaiFujita and others, 2008, Reference Fujita, Sakai, Nuimura, Yamaguchi and Sharma2009; Reference Wang, Liu, Guo and XuWang and others, 2008) and to address the formation (Reference QuinceyQuincey and others, 2007; Reference Frey, Haeberli, Linsbauer, Huggel and PaulFrey and others, 2010; Reference Sakai and FujitaSakai and Fujita, 2010) and variation (Reference Chen, Peng, Li, Zhong and QiChen and others, 2007; Reference KomoriKomori, 2008) of glacial lakes.
In the southeastern Tibetan Plateau, where the warm and humid moisture from the Indian Ocean significantly influences the local climate, there are a large number of marine-type glaciers (Reference Yao, Ren and XuYao and others, 2008) the retreat of which has given rise to many glacial lakes in front of glacier termini (Reference YaoYao, 2010). In situ mass-balance measurements indicate that these glaciers melted much faster than the continental or subcontinental glaciers (Reference Yang, Yao, Xu, Wu, Ma and XinYang and others, 2008, Reference Yang, Yao, Xu, Ma, Wang and Wan2010), which, in combination with abundant precipitation during the summer monsoon, makes southeast Tibet one of the most GLOF-affected regions of Asia (Reference Ding and LiuDing and Liu, 1992;Reference Wang and LiuWang and Liu, 2007; Reference Cheng, Zhu, Dang and LiuCheng and others, 2008; Reference Liu, Cheng, Li and SuLiu and others, 2008). For example, China’s authoritative websites reported two severe GLOFs in Chamdo and Lhoka, respectively, in Tibet in July 2009. There is therefore an urgent need to monitor the variations of glaciers and glacial lakes in southeast Tibet, so as to assess potential GLOF hazards in the region.
Remote sensing makes it possible to investigate simultaneously a large number of glaciers and glacial lakes in inaccessible mountainous areas (Reference Huggel, Kääb, Haeberli, Teysseire and PaulHuggel and others, 2002; Reference KääbKääb and others, 2005; Reference Quincey, Lucas, Richardson, Glasser, Hambrey and ReynoldsQuincey and others, 2005; Reference Bolch, Buchroithner, Peters, Baessler and BajracharyaBolch and others, 2008). However, high-quality remote-sensing images are difficult to obtain for southeast Tibet, as both cloud cover during the monsoon season (mid-May to mid-September; Reference Yang, Xu, Yang, Qu and LinYang and others, 2009) and snow cover during winter (November to March; Reference Yang, Xu, Yang, Qu and LinYang and others, 2009) hinder precise image interpretation. Therefore, in this study we chose only the satellite images acquired during September– November in the Boshula mountain range (Reference Xin, Yao, Ye, Guo and YangXin and others, 2009) and used 1 : 50 000 topographic maps for the 1970s, high-quality Landsat images for 1988 and 2001, and Advanced Land Observing Satellite (ALOS) Advanced Visible and Near Infrared Radiometer type 2 (AVNIR-2) satellite data for 2009 (Table 1). By combining meteorological data from stations near our study area, we also address the variations of glaciers and glacial lakes in response to possible climate variation.
Table 1. Data used in this study and corresponding applications
Study Area
The Boshula mountain range (29.5–30.0˚ N, 96.25–96.75˚ E) is located at the boundary of Bomi county and Paksho county in the southeastern Tibetan Plateau (Fig. 1). The elevation ranges from 3100 to 6200 m and the topography is characterized by huge relief differences and steep slopes, with National Highway No. 318 (Sichuan–Tibet Highway) winding through this area. Within the domain of the Indian monsoon, the Boshula mountain range is significantly influenced by warm and humid moisture in summer.
Fig. 1. (a) Location of the Boshula mountain range on the Tibetan Plateau. (b) Glaciers, glacial lakes and villages in the study area. See Figures 5a and b, 3 and 7b and d for detail in frames A, B, C, D, E, respectively.
Data and Methods
Data
Topographic maps of 1 : 50 000 scale constructed from 1970s aerial photography and corresponding digital elevation models (DEMs) with 20 m equidistant contour lines and 25 m cell size were used as the benchmark data to detect the variations of glaciers and glacial lakes since that time. Topographic maps were scanned into digitized products at a resolution of 300 dots per inch (dpi) followed by geometric calibration using ERDAS 9.1 software. Remote-sensing images presented in this paper were all acquired from different sensors during September and October with low cloud cover and snow cover, so they are conducive to analyzing glacier and glacial-lake variation in this area. In data processing, Landsat images were co-registered with the 1 : 50 000 scale topographic maps. The ALOS AVNIR-2 data were then orthorectified with DEM and topographic maps using the PCI Geomatica OrthoEngine module. All satellite images were converted into the Universal Transverse Mercator projection and Krasovsky 1940 spheroid, which is the same projected coordinate system as the topographic maps. Image co-registration was based on ground-control points, which were collected from the topographic maps and identified on the satellite images. The root-mean-square errors for Landsat Thematic Mapper (TM) (1988), Enhanced TM Plus (ETM+) (2001) and ALOS (2009) were 21.62, 19.36 and 5.06 m, respectively.
Identification of glaciers and glacial lakes
Glacial lakes and glaciers were delineated manually from digitized topographic maps and/or false-color composite (FCC) satellite images pixel by pixel in ArcGIS 9.2. Though relatively labor-intensive and time-consuming, manual delineation presents more accurate surface-feature mapping than band algebraic approaches such as the normalized-difference snow index or the normalized-difference water index (Reference Paul, Kääb, Maisch, Kellenberger and HaeberliPaul and others, 2002). Owing to the relatively low resolution of Landsat imagery, we only analyzed the changes of glacial lakes when they were larger than 0.02 km2 (Reference Chen, Peng, Li, Zhong and QiChen and others, 2007). After digitizing the individual images periodically, vector layers of glacial lakes and glaciers were obtained for four periods: the 1970s, 1988, 2001 and 2009. For vector layers of glacial lakes, the type, area, altitude and perimeter of lakes were presented by way of eye-witness interpretation and geographic calculation as attribute data. For vector layers of glaciers, glacier catchments were segmented based on the Global Land Ice Measurements from Space (GLIMS) algorithm (http://www.glims.org/MapsandDocs/assets/GLIMS_Analysis_Tutorial_a4.pdf) to identify the mother glaciers in the upper reaches that feed corresponding glacial lakes. Detailed procedures include: (1) filling DEM sinks; (2) creating a flow direction grid; (3) finding glacier toes; (4) producing watersheds; and (5) separating ice polygons into different glacier individuals.
Area uncertainties
Since we use multitemporal images with different spatial resolutions to study the changes of glaciers and glacial lakes, our results may contain two types of uncertainty which are evaluated quantitatively below.
Error in co-registration
According to Reference Hall, Bayr, Schöner, Bindschadler and ChienHall and others (2003) and Reference Ye, Kang, Chen and WangYe and others (2006), the uncertainty in co-registration of multitemporal images can be quantified as
where U L is the linear uncertainty (m), U A is the area uncertainty (m2), λ is the original pixel resolution of each individual image (m) and σ is the co-registration error of each individual image to the topographic maps (m). Accordingly, the maximum error in co-registration (U A) for changes in glaciers and glacial lakes from the 1970s to 2009 was calculated as ±0.008km2.
Error introduced by glacier/glacial lake outline delineation
According to the spatial resolution and quality of each image, we estimate the average accuracy of the glacier/ glacial-lake outlines derived from manual digitizing to be 1.0, 1.5, 1.0 and 1.0 pixels, respectively, for the 1970s (topographic maps), 1988 (Landsat TM), 2001 (ETM+) and 2009 (ALOS) datasets. Note that owing to the remaining snow cover during remote sensing in 1988, images for that period have a relatively large estimation uncertainty. The uncertainties of glacier/glacial-lake area were thus determined as the length of shorelines multiplied by the estimated error of outline delineation (Reference Racoviteanu, Arnaud, Williams and OrdonezRacoviteanu and others, 2008; Reference Fujita, Sakai, Nuimura, Yamaguchi and SharmaFujita and others, 2009; Reference Wang, Hou and LiuWang and others, 2009). The total areal uncertainty was estimated as the sum of the aforementioned two errors.
Results
Glacial lake variation
Glacial lakes can be categorized by their causes into three types: moraine-dammed, trough-valley and landslide-dammed lakes. Table 2 shows the number and area of glacial lakes of different types during four periods. From the 1970s to 2009, we calculated an 18.6% expansion in the area of glacial lakes, with the expansion rate reaching 0.55% a–1 (assuming the beginning year is 1975). This expansion rate is similar to those of the Ranwu lake region (0.47% a–1; Reference Xin, Yao, Ye, Guo and YangXin and others, 2009) and the Pumqu river basin (0.52% a–1; Reference Che, Jin, Li and WuChe and others, 2004), but smaller than those of the Poiqu river basin (3.33% a–1; Reference Chen, Peng, Li, Zhong and QiChen and others, 2007) and the Luozha area (1.14% a–1; Reference LiLi, 2010). From the 1970s to 2001, glacial lakes expanded by 0.05 km2 a–1, increasing to 0.07 km2 a–1 thereafter until 2009, indicating the acceleration of glacial lake expansion in the last decade. Moraine-dammed lakes were predominant in this area both in number and area, also contributing mainly to the expansion of glacial lakes. From the 1970s to 2009, moraine-dammed lakes expanded by 1.55 km2, a 26.8% increase, while trough-valley lakes showed a slight decrease in area (Table 2).
Table 2. Variation of number and area (km2) of different types of glacial lake
Glacial lakes can also be categorized by area into four classes: small (lake area 0.02–0.05km2), medium (0.05– 0.10km2), large 0.10–0.50 km2) and giant (≥0.50km2). Nearly half of the glacial lakes in this area are small and the number of small glacial lakes has been increasing since the 1970s (Fig. 2a). The emergence of new small lakes in recent decades is clearly observable in Figure 3, and numerous lakes are seen in the 2009 ALOS image that were not present when the topographic maps were produced in the 1970s. In comparison to the increasing number of small moraine-dammed lakes, no apparent changes occurred in trough-valley lakes (Fig. 2a). This suggests that the new emerging lakes in southeast Tibet were primarily moraine-dammed. In addition, glacial lakes are located at different elevations, with the distribution of trough-valley lakes being more concentrated within the range 5000–5200ma.s.l. Moraine-dammed lakes are more widespread and are evenly distributed within three major altitudinal ranges: <4800 m, 4800–5000m and 5000–5200m (Fig. 2b). The number of moraine-dammed lakes in each altitudinal range increased from the 1970s to 2009, while the number of trough-valley lakes remained steady or even slightly decreased over these decades (Fig. 2b).
Fig. 2. Numbers of different types of glacial lake from the 1970s to 2009, by (a) area class and (b) altitudinal range.
Fig. 3. New emergences of glacial lakes in the study area. (a) Topographic map of 1975; (b) ALOS image of 2009.
Glacier variation
Careful inspection of the 1970s images revealed seasonal snow cover in the northern part of the Boshula mountain range, so we eliminated this part from the study area when analyzing glacier variation. Glacier variation in the study area is dominated by significant recession (Fig. 4), which coincides with glacier variations since the early 20th century in the Kangri Karpo mountains (Reference LiuLiu and others, 2005), approximately 100 km south of the Boshula mountain range. Calculations from topographic maps and remote-sensing images show the glacierized area in the 1970s (167.5±0.5km2) decreasing consecutively to 162.8±3.2 and 155.0±2.1km2 in 1988 and 2001, respectively, before reaching 146.3 ±1.0km2 in 2009 (Table 3). Thus, from the 1970s to 2009, the glacierized area in the Boshula mountain range decreased by 21.2 km2 (or 12.7% of the total glacierized area), retreating at a rate of 0.37% a–1 (Fig. 4; Table 3).
Fig. 4. Glacier area changes in the Boshula mountain region between the 1970s and 2009.
Table 3. Variation of glaciers in the Boshula mountain range from the 1970s to 2009
Different temporal intervals witness different retreat rates: 0.22% a–1 from the 1970s to 1988, increasing to 0.37% a–1 from 1988 to 2001 and further increasing to 0.70% a–1 from 2001 to 2009. This trend is consistent with the study of Reference Xin, Yao, Ye, Guo and YangXin and others (2009) in the Ranwu lake region, southeast Tibet, which also found an accelerated glacier retreat. The retreat of glaciers corresponds closely to the expansion of glacial lakes, especially those at the glacier terminus. As shown in Figure 5, there are two mother glaciers in the center of our study area that underwent rapid melting and shrinkage from the 1970s to 2009, leading to significant expansion of glacial lakes by approximately 180% and 1400% respectively at the glacier terminus. Moreover, the disappearance of small glaciers and the fragmentation of large glaciers observed in other mountainous glacierized areas (Reference Paul, Kääb, Maisch, Kellenberger and HaeberliPaul and others, 2004; Reference Niederer, Bilenko, Ershova, Hurni, Yerokhin and MaselliNiederer and others, 2008; Reference Bolch, Menounos and WheateBolch and others, 2010; Reference Narama, Kääb, Duishonakunov and AbdrakhmatovNarama and others, 2010) is also common in our study area (e.g. two cases of glaciers splitting into separate glaciers are seen in Fig. 5).
Fig. 5. Two examples (a, b) of rapid glacier retreat and glacial lake expansion in the study area. The false-color composite ALOS AVNIR-2 image is used as the background.
Discussion and Conclusions
The role of temperature and precipitation in the variation of glaciers and glacial lakes
From the 1970s to 2009, the area of glacial lakes in our study area has increased by 18.6%, while the glaciers have shrunk by 12.7%. Temperature and precipitation are considered as two major climatic factors controlling the changes of glaciers and glacial lakes. As the glaciers in this region are summer-accumulation type and glacier mass balance depends on temperature and precipitation during the summer, we analyzed the fluctuations of temperature and precipitation during May–September. Owing to the lack of meteorological stations in the study area, meteorological data from Bomi station (29.52˚ N, 95.46˚ E, 2730ma.s.l.; the nearest station, 70 km west of our study area) were used (Fig. 6). Also dominated by the Indian monsoon, data from the Bomi station can approximate the trend of climate change in the study area. The mean summer air temperature at Bomi station increased significantly (P < 0.0001, n = 42) during the period 1968–2009, at a rate of 0.27±0.04 ˚C(10 a)–1 (R = 0.69). In the most recent decade (1999–2009), the rate of mean summer air-temperature increase of 0.95±0.29 ˚C(10 a)–1 (R = 0.74, P < 0.01, n = 11) was nearly 3.5 times higher than during the full period of meteorological data collection.
Fig. 6. Variations of summer (May–September) temperature and precipitation from 1968 to 2009 recorded at Bomi meteorological station.
Although the summer precipitation trend during the period 1968–2009 was not statistically significant at the 5% level, it demonstrates a sharp diversion in 1998, with the former period (1968–1998) witnessing a slight increase and the latter period (1998–2009) an obvious decrease. Thus, glacier shrinkage and concomitant expansions of glacial lakes in the former period may be attributed to the severe summer air-temperature rise, which counter-balanced and overcame the effect of the slight increase in precipitation. In comparison, intensified warming in the last decade, together with decreasing precipitation, resulted in accelerated glacier retreat in this period.
Impact of glacier shrinkage on glacial lake expansion
As mentioned above, the moraine-dammed lakes expanded significantly from the 1970s to 2009, whereas the trough-valley lakes slightly decreased. This contrast in types of glacial lake within the same climate zone may be attributed to the difference in lake water supply. Compared with the trough-valley lakes mainly supplied by precipitation, the moraine-dammed lakes are fed jointly by precipitation and glacial meltwater. If we consider that the outflows of lake water from these two types of glacial lake are generally similar, with mother glacier melting the moraine-dammed lakes will receive more water supply than the trough-valley lakes, resulting in a general expansion of moraine-dammed lakes. For instance, of the 78 moraine-dammed lakes under study, whereas 16 lakes shrank between the 1970s and 2009, 62 other glacial lakes, including 27 new emergences, expanded at a rate of 35 m2 a–1 to >20 000 m2 a–1. In addition, glacier shrinkage will have another impact on moraine-dammed lakes: with the retreat of glacier ice, they may expand to flat areas left by retreating glaciers. This also explains the difference between the variations of moraine-dammed lakes and trough-valley lakes and shows the impact of glacier shrinkage on glacial lake expansion.
Potential threats of GLOF
Settlements in the Boshula mountain range are generally located on the terrace in the upper reaches of rivers with glacial meltwater as the major water supply. Though river runoff will increase for a short time with intensified glacier shrinkage, the water supply to support daily livelihoods is no longer sustainable, which has dramatic consequences for long-term well-being in the region. In addition, the more important effect of glacier retreat is the increased frequency of GLOFs or other glacial hazards. According to local historical documents at Chamdo, a devastating GLOF occurred on 12 June 1991, destroying the dams of a local hydrological power station, washing away 11 houses and 144 livestock and destroying extensive sections of the Sichuan–Tibet Highway. This GLOF also led to a severe landslide, forming a landslide-dammed lake measuring 0.78 km2 from the remote-sensing images for 2001 (Fig. 7a and b). The glacial lake shown in Figure 7c and d also caused GLOFs during 2005–09, carrying large glacial moraines downstream, with the deposit clearly identifiable from remote-sensing images. There is therefore an urgent need to intensify monitoring of glacial lakes and glaciers in the region and to set up early-warning systems in addition to assessing the risks posed by glacial lakes (Reference ShresthaShrestha, 2008).
Fig. 7. Two outburst floods revealed by satellite images. (a, b) A landslide-dammed lake was formed after outburst of a glacial lake: (a) Landsat TM image of 1988; (b) Landsat ETM+ image of 2001. (c, d) the deposited sediment bed can be observed as a white trace on the satellite image: (c) Landsat TM image of 2005; (d) ALOS image of 2009.
This study also revealed a potentially dangerous lake, Longlikum Co (29˚43'55" N, 96˚35'23" E; 4800ma.s.l.), which underwent continuous expansion from 0.17 to 0.48 km2 during 1968–2009, corresponding to the mother glacier decreasing from 11.9 to 10.9km2 (Fig. 5a). The terminus of the mother glacier with a precipitous glacial tongue is connected with the glacial moraine-dammed lake. Given the fact that Longlikum Co is only 4 km away from the outburst lake shown in Figure 7a, it poses a potential threat to the people and economy of the region. For these extremely dangerous lakes, appropriate engineering countermeasures should be adopted to lower the lake level as soon as possible (Reference WangWang, 2008).
Acknowledgements
We thank two anonymous reviewers and the scientific editor, D.R. MacAyeal, for constructive comments on the manuscript. We also thank D.R. Joswiak for kindly polishing the English. This work was supported by the Chinese Academy of Sciences (CAS)/State Administration of Foreign Experts Affairs (SAFEA) International Partnership Program for Creative Research Teams, the External Cooperation Program of the CAS (GJHZ0960), the Knowledge Innovation Program of the CAS(KZCX2-YW-145), the National Natural Science Foundation of China (40810019001) and the Program of Tibet Science and Technology Agency. 1 : 50 000 scale topographic maps and DEMs were supplied by the National Geomatics Center of China (NGCC). Meteorological data from Bomi station were kindly supplied by the China Meteorological Data Sharing Service System.
