ArticlePDF Available

Stagnation and mass loss on a Himalayan debris-covered glacier: Processes, patterns and rates

Authors:

Abstract and Figures

The ablation areas of debris-covered glaciers typically consist of a complex mosaic of surface features with contrasting processes and rates of mass loss. This greatly complicates glacier response to climate change, and increases the uncertainty of predictive models. In this paper we present a series of high-resolution DEMs and repeat lake bathymetric surveys on Ngozumpa Glacier, Nepal, to study processes and patterns of mass loss on a Himalayan debris-covered glacier in unprecedented detail. Most mass loss occurs by melt below supraglacial debris, and melt and calving of ice cliffs (backwasting). Although ice cliffs cover only ~5% of the area of the lower tongue, they account for 40% of the ablation. The surface debris layer is subject to frequent re-distribution by slope processes, resulting in large spatial and temporal differences in debris-layer thickness, enhancing or inhibiting local ablation rates and encouraging continuous topographic inversion. A moraine-dammed lake on the lower glacier tongue (Spillway Lake) underwent a period of rapid expansion from 2001 to 2009, but later experienced a reduction of area and volume as a result of lake level lowering and sediment redistribution. Rapid lake growth will likely resume in the near future, and may eventually become up to 7 km long.
Content may be subject to copyright.
Journal of Glaciology
http://journals.cambridge.org/JOG
Additional services for Journal of Glaciology:
Email alerts: Click here
Subscriptions: Click here
Commercial reprints: Click here
Terms of use : Click here
Stagnation and mass loss on a Himalayan debris-covered glacier: processes, patterns and rates
SARAH THOMPSON, DOUGLAS I. BENN, JORDAN MERTES and ADRIAN LUCKMAN
Journal of Glaciology / FirstView Article / April 2016, pp 1 - 19
DOI: 10.1017/jog.2016.37, Published online: 19 April 2016
Link to this article: http://journals.cambridge.org/abstract_S002214301600037X
How to cite this article:
SARAH THOMPSON, DOUGLAS I. BENN, JORDAN MERTES and ADRIAN LUCKMAN Stagnation and mass loss on a
Himalayan debris-covered glacier: processes, patterns and rates. Journal of Glaciology, Available on CJO 2016 doi:10.1017/
jog.2016.37
Request Permissions : Click here
Downloaded from http://journals.cambridge.org/JOG, IP address: 80.12.58.119 on 25 Apr 2016
Stagnation and mass loss on a Himalayan debris-covered glacier:
processes, patterns and rates
SARAH THOMPSON,
1
DOUGLAS I. BENN,
1,2
JORDAN MERTES,
1,3
ADRIAN LUCKMAN
1,4
1
Department of Arctic Geology,University Centre in Svalbard (UNIS),Longyearbyen,Norway
2
Department of Geography and Sustainable Development,University of St Andrews,UK
3
Department of Geological and Mining Science,Michigan Technological University,Houghton,MI,USA
4
Department of Geography,Swansea University,Swansea,UK
Correspondence: Sarah Thompson <sarah.thompson@unis.no>
ABSTRACT. The ablation areas of debris-covered glaciers typically consist of a complex mosaic of
surface features with contrasting processes and rates of mass loss. This greatly complicates glacier re-
sponse to climate change, and increases the uncertainty of predictive models. In this paper we
present a series of high-resolution DEMs and repeat lake bathymetric surveys on Ngozumpa Glacier,
Nepal, to study processes and patterns of mass loss on a Himalayan debris-covered glacier in unprece-
dented detail. Most mass loss occurs by melt below supraglacial debris, and melt and calving of ice cliffs
(backwasting). Although ice cliffs cover only 5% of the area of the lower tongue, they account for 40%
of the ablation. The surface debris layer is subject to frequent re-distribution by slope processes, resulting
in large spatial and temporal differences in debris-layer thickness, enhancing or inhibiting local ablation
rates and encouraging continuous topographic inversion. A moraine-dammed lake on the lower glacier
tongue (Spillway Lake) underwent a period of rapid expansion from 2001 to 2009, but later experienced
a reduction of area and volume as a result of lake level lowering and sediment redistribution. Rapid lake
growth will likely resume in the near future, and may eventually become up to 7 km long.
KEYWORDS: debris-covered glaciers, glacier hazards, glacier mass balance, moraine, remote sensing
1. INTRODUCTION
Glaciers in many parts of the Himalaya have lost mass in
recent decades in response to warming climate (Berthier
and others, 2007; Bolch and others, 2008,2011; Quincey
and others, 2009; Gardelle and others, 2012; Kääb and
others, 2012). The summer accumulation type glaciers of
the eastern Himalaya are particularly sensitive to climate
change, because a rise in temperature can both increase ab-
lation and reduce the proportion of monsoon precipitation
falling as snow (Fujita and Ageta, 2000; Benn and others,
2012). Glacier response to climate change is also affected
by debris cover, which is widespread on many glaciers in
the region. Debris cover can either increase or decrease ab-
lation rates depending on its thickness (Østrem, 1959;
Nicholson and Benn, 2006), and the formation of supragla-
cial lakes on low-gradient debris-covered glacier tongues
can locally increase ablation by orders of magnitude (Sakai
and others, 2000a,b,2002,2009; Benn and others, 2001;
Röhl, 2008). Debris cover is therefore an important control
on regional variations in glacier mass balance (Scherler and
others, 2011; Kääb and others, 2012). The response of
debris-covered glaciers to climate warming can have a sub-
stantial impact at the local and regional catchment scales, in-
cluding affecting long-term water availability and frequency
of glacier related hazards, such as glacial lake outburst floods
(GLOFs).
Numerous studies have investigated processes of mass
loss on debris-covered glaciers, including sub-debris
melting (Nakawo and Rana, 1999; Nicholson and Benn,
2006,2013; Reznichenko and others, 2010), backwasting
of exposed ice cliffs (Sakai and others, 1998; Benn and
others, 2001), growth of supraglacial ponds and lakes
(Sakai and others, 2000a,b; Benn and others, 2001; Röhl,
2008) and internal ablation in englacial conduits (Gulley
and Benn, 2007; Gulley and others, 2009). In a few cases,
spatial patterns of mass loss have been determined, allowing
the relative contribution of different processes (e.g. sub-
debris melting vs. ice cliff retreat) to be established (Sakai
and others, 2002; Han and others, 2010; Immerzeel and
others, 2014; Juen and others, 2014; Reid and Brock,
2014). Such studies, however, require labor-intensive field
campaigns, placing limits on spatial and temporal coverage
(Sakai and others, 1998,2009; Nuimura and others, 2011;
Immerzeel and others, 2014). Greater coverage is possible
using remote sensing techniques, although the relatively
low resolution of readily available DEM products (e.g.
ASTER) limits the level of detail that can be achieved.
There is thus a need for studies that bridge the gap between
detailed, local-scale process studies and glacier-wide assess-
ments of mass change.
In this paper, we use a series of high-resolution DEMs and
repeat lake bathymetric surveys to study processes and pat-
terns of mass loss on a Himalayan debris-covered glacier in
unprecedented detail. The study focuses on Ngozumpa
Glacier, Nepal, which has been the subject of a series of
process studies dating from 1998 (Benn and others, 2001;
Thompson and others, 2012; Nicholson and Benn, 2013).
We use stereo GeoEye-1 and WorldView-3 imagery from
June 2010, December 2012 and January 2015 to construct
three DEMs with a spatial resolution of 1 m for 17.4 km
2
of
the ablation area of the glacier. In addition, volume
changes in a lake complex on the lower glacier were
Journal of Glaciology (2016), Page 1 of 19 doi: 10.1017/jog.2016.37
©The Author(s) 2016. This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.
org/licenses/by/4.0/), which permits unrestricted re-use, distribution, and reproduction in any medium, provided the original work is properly cited.
obtained by field surveys over a similar period. Velocities on
the glacier tongue and the extent of stagnant ice were deter-
mined by feature tracking using TerraSAR-X images. In com-
bination, these data allow us to investigate ablation processes
on the glacier and assess their relative contribution to mass
loss. Our results provide a high-resolution snapshot of
Ngozumpa Glacier at a critical stage of its evolution, and
data necessary for modeling the future response of
Himalayan debris-covered glaciers to climate change
(Rowan and others, 2015).
2. STUDY AREA
Ngozumpa Glacier is located in the Khumbu Himal of
Eastern Nepal (27°57N, 85°42E; Fig. 1). The glacier flows
southward from cirques below Cho Oyu (8188 m a.s.l.) and
Gyachung Kang (7922 m a.s.l), where most accumulation
occurs by ice and snow avalanching. Ngozumpa Glacier is
18 km long, extending down to 4659 m a.s.l, 500 m
lower than clean-ice glaciers in the area. The lower 15 km
is extensively debris covered, with debris increasing in thick-
ness from zero at the equilibrium line (c. 5500 m a.s.l) to 13
m towards the terminus (Nicholson, 2005). In the ablation
zone, the glacier surface lies <150 m below the Little Ice
Age lateral moraines (Benn and others, 2001; Sakai and
Fujita, 2010) due to surface lowering. Mass loss rates are
lower near the terminus than up-glacier, reflecting the influ-
ence of debris cover on the mass-balance gradient. This
spatial pattern of surface lowering has resulted in an overall
reduction of the surface gradient in the ablation area,
leading to a reduction in driving stress and flow speed
(Quincey and others, 2009). Feature tracking in the late
1990s has shown the lower 6.5 km of the Ngozumpa to
be stagnant (Quincey and others, 2009).
Surface topography in the glacial ablation area is highly ir-
regular, with a typical local relief of 3050 m. Numerous
hollows are occupied by so called perched lakes, because
they are perched above the elevation of the terminal
moraine that acts as the hydrological base level of the
glacier. The majority of these lakes are ephemeral, typically
persisting for 23 a before drainage via englacial conduits
(Benn and others, 2001). The englacial drainage system of
Ngozumpa Glacier has been investigated by speleological
techniques (Gulley and Benn, 2007; Gulley and others,
2009), and largely consists of cut and closureconduits
formed by the incision of surface streams followed by roof
closure. Many of these are relict features that are only inter-
mittently active during lake drainage events. Enlargement
of conduits during the discharge of warm lake water can
create large voids within the glacier, which can collapse
and form new hollows on the glacier surface. Englacial con-
duits are therefore important for both the drainage and forma-
tion of supraglacial lakes (Gulley and Benn, 2007; Thompson
and others, 2012).
In the early 1990s a system of lakes began to form close to
the terminus of Ngozumpa Glacier. The level of these lakes
(informally known as Spillway Lake) is controlled by the ele-
vation of a spillway through the western frontal-lateral
moraine (Benn and others, 2001; Thompson and others,
2012). Unlike the perched lakes, this system is not suscep-
tible to drainage via englacial conduits and can continue to
expand while the spillway remains at its current level. The
lake system entered a major period of expansion after 2001
when it increased in area by 10% a
1
(Thompson and
others, 2012).
3. DATA AND METHODS
High-resolution stereo multispectral satellite imagery (0.5 m
panchromatic) was acquired in June 2010 and December
2012 (GeoEye-1) and January 2015 (WorldView-3), for a
region of 100 km
2
, covering 17.4 km
2
of the ablation area
of Ngozumpa Glacier (Table 1). These data were used to con-
struct three DEMs, which were then differenced to determine
spatial patterns of elevation change. In addition, three field
bathymetric surveys were conducted over Spillway Lake in
December 2009, 2012 and 2014. The results of the 2009
bathymetric survey are published in Thompson and others
(2012). Finally, ice velocities were determined by feature
tracking a pair of TerraSAR-X images, acquired in
September 2014 and January 2015 (Table 1).
Fig. 1. Study of Ngozumpa Glacier, Nepal, insert shows the regional
location. Image is orthorectified GeoEye-1, December 2012 with
Spillway Lake highlighted.
2Thompson and others: Stagnation and mass loss on a Himalayan debris-covered glacier
3.1. DEM construction
The PCI Geomatica Software Package was used to automat-
ically construct three stereoscopic DEMs from the stereo
multispectral imagery. The Rational Functions math model
was employed to transform to epipolar geometry and
convert parallaxes to elevations, building a correlation
between pixels and their ground location and elevation.
The technique is based on rational polynomial coefficients
(RPCs) supplied with the imagery and does not require the
input of ground control points. A correlation value between
each of the images in the stereo pair was generated for
each DEM pixel, with a value of between 0 (no data) and
100 (full correlation). This correlation value allowed the
identification of areas of no or poor correlation between
each stereo pair. All areas with a correlation value <50
were classified as poor correlation and used to create a
mask for the relevant DEM. At a later stage any areas of
cloud cover or shadow still evident in the data were added
to the DEM mask. All pixel values falling within the relevant
masks were removed from the datasets and any calculations
(Fig. 2). However, no major areas of poor correlation oc-
curred on the glacier surface (Fig. 2).
Each DEM was resampled to provide a final pixel reso-
lution of 1 m. Kääb (2005) showed that producing a DEM
of coarser resolution than the original stereo image can de-
crease errors and data voids within the DEM. Initially
DEMs at 0.5 m resolution were also generated but the level
of detail on such a heterogeneous surface made classification
of processes and identification of features more difficult than
in the 1 m DEM, due to the increased soothing of the 1 m
DEM. The DEMs were manually edited prior to geocoding,
any erroneous elevations from areas delineated as lakes.
Available check-points, collected in 2006, 2010 and 2014
using a Leica Geosystems 1200 differential GPS (dGPS),
were not used in DEM generation due to a locational bias
of points (Fig. 3), but were used subsequently in an
Table 1. Satellite images and acquisition information
Sensor Product type Resolution Acquisition date Collection azimuths Cloud cover
%
GeoEye-1 GeoStereo PAN/MSI PAN 0.46 MSI 1.84 9 June 2101 155.5989 2
50.3696 3
GeoEye-1 GeoStereo PAN/MSI PAN 0.46 MSI 1.84 23 December 2012 197.6473 0
11.0241 0
World View-3 Stereo OR2A PAN 0.46 MSI 1.84 5 January 2015 3.2 0
303.04 0
Terra SAR-X StripMap Mode SAR 2 (1.36 × 1.99 range × azimuth in
slant range geometry)
19 September 2014 ––
Terra SAR-X StripMap Mode SAR 2 (1.36 × 1.99 range × azimuth in
slant range geometry)
18 November 2015 ––
Fig. 2. Edited DEM constructed from each of the three sets of stereo imagery, including the areas of no data, poor correlation, cloud and
shadow shown in black and classified as no data and the location of all delineated ice cliffs and lake perimeters.
3Thompson and others: Stagnation and mass loss on a Himalayan debris-covered glacier
assessment of the generated DEMs. The DEMs were geo-
coded based on the supplied RPCs, yielding a geopositioning
accuracy of close to 0.10 m (0.2 pixels) in planimetry and
0.25 m (0.5 pixels) in height in tests (Frazer and
Ravanbakhsh, 2009).
3.2. Planimetric and vertical adjustments
A series of steps was followed to accurately detect glacier
surface elevation changes for the periods 20102012 and
20122015, largely following the procedures set out by
Nuth and Kääb (2011). Firstly, inconsistencies in the geoloca-
tion of the DEMs were detected and corrected. This step is
vital as a small horizontal offset between two DEMs can
produce a large elevation error where the local slope is
steep, as is often the case on debris-covered glaciers
(Berthier and others, 2004). Both the correlation analysis
and check-point comparison gave better results for the
2012 DEM than the others, so we chose to adjust the position
of the 2010 and 2015 DEMS to the 2012 DEM. The adjust-
ment is based on a characteristic relationship between the
residuals of the elevation difference and the elevation deriva-
tives of slope and aspect of two DEMs that are not perfectly
aligned (Kääb, 2005). This relationship is precisely related
to the xyshift vector between them, such that
dh¼acos ðbΨÞtan ðαÞþdhð1Þ
where dhis the individual elevation difference, ais the mag-
nitude of horizontal shift, bis the direction of shift, αis the
terrain slope, Ψis the terrain aspect and dhis the overall ele-
vation bias between the two DEMs (Kääb, 2005). The equa-
tion is applied to off-glacier areas, assumed to be areas of
little or no change in surface elevation. To provide a success-
ful correction, off-glacier areas were carefully identified to
exclude zones of debris slumping, rock fall or erroneous
data values, most common in steeply sloping areas. The hori-
zontal shift was determined by minimising the RMSE of the
chosen elevation differences using iterative shifting of the
DEM until improvement in the SD of the off-glacier pixel
values was <2% (Nuth and Kääb, 2011). The correction
improved the SD of the difference value of the off-glacier
area by 24% (20102012) and 74% (20122015). We inves-
tigated the elevation dependent bias but there was no clear
relationship between bias and elevation, and the magnitudes
are not enough to attempt a correction (Nuth and Kääb,
2011). In addition, we investigated the presence of along/
cross track bias, associated with the satellite acquisition
geometry but no bias could be identified.
The three optical stereo images pairs were acquired in
June 2010, December 2012 and January 2015. Changes in
surface elevation or ice mass derived from the three DEMs
are reported over their relevant time periods. Annual rates
of change were calculated only for the December 2012 to
January 2015 period, assumed to represent 2 balance
years. Glacier perimeters and ponds were manually digitised
in each of the images. The bathymetry data were all collected
in early December. Because little lake expansion occurs in
the winter months, each survey is expected to be representa-
tive of the lake at the end of the preceding ablation period.
The DEMs also allowed automatic orthorectification of the
multispectral imagery. This was used to delineate a precise
glacier area and identify other features of interest on the
glacier surface. Identifying the margin of debris-covered gla-
ciers is a common issue in both remote sensing and field
surveys, as debris layers can obscure the extent of underlying
ice (Bolch and others, 2008). The ablation zone of Ngozumpa
Glacier is flanked by lateral moraines, and the topographic
lows at the base of the ice-proximal slopes of the moraines
were adopted as a convenient and consistent choice of
glacier limit. The GeoEye-1 (9 June 2010 and 23 December
2012) and WorldView-3 (5 January 2015) optical imagery
(Table 1) were used in combination with the corresponding
DEM to estimate the ice-moraine boundary for the whole ab-
lation area. For each dataset, classification of supraglacial
lakes was conducted manually, because variability in sus-
pended sedimentconcentration, frozen and unfrozen lake sur-
faces and variable snow cover both within and between
images precluded automatic classification. In addition, lake-
perimeter length, lake area and their evolution through the
Fig. 3. Initial difference map from 2010 to 2012 used to identify
areas of off-glacier stable ground for DEM co-registration; areas
classified as no data are shown in white. The glacier area and
suitable stable areas are highlighted, with the additional location
of the >6000 off-glacier check points. Adjustments for 20102012
were x=+1.2, y=3.8, z=0.1 and based on the initial
difference map for 20122015, x=0.4, y=+5.3, z=+3.0.
4Thompson and others: Stagnation and mass loss on a Himalayan debris-covered glacier
intervening period (areal expansion/reduction, drainage/
filling) were recorded. Where possible, the top of the ice
cliffs was delineated using a combination of DEM elevations
and slope and aspect characteristics (Fig. 4). Slope measured
in degrees was calculated for each cell deriving the
maximum rate of change in elevation from that cell to its
neighbors, the maximum change in elevation over the dis-
tance between the cell and its eight neighbors identified the
steepest downhill descent from the cell. Aspect was calculated
in a similar way, identifying the downslope direction of the
maximum rate of change in value from each cell to its neigh-
bors. Aspect was measured clockwise (°) from 0° (due north) to
360° (again due north), flat areas have no downslope direc-
tion. Aspect of an individual ice cliff was then classified as
North (315°45°), East (45°135°), South (135°225°) or
West (225°315°) based on the mean aspect of the area iden-
tified. Length, maximum height, aspect and persistence
through consecutive datasets were recorded.
3.3. Assessment of errors
Error or uncertainty in surface elevation changes depends on
both the quality of the DEMs and their co-registration. Over
6000 off-glacier dGPS points collected between 2006 and
2014 (Fig. 3) were used to estimate the RMSE
z
. by extracting
the elevation values from each DEM at the location of each
point. The nominal accuracy of the post-processed dGPS
data is ±1 cm in the horizontal and ±2 cm in the vertical.
This gives RMSE
z
values of 1.9, 1.4 and 2.3 m for the
2010, 2012 and 2015 DEMs, respectively. However, we
cannot take into account the error in the x, y location. In
light of this, we focus on an accurate assessment of the uncer-
tainty in the elevation difference values derived from the co-
registered DEMs. A first estimate of the uncertainty of the
DEM differencing results can be derived from the SDof the
off-glacier areas. However previous studies suggest this will
likely overestimate errors (Berthier and others, 2007; Bolch
and others, 2011). We follow Bolch and others (2011)in
using the standard error (SE) and the mean elevation differ-
ence (MED) of the non-glaciated area as an estimate of
error, where SE is calculated from the SD of the off-glacier dif-
ferences σ
sg
and the number of pixels nused
SE ¼σsg
ffiffiffi
n
pð2Þ
and the error Eis calculated by
E¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
SE2þMED2
p:ð3Þ
Surface elevation changes on the glacier were converted into
ice volume change by summing the difference between
DEMs for areas of interest. Volume change errors were calcu-
lated by multiplying the error Eby the area over which the
volume change occurs (Barrand and others, 2010; Bolch
and others, 2011).
3.4. Sonar data 2009, 2012 and 2014
Bathymetric surveys were conducted on Spillway Lake in
early December 2009, 2012 and 2014 when the lake
surface was frozen. In each case the lake perimeter and
any adjoining ice cliffs were mapped using Leica
Geosystems 1200 dGPS, recording points at 2 m intervals
in kinematic mode. All data were post processed to a base
station located on the eastern side of the glacier (27.941 N,
86.719 E; 4648 m a.s.l.). Bathymetric measurements were
made using a Hummingbird 385ci echo-sounder deployed
through the frozen lake surface and located using the
dGPS. Good contact between the sonar transceiver and the
Fig. 4. Identification of ice cliffs from the 2012 DEM and derived DEM slope and aspect. Maximum height was derived from the DEM, length
from the DEM and slope. Aspect was identified from the derived aspect and classified as flat, north, south, east or west based on the mean
value (°).
5Thompson and others: Stagnation and mass loss on a Himalayan debris-covered glacier
lake surface was achieved by wetting the ice prior to each
measurement. The accuracy of the sonar measurements
was regularly assessed by chipping through the frozen lake
surface and comparing the depth measurement from the
echo sounder at the ice surface and the water surface (includ-
ing the ice thickness) with that of a weighted line. Maximum
difference between the results of the two methods was 1%.
Each survey consisted of a quasi-regular grid of 15 m
spacing, but with closer spacing in areas with high-depth
variability. A bathymetric map was derived for each survey
by interpolating the depth measurements in a GIS package
using a natural neighbor interpolation, which ensures the
preservation of all measurement points. The interpolations
were constrained by the relevant lake perimeter outline set
to a depth of 0 m. The interpolated bathymetry maps were
each re-sampled to 1 m pixel size to allow better incorpor-
ation into the relevant DEM by mosaic.
3.5. Velocity measurements
Glacier surface velocities were derived using feature tracking
between synthetic aperture radar images acquired by the
TerraSAR-X satellite on 19 September 2014 and 18 January
2015 (Table 1) (e.g. Luckman and others, 2007). The
method searches for a maximum correlation between
evenly-spaced subsets (patches) of each image giving the dis-
placement of glacier surface features, which are converted to
speed using time delay between images. Image patches were
400 m in size and sampled every 40 m producing a spatial
resolution of between 40 and 400 m depending on feature
density. Estimated precision is 0.015 m d
1
.
4. RESULTS
Between 2010 and 2015 substantial changes in surface ele-
vation were measured over the area covered by the three
DEMs (Fig. 5). Between June 2010 and December 2012,
overall volume losses were 50.8 ± 2.2 × 10
6
m
3
; and
between December 2012 and January 2015, 31.8 ± 5.7 ×
10
6
m
3
. When aggregated over the whole area, this indicates
mean surface lowering of 2.9 ± 0.13 m and 1.8 ± 0.33 m, re-
spectively, implying an annual rate of 0.92 ± 0.33 m a
1
for
the latter period. However substantial variability in both
spatial pattern and magnitude of elevation change is
evident in both DEM difference maps. On the basis of the
visual appearance in optical imagery and the DEM difference
maps, we define six morphometric categories: (1) areas
where surface elevation change is influenced by glacier
movement (2) ice cliffs, (3) debris-covered ice, (4) lateral
moraines and ice-marginal troughs, (5) perched lakes (6)
Spillway Lake. These are discussed in turn below.
4.1. Dynamic vs. mass change
Ice velocities on the glacier tongue for September 2014
January 2015 are shown in Figure 6a. Both the western
(Cho Oyu) and eastern (Gyachung Kang) branches of the
glacier have maximum velocities of 0.1 m d
1
for the
Fig. 5. Final DEM differences between (a) June 2010 and December 2012 and (b) December 2012 and January 2015, regions classified as no
data are shown in white. The scale shows the difference in elevation between the two images with the blue positive numbers indicating areas
of elevation increase and red negative numbers indicating elevation decrease. The location of delineated ice cliffs and lake perimeters are also
illustrated.
6Thompson and others: Stagnation and mass loss on a Himalayan debris-covered glacier
2014/15 autumnearly winter period, equivalent to annual
speeds of 36 m a
1
. The two branches appear to be dy-
namically disconnected, and the main tongue is fed by
the western tributary only. Velocities diminish down-
glacier, falling to <0.015 m d
1
(5ma
1
)7kmfrom
the terminus. Below this point, measured speeds are indis-
tinguishable from error and the glacier is essentially
stagnant.
Both DEM difference maps show a marked transverse
striped pattern on the dynamically active part of the glacier
tongue at distances >7 km from the terminus (Fig. 5). The
striping consists of areas of surface lowering alternating
with regions of apparent uplift, with values up to ±50 m.
We interpret the striped pattern as the result of displacement
of surface topography down-glacier by ice flow, superim-
posed on elevation changes due to mass loss. The striping
is absent from the lower, stagnant part of the tongue. On
the basis of the velocity map and the striping pattern, we
define a boundary between active and inactive parts of the
glacier (Fig. 6a). For the lower stagnant part, we calculated
volume changes associated with different processes, such
as ice cliff retreat and melt of debris-covered ice. We did
not attempt this for the dynamically active part, due to the dif-
ficulty of accurately separating dynamic and mass-balance
components of elevation change.
Total volume loss in the area affected by ice motion is
40.1 ± 1.5 × 10
6
m
3
and 24.7 ± 3.3 × 10
6
m
3
for 201012
and 201215, respectively. The lower stagnant region
covers an area of 5.6 km
2
, and has undergone volume
losses of 10.6 ± 0.7 × 10
6
m
3
and 7.1 ± 1.8 × 10
6
m
3
for the
periods 20102012 and 20122015, respectively. When
aggregated over the areas, mean surface lowering for the
entire period in the upper portion of the glacier is substantial-
ly greater (5.4 m) than in the lower region (3.1 m). In the fol-
lowing paragraphs, we focus on the stagnant, lower 7 km of
the ablation area (Fig. 6b).
Fig. 6. (a) Velocity map derived using feature tracking between synthetic aperture radar images acquired by the TerraSAR-X satellite on 19
September 2014 and 18 January 2015. (b) The inactive lower ablation area, with the location of delineated ice cliffs and lake perimeters from
each of the 3 years of investigation, highlighting patterns of elevation change associated with: a ice cliff retreat, b distance from terminus,
cthe glacier margins, d apparent uplift of debris covered areas, e changes in lake level.
7Thompson and others: Stagnation and mass loss on a Himalayan debris-covered glacier
4.2. Ice cliffs
Cliffs of bare ice were manually delineated on the basis of
their optical characteristics and steep slope. Over 100 indi-
vidual ice cliffs were identified in the lower ablation area
in 2010, with a total top of ice cliff length of 10.5 km. Both
number and total length reduced to 65 individual cliffs in
2012 with a total top of ice cliff length of 9 km, and to 42
cliffs with a total top of ice cliff length of 7 km by 2015.
75.3% of ice cliffs mapped in 2010 bordered a supraglacial
lake or pond. The mean height of ice faces was 15.5 m but
the maximum was substantially higher at 45 m. 44.8% of
the identified ice cliffs in 2010 were north-facing and only
10.5% south-facing. East- and west-facing accounted for
20.9 and 23.8%, respectively. Little difference was measured
in mean height across all aspects but the maximum height of
45 m was measured on a south-facing cliff (all other aspects
have a maximum of 30 m). In addition, 40% of the ice cliffs
facing north, east and west persist in some form and were
identifiable in both the 2012 and 2015 orthorectified
images, but only 18.2% of those facing south persisted
throughout the study period.
The DEM difference maps show that the largest observed
surface elevation changes occur at ice cliffs, with typical
surface lowering of 2050 m (Figs 6b, 7). This lowering
results from rapid backwasting of exposed ice faces, often
but not always in contact with supraglacial ponds or lakes.
Throughout the lower ablation area, backwasting of ice
cliffs accounted for volume losses of 4.3 ± 0.03 × 10
6
m
3
and 3.2 ± 0.07 × 10
6
m
3
in 20102012 and 20122015, re-
spectively. Backwasting affected 5% of the glacier area
but contributed over 40% of the total volume lost.
4.3. Debris-covered ice
Areas of debris-covered ice are characterized by rough, het-
erogeneous surface textures and slopes <25° (based on DEM
derived slopes). These areas typically experienced low mag-
nitude elevation lowering during both periods, interpreted as
sub-debris ice melt or surface lowering (Fig. 6b). These areas
cover much of the lower ablation area outside of the areas
classified as rapidly retreating ice cliffs, lakes formation and
drainage and channel subsidence. Total surface lowering of
0.55 m occurred over the 4.5 a period of investigation,
with increasing magnitude with distance from the terminus
(Figs 6b, 8). Within 1 km of the terminal moraine lowering
was between 0.5 and 1 m (Fig. 8, profile c), increasing to a
maximum of 5 m close to the northern extent of the study
area (Fig. 8, profile a).
Areas of elevation increase were also identified on the
stagnant lower tongue of the glacier (Fig. 6b). These are
Fig. 7. High-magnitude elevation changes over small area changes are observed between ice cliffs of consecutive datasets. Elevation profiles
(a) and (b), extracted from each of the three DEMs, illustrate the magnitude of ice cliff retreat.
8Thompson and others: Stagnation and mass loss on a Himalayan debris-covered glacier
of two types: (1) increases in the elevation of debris-
covered ice in a topographic low (Fig. 9); and (2) increase
in the level of a supraglacial lake (Section 4.5). On active-
ly flowing glaciers, uplift can also be caused by ice-
dynamic factors such as compressive flow. However, the
observed areas of uplift are in stagnant ice far down-
glacier of the actively-flowing part (Fig. 6b), so this
process can be ruled out.
Areas of elevation increase in debris-covered areas are
typically bounded by relatively steep slopes exhibiting
surface lowering. Elevation profiles from the three DEMs
show that the pattern of elevation change is consistent with
re-distribution of surface debris (Fig. 9). In places, downslope
translation of transverse ridge-trough features could be
detected in areas of elevation increase, supporting the inter-
pretation that they record redistribution of surface debris.
Debris redistribution is particularly evident below retreating,
non-calving ice cliffs and in some cases, infilling of topo-
graphic lows resulted in >6 m of elevation increase (Fig. 9,
profile b). Areas of elevation increase cover 10% of the
lower ablation area, suggesting that even where the debris
layer is continuous and relatively thick there is noticeable
re-distribution of surface debris. Areas of elevation increase
were included in the estimate of volume loss due to surface
lowering and reduced the total volume loss by sub-debris
melt by 14%. Debris-covered ice occupied 7589% of
the lower ablation area, and sub-debris melt contributed
5.2 ± 0.5 × 10
6
m
3
and 4.3 ± 1.5 × 10
6
m
3
to volume loss
over the two consecutive periods.
4.4. Lateral moraines and ice-marginal troughs
Most of the lower tongue is bounded by lateral moraines that
stand up to 150 m above the adjacent glacier surface. The
ice-distal slopes of the moraines exhibit little or no elevation
change, consistent with their stable, vegetated appearance in
the field. In contrast, the ice-proximal slopes typically show
large reductions in elevation, with localized areas of uplift
(Fig. 5). In the field, these slopes consist of unvegetated, un-
consolidated till, which undergoes active erosion by a range
of processes including rockfall and rotational landslipping
(Benn and others, 2012). The observed patterns of elevation
change are consistent with slope retreat and block displace-
ment. Crestline retreat rates are up to 20 m in isolated loca-
tions, with an overall mean of 0.5 m a
1
. The total volume
lost (20102015) from the moraines is 2.8 ± 0.6 × 10
6
m
3
.
Despite this large loss of mass from the ice-proximal slopes
of the moraines, we do not observe a corresponding increase
in surface elevation at the base of the slopes. Instead, linear
troughs extend along both margins of the glacier at the base
of the moraine slopes, with surface lowering of 69 m over
the whole period of study (Figs 6b, 10). Elevation profiles
across the trough and moraine slope show a steep slope
from the glacier towards the margin, with the greatest
surface lowering at the lowest point of the trough (Fig. 10).
Debris eroded from the overlying moraine slopes must have
been transferred into the trough, implying that subsidence
has occurred along both margins of the glacier. Linear
zones of subsidence have been observed on Ngozumpa
Glacier prior to our study, and shown to be associated with
surface collapse into englacial conduits (Gulley and others,
2009; Benn and others, 2012; Thompson and others, 2012).
We conclude that sub-marginal conduits extend along both
margins of Ngozumpa Glacier, and that these undergo epi-
sodic enlargement and collapse. Lateral streams could also
transport sediment, contributing to mass loss from the troughs.
We estimate the implied internal ablation along the lateral
margins by summing the volume losses in the trough and ad-
jacent moraine, for the area between 1.5 and 7 km from the
terminus on both margins where the troughs are most clearly
defined. For the whole period (20102015), moraine volume
loss was 2.8 ± 0.6 × 10
6
m
3
and trough volume losses were
0.36 ± 0.05 × 10
6
m
3
. Losses for 20122015 are 0.7 ±
0.32 × 10
6
m
3
(moraine) and 0.11 ± 0.03 × 10
6
m
3
(trough),
with a total implied annual mass loss of 0.4 ± 0.035 × 10
6
m
3
. Because this figure encompasses any sediment flushed
out of the system via the lateral streams, it represents a
maximum estimate of internal ice ablation.
4.5. Perched lakes
Over the period of investigation, the total area classified as
perched supraglacial lakes in the lower ablation area under-
went a decrease from 1.8 × 10
5
m
2
in 2010 through 1.6 ×
10
5
m
2
in 2012 to 1.4 × 10
5
m
2
by 2015. The June acquisition
date of the 2010 image, falling within the early monsoon and
ablation season, makes it difficult to assess the significance of
this trend, however, because there may be seasonal cycles of
lake expansion and contraction (Fig. 11). Of the 120 supragla-
cial lakes identified in the lower ablation area in 2010, 74
were evident in some form in 2012, and 32 new lakes were
added. In 2015, 64 of the 106 lakes identified in 2012 were
visible, 54 lakes remained and 12 additional lakes were iden-
tified. It is likely that lakes also formed and drained in
Fig. 8. Elevation profiles a, b and c, located in Figure 6b at b1, b2
and b3, respectively, extracted from the 3 DEMs illustrate that this
pattern of mass loss typically increases with distance from the
terminus.
9Thompson and others: Stagnation and mass loss on a Himalayan debris-covered glacier
between the image acquisitions. Surface elevation changes in
areas identified as lakes or ponds (excluding area of change
attributed to backwasting at the perimeter of lakes or ponds)
indicate a net loss of stored surface water of 0.58± 0.02 ×
10
6
m
3
in the period 20102012 and 0.53 ± 0.05 × 10
6
m
3
between 2012 and 2015.
Fig. 9. Areas of elevation increase, in the absence of glacier lakes, were identified to persist across the difference map 20102015. The
extracted DEM values across profile a illustrates the process of debris re-distribution downslope, while profile b suggests an increase due
to debris infilling of a topographic low point.
Fig. 10. An enhanced surface lowering signal is observed towards the margins of the glacier over the entire period. Extracted elevation profiles
from each of the DEMs illustrate surface lowering at the topographic low point >10 m. Substantial displacement in the lateral moraine is also
evident particularly in profile b.
10 Thompson and others: Stagnation and mass loss on a Himalayan debris-covered glacier
4.6. Spillway Lake
The three Spillway Lake surveys illustrate a reduction in the
areal extent of the base level Spillway Lake, from 3.0 × 10
5
m
2
in 2009 to 2.5 × 10
5
m
2
in 2012 and 2.4 × 10
5
m
2
in
2014 (Fig. 12). Some areas of lake expansion occurred (e.g.
in the chain of lakes in the western part of the complex (1
in Fig. 12), and towards the elongated lake at the eastern
lateral margin (3 in Fig. 12), but these were more than
offset by loss of lake area in the central part (2 in Fig. 12)).
This net loss of lake area is marked in contrast to the 2001
2009 period, during which lake area increased by 10%
a
1
(Thompson and others, 2012). In the area surveyed in
all 3 years, the lake also underwent a reduction in volume,
from 11.3 × 10
4
m
3
in 2009 to 10.3 × 10
4
m
3
in 2012 and
9.5 × 10
4
m
3
in 2014.
The bathymetry surveys from 2012 and 2014 show a
complex pattern of lake evolution, with areal expansion
and deepening in some locations but reductions in both
area and depth in others. Contrasting patterns of depth
change are illustrated using four cross profiles through the
Spillway Lake complex (Fig. 12).
(1) The small lake to the southeast of the main Spillway Lake
(Blue Lake) has no identifiable inflow and stable low-
gradient shorelines, and has remained stable in areal
extent since, first identified in 1984 (Thompson and
others, 2012). The bathymetric profiles for 2009 and
2012 are essentially identical (Fig. 12, profile a), indicat-
ing no substantial change in lake depth. Small differences
between the profiles may have resulted from interpol-
ation errors. (Only a partial survey was conducted in
2014 due to unfavorable ice conditions.)
(2) Profile b (Fig. 12) is in the southern part of the Spillway
Lake complex, most of which underwent little change
in shoreline position from 2009 to 2014. Along most of
the profile, depth differences are small and may be due
to interpolation errors. At the northern end of the
profile, however, two basins underwent shallowing by
up to 4 m between 2012 and 2014. The pattern of shal-
lowing is consistent with sediment infilling. Because
there are no identifiable local sources of sediment on
the stable lake shorelines, the most likely source is sus-
pended silt and sand transported by through-passing
meltwater.
(3) Profile c (Fig. 12) crosses the central basin of the lake
system from west to east, and illustrates an area of sub-
stantial change in both the subaerial and subaqueous sur-
faces. Backwasting of large ice cliffs caused slope retreat
at each end of the profile, and the disappearance of a
promontory in the center. Conversely, infilling of lake
basins occurred adjacent to the former ice cliffs. The
pattern of infilling is consistent with deposition of local
supraglacial debris from retreating ice cliffs. In addition,
a drop in lake level, on the order of 1.5 m was observed
between 2009 and 2012.
(4) Profile d (Fig. 12) traverses the northern and central parts
of the Spillway Lake complex from northwest to south-
east. Backwasting of subaerial ice cliffs is evident at
Fig. 11. Differences in elevation relating to supraglacial lake formation and drainage were identified. The 20102012 difference map and
extracted DEM profile shows an increase in surface elevation associated with the filling of a topographic low with melt water; the
resulting lake is evident in 2012. By 2015 this lake has largely drained and the 20122015 difference map shows a region of elevation
lowering. The elevation difference between 2010 and 2015 in the DEM profile also suggests melting at the bed during lake occupation.
11Thompson and others: Stagnation and mass loss on a Himalayan debris-covered glacier
Fig. 12. The evolution of the base level Spillway Lake from 2009 to 2014, illustrating lake area, lake bathymetry and ice cliff position. The
inflow close to profile d in the 2009 map is the location of an upwelling identified in 2001 (Benn and others, 2001). The inflow into the north-
eastern basin evident in all three maps evolved from a conduit exiting into the lake in 2009 to a sub-aerial meltwater stream in 2010 and 2014.
The inflow into the western margin of the lake in 2014 is the site of an upwelling. The photo insert from 2012 provides evidence for lake level
change, the location of the image is shown by the green star in the 2012 bathymetric map. Each of the four profiles a, b, c and d corresponds to
the arrows on the map and are represented in the same directions. The dashed line illustrates the water level of the subaqueous areas: 1
Northward expansion and coalescence of ponds, 2 Substantial reduction of lake area. 3 Lateral pond connection to the main lake system.
12 Thompson and others: Stagnation and mass loss on a Himalayan debris-covered glacier
both ends of the profile. In addition, a high point on the
lake floor in the southern part of the profile disappeared
between 2009 and 2012, possibly reflecting subaqueous
calving at the base of the subaerial ice cliff. The northern
lake basin displays a localised area of shallowing, again
consistent with deposition of local supraglacial debris
from retreating ice cliffs. This basin also experienced a
drop in lake level of 1 m between 2009 and 2012 but
regained the 2009 level by 2014. This is linked to the
connection between two lake basins, which narrowed
from 25 to 35 m in 2009 to a rocky channel <5m
wide in 2012, then vanished completely by 2014. The
central part of the profile displays emergence of the
lake bed above lake level, relating to a lake level drop
of 1 m between 2009 and 2012 and a further 1.5 m
drop towards the south eastern extent between 2012
and 2014.
5. ANALYSIS AND DISCUSSION
The results presented above reveal patterns and rates of mass
loss on a stagnant, debris-covered Himalayan glacier tongue
in unprecedented detail. Rates of mass change have been
quantified for well-known processes, such as backwasting
and sub-debris melt of debris-covered ice, and also for hith-
erto poorly known processes, such as sediment redistribu-
tion, internal ablation and depth change in supraglacial
lakes. Taken together, these data provide a detailed portrait
of current state of the glacier, and a greatly improved under-
standing of the key processes that will determine its future
evolution.
5.1. Mass-balance gradients
Inverted ablation gradients, in which melt rates decrease with
elevation in the lower ablation zone, have long been recog-
nised as a fundamental factor underpinning the unique re-
sponse of debris-covered glaciers to climate change (Benn
and Lehmkuhl, 2000; Berthier and others, 2007; Scherler
and others, 2011; Benn and others, 2012; Rowan and
others, 2015). The data presented in this paper allow the
first detailed measurements of this important glaciological
variable for a Himalayan debris covered glacier.
To assess the relationship between surface elevation
change and glacier altitude the lower ablation area was
divided into 50 m long sections, extending across the width
of the glacier. The total volume change by both melting of
debris-covered ice and ice cliff backwasting was calculated,
together with the associated glacier areas (Fig. 13). For the
actively flowing part of the ablation area, total mass loss
was calculated for 300 m long sections; this is approximately
two wavelengths of the striping pattern, and thus should elim-
inate the effect of glacier movement on patterns of elevation
change (Fig. 5).
For the period December 2012January 2015 (assumed to
represent 2 balance years), the mass losses by sub-debris melt
were converted into mean annual rates and a mass-balance
gradient calculated from linear regression (Cogley and
others, 2012). Even though the surface lowering shows an in-
crease with elevation (Fig. 13), total surface lowering exhibits
a much more complex and heterogeneous pattern, with no
clear relationship between mass loss and elevation or dis-
tance from the terminus (Fig. 14b). At higher elevations, the
addition of backwasting processes results in a highly irregular
balance gradient, with local peaks of up to 2.2 m w.e. a
1
of
ice loss (assuming an ice density of 900 kg m
3
) (glacier-
wide mean). It is clear that the overall distribution of ablation
on the glacier is very sensitively dependent on the distribu-
tion of ice faces and their persistence through time.
5.2. Backwasting of ice cliffs
The heterogeneity in surface elevation change and ablation
rate is a direct result of rapid backwasting of exposed ice
cliffs, mostly around the margins of supraglacial lakes (Figs
13,14c). In the lower ablation area this process accounted
for 40% of the observed volume loss over both periods of
investigation, although ice cliffs occupy only 5% of the
glacier area. It is instructive to compare these results with
those of other studies of ablation on debris-covered glaciers.
To allow direct comparison we convert reported results to a
backwasting-sub-debris melt ratio BDR, which measures the
relative rates of the two processes:
BDR ¼
ð% total ablation by backwasting=% ice cliff areaÞ
% total ablation by sub debris melt
% debris covered area
:
ð4Þ
The results are shown in Table 2, and indicate (1) inter-
glacier variability in the area covered by ice cliffs and (2)
large variations in the ratio between backwasting and sub-
debris melt rates. Variability in ice cliff area reflects processes
of ice cliff formation and factors determining their persist-
ence. Ice cliffs on debris-covered glaciers have been
Fig. 13. Relative contributions to volume change from ice cliff retreat, lake change and sub-debris melt for the whole investigation period
across the lower ablation area. Note there is a clear increase in the contribution from sub-debris melt with distance from the terminus, this
is noticeably lacking in the contribution from lakes of backwasting.
13Thompson and others: Stagnation and mass loss on a Himalayan debris-covered glacier
observed to form by ice exposure from debris slumping,
surface subsidence into englacial voids and calving into
supraglacial ponds and lakes (Kirkbride, 1993; Benn and
others, 2001; Sakai and others, 2002). On Ngozumpa
Glacier >75% of ice cliffs in the lower ablation area bor-
dered supraglacial lakes or ponds in 2010. Factors influen-
cing the development and persistence of ice cliffs on
Lirung Glacier were investigated by Sakai and others
(2002). They found that melt rates on non-calving ice cliffs
depended predominantly on solar radiation and therefore
on the orientation of the ice cliff. South-facing cliffs received
more shortwave radiation at the top of the cliff than the
bottom due to shading. North-facing cliffs receive less short-
wave radiation and their energy balance is dominated by
long wave receipts, which are greater towards the base. As
a result, north-facing ice cliffs tend to be larger, steeper, gen-
erally debris free and therefore longer lived than south-facing
cliffs. Although initial backwasting could be rapid on south-
facing cliffs, ice faces become less steep and, once an angle
of <30° was achieved, faces became increasingly debris-
covered (Sakai and others, 2002). On Ngozumpa Glacier,
north-facing ice faces are most common and also persist for
much longer than south-facing ice faces, probably for
similar reasons.
The BDR on Ngozumpa Glacier is similar to that on Lirung
Glacier found by Sakai and others (2002), but is higher than
values determined on Koxkar and Miage Glaciers (Juen and
others, 2014; Reid and Brock, 2014;Table 2). High ratios
likely reflect the low background sub-debris melt rates in
the lower ablation area of Ngozumpa Glacier, in contrast
with sites where debris layers are thinner and less continuous
(cf. Reid and Brock, 2014). However, high ratios also reflect
high backwasting rates, reflecting the importance of calving
into supraglacial lakes (Figs 5,6b, 7). At Koxkar Glacier,
China, Han and others (2010) reported a mean ice cliff back-
wasting rate of 7.4 m a
1
. A similar value was reported for
Tasman Glacier (11 m a
1
) before the onset of calving,
which increased to 34 m a
1
after full height slab calving
onset (Röhl, 2008). At a number of ice cliffs on Ngozumpa
glacier, generally those <15 m in height, annual backwast-
ing rates are 10 m a
1
. However, backwasting rates 23
times greater were measured in a number of areas (Fig. 5;
Table 3).
Previous work on Ngozumpa Glacier found the threshold
for full height slab calving to be an ice cliff height of 15 m,
possibly related to the minimum stress gradients required to
reactivate relict crevasses (Benn and others, 2001). Thermal
under-cutting of ice cliffs at the water line has also been
observed to increase stress concentrations and increase ice
cliff calving rates (Kirkbride and Warren, 1997; Röhl, 2006,
2008; Benn and others, 2007; Sakai and others, 2009).
Under-cutting is accelerated by wind-driven currents, and
modeling carried out using ideal landform conditions con-
cluded the onset of calving due to undercutting is controlled
Fig. 14. (a) The theoretical modeled ablation balance for Ngozumpa Glacier (from Benn and others, 2012). The blue box highlights the zone
of the gradient shown in (b). (b) The total annual ablation gradient for the lower ablation area is calculated from the DEM difference maps. The
ablation gradients shown in (b) are calculated assuming an ice density of 900 kg m
3
.
Table 2. Volume loss and mean surface lowering for the periods 20102012 and 20122015
Glacier region Area Volume loss × 10
6
m
3
Surface lowering
m
km
2
20102012 20122015 20102012 20122015
Upper flowing area 12 40.1 ± 1.5 24.7 ± 3.3 3.3 ± 0.12 2.1 ± 0.28
Lower stagnant area 5.6 10.6 ± 0.7 7.1 ± 1.8 1.8 ± 0.11 1.3 ± 0.32
Whole area 17.6 50.8 ± 2.2 31.8 ± 5.7 2.9 ± 0.13 1.8 ± 0.33
Values for the whole area of investigation, the upper section experiencing ice flow and the lower area of stagnant ice are all included.
14 Thompson and others: Stagnation and mass loss on a Himalayan debris-covered glacier
by wind driven currents and occurs when the fetch is >80 m
(Sakai and others, 2009). This criterion is met by many of the
supraglacial lakes in Ngozumpa Glacier, particularly the
base level Spillway Lake, which has potential fetches of
several hundreds of meters (Fig. 10). The high backwasting
to sub-debris melt ratio at Ngozumpa Glacier, therefore,
likely reflects a high proportion of lake-contact ice cliffs,
which have higher backwasting rates than non-lake-contact
cliffs (Fig. 15). In turn, this may be symptomatic of the stage
of retreat of Ngozumpa Glacier, in which thermokarst pro-
cesses are well advanced.
5.3. Internal ablation
The measured volume changes along the lateral margins of
the glacier imply annual losses by internal ablation and sedi-
ment evacuation of 0.4 × 10
6
m
3
(Section 4.4). Here, we
compare this figure with calculated internal ablation rates
due to drainage of supraglacial lakes and loss of potential
energy by runoff of meltwater.
Over the 201215 period, there was an annual net loss of
water from supraglacial lakes of 0.26 × 10
6
m
3
. The volume
of internal melt caused by drainage of this water through en-
glacial conduits can be calculated, on the assumption that all
Table 3. Reported values for ablation on debris-covered glaciers converted to the backwasting-sub-debris melt ratio by Eqn (4).
Glacier % of glacier area % of ablation by backwasting Backwasting/sub-debris melt ratio Reference
Lirung 2 20 12.25 Sakai and others (1998,2002)
Lirung 8 24 3.63 Immerzeel and others (2014)
Miage 1.3 7.4 6.07 Reid and Brock (2014)
Koxkar 1.7 12 7.9 Juen and others (2014)
Ngozumpa 5 40 12.67 This work
Fig. 15. The location of the top of delineated ice cliffs surrounding the upper basin of Spillway Lake in (a) 2010, (a) 2012 and (a) 2015, overlain
on the orthorectified image in each instance. The ice cliff position of previous years is shown in both (b) and (c) to illustrate the change. (d) The
evolution of an individual ice cliff from 2009 through 2012, the location of which is marked by the red star in (a) and (b).
15Thompson and others: Stagnation and mass loss on a Himalayan debris-covered glacier
water was reduced from its initial temperature to 0°C during
its passage through the glacier. Observed surface tempera-
tures of Himalayan supraglacial ponds lie in the range,
0.76.6°C, with a mean of 3°C (summer and early
autumn; Wessels and others, 2002; Sakai and others,
2009). The ice mass M
l
that draining lake water can melt is:
Ml¼MwdTC
L

;ð5Þ
where M
w
is the water mass, dTthe temperature drop to
the melting point, Cthe specific heat capacity of water
(4.2 kJ kg
1
K
1
)andLthe latent heat of melting (334 kJ
kg
1
K
1
). This shows that 0.26 × 10
6
m
3
(2.6 × 10
8
kg) of
water can melt 7300 m
3
(6.5 × 10
6
kg) of ice for a tem-
perature drop of 2°C, and 18 200 m
3
(16.3 × 10
6
kg) of
ice for a temperature drop of 5°C. This calculation repre-
sents a minimum for internal ablation by this process
during the study period because it only includes the net
volume of surface lake water lost from the glacier, and
not the total flux of water through the supraglacial lakes.
This could be much larger than the net change, as some
lakes may have experienced episodes of filling and drain-
age between the dates of our images.
Internal ablation by potential energy losses during melt-
water runoff M
r
can be calculated from:
Mr¼PMmgdH
ρiL;ð6Þ
where M
m
and dHare the meltwater mass and elevation
above base level for successive sections of the glacier (cf.
Oerlemans, 2013). Taking the annual mass losses for
20122015, this yields an annual internal melt volume of
115 000 m
3
. Combining this with the previous values of
7300 to 18 200 m
3
for internal ablation by lake drainage,
the total annual internal ablation is 0.120.13 × 10
6
m
3
.
This compares well with the measured volume loss,
0.4 × 10
6
m
3
in the lateral moraine/trough systems, which
represents a combination of internal ablation and sediment
evacuation.
The estimated internal ablation is small, compared with
the total volume losses across the lower ablation area, but
this process exerts a major influence on the evolution of
the glacier. The collapse of englacial voids can form new
supraglacial lake basins and determine patterns of lake ex-
pansion (Benn and others, 2012; Thompson and others,
2012). Furthermore, sediment evacuation may have import-
ant consequences for the evolution of Spillway Lake, as dis-
cussed in the following section.
5.4. Spillway Lake: implications for evolution of base
level lakes
Spillway Lake underwent a period of dramatic expansion
between 2001 and 2009, when an area of 3 × 10
5
m
2
and
a volume of 2.2 × 10
6
m
3
(covering 85% of the area) was
attained (Thompson and others, 2012). In comparison, the
period covered by the present study (20092014) is one of
relative quiescence (Fig. 16b). Interruptions in the growth
of base-level lakes have been observed on other
Himalayan glaciers. For example, both Tsho Rolpa and
Lugge Tsho went through similar cycles of rapid areal expan-
sion punctuated by a period of shrinkage or very slow expan-
sion (Sakai and others, 2009;Fig. 16a). Our data allow the
causes of the reversal at Spillway Lake to be examined in
detail. Two main processes can be identified: (1) reduction
in the number of ice cliffs around the lake periphery and
(2) deposition into the lake. These two processes appear to
have been closely linked, and together acted to reduce
both the expansion rate and volume increase of the lake.
Mean debris cover thickness on the lower Ngozumpa
Glacier is 1.8 m with a maximum of >7 m (Nicholson,
2005). Thus, melting and calving of ice cliffs around
Spillway Lake transfers substantial volumes of debris from
Fig. 16. (a) shows the areal extent of Spillway Lake in the context of a number of other glacial lakes in the region. The pattern of rapid
expansion punctuated by periods of relative quiescence is evident on a number of other lakes adapted from Thompson and others (2012);
Sakai and others (2009). (b) The rapid increase in Spillway Lake area from the late 1990s to 2010 and more recent period of quiescence
from 2010 to 2014. The reduction in area is largely related to a drop in lake levels.
16 Thompson and others: Stagnation and mass loss on a Himalayan debris-covered glacier
the top of the cliffs into the lake. Accumulation of debris at
the base of ice cliffs can separate the ice cliff from the lake,
switching off calving and halting lake expansion. Spillway
Lake is an important sink for fine-grained sediment trans-
ported from up-glacier. The bathymetric profiles shown in
Figure 12 suggest sedimentation rates up to 1ma
1
in
basins that do not receive direct debris input from the adja-
cent slopes. High-sedimentation rates are consistent with
the evacuation of sediment from the lateral margins of the
glacier, as noted in Section 4.4. We conclude that sediment
redistribution can act as an important brake on the growth of
supraglacial base-level lakes, delaying their transition to full-
depth lakes. In addition, parts of Spillway Lake dropped in
elevation during the study period. The pattern and magnitude
of lake level change indicated the adjustment of lake basins
more recently integrated into the Spillway Lake complex to
the hydrological base level.
5.5. Evolution of the glacier
Our results catch Ngozumpa Glacier at a key moment in its
evolution. Benn and others (2012) proposed that, during
periods of negative mass balance, large debris-covered gla-
ciers cross thresholds between three regimes. In Regime 1,
the whole glacier is dynamically active, ablation is domi-
nated by sub-debris melting and meltwater is readily evacu-
ated via efficient drainage systems. The transition to Regime
2 occurs when ablation on the lower tongue exceeds influx
of ice from up-glacier. Progressive reduction of the ice
surface gradient (due to the inverted ablation gradient)
results in ice stagnation and retention of meltwater in
perched lakes. Melting and calving of bare ice faces
becomes increasingly dominant. Regime 3 occurs when
the lowering ice surface intersects the hydrologic base
level, and rapid ablation occurs by the growth of a base-
level lake. Initially, base-level lakes occupy supraglacial
positions, but can transition into full-depth lakes that can
expand up-glacier by calving (Sakai and others, 2009).
According to this scheme, Ngozumpa Glacier crossed the
threshold into the initial phase of Regime 3 in the 1990s,
when Spillway Lake began to form at base level. The lake
underwent a period of rapid expansion from 2001 to 2009
(Thompson and others, 2012), but since then has undergone
a reduction in both area and volume. Such hiatuses in the
growth of base-level lakes have been noted before (e.g.
Sakai and others, 2009), but the data presented in this
paper provide the first detailed view of their character and
possible cause. For Spillway Lake, the key process appears
to be redistribution of sediment, both dumping of debris at
the base of retreating ice cliffs and deposition of fine sedi-
ment transported from up-glacier.
By analogy with other debris-covered glaciers in the
Himalaya, it can be expected that Spillway Lake will transi-
tion into a full depth lake in the coming years (Sakai and
others, 2009; Thompson and others, 2012). Observations
on similar glacier systems in New Zealand suggest that
buoyant calving may play a major role in this transition, con-
sequent upon a reduction of overburden pressures by surface
lowering (Dykes and others, 2010).
6. CONCLUSIONS
The combination of high-resolution optical stereo imagery
and lake bathymetry have allowed a detailed assessment of
the patterns of mass loss occurring on the lower ablation
area of Ngozumpa Glacier. Evolution of the lower ablation
area results from a complex suite of processes and feedbacks.
Our principal conclusions are as follows.
(1) Most mass loss occurs by melt below supraglacial debris,
and melt and calving of ice cliffs (backwasting). For the
period 20122015, annual losses by these processes on
the stagnant lower tongue are 2.15 ± 1.5 × 10
6
m
3
and
1.6 ± 0.07 × 10
6
m
3
, respectively, accounting for 52
and 39% of total volume loss. Ice cliffs cover 5% of
the area of the lower tongue, but account for almost
40% of the ablation. The implied ratio of backwasting
and sub-debris melt rates is at the upper end of values
reported in the literature, probably as the result of thick
debris cover (low sub-debris melt rates) and rapid
calving around lake margins.
(2) Subsidence of lateral troughs along both margins of the
glacier indicates internal ablation in association with
sub-marginal drainage channels. Calculated melt rates
yield internal ablation rates >0.1 × 10
6
m
3
a
1
, while
measured volume changes indicate that internal ablation
plus sediment evacuation is 0.4 × 10
6
m
3
a
1
, providing
an upper bound for internal ablation, accounting for
9% of the total annual glacial ablation.
(3) The surface debris layer is subject to frequent re-distribution
by slope processes. This can result in large differences in
debris-layer thickness, enhancing or inhibiting local
ablation rates and encouraging continuous topographic
inversion.
(4) The base level lake, Spillway Lake is in a period of rela-
tive quiescence, following a period of rapid expansion
from 2001 to 2009. Reduction in lake area and volume
between 2009 and 2015 resulted from lake level lower-
ing and redistribution of sediment from both local and
non-local sources (backwasting ice cliffs and suspended
sediment transported by melt streams, respectively).
Sediment redistribution therefore acts as an important,
though likely temporary, brake on lake expansion.
(5) It is likely that rapid lake growth will resume in the near
future, although it is not possible to make precise predic-
tions due to the complexity of the system and the exist-
ence of both positive and negative feedbacks between
key processes. Mass loss on the lower glacier averages
1 m w.e. a
1
(Fig. 12c). Approximately 7.0 × 10
5
m
2
of
the glacier has an elevation within 10 m of the hydro-
logical base level. Combined with the 2014 lake area
of 2.4 × 10
5
m
2
, this suggests lake expansion to an areal
extent of 1 km
2
within the next 10 a. Because of its low
gradient and stagnant nature, if a fully formed moraine
dammed lake does develop in the future, it is possible
that the lake may eventually become up to 7 km long if
the moraine dam remains in place.
(6) The lowermost 1 km of the glacier (below Spillway Lake)
is largely stable, and very little change was identified
over the entire period of investigation. The zone of stabil-
ity includes the spillway through the western lateral
moraine, which has maintained a constant elevation
since our first survey in 2001 (Benn and others, 2001).
While the developing base level lake remains separated
from the terminal moraine by 1 km, there is unlikely
to be any major risk of GLOFs from the glacier.
(7) High-resolution stereo imagery is a powerful tool for
monitoring volume change on large debris-covered
17Thompson and others: Stagnation and mass loss on a Himalayan debris-covered glacier
glaciers, allowing quantification of processes not visible
on lower resolution imagery and difficult to observe in
the field due to the scale and complexity of the terrain.
ACKNOWLEDGEMENTS
This research was supported financially by the University
Centre in Svalbard (UNIS), National Geographic Society
GRANT #W135-10, The Natural Environmental Research
Council and the European Commission FP7-MC-IEF.
Thanks to Duncan Quincey and Akiko Sakai for their thor-
ough reviews that greatly improved the manuscript. We
thank Joe Alexander, Alia Kahn and Wes Farnsworth for
help in the field, Rijan Bhakta Kayastha for invaluable assist-
ance in obtaining the correct permits for work in Nepal,
Endra Rai Bahing, Ani Bhattarai and Sujan Bhattarai for logis-
tical support and Lhakpa Nuru Sherpa and all of the staff at
the Cho La Pass Resort for fantastic hospitality, logistical
support and general assistance in all fieldwork.
REFERENCES
Barrand NE, James TD and Murray T (2010)Spatiotemporal variability
in elevation changes of two high-Arctic valley glaciers. J. Glaciol.,
56(199), 771780 (doi: 10.3189/002214310794457362)
Benn DI and Lehmkuhl F (2000) Mass balance and equilibrium-line
altitudes of glaciers in high mountain environments. Quatern.
Int.,65/66,1529 (doi: 10.1016/S1040-6182(99)00034-8)
Benn DI, Wiseman S and Hands KA (2001) Growth and drainage of
supraglacial lakes on the debris-mantled Ngozumpa Glacier,
Khumbu Himal, Nepal. J. Glaciol.,47, 626638 (doi: 10.3189/
172756501781831729)
Benn DI, Warren CR and Mottram RH (2007) Calving processes
and the dynamics of calving glaciers, Earth-Sci. Rev.,82, 143
179 (doi: 10.1016/j.earscirev.2007.02.002)
Benn DI and 9 others (2012) Response of debris-covered glaciers in
the Mount Everest region to recent warming, and implications for
outburst flood hazards. Earth-Sci. Rev.,114, 156174 (doi:
10.1016/j.earscirev.2012.03.008)
Berthier E, Arnaud Y, Baratoux D, Vincnet C and Rémy F (2004)
Recent rapid thinning of the Mer de Glaceglacier derived
from satellite optical images, Geophys. Res. Let.,31, L17401
(doi: 10.1029/2004GL020706)
Berthier E and 5 others (2007) Remote sensing estimates of glacier
mass balances in the Himachal Pradesh (Western Himalaya,
India). Remote Sens. Environ.,108(3), 327338 (doi: 10.1016/
j.rse.2006.11.017)
Bolch T, Buchroithner MF, Pieczonka T and Kunert A (2008)
Planimetric and volumetric glacier changes in Khumbu Himalaya
since 1962 using Corona, Landsat TM and ASTER data. J.
Glaciol.,54,592600 (doi: 10.3189/002214308786570782)
Bolch T, Pieczonka T and Benn DI (2011) Multi-decadal mass loss of
Glaciers in the Everest area (Nepal Himalaya) derived from stereo
imagery. Cryosphere,5, 349358 (doi: 10.5194/tc-5-349-2011)
Cogley JG and 10 others (2012) Glossary of Glacier mass balance
and related terms. IHP-VII Technical Documents in Hydrology
No. 86, IACS Contribution No. 2. UNESCO-IHP, Paris
Dykes RC, Brook MS and Winkler S (2010) The contemporary retreat
of Tasman Glacier, Southern Alps, New Zealand, and the evolu-
tion of Tasman proglacial lake since AD 2000. Erdkunde,64
(H. 2), 141154 (doi: 10.3112/erdkunde.2010.02.03)
Frazer CS and Ravanbakhsh M (2009) Georeferencing accuracy of
GeoEye-1 stereo imagery. Photogramm. Eng. Remote Sens.,75
(6), 634638
Fujita K and Yutaka A (2000) Effect of summer accumulation on
glacier mass balance on the Tibetan Plateau revealed by mass
balance model, J. Glaciol.,46, 153
Gardelle J, Berthier E, Arnaud Y and Kääb A (2012) Region-wide
glacier mass balances over the Pamir-Karakoram-Himalaya
during 19992011. Cryosphere,7, 12631286 (doi: 10.5194/tc-
7-1263-2013)
Gulley J and Benn DI (2007) Structural control of englacial drainage
systems in Himalayan debris-covered glaciers. J. Glaciol.,53
(182), 399412 (doi: 10.3189/002214307783258378)
Gulley J, Benn DI, Luckman A and Müller D (2009) A cut-and-
closure origin for englacial conduits on uncrevassed parts of
polythermal glaciers. J. Glaciol.,55(189), 6680 (doi: 10.3189/
002214309788608930)
Han H, Wang J, Wei J and Liu S (2010) Backwasting rate on debris-
covered Koxkar glacier, Tuomuer Mountain, China. J. Glaciol.,
56(196), 287296 (doi: 10.3189/002214310791968430)
Immerzeel WW and 6 others (2014) High-resolution monitoring of
Himalayan glacier dynamics using unmanned aerial vehicles.
Remote Sens. Environ.,150,93103 (doi: 10/1016/j.
rse.2014.04.025)
Juen M, Mayer C, Lambrecht A, Han H and Liu S (2014) Impact of
varying debris cover thickness on ablation: a case study for
Koxkar Glacier in the Tien Shan. Cryosphere,8, 377368 (doi:
10.5194/tc-8-377-2014)
Kirkbride MP (1993) The temporal significance of transitions from
melting to calving termini at glaciers in the central Southern
Alps of New Zealand, Holocene,3, 232240.
Kirkbride MP and Warren CR (1997) Calving processes at a
grounded ice cliff, Ann. Glaciol.,24, 116121.
Kääb A (2005) Remote sensing of mountain glaciers and permafrost
creep. Geographisches Institut der Universitürich, Zürich
Kääb A, Berthier E, Nuth C, Gardelle J and Arnaud Y (2012)
Contrasting patterns of early twenty-first-century glacier mass
change in the Himalayas. Nature,488, 495498 (doi: 10.1038/
nature11324)
Luckman A, Quincey DJ and Bevan S (2007) The potential of satel-
lite radar interferometry and feature tracking for monitoring flow
rates of Himalayan glaciers. Remote Sens. Environ.,111, 172
181 (doi: 10.1016/j.rse.2007.05.019)
Nakawo M and Rana B (1999) Estimate of ablation rate of glacier ice
under a supraglacial debris layer. Geogr. Ann.,81(A), 695701
(doi: 10.1111/1468-0459.00097)
Nicholson LA (2005) Modelling melt beneath supraglacial debris:
implications for the climatic response of debris-covered glaciers.
(PhD thesis, University of St. Andrews, UK)
Nicholson LA and Benn DI (2006) Calculating ice melt beneath a
debris layer using meteorological data. J. Glaciol.,52(178),
463470 (doi: 10.3189/172756506781828584)
Nicholson LA and Benn DI (2013) Properties of natural supraglacial
debris in relation to modelling sub-debris ice ablation. Earth Surf.
Proc. Land.,38, 490501 (doi: 10.1002/esp.3299)
Nuimura T and 5 others (2011) Temporal changes in elevation of the
debris-covered ablation area of Khumbu Glacier in the Nepal
Himalaya since 1978. Arct. Antarc. Alp. Res.,43(2), 246255
(doi: 10.1657/1938-4246-43.2.246)
Nuth C and Kääb A (2011) Co-registration and bias corrections of
satellite elevation data sets for quantifying glacier thickness
change. Cryosphere,5, 271290 (doi: 10.5194/tc-5-271-2011)
Oerlemans J (2013) A note on the water budget of temperate gla-
ciers. Cryosphere,7, 15571564 (doi: 10.5194/tc-7-1557-2013)
Østrem G (1959) Ice melting under a thin layer of moraine, and theex-
istence of ice cores in moraine ridges. Geogr. Ann.,41(4), 228230
Quincey DJ, Luckman A and Benn DI (2009) Quantification of
Everest region glacier velocities between 1992 and 2002, using
satellite radar interferometry and feature tracking. J. Glaciol.,55
(192), 596606 (doi: 3189/002214309789470987)
Reid TD and Brock BW (2014) Assessing ice-cliff backwasting and
its contribution to total ablation of debris-covered Miage
glacier, Mont Blanc massif. Italy, J. Glaciol.,60(219), 313 (doi:
10.3189/2014JoG045)
Reznichenko N, Davies T, Shulmeister J and McSaveney M (2010)
Effects of debris on ice-surface melting rates: an experimental
18 Thompson and others: Stagnation and mass loss on a Himalayan debris-covered glacier
study. J. Glaciol.,56(197), 385394 (doi: 10.3189/002214310
792447725)
Röhl K (2006) Thermo-erosional notch development at fresh-water-
calving Tasman Glacier, New Zealand. J. Glaciol.,52(177), 203
213 (doi: 10.3189/172756506781828773)
Röhl K (2008) Characteristics and evolution of supraglacial ponds on
debris-covered Tasman Glacier, New Zealand. J. Glaciol.,54
(188), 867880 (doi: 10.3189/002214308787779861)
Rowan AV, Egholm DL, Quincey DL and Glasser NF (2015)
Modelling the feedbacks between mass balance, ice flow and
debris transport to predict the response to climate change of
debris-covered glaciers in the Himalaya. Earth Planet. Sci. Lett.,
430, 427438 (doi: 10.1016/j.epsl.2015.09.004)
Sakai A and Fujita K (2010) Formation conditions of supraglacial
lakes on debris-covered glaciers in the Himalaya. J. Glaciol.,
56(195), 177181 (doi: 10.3189/002214310791190785)
Sakai A, Nakawo M and Fujita K (1998) Melt rate of ice cliffs on the
Lirung Glacier, Nepal Himalaya. Bull. Glacier Res.,16,5766
Sakai A, Takeuchi N, Fujita K and Nakawo M (2000a) Role of sup-
raglacial ponds in the ablation processes in debris-covered
glaciers in the Nepal Himalayas. In Nakawo N, Fountain A and
Raymond C eds. Debris-covered glaciers. IAHS-AISH
Publication 264, Wallingford, 119130
Sakai A, Chikita K and Yamada T (2000b) Expansion of a moraine-
dammed glacial lake, Tsho Rolpa, in Rolwaling Himal, Nepal
Himalaya. Limnol. Oceanogr.,45, 14011408 (doi: 10.4319/
lo.2000.45.6.1401)
Sakai A, Nakawo M and Fujita K (2002) Distribution, characteristics
and energy balance on ice cliffs in debris-covered glaciers, Nepal
Himalayas. Arct. Antarct. Alp. Res.,34(1), 1219 (doi: 10.2307/
1552503)
Sakai A, Nishimura K, Kadota T and Tekeuchi N (2009) Onset of
calving at supraglacial lakes on debris-covered glaciers of the
Nepal Himalaya. J. Glaciol.,55(193), 909917 (doi: 10.3189/
002214309790152555)
Scherler D, Bookhagen B and Strecher MR (2011) Spatially variable
response of Himalayan glaciers to climate change affected
by debris-cover. Nat. Geosci.,4, 156159 (doi: 10.1038/
ngeo1068)
Thompson SS, Benn DI, Dennis K and Luckman A (2012) A rapidly
growing moraine-dammed glacial lake on Ngozumpa Glacier,
Nepal. Geomorphology,145146,1
11 (doi: 10.1016/
j.geomorph.2011.08.015)
Wessels RL, Kargel JS and Kieffer HH (2002) ASTER measurement of
supraglacial lakes in the Mount Everest region of the Himalaya,
Ann. Glaciol.,34,399408 (doi: 10.3189/172756402781817545)
MS received 22 October 2015 and accepted in revised form 6 January 2016
19Thompson and others: Stagnation and mass loss on a Himalayan debris-covered glacier
... ice cliffs) and supraglacial ponds are often scattered across the surface of debris-covered glaciers (Steiner et al 2019, Kneib et al 2021a; these features are disproportionately responsible for these glaciers' mass loss (e.g. Immerzeel et al 2014, Thompson et al 2016, Salerno et al 2017, Brun et al 2018, Mölg et al 2019 and are difficult to constrain due to their variability in space and time (Miles et al 2017a, Steiner et al 2019. ...
... Sakai et al 2002, Thompson et al 2016, Salerno et al 2017. However, these features typically account for less than 15% of the glacier's debris-covered area (e.g. ...
Article
Full-text available
Supraglacial debris covers 7% of mountain glacier area globally and generally reduces glacier surface melt. Enhanced energy absorption at ice cliffs and supraglacial ponds scattered across the debris surface leads these features to contribute disproportionately to glacier-wide ablation. However, the degree to which cliffs and ponds actually increase melt rates remains unclear, as these features have only been studied in a detailed manner for selected locations, almost exclusively in High Mountain Asia. In this study we model the surface energy balance for debris-covered ice, ice cliffs, and supraglacial ponds at a set of automatic weather station records representing the global prevalence of debris-covered glacier ice. We generate 5000 random sets of values for physical parameters using probability distributions derived from literature, which we use to investigate relative melt rates and to isolate the melt responses of debris, cliffs and ponds to the site-specific meteorological forcing. Modelled sub-debris melt rates are primarily controlled by debris thickness and thermal conductivity. At a reference thickness of 0.1 m, sub-debris melt rates vary considerably, differing by up to a factor of four between sites, mainly attributable to air temperature differences. We find that melt rates for ice cliffs are consistently 2-3x the melt rate for clean glacier ice, but this melt enhancement decays with increasing clean ice melt rates. Energy absorption at supraglacial ponds is dominated by latent heat exchange and is therefore highly sensitive to wind speed and relative humidity, but is generally less than for clean ice. Our results provide reference melt enhancement factors for melt modelling of debris-covered glacier sites, globally, while highlighting the need for direct measurement of debris-covered glacier surface characteristics, physical parameters, and local meteorological conditions at a variety of sites around the world.
... As a significant agent of landscape changes, glacier movement or stagnation is of notable interest to the geomorphologist (Watson and Quincey 2015). The implication of glacier movement includes reconstructing the past glacier extent (Benn and Evans 2010), assessing glacial lake formation due to recession (Shugar et al. 2020) and associated hazards (Westoby et al. 2014), and examining glacier morphology and dynamics due to surge events (Quincey et al. 2011;Bhambri et al. 2020) or stagnation and surface lowering of debris-covered glaciers (Thompson et al. 2016;Steiner et al. 2021). Velocity assessment of debris-covered glaciers is important due to their complex response to climate change (Watson and Quincey 2015;Miles et al. 2018). ...
... Velocity assessment of debris-covered glaciers is important due to their complex response to climate change (Watson and Quincey 2015;Miles et al. 2018). Debris-covered glaciers that are stagnating and losing mass primarily through surface lowering can be more prone to the development of supraglacial ponds, which can accelerate mass loss (Thompson et al. 2016;Miles et al. 2020;Steiner et al. 2021). Therefore, it can be assumed that the surface velocity has a significant impact on the health and sustenance of a glacier. ...
Article
Full-text available
Spatiotemporal surface velocity measurements of the alpine valley type debris-covered Miyar Glacier of the Chandrabhaga (Chenab) basin, western Himalaya, were assessed based on the cross-correlation of Landsat images spanning nearly three decades (1992-2019). Long-term (1950-2015) temperature and precipitation trends were evaluated using Asian Precipitation Highly Resolved Observational Data Integration Towards Evaluation (APHRODITE) datasets. The mean velocity (1992-2019) of the Miyar Glacier is ∼29 m/yr, with spatial patterns revealing that the debris-covered tongue is nearly stagnant (∼5 m/yr) compared to the debris-free up-glacier zone (∼35 m/yr). The transition zone from clean to debris-covered ice in the mid-ablation area shows the highest long-term mean velocities of ∼60 m/yr during the observation period, likely resulting from a steep surface gradient and greater ice thickness than the other regions of this glacier. The slow-moving and nearly stagnant debris-covered area reveals the highest amount of surface lowering due to the expansion of supraglacial ponds. Miyar Glacier experiences summer speed-up of ∼67-80% in seasonal velocity compared to winter, interpreted as a result from enhanced basal sliding during summer months due to warmer temperatures inputting more meltwater into the subsurface drainage system. Inