The Cryosphere, 9, 849–864, 2015
www.the-cryosphere.net/9/849/2015/
doi:10.5194/tc-9-849-2015
© Author(s) 2015. CC Attribution 3.0 License.
The GAMDAM glacier inventory: a quality-controlled
inventory of Asian glaciers
T. Nuimura1,a,* , A. Sakai1,* , K. Taniguchi1,b , H. Nagai1,c , D. Lamsal1 , S. Tsutaki1,d,e , A. Kozawa1 , Y. Hoshina1 ,
S. Takenaka1 , S. Omiya1,f , K. Tsunematsu1,g , P. Tshering1,h , and K. Fujita1
1 Graduate
School of Environmental Studies, Nagoya University, Nagoya, Japan
at: Chiba Institute of Science, Choshi, Japan
b now at: Fukushima Prefecture, Fukushima, Japan
c now at: Japan Aerospace Exploration Agency, Tsukuba, Japan
d now at: National Institute of Polar Research, Tachikawa, Japan
e now at: Institute of Low Temperature Science, Hokkaido University, Sapporo, Japan
f now at: Civil Engineering Research Institute for Cold Region, Sapporo, Japan
g now at: Yamanashi Institute of Environmental Sciences, Fujiyoshida, Japan
h now at: Department of Geology and Mines, Thimphu, Bhutan
* These authors contributed equally to the manuscript.
a now
Correspondence to: T. Nuimura (tnuimura@cis.ac.jp) and A. Sakai (shakai@nagoya-u.jp)
Received: 29 April 2014 – Published in The Cryosphere Discuss.: 3 June 2014
Revised: 9 April 2015 – Accepted: 9 April 2015 – Published: 6 May 2015
Abstract. We present a new glacier inventory for highmountain Asia named “Glacier Area Mapping for Discharge
from the Asian Mountains” (GAMDAM). Glacier outlines
were delineated manually using 356 Landsat ETM+ scenes
in 226 path-row sets from the period 1999–2003, in conjunction with a digital elevation model (DEM) and highresolution Google EarthTM imagery. Geolocations are largely
consistent between the Landsat imagery and DEM due to
systematic radiometric and geometric corrections made by
the United States Geological Survey. We performed repeated
delineation tests and peer review of glacier outlines in order
to maintain the consistency and quality of the inventory. Our
GAMDAM glacier inventory (GGI) includes 87 084 glaciers
covering a total area of 91 263 ± 13 689 km2 throughout
high-mountain Asia. In the Hindu Kush–Himalaya range, the
total glacier area in our inventory is 93 % that of the ICIMOD
(International Centre for Integrated Mountain Development)
inventory. Discrepancies between the two regional data sets
are due mainly to the effects of glacier shading. In contrast, our inventory represents significantly less surface area
(−24 %) than the recent global Randolph Glacier Inventory,
version 4.0 (RGI), which includes 119 863 ± 9201 km2 for
the entirety of high Asian mountains. Likely causes of this
disparity include headwall definition, effects of exclusion of
shaded glacier areas, glacier recession since the 1970s, and
inclusion of seasonal snow cover in the source data of the
RGI, although it is difficult to evaluate such effects quantitatively. Further rigorous peer review of GGI will both improve the quality of glacier inventory in high-mountain Asia
and provide new opportunities to study Asian glaciers.
1 Introduction
The state and fate of Asian glaciers (Bolch et al., 2012) have
important implications for both regional water resources
(e.g. Immerzeel et al., 2010; Kaser et al., 2010) and future sea
level rise (e.g. Radić and Hock, 2011; Gardner et al., 2013).
Changes in glacier mass have been documented and/or estimated using a variety of approaches, such as in situ measurements (Fujita and Nuimura, 2011; Yao et al., 2012), numerical modelling (Immerzeel et al., 2010; Radić and Hock,
2011), and remote sensing (Matsuo and Heki, 2010; Jacob et
al., 2012; Kääb et al., 2012; Gardelle et al., 2013; Gardner et
al., 2013), in order to understand modern spatial variability in
high-mountain Asia. However, discrepancies exist among es-
Published by Copernicus Publications on behalf of the European Geosciences Union.
850
T. Nuimura et al.: The GAMDAM glacier inventory: a quality-controlled inventory of Asian glaciers
timates based on these different methods (e.g. Cogley, 2012;
Gardner et al., 2013).
A glacier inventory is a fundamental component of regional projections of mass balance and glacier discharge. For
example, glacier hypsometry (area–elevation distribution) directly affects estimates of mass balance, discharge, and modelled contribution to sea level rise (Raper and Braithwaite,
2005), while uncertainty in glacier outline influences estimates of mass changes using laser altimetry (Kääb et al.,
2012; Gardner et al., 2013). To support the Fifth Assessment
of the Intergovernmental Panel on Climate Change (IPCC),
the global Randolph Glacier Inventory (RGI) was published
(Pfeffer et al., 2014). However, while the majority of glacieroutline data used in that study were derived from recent satellite imagery, glacier extents in China were incorporated from
an inventory dating from 1956 to 1983. For brevity, we refer to this Chinese inventory as being from the 1970s (Shi,
2008). In December 2014, the second Chinese glacier inventory was released. However, this new data set has not
been incorporated into the RGI ver. 4.0 (Arendt et al., 2014)
employed in this study. Furthermore, a small number of the
glaciers used in the RGI are undated (Pfeffer et al., 2014).
In 2011, we launched a project, entitled Glacier Area Mapping for Discharge in Asian Mountains (GAMDAM), with
the goal of investigating the contribution of glacier meltwater to Asian river systems. Our initial and main purpose
for creating the glacier inventory is to estimate the elevation change of glaciers in Asian mountain areas, which is
equivalent to evaluating the effect of glacier volume change
on river run-off (Kääb et al., 2012). Here, we describe
the materials and procedures used to delineate glacier outlines over high-mountain Asia and show preliminary comparisons of our GAMDAM glacier inventory (GGI) to the
RGI and a glacier inventory produced by Bajracharya and
Shrestha (2011) (ICIMOD inventory; ICIMOD: the International Centre for Integrated Mountain Development, Kathmandu, Nepal) for the Hindu Kush–Himalayan (HKH) region.
Our target region covers high-mountain Asia between
67.4 and 103.9◦ E longitude and 27.0 and 54.9◦ N latitude,
which corresponds to the regions of central Asia, southwestern Asia, southeastern Asia, and Altay and Sayan of northern Asia in the RGI (Arendt et al., 2014; Pfeffer et al., 2014).
Pfeffer et al. (2014) have provided 62 606 km2 with 8.4 %
error, 33 859 km2 with 7.7 % error, 21 799 km2 with 8.3 %
error, and 1803 km2 with 10.3 % (< 54.9◦ N) error in these
regions, respectively.
2 Data sets
We analysed 356 Landsat level 1 terrain-corrected (L1T)
scenes in 226 path-row sets available from USGS EarthExplorer (http://earthexplorer.usgs.gov/), for the period 1999–
2003 (Table S1 in the Supplement), prior to the 2003 failThe Cryosphere, 9, 849–864, 2015
Russia
50˚N
Kazak
hstan
lia
Mongo
China
40˚N
30˚N
India
[N]
0
70˚E
1
80˚E
2
3
4
90˚E
5
100˚E
Figure 1. Footprints of Landsat scenes used in this study to delineate glaciers over high-mountain Asia. Colours refer to the number [N] of scenes used, while zero values (black squares) indicate
that no glaciers exist in that scene.
ure of the scan line corrector (SLC). Systematic radiometric and geometric corrections were performed for the L1T
imagery using the Global Land Survey digital elevation
model (DEM) 2000, which is a merged product comprising the Shuttle Radar Topography Mission (SRTM) DEM
(http://landsat.usgs.gov/LandsatProcessing_Details.php) and
other DEMs. We selected Landsat scenes with minimal cloud
and snow cover from paths 130–154 and rows 22–41 in the
Worldwide Reference System 2. In regions where seasonal
snow and cloud cover frequently hamper the identification
of glacier limits (e.g. Karakoram, Himalayas, and Hengduan
Shan), we used multiple scenes to increase accuracy (Fig. 1).
If we were unable to obtain perfect (i.e. free of both seasonal snow and cloud cover) imagery for a certain pathrow scene, we searched other partially clear images to obtain clear glacier outlines for whole glaciers. In addition, we
utilised both wintertime and summertime imagery, since the
former are unaffected by monsoon cloud or seasonal snow in
the monsoon-affected area and therefore can be used for the
delineation of glaciers on south-facing slopes. Details of this
methodology are given in Sect. 3.3. Images lacking glaciers
are shown in Fig. 1 as “zero scene”. Where appropriate Landsat ETM+ scenes were unavailable, we utilised Landsat TM
scenes collected prior to 1999 (two scenes, Table S1).
To delineate glacier outlines topographically, we used contours (20 m intervals) and slope distribution overlain on the
satellite scenes. These topographic data were generated using a gap-filled DEM from the SRTM (Jarvis et al., 2014)
and are compatible with the L1T imagery because the latter
is corrected using the SRTM. Although a recent report aswww.the-cryosphere.net/9/849/2015/
T. Nuimura et al.: The GAMDAM glacier inventory: a quality-controlled inventory of Asian glaciers
serts the ASTER GDEM has superior accuracy to the SRTM
(Hayakawa et al., 2008), that evaluation was made over a
non-glaciated region. Therefore, in our analysis of median
glacier elevation, we compared the SRTM and the most recent version of the ASTER GDEM version 2 (GDEM2, released in 2011) using the laser-altimetry product ICESat
GLA14 (Kääb, 2008), as described in Sect. 3.2.
We compared the GGI to both the RGI (Pfeffer et al.,
2014) and the ICIMOD glacier inventory (Bajracharya and
Shrestha, 2011). The RGI is a collection of digital outlines of
the world’s glaciers. Although the inventory includes some
misinterpreted polygons and limited attribute data, the RGI
remains the only glacier inventory with global coverage (excluding the ice sheets in Greenland and in Antarctica). Furthermore, it is the only data set comparable to our glacier
inventory. For our comparison here, we used version 4.0 of
the RGI (released 1 December 2014) (Arendt et al., 2014).
We also compared the GGI with the ICIMOD inventory
(Bajracharya and Shrestha, 2011), which covers the HKH
region (the Amudarya, Indus, Ganga, Brahmaputra and Irrawaddy basins) and Chinese region (the Salween, Mekong,
Yangtze, Yellow, and Tarim Interior basins, and Qinghai–
Tibetan Plateau). The ICIMOD inventory was generated
semi-automatically using more than 200 Landsat 7 ETM+
images taken between 2002 and 2008. Polygon data for the
HKH Region are available at http://apps.geoportal.icimod.
org/HKHGlacier/#. We employed these data to make detailed
inter-inventory comparisons of total glacier area for the HKH
region (Table 2).
3 Methods
3.1
Pre-processing
We used the Landsat scenes to generate both true-colour
(bands 3, 2, and 1 as RGB) and false-colour (bands 7, 4,
and 2 as RGB) composite images at 30 m resolution. Composite colour-band weight was adjusted automatically using
image contrast and GIS software. True-colour composite images were used primarily for glacier delineation. False-colour
images enabled us to differentiate ice from cloud owing to the
strong absorption of ice/snow in the SWIR compared with
clouds. Additionally, we employed thermal infrared (band 6)
at 60 m resolution to identify ice with a thin debris cover. Due
to the time-intensive nature of manually delineating glaciers
in high-resolution imagery (Bhambri et al., 2011), we did not
adopt a pan-sharpening method using 15 m resolution images
(band 8).
For debris-free glaciers, automated delineation using the
spectral ratio is more consistent and reproducible than manual delineation (Paul et al., 2013). For example, Fig. 2
compares manual and automated delineations of debris-free
glacier area using Landsat imagery that is free of cloud and
seasonal snow cover. It shows that glacier outlines generwww.the-cryosphere.net/9/849/2015/
851
Figure 2. Comparison of debris-free glacier delineation at
28.380◦ N, 86.472◦ E, using automated mapping derived from the
band ratio method (grid cells with band3/band5 > 1.8 are glacier
ice; Paul et al., 2013) and manual delineation. Background imagery
is Landsat false-colour (bands 7, 4, and 2 as RGB) composite imagery, taken on 5 January 2002, at path 145 row 039.
ated manually exhibit a difference of approximately ±1–
2 grid cells from those generated through automated mapping (Fig. 2). Furthermore, manual delineation often failed
to identify small glaciers. However, we did not employ automated mapping for the GGI for the following reason: in
high-mountain Asia there is an abundance of debris-covered
glaciers, particularly in the Himalaya and the Karakoram
ranges.
We generated contour lines, basin polygons, and slope distribution from SRTM data. Contour lines were then used
to delineate the termini of debris-covered glaciers and outlines of shaded glacier sections (see Sect. 3.2), and to divide
glacier polygons. To avoid misinterpretation of ice divides
due to potentially erroneous interpolation of the gap-filled
SRTM (Frey et al., 2012), we chose not to use basin polygons
to separate ice divides automatically. Instead, we referred to
contour lines in order to identify glacier divides.
3.2
Digital elevation models
We tested the SRTM output to that of the GDEM2, focusing on glacier polygons exhibiting inter-model elevation differences of > 100 m. Upon comparing the two DEMs to the
ICESat GLA 14 (Fig. 3a), we found that elevations in the
GDEM2 are consistent with those of ICESat, with a slight
bias of +40 m relative to ICESat. In contrast, elevations derived from the SRTM show a significantly negative bias of
−99 m relative to ICESat, as well as a larger analytical uncertainty (Fig. 3b).
The Cryosphere, 9, 849–864, 2015
852
T. Nuimura et al.: The GAMDAM glacier inventory: a quality-controlled inventory of Asian glaciers
a)
b)
Figure 3. Evaluation of DEMs based on ICESat GLA14. Where large (> 100 m) differences exist between SRTM and ASTER GDEM
(version 2) data sets, we compared modelled elevations to those of ICESat GLA14: (a) scattergram, (b) histogram, and (c) spatial distribution.
The distribution of elevation differences indicates that significant error in the SRTM occurs along the Karakoram and
Himalaya ranges and in the central Tien Shan, while significant error in the GDEM2 occurs locally throughout the
central Tibetan Plateau (Fig. 3c). In the Karakoram and Himalayas, high-relief topography resulted in numerous voids
in the original SRTM-3 product (Frey et al., 2012), thereby
resulting in the considerable errors observed there. Meanwhile, the low relief and decreased colour contrast of snowfields on the Tibetan Plateau may be responsible for the
large uncertainty in the GDEM2, which was created by optical stereo photogrammetry. Therefore, considering the relatively small uncertainty in the GDEM2 for the entire highmountain Asia region (Fig. 3a and b), we conclude that the
GDEM2 is more appropriate for glacier–altitude analysis in
high-mountain Asia.
3.3
Criteria for manual delineation
According to the Global Land Ice Measurements from Space
(GLIMS) protocol (Raup and Khalsa, 2007; Racoviteanu et
al., 2009), all perennial snow masses must be included as
glaciers, and only exposed ground can be excluded. Above
the bergschrund, ice bodies that are connected to the glacier
below shall also be considered part of the glacier. In our
study, however, we excluded steep headwalls even where
snow covered, since avalanching precludes development of
a permanent ice cover there. Although this avalanching is
an important source of glacier nourishment, steep headwalls
generally do not experience changes in surface elevation related to glacier mass fluctuations.
The Cryosphere, 9, 849–864, 2015
As satellite imagery documents only a single point in time,
distinguishing between glacier ice and snow-covered rock
headwalls and valley sides can be difficult. Consequently,
previous studies have delineated glacier outlines differently
at upper headwalls depending on the image source utilised.
On the Khumbu Glacier in Nepal, for example, variable
glacier-outline delineations along steep headwalls are the result of varying surface snow/ice conditions among the images
used (e.g. Salerno et al., 2008; Bolch et al., 2011; Thakuri et
al., 2014). In addition, dry slab avalanches are common on
headwalls steeper than 40◦ (McClung and Schaerer, 2006).
Therefore, where a headwall gradient exceeds 40◦ (coloured
in yellow to brown in Fig. 4b), we checked the surface condition of the wall in Google EarthTM and excluded those slopes
with a longitudinal plicate surface (Fig. 4c, purple) or thinly
snow-covered rock walls (Fig. 4c, orange). Figure 4 shows an
example of the steep headwalls excluded from our inventory.
Where glacier surfaces are largely free of debris, delineation of the ice surface was possible using false-colour composite imagery, which can distinguish glacier surfaces from
cloud cover (Fig. 5c and d). Similarly, we employed falsecolour imagery to identify boundaries of thinly dust-covered
glaciers (Fig. 6). By contrast, we used contour lines to delineate indistinct boundaries of debris-covered ablation zones
(Fig. 7a), since contour lines tend to exhibit clear inflections at their intersection with glacier outlines. On debrismantled glacier surfaces, areas of relatively thin debris cover,
which have relatively low surface temperature, were delineated using thermal infrared band (Fig. 7b). Identification of
thermokarst features, such as rugged surface topography, was
verified with high-resolution Google EarthTM images, which
www.the-cryosphere.net/9/849/2015/
T. Nuimura et al.: The GAMDAM glacier inventory: a quality-controlled inventory of Asian glaciers
853
Figure 4. Example of an excluded steep headwall of the Khumbu Glacier. The background is false colour (bands 7, 4, and 2 as RGB)
composite Landsat imagery, taken on 30 October 2000, at path 140 row 41 (99.38◦ E, 35.70◦ N) (a, b). Steep (> 40◦ ) headwalls (c, d, e)
were not included as glacier area, since accumulation cannot occur on longitudinally plicate surfaces (d) or where rock surfaces are only
thinly snow-covered (e). Not all slopes with > 40◦ inclination were excluded from the GGI: gradient was used as a guide only. Glacier
outlines of RGI ver. 4.0 at the Khumbu Glacier equate to those of the ICIMOD glacier inventory.
can identify exposed ice cliffs on the debris-covered glacier
(Fig. 7c). Non-glacial lakes surrounded by smooth terrain can
also be identified in Google EarthTM imagery (Fig. 7d). This
method is effective for the delineation of terminus outlines
on debris-covered glaciers.
We used both winter and summer Landsat images for one
path-row scene to avoid shade, cloud, and seasonal snow
cover. Examples of glacier-outline delineations made using
these two types of imagery are shown in Fig. 5. The Landsat imagery exhibits greater seasonal snow cover on southfacing slopes (Fig. 5a), whereas imagery collected on 2 August 2002 shows shading on north-facing slopes (Fig. 5b).
Therefore, glacier outlines in shaded areas are delineated
based on the image of Fig. 5a, while glaciers on south-facing
slopes are delineated using the image of Fig. 5b. Landsat
imagery taken on 20 October 2001 contains partial cloud
cover but less shading (Fig. 5c), whereas imagery taken on
1 August 2001 contains no cloud cover but greater shading
(Fig. 5d). In this case, the cloud-obscured glacier area shown
in Fig. 5c was delineated using the image shown in Fig. 5d
(pink line), while shaded areas in Fig. 5d were delineated
using the image shown in Fig. 5c (yellow line). In this delineation phase, we made different polygon files for each image
source (i.e. one path-row scene comprises multiple polygon
www.the-cryosphere.net/9/849/2015/
file sets). We then added the Landsat image ID as attribute
data for each glacier when merging these polygon data.
Furthermore, where we could obtain clear (i.e. free of seasonal snow and cloud) wintertime but not summertime imagery, slope transition zones (indicated by a change in the
spacing of contour lines) are used to indicate the glacier outline (Paul et al., 2004) in areas of shade, as shown in Fig. 8.
Additionally, SLC-off scenes (Landsat ETM+ post-dating
May 2003) were used to identify ambiguous glacier boundaries when clear Landsat L1T imagery or Google EarthTM
imagery was unavailable, though we note their acquisition
dates are different from those of L1T scenes. Some glacierlike areas visible on Landsat scenes (Fig. 9a) were identified
later as seasonal snow in Google EarthTM images (Fig. 9b).
In the final aggregation process, we excluded small
glaciers (< 0.05 km2 ), which is the same as the threshold
employed by Rastner et al. (2012). The minimum area of
0.05 km2 corresponds to about 55 grids of Landsat images
(band 1–5, 7) with 30 m resolution.
3.4
Quality control
Considerable variability among measurements of glacier
area is possible owing to different interpretations of glacier
boundaries (Paul et al., 2013), as well as personnel changes
The Cryosphere, 9, 849–864, 2015
854
T. Nuimura et al.: The GAMDAM glacier inventory: a quality-controlled inventory of Asian glaciers
Figure 5. Two examples where multiple images were required to
delineate glacier outline for a single path-row scene because of
seasonal cover/partial cloud cover or shade: (a, b) at 76.856◦ E,
32.512◦ N (path 147 row 38); (c, d) at 79.357◦ E, 30.824◦ N
(path 145 row 039). All background imagery is false colour
(bands 7, 4, and 2 as RGB). The Landsat imagery, taken on 15 October 2001.
Figure 6. Thinly dust-covered glaciers located at 42.316◦ N,
78.833◦ E (path 148 row 31). Identification of such glaciers is
problematic (i.e. they show only black surfaces) using true-colour
(bands 3, 2, and 1 as RGB) composite imagery (a) but relatively
straightforward using false-colour (bands 7, 4, and 2 as RGB) composite imagery (b). Background imagery was acquired on 25 August 2002.
The Cryosphere, 9, 849–864, 2015
Figure 7. Example of glacier outlines generated for the GGI using
contour lines at 20 m intervals. The full extent of debris-covered
glacier surfaces can be identified using both the inflections of contour lines (a) and thermal band imagery (band 6) (b). Background
imageries are false colour (bands 7, 4, and 2 as RGB) (a) and thermal band (band 6) (b) Landsat imagery (30.911◦ N, 79.088◦ E) acquired on 1 August 2001, at path 145 row 39. Thermokarst features
and supra-glacial lakes with ice cliffs (27.911◦ N, 86.949◦ E) (c)
and a non-glacial lake surrounded by smooth terrain (28.083◦ N,
86.471◦ E) (d) are used to differentiate between debris-covered
glacier surfaces and ice-free areas. Both (c) and (d) are screenshots
from Google EarthTM , © 2014 DigitalGlobe.
over the course of the project. Figure 10 depicts several examples where glacier boundaries were delineated differently.
For example, orange lines depict the erroneous inclusion of
steep rock walls (indicated by yellow arrow) in an accumulation zone at 28.74◦ N, 84.39◦ E. Google EarthTM imagery
reveals partially exposed bedrock on steep headwalls, which
were not included as glacier area according to our criteria (Fig. 10a). In a debris-covered ablation zone (28.78◦ N,
84.32◦ E), yellow dotted circles indicate areas misidentified
as glacier ice. Red, blue, and light-green lines represent
correctly delineated debris-covered glacier area (Fig. 10b).
Therefore, we conducted a total of five delineation tests (Table S2) in order to ensure adherence to the delineation criteria
and to homogenise the quality of our inventory. In these five
tests, we evaluated delineations made by each operator and
provided feedback in order to minimise differences among
output and to improve delineation accuracy. Accordingly, the
errors described above were corrected and the operators were
advised of these problems.
www.the-cryosphere.net/9/849/2015/
T. Nuimura et al.: The GAMDAM glacier inventory: a quality-controlled inventory of Asian glaciers
855
Figure 9. Glacier-like seasonal snow cover seen in false-colour
(bands 7, 4, and 2 as RGB) composite imagery at path 140 row 41
(27.984◦ N, 87.657◦ E), taken on 17 October 2001 (a), and Google
Earth, © 2014 DigitalGlobe screenshots of the same location (b). We
can distinguish between such snow cover and glacier ice using highresolution Google EarthTM imagery, which reveals that the surface
is undulating and has the appearance of thin snow on a rock surface.
3.5
Figure 8. Example of glacier outlines generated with contour lines
at path 140 row 41 of Landsat imagery (27.991◦ N, 86.730◦ E),
taken on 30 October 2000 (a), and Google Earth, © 2015 DigitalGlobe screenshots of the same location (b). When summertime
(high solar angle) Landsat imagery lacking seasonal snow cover was
unavailable, we employed wintertime (low solar angle) imagery. In
that case, glacier outlines in shaded areas were delineated by reference to slope-change boundaries indicated by contour intervals.
Initial delineation of glacier outlines was carried out by
11 operators over a period of 20 months, during which time
the quality of delineation might have varied significantly. Operators can be classified as those with field experience on
glaciers (e.g. with glaciological knowledge and experience
of remote sensing) and those without. Consequently, glacier
polygons delineated by non-experienced operators were reviewed by field-experienced operators. Figure 11 shows an
example where the second operator corrected the polygon
delineated by the first, by using different source imagery.
Whereas the first operator delineated glacier outlines using Landsat imagery with a low solar angle and seasonal
snow cover (Fig. 11a), the second employed summertime imagery containing less seasonal snow cover (Fig. 11b), thereby
enabling shaded glacier areas to be incorporated. Following this peer review of glacier outlines, topological properties were checked. For example, overlapping polygons may
cause overestimation of glacier area (Fig. S1a in the Supplement), while irregular polygons (e.g. self-intersecting polygons; Fig. S1b) cannot represent the glacier area accurately.
Such mis-delineations were detected automatically by GIS
functions and then corrected.
www.the-cryosphere.net/9/849/2015/
Attribute data
We attached 15 attributes to every glacier analysed. Each
glacier is assigned a unique ID consisting of a sequential 6digit number, beginning with id = 000001 in p130r037 and
ending with id = 087084 in p154r033. The highest ID corresponds to the total number of glaciers in the GGI. Path, row,
granule ID, and acquisition date of the Landsat scene, as well
as the name of the operator, are included to enable traceability and validation by others. In addition, basic geographic information – such as longitude, latitude, and area – is provided
together with elevation data (mean, median, maximum, minimum, range, and mid-range elevation), which were derived
from GDEM2 (Table S3). We also provided records of the
peer review and revision of glacier outlines (reviewer name
and date) that were performed on each scene (Table S4).
These records will permit others to validate our inventory and
analyse changes in glacier extent over time using another inventory.
3.6
Evaluation of uncertainties
We evaluated uncertainty in glacier delineation using the results of five separate delineation tests (Fig. 12). Here, uncertainty is defined as one normalised standard deviation, calculated as the standard deviation of the glacier area measured by different operators divided by the mean value of
the glacier area measured by all operators. Figure 12 shows
that the normalised standard deviation decreases with increasing glacier area. Specifically, large glaciers (> 2.5 km2 )
exhibit lower normalised standard deviations (< 15 %) than
smaller glaciers (< 2.5 km2 area; > 25 % standard deviation).
A debris-covered glacier gives a normalised standard deviation of approximately 10 %. In summary, the uncertainty of
delineated glacier areas in the GGI is less than 25 % for small
glaciers and ∼ 15 % for large glaciers. Therefore, we expect
The Cryosphere, 9, 849–864, 2015
856
T. Nuimura et al.: The GAMDAM glacier inventory: a quality-controlled inventory of Asian glaciers
Figure 10. Examples of delineation tests, in which coloured lines represent glacier outlines delineated by different operators. Both background images (left panels) are true-colour composites of the Landsat ETM+ scenes. Right-hand panels in both (a) and (b) are Google
EarthTM screenshots, © 2014 DigitalGlobe.
Figure 11. Example of glacier-outline retrieval by a second operator using Landsat imagery path 133 row 035 (35.70◦ N, 99.38◦ E). Background images are false-colour (bands 7, 4, and 2 as RGB) composites taken on (a) 7 January 2003 and (b) 12 July 2001, in addition to
Google EarthTM imagery (c).
approximately 15 % uncertainty in our glacial area computation. In its current form, the GGI has a relatively large uncertainty, which incorporates all differences in glacier outlines
delineated by 5–8 operators. We anticipate that rigorous peer
review by field operators will reduce this uncertainty in the
future.
The Cryosphere, 9, 849–864, 2015
4 Results
4.1
Distribution of glaciers and their median elevations
We delineated a total of 87 084 glaciers with a total area of
91 263 ± 13 689 km2 in high-mountain Asia (Table 1). Figure 13 shows the distribution of median glacier elevations
www.the-cryosphere.net/9/849/2015/
T. Nuimura et al.: The GAMDAM glacier inventory: a quality-controlled inventory of Asian glaciers
857
Table 1. Summary of glaciers in the GGI, ICIMOD inventory, and the RGI, excluding glaciers smaller than 0.05 km2 . The uncertainty in
RGI ver. 4.0 was calculated using the error estimation equation (Eq. 1) in Pfeffer et al. (2014).
GGI
Amudarya, Indus, Ganges,
Brahmaputra, and Irrawaddy basins
Total area [km2 ]
Excluded small glaciers
43 570 ± 6536
6623
High-mountain Asia
Total area [km2 ]
Excluded small glaciers
91 263 ± 13 689
11 181
Normalised standard deviation [ % ]
100
●
ICIMOD
46 826
4060
–
–
RGI 4.0
57 285 ± 4212
4495
119 878 ± 9201
6149
Debris−covered
Debris−free
80
60
●
40
●
●
●
●
●● ●
●
● ●● ● ● ●
●
●
●
●●
●●
●
●
●
●
● ●●
● ●
● ●●●●
● ●● ● ●
●● ● ●●●
●
●
●
● ●●
●
●●
●●
●
●
●
●
●●●● ●
●
● ● ●●●●●
●
●●
●●
●
● ● ●●●
● ● ●●
●
●
●
●
●
●●
●
● ●●●●●●
●●●●●
●
●●
●● ●●
●●●●
● ●
●●● ●●● ●● ●
● ●
●
●●
●
●●
●
●
●
20
0
0.01
0.10
1.00
10.00
2
Mean glacier area [ km ]
Figure 12. Normalised standard deviation of glacier area, based on
delineations by different operators divided by the mean glacier area
for all operators.
based on the GDEM2 and contour lines. Contour values represent the area-weighted average of median elevations within
each 0.5◦ grid cell. The area-weighted average of median
elevations was based on the concept that the median elevation of larger glaciers is more representative of each region, because the mass balance (particularly accumulation)
of smaller glaciers is affected by local topographic effects,
such as snow drifting. This figure also shows the distribution of snow-line elevations estimated by Shi et al. (1980)
and Shi (2008). The estimation method is described in Shi
et al. (1980) as follows: “some firn line elevations were determined on the spot, while most were diagnosed according to topographical maps or calculated by Hôfer’s method”.
Large-scale features evident in the distribution of snow-line
elevations are consistent with our median-glacier elevations.
These include a pronounced trough in southeastern Tibet,
caused by intense precipitation along the Brahmaputra River
(Liu et al., 2006; He, 2003), and a crest in western Tibet
resulting from the prevailing arid, cold climate (Shangguan
et al., 2007). Median elevation increases with distance from
the moisture source, while areas of low median elevation
are shown in the northwest, in the Himalaya and Karakoram
ranges, as reported by Bolch et al. (2012).
www.the-cryosphere.net/9/849/2015/
Figure 13. Distribution of glaciers in the GGI coloured by median elevation. Purple dashed contours indicate snow-line elevations
used in the Chinese glacier inventory (Shi, 2008).
4.2
Comparison of inventories in the HKH range
We compared our GGI with the ICIMOD inventory (Bajracharya and Shrestha, 2011) in the HKH region, excluding from our assessment glaciers with an area of < 0.05 km2
so as to standardise the delineation of minimum glacier size
among operators. In the following analysis, altitude data
for both glacier inventories were derived from the GDEM2.
Glacier area for each basin is given in Table 2. In addition, we
compared the area for each area class and altitude (Fig. 14a
and c). Although the total glacier area in the HKH range was
less (−7 %) in the GGI than in the ICIMOD inventory, totals
for each area class are strongly consistent between the inventories, with the exception of glaciers with areas between
16 and 32 km2 (Fig. 14a). In contrast, glacier hypsometry for
the HKH range is less in the GGI than in the ICIMOD inventory for elevations between 5000 and 7000 m (Fig. 14c).
The Cryosphere, 9, 849–864, 2015
858
T. Nuimura et al.: The GAMDAM glacier inventory: a quality-controlled inventory of Asian glaciers
Table 2. Summary of glaciers in the GGI, ICIMOD inventory, and the RGI 4.0. The uncertainty in the RGI ver. 4.0 was calculated using the
error estimation equation (Eq. 1) in Pfeffer et al. (2014).
GGI
ICIMOD inventory
area
Area
[km2 ]
[km2 ]
[km2 ]
Amu Darya
Indus
Ganges
Brahmaputra
Irrawaddy
Salween
Mekong
Yangtze
Yellow
Tarim Interior
Qinghai–Tibetan Interior
2498
23 668
7537
9803
64
1318
225
1574
132
2640
7747
2566
21 193
9012
14 020
35
1352
235
1660
137
2310
7535
Total
57 204
60 054
The glacier number, area, and median elevation for both
inventories were compared for each 0.5◦ grid cell (Fig. 15).
Root mean square differences for these values are 28 %,
26 %, and 77.9 m, respectively, and the inclinations of the
fitted lines are close to 1. We also evaluated the spatial distributions of glacier number, area, and the area-weighted
mean of median elevation for each 0.5◦ grid cell to identify differences between the GGI and the ICIMOD inventories (Fig. 16). We found that glacier area and number are
greater in the GGI for the southern Karakoram and western Himalaya, but lesser in the northern Karakoram and central Himalaya (Fig. 16b). Moreover, while the total glacier
area is less in the GGI than in the ICIMOD inventory, the
number of glaciers in the Hengduan Shan is greater in the
GGI. The median elevation of glaciers is considerably lower
(200–300 m) in the GGI than in the ICIMOD inventory for
the northern Hindu Kush and northern Karakoram (Fig. 16c),
whereas in the central Himalaya the discrepancy is approximately 100 m. Such inconsistency in median elevation for
the northern Karakoram may be the product of inaccurate delineation in the shaded upper portions of glaciers (details of
required GGI revisions are given in Table S5), whereas the
discrepancy in the central Himalaya is probably due to our
exclusion of steep headwalls.
4.3
Comparison of inventories in high-mountain Asia
To evaluate our entire inventory, we compared glacier area
in the GGI and RGI across high-mountain Asia (27.0–
54.9◦ N, 67.4–103.9◦ E), focusing on glaciers > 0.05 km2 in
area. Whereas total glacier area in the GGI is comparable
to the ICIMOD inventory for the HKH range, this value is
significantly lower (by 28 615 ± 22 890 km2 , or −24 ± 19 %)
relative to the RGI for high-mountain Asia. Glaciers in the
The Cryosphere, 9, 849–864, 2015
RGI 4.0
Difference
Area
Difference
[%]
[km2 ]
[km2 ]
[%]
68
−2475
1475
4217
−29
34
10
86
5
−330
−212
3
−10
20
43
−45
3
4
5
4
−13
−3
3154 ± 256
26 018 ± 1750
10 621 ± 824
17 419 ± 1373
73 ± 9
2198 ± 210
586 ± 49
2441 ± 183
189 ± 16
2768 ± 159
10 000 ± 796
656
2350
3084
7616
9
880
361
867
57
128
2253
26
10
41
78
14
67
160
55
43
5
29
2850
5
75 466 ± 5625
18 262
32
RGI are larger than those in the GGI (Fig. 14b). Furthermore, glacier area between 4000 and 6000 m elevation is significantly greater in the RGI hypsometry than in the GGI
(Fig. 14d). We suggest that these differences between inventories are due to four potential factors: (1) the result of
real changes in glacier extent on the Tibetan Plateau since
the 1970s (Ding et al., 2006; Li et al., 2008); (2) the omission of shaded glacier areas in the GGI; (3) the exclusion of
steep headwalls in the GGI; and (4) the inclusion of seasonal
snow cover at Hengduan Shan (Gardelle et al., 2013) and at
Western Nyainqentanglha (Bolch et al., 2010) in the RGI, for
which the data source is the first Chinese glacier inventory
(Shi, 2008).
We also performed area comparison tests between the GGI
and the GlobGlacier inventory (Frey et al., 2012) for the region covered by the latter. The GlobGlacier inventory was
the source data for the RGI and so has already been integrated into the RGI with minor modification. The GGI and
GlobGlacier give glacier areas of 8007 and 9270 km2 , respectively, corresponding to a difference of 1263 km2 , or
15 %. This area difference is consistent with the glacier definition employed by the GlobGlacier, which, like the RGI,
includes upper steep headwall areas. This comparison shows
that the considerable disparity in area between the GGI and
RGI is due largely to differences in glacier definition in the
western part of Himalaya that is covered by the GlobGlacier
inventory.
Additionally, we compared total glacier area for the
HKH regions according to the GGI against values from the
RGI, the ICIMOD inventory (including Chinese basins (Bajracharya and Shrestha, 2011); Table 2), the inventory of
Bolch et al. (2012), and GlobGlacier (Frey et al., 2012)
(Table 3). The data sources for the inventory of Bolch et
al. (2012) include the ICIMOD, GlobGlacier, and Chinese
www.the-cryosphere.net/9/849/2015/
T. Nuimura et al.: The GAMDAM glacier inventory: a quality-controlled inventory of Asian glaciers
859
Table 3. Comparison of regionally aggregated total glacier areas from the GGI, Bolch et al. (2012) inventory, ICIMOD inventory, and the
RGI. The uncertainty in the RGI ver. 4.0 was calculated using the error estimation equation (Eq. 1) in Pfeffer et al. (2014).
GGI
Bolch et al. (2012) inventory
area
Area
[km2 ]
[km2 ]
[km2 ]
Difference
Karakoram
Western Himalaya
Central Himalaya
Eastern Himalaya
17 385
8402
8221
2836
17 946
8943
9940
3946
Total
36 845
40 775
ICIMOD inventory
Area
RGI 4.0
Difference
[%]
[km2 ]
[km2 ]
561
541
1719
1110
3
6
21
39
13 646
7696
9575
3008
3930
11
33 924
Area
Difference
[%]
[km2 ]
[km2 ]
[%]
−3739
−706
1354
172
−22
−8
16
6
19 680 ± 1052
9585 ± 869
11 502 ± 899
4605 ± 362
2295
1183
3281
1769
13
14
40
62
−2921
−8
45 372 ± 3182
8527
23
glacier inventories, as well as their own mapping. In the
Karakoram, most of the data are derived from the ICIMOD inventory, with smaller contributions from the Chinese
glacier inventory and their own mapping. For the western Himalaya, source data are derived largely from GlobGlacier,
with contributions from the ICIMOD inventory, whereas data
for the central Himalaya are sourced primarily from the ICIMOD inventory, with additional data from GlobGlacier. Similarly, the ICIMOD inventory is the primary data source for
the eastern Himalaya, with additional input from the Chinese
glacier inventory. Regional summaries for each inventory are
given in Tables 2 and 3, and are shown in Fig. 17. Source
satellite data for each were Landsat ETM+ images taken after 2000, meaning any time difference among the inventories is minor. Discrepancies in glacier area between the GGI
and the ICIMOD inventory (including China) and Bolch et
al. (2012) inventory are 7 and 11 %, respectively. As above,
we suggest these inconsistencies result from the omission of
shaded glacier areas and the elimination of high-angle glacier
areas from the GGI, as well as different interpretations of
debris-covered glaciers.
5 Discussion
5.1
Figure 14. Size distributions of glacier area in the Hindu Kush–
Himalaya range from the GGI and the ICIMOD inventories (a),
and in high-mountain Asia from the RGI and GGI (b). Glacier hypsometries for the Hindu Kush–Himalaya range from GGI and ICIMOD (c) and high-mountain Asia derived from the GGI and RGI
in 100 m bins (d). Only glaciers > 0.05 km2 in area are included in
the calculation for each inventory. All hypsometries were calculated
using the GDEM2.
www.the-cryosphere.net/9/849/2015/
Intended purpose of the GGI
We have excluded seasonally snow-covered areas and steep
headwalls from our glacier delineations because our objective is to estimate total elevation change of glaciers. Kääb
et al. (2012) reported that the inclusion of steep flanks,
ice patches, ice-cored moraines, and rock glaciers can result in considerable differences among estimates of elevation change, particularly in the Himachal Pradesh, Nepal, and
Bhutan Himalayas. Thus, in excluding such glacier ice-free
areas, the GGI is well suited for estimating glacier elevation
change.
While our strict criteria for the exclusion of steep upper
headwalls will allow reliable elevation change of glaciers, we
note that changes in glacier area cannot be estimated by comparison of the GGI to other glacier inventories, since each
The Cryosphere, 9, 849–864, 2015
860
T. Nuimura et al.: The GAMDAM glacier inventory: a quality-controlled inventory of Asian glaciers
Figure 15. Scattergrams of (a) glacier number, (b) glacier area, and (c) area-weighted mean of median glacier elevation in each 0.5◦ grid
cell of the ICIMOD inventory, plotted against the GGI in the Hindu Kush–Himalaya range. The dashed lines indicate 1 : 1 correspondence
between ICIMOD and GGI. Also shown is the root mean square number (or area) difference ratio (%) against average number (or area) of
ICIMOD. The solid lines are the best-fitting linear equations. All median elevations were calculated using the GDEM2.
will employ different criteria for delineating glacier boundaries (e.g. including all snow- or ice-covered walls). Assessment of area change, therefore, should only be made using
the same definition criteria.
5.2
Required revision for GGI by comparison with
other inventories
Our analysis shows that the total glacier area in the GGI is
only 7 % less than that of the ICIMOD inventory for the HKH
ranges (Table 1). However, we note that considerable differences in the spatial distribution of glacier area and median
elevation exist between the two inventories (Fig. 16b and c).
We also analysed the distributions of area difference in both
the upper and lower zones of glaciers, distinguished by the
median elevation, for each 0.5◦ grid cell. The normalised difference (%) is calculated as follows:
normalized difference of glacier area =
VICIMOD − VGGI
,
VGGI
(1)
where the variable (V ) is the glacier area in each 0.5◦ grid
cell, and the subscript denotes the inventory. Area-weighted
means of median elevation of GGI in each 0.5◦ grid cell were
used to distinguish the upper and lower zones for both inventories (Fig. 18).
Here, we investigate the differences in glacier area and median elevation between the ICIMOD inventory and GGI, focusing on several regions (Fig. S2). We also summarise the
considerable revisions required for both glacier inventories
in Table S5. The disparity in regional glacier area between
the GGI and ICIMOD inventories cannot be explained by
long-term changes in glacier area, since the acquisition dates
of the source Landsat imagery are similar for both. Instead,
we note that both inventories include topography where areas of shaded glacier ice have been omitted, and that this
effect is highly variable regionally. For example, the GGI exhibits a smaller total glacier area than the ICIMOD inventory as a result of our inclusion of wintertime (and thereThe Cryosphere, 9, 849–864, 2015
fore low solar angle) Landsat imagery (see Sects. 2 and 3.1).
Similarly, median elevations for the eastern Pamir are notably lower (> 200 m) in the GGI than in the ICIMOD inventory (Fig. 16c), owing to the erroneous exclusion of shaded
glacier areas.
Further discrepancy between the two inventories is caused
by the variable identification of debris-covered glaciers. For
example, the ICIMOD inventory identified debris-covered
glaciers in the Hengduan Shan that are absent from our inventory. Such inconsistencies indicate that future revisions
of glacier outlines must focus on (1) shaded glacier area and
(2) debris-covered glaciers. Specifically, we will incorporate summertime Landsat images in order to delineate those
glacier surfaces obscured by shade and use high-resolution
Google EarthTM imagery to conduct a closer investigation of
debris-covered glaciers. Finally, our exclusion of steep headwalls that are unaffected by glacier mass balance potentially
discounts glaciers located on steep ground, resulting in an
underestimation of total ice volume and median elevations
in the GGI. In Landsat scenes where clear summertime imagery was unavailable, we employed heavily shaded wintertime imagery. Glacier outlines were then delineated with reference to contour lines derived from SRTM (see Sect. 3.3)
(Fig. 8). However, differences in resolution between Landsat imagery (30 m) and SRTM data (90 m) mean that shaded
glacier delineations based on contours are inherently less precise. To minimise this limitation in future revisions, the use
of both simple band ratios (band 3/band 5) and additional
thresholds in band 1 (blue) will help delineate shaded portions of debris-free glaciers (Rastner et al., 2012). Ultimately,
revision of shaded glacier boundaries will reinforce our confidence in the quality of glacier outlines incorporated in the
GGI.
www.the-cryosphere.net/9/849/2015/
T. Nuimura et al.: The GAMDAM glacier inventory: a quality-controlled inventory of Asian glaciers
861
40˚N
a)
30˚N
−150−100−50 0
70˚E
[N]
50 100 150
80˚E
100˚E
90˚E
40˚N
b)
30˚N
Figure 17. Overview of area comparisons and catchment outlines
for each sub-region.
−100 −60 −20
70˚E
20
80˚E
60
40˚N
[ km2 ]
100
90˚E
a)
100˚E
40˚N
c)
30˚N
30˚N
[%]
−50 −30 −10
70˚E
−250 −150 −50
70˚E
80˚E
50
100˚E
90˚E
b)
30˚N
[%]
−50 −30 −10
70˚E
Comparison between SRTM DEM and ASTER
GDEM ver. 2
As described above, we used the gap-filled SRTM DEM to
support our delineation of glacier outlines and the GDEM2 to
calculate median elevation (Fig. 13). Here, we compare areaweighted median elevations of glaciers derived using the two
models for each 0.5◦ grid cell (Fig. 19). In comparing the
SRTM DEM with the GDEM2 data, we identified zones of
lower median elevation in the SRTM at the southern edge
of the Tibetan Plateau (30–31◦ N, 78.5–90.0◦ E), the western Himalaya, and parts of Hengduan Shan and the central
www.the-cryosphere.net/9/849/2015/
50
100˚E
Figure 16. Differences among (a) glacier number, (b) glacier area,
and (c) area-weighted mean median elevation in the ICIMOD inventory and GGI (i.e. ICIMOD − GGI) for each 0.5◦ grid cell in
the Hindu Kush–Himalaya range. Calculations were based on the
GGI in the same area as the ICIMOD glacier inventory. Median
elevations of glaciers for both inventories were derived from the
GDEM2.
5.3
80˚E
30
40˚N
[m]
150 250
90˚E
10
80˚E
10
30
90˚E
50
100˚E
Figure 18. Normalised differences between glacier area in the
(a) upper and (b) lower zones of the ICIMOD inventory and GGI
for each 0.5◦ grid cell in the Hindu Kush–Himalaya range.
Tien Shan (Fig. 19a). For both models, these regions show a
larger standard deviation (40–280 m) in the difference in median glacier elevation (Fig. 19b). Evaluations by ICESat also
suggest significant error in the SRTM DEM (Fig. 3c), which,
if true, indicates regions of incorrectly interpolated data in
the model. In the context of the GGI, application of invalid
data to the Global Land Survey DEM during our geometric
The Cryosphere, 9, 849–864, 2015
862
T. Nuimura et al.: The GAMDAM glacier inventory: a quality-controlled inventory of Asian glaciers
a)
50˚N
40˚N
30˚N
−100 −60 −20
70˚E
20
80˚E
60
[m]
100
100˚E
90˚E
b)
50˚N
glacier number, area, and median elevation between the two
inventories suggest significant regional variability. We propose that this variability is due primarily to the omission of
shaded glacier areas from the GGI, resulting from our inclusion of wintertime Landsat imagery.
Our comparison of the entire GGI and RGI in highmountain Asia revealed that total glacier area is significantly
less (−24 %) in the GGI than in the RGI (Table 1). The
large discrepancies in glacier area between the two inventories are probably due to area change since the 1970s (e.g. the
1950s to 1970s in most of China in the RGI), the exclusion of
shaded glacier areas from the GGI, and the inclusion of seasonal snow cover in the source data of the RGI. The definition of glacier extent, in particular the inclusion or exclusion
of upper steep headwalls, is another potential cause of differences in total glacier area between the two inventories. This
interpretation is supported by our comparison of the GGI and
the GlobGlacier inventory in the western Himalaya, where
the greater (15 %) glacier area in the GlobGlacier inventory
reflects the inclusion of steep upper headwalls as glacier area.
To evaluate the contribution of these potential causes, further rigorous peer review by field-experienced operators is
required before we can quantify the effects of recent changes
in glacier area or differences in the criteria used to identify
glacier area (e.g. steep headwalls).
40˚N
The Supplement related to this article is available online
at doi:10.5194/tc-9-849-2015-supplement.
30˚N
0
70˚E
50
[m]
100 150 200 250
80˚E
90˚E
100˚E
Figure 19. (a) Differences between area-weighted means of median
elevations in the GGI derived from SRTM and those from GDEM2
(i.e. SRTM − GDEM2). (b) Standard deviations of the difference
in median elevation of each glacier derived by SRTM and GDEM2
models. Grid cell size is 0.5◦ for both.
correction of Landsat imagery would result in erroneous orthorectification and potentially imprecise glacier delineation.
6 Conclusions
We present a new glacier inventory for high-mountain Asia
based primarily on ortho-calibrated Landsat ETM+ scenes
from the period 1999–2003. The total glacier area determined by the GGI for the HKH range is similar to that of
the ICIMOD inventory. Nonetheless, spatial differences in
The Cryosphere, 9, 849–864, 2015
Author contributions. The two first authors, T. Nuimura and
A. Sakai, contributed equally and shared the responsibilities for
this paper. K. Fujita contributed to discussions and writing the paper. H. Nagai contributed to designing the delineation methodology.
S. Takenaka contributed to the management and synchronisation of
the inventory data. All other co-authors contributed to delineating
glacier outlines.
Acknowledgements. We thank T. Bolch, G. Cogley, M. Pelto,
S. R. Bajracharya, and F. Paul for their helpful comments that led to
a considerably improved manuscript. We thank S. R. Bajracharya
and B. Shrestha, without whom we could not have made a detailed
comparison of our GAMDAM glacier inventory with the ICIMOD
inventory. We also thank the RGI consortium for use of the RGI,
the USGS for Landsat imagery, and CGIAR-CSI for gap-filled
SRTM DEMs. We are grateful to S. Okamoto for assistance in
selecting Landsat imagery. This project was supported by a grant
from the Funding Program for Next Generation World-Leading
Researchers (NEXT Program, GR052) and Grants-in-Aid for
Scientific Research (26257202) of the Japan Society for the
Promotion of Science.
Edited by: T. Bolch
www.the-cryosphere.net/9/849/2015/
T. Nuimura et al.: The GAMDAM glacier inventory: a quality-controlled inventory of Asian glaciers
References
Arendt, A., Bliss, A., Bolch, T., Cogley, J. G., Gardner, A. S., Hagen, J.-O., Hock, R., Huss, M., Kaser, G., Kienholz, C., Pfeffer, W. T., Moholdt, G., Paul, F., Radić, V., Andreassen, L., Bajracharya, S., Barrand, N., Beedle, M., Berthier, E., Bhambri,
R., Brown, I., Burgess, E., Burgess, D., Cawkwell, F., Chinn, T.,
Copland, L., Davies, B., de Angelis, H., Dolgova, E., Filbert, K.,
Forester, R., Fountain, A., Frey, H., Giffen, B., Glasser, N., Gurney, S., Hagg, W., Hall, D., Haritashya, U., Hartmann, G., Helm,
C., Herreid, S., Howat, I., Kapustin, G., Khromova, T., König,
M., Kohler, J., Kriegel, D., Kutuzov, S., Lavrentiev, I., LeBris,
R., Lund, J., Manley, W., Mayer, C., Miles, E., Li, X., Menounos,
B., Mercer, A., Mölg, N., Mool, P., Nosenko, G., Negrete, A.,
Nuth, C., Pettersson, R., Racoviteanu, A., Ranzi, R., Rastner, P.,
Rau, F., Raup, B., Rich, J., Rott, H., Schneider, C., Seliverstov,
Y., Sharp, M., Sigurðsson, O., Stokes, C., Wheate, R., Winsvold,
S., Wolken, G., Wyatt, F., and Zheltyhina, N.: Randolph Glacier
Inventory – A Dataset of Global Glacier Outlines: Version 4.0,
Global Land Ice Measurements from Space, National Snow and
Ice Data Center, Boulder, USA, 2014.
Bajracharya, S. R. and Shrestha, B. (Eds.): The status of glaciers
in the Hindu Kush–Himalayan region, International Centre for
Integrated Mountain Development, Kathmandu, Nepal, 2011.
Bhambri, R., Bolch, T., and Chaujar, R.: Mapping of debris-covered
glaciers in the Garhwal Himalayas using ASTER DEMs and
thermal data, Int. J. Remote Sens., 32, 8095–8119, 2011.
Bolch, T., Yao, T., Kang, S., Buchroithner, M. F., Scherer, D., Maussion, F., Huintjes, E., and Schneider, C.: A glacier inventory for
the western Nyainqentanglha Range and the Nam Co Basin, Tibet, and glacier changes 1976–2009, The Cryosphere, 4, 419–
433, doi:10.5194/tc-4-419-2010, 2010.
Bolch, T., Pieczonka, T., and Benn, D. I.: Multi-decadal mass loss
of glaciers in the Everest area (Nepal Himalaya) derived from
stereo imagery, The Cryosphere, 5, 349–358, doi:10.5194/tc-5349-2011, 2011.
Bolch, T., Kulkarni, A. Kääb, A., Huggel, C., Paul, F., Cogley, J. G.,
Frey, H., Kargel, J. S., Fujita, K., Scheel, M., Bajracharya, S., and
Stoffel, M.: The state and fate of Himalayan Glaciers, Science,
336, 310–314, 2012.
Cogley, J. G.: Climate Science: Himalayan glaciers in the balance,
Nature, 488, 468–469, doi:10.1038/488468a, 2012.
Ding, Y., Liu, S., Li, J., and Shangguan, D.: The retreat of glaciers
in response to recent climate warming in western China, Ann.
Glaciol., 43, 97–105, 2006.
Frey, H., Paul, F., and Strozzi, T.: Compilation of a glacier inventory for the western Himalayas from satellite data: methods,
challenges, and results, Remote Sens. Environ., 124, 832–843,
doi:10.1016/j.rse.2012.06.020, 2012.
Fujita, K. and Nuimura, T.: Spatially heterogeneous wastage of Himalayan glaciers, P. Natl. Acad. Sci. USA, 108, 14011–14014,
2011.
Gardelle, J., Berthier, E., Arnaud, Y., and Kääb, A.: Region-wide
glacier mass balances over the Pamir-Karakoram-Himalaya during 1999–2011, The Cryosphere, 7, 1263–1286, doi:10.5194/tc7-1263-2013, 2013.
Gardner, A., Moholdt, G., Cogley, J., Wouters, B., Arendt, A.,
Wahr, J., Berthier, E., Hock, R., Pfeffer, W., Kaser, G., Ligtenberg, S., Bolch, T., Sharp, M., Hagen, J., van den Broeke,
M., and Paul, F.: A reconciled estimate of glacier contribu-
www.the-cryosphere.net/9/849/2015/
863
tions to sea level rise: 2003 to 2009, Science, 340, 852–857,
doi:10.1126/science.1234532, 2013.
Hayakawa, Y., Oguchi, T., and Lin, Z.: Comparison of new
and existing global digital elevation models: ASTER GDEM and SRTM-3, Geophys. Res. Lett., 35, L17404,
doi:10.1029/2008GL035036, 2008.
He, Y.: Changing features of the climate and glaciers in China’s
monsoonal temperate glacier region, J. Geophys. Res., 108, 1–7,
doi:10.1029/2002JD003365, 2003.
Immerzeel, W., van Beek, L. P. H., and Bierkens, M.: Climate
change will affect the Asian water towers, Science, 328, 1382–
1385, doi:10.1126/science.1183188, 2010.
Jacob, T., Wahr, J., Pfeffer, W., and Swenson, S.: Recent contributions of glaciers and ice caps to sea level rise, Nature, 482,
514–518, 2012.
Jarvis, A., Reuter, H., Nelson, A., and Guevar, E.: Hole-filled
SRTM for the globe Version 4, available from the CGIARCSI SRTM 90 m Database, http://srtm.csi.cgiar.org/, last access:
28 March 2014.
Kääb, A.: Glacier volume changes using ASTER satellite stereo and
ICESat GLAS laser altimetry. A Test Study on Edgeøya, Eastern
Svalbard, IEEE T. Geosci. Remote, 46, 2823–2830, 2008.
Kääb, A., Berthier, E., Nuth, C., Gardelle, J., and Arnaud,
Y.: Contrasting patterns of early twenty-first-century glacier
mass change in the Himalayas, Nature, 488, 495–498,
doi:10.1038/nature11324, 2012.
Kaser, G., Grosshauser, M., and Marzeion, B.: Contribution potential of glaciers to water availability in different climate regimes,
P. Natl. Acad. Sci. USA, 107, 20223–20227, 2010.
Li, X., Cheng, G., Jin, H., Kang, E., Che, T., Jin, R., Wu, L., Nan,
Z., Wang, J., and Shen, Y.: Cryospheric change in China, Global
Planet. Change, 62, 210–218, 2008.
Liu, S., Shangguan, D., Ding, Y., Han, H., Xie, C., Zhang,
Y., Li, J., Wang, J., and Li, G.: Glacier changes during the
past century in the Gangrigabu mountains, southeast QinghaiXizang (Tibetan) Plateau, China, Ann. Glaciol., 43, 187–193,
doi:10.3189/172756406781812348, 2006.
Matsuo, K. and Heki, K.: Time-variable ice loss in Asian high
mountains from satellite gravimetry, Earth Planet. Sc. Lett., 290,
30–36, 2010.
McClung, D. and Schaerer, P.: The Avalanche Handbook, 3rd Edn.,
The Mountaineers Books, Seattle, 342 pp., 2006.
Paul, F., Huggel, C., and Kääb, A.: Combining satellite multispectral image data and a digital elevation model for mapping debriscovered glaciers, Remote Sens. Environ., 89, 510–518, 2004.
Paul, F., Barrand, N., Baumann, S., Berthier, E., Bolch, T., Casey,
K., Frey, H., Joshi, S., Konovalov, V., Le Bris, R., Mölg, N.,
Nosenko, G., Nuth, C., Pope, A., Racoviteanu, A., Rastner, P.,
Raup, B., Scharrer, K., Steffen, S., and Winsvold, S.: On the
accuracy of glacier outlines derived from remote-sensing data,
Ann. Glaciol., 54, 171–182, doi:10.3189/2013AoG63A296,
2013.
Pfeffer, W. T., Arendt, A. A., Bliss, A., Bolch, T., Cogley, J. G.,
Gardner, A. S., Hagen, J. O., Hock, R., Kaser, G., Kienholz, C.,
Miles, E. S., Moholdt, G., Mölg, N., Paul, F., Radić, V., Rastner,
P., Raup, B. H., Rich, J., Sharp, M. J., and the Randolph Consortium: The Randolph Glacier Inventory: a globally complete
inventory of glaciers, J. Glaciol., 60, 537–552, 2014.
The Cryosphere, 9, 849–864, 2015
864
T. Nuimura et al.: The GAMDAM glacier inventory: a quality-controlled inventory of Asian glaciers
Racoviteanu, A., Paul, F., Raup, B., Khalsa, S., and Armstrong, R.:
Challenges and recommendations in mapping of glacier parameters from space: results of the 2008 Global Land Ice Measurements from Space (GLIMS) workshop, Boulder, Colorado, USA,
Ann. Glaciol., 50, 53–69, doi:10.3189/172756410790595804,
2009.
Radić, V. and Hock, R.: Regionally differentiated contribution of
mountain glaciers and ice caps to future sea-level rise, Nat.
Geosci., 4, 91–94, doi:10.1038/ngeo1052, 2011.
Raper, S. C. B. and Braithwaite, R.: The potential for sea
level rise: New estimates from glacier and ice cap area
and volume distributions, Geophys. Res. Lett., 32, L05502,
doi:10.1029/2004GL021981, 2005.
Rastner, P., Bolch, T., Mölg, N., Machguth, H., Le Bris, R., and
Paul, F.: The first complete inventory of the local glaciers
and ice caps on Greenland, The Cryosphere, 6, 1483–1495,
doi:10.5194/tc-6-1483-2012, 2012.
Raup, B. and Khalsa, S.: GLIMS analysis tutorial, Boulder, CO,
University of Colorado, National Snow and Ice Data Center, http://www.glims.org/MapsAndDocs/guides.html (last access: 28 January 2014), 2007.
The Cryosphere, 9, 849–864, 2015
Salerno, F., Buraschi, E., Bruccoleri, G., Tartari, G., and Simiraglia,
C.: Glacier surface-area changes in Sagarmatha national park,
Nepal, in the second half of the 20th century, by comparison of
historical maps, J. Glaciol., 54, 738–752, 2008.
Shangguan, D., Liu, S., Ding, Y., Li, J., Zhang, Y., Ding, L.,
Wang, X., Xie, C., and Li, G.: Glacier changes in the west Kunlun Shan from 1970 to 2001 derived from Landsat TM/ETM+
and Chinese glacier inventory data, Ann. Glaciol., 46, 204–208,
doi:10.3189/172756407782871693, 2007.
Shi, Y. (Ed.): Concise Glacier Inventory of China, Shanghai Popular
Science Press, China, 2008.
Shi, Y., Hsieh, T., Cheng, P., and Li, C.: Distribution, features and
variations of glaciers in China, IAHS Publ., 126, 111–116, 1980.
Thakuri, S., Salerno, F., Smiraglia, C., Bolch, T., D’Agata, C., Viviano, G., and Tartari, G.: Tracing glacier changes since the
1960s on the south slope of Mt. Everest (central Southern
Himalaya) using optical satellite imagery, The Cryosphere, 8,
1297–1315, doi:10.5194/tc-8-1297-2014, 2014.
Yao, X., Liu, S., Sun, M., Wei, J., and Guo, W.: Volume calculation and analysis of the changes in moraine-dammed lakes in the
north Himalaya: a case study of Longbasaba lake, J. Glaciol., 58,
753–760, 2012.
www.the-cryosphere.net/9/849/2015/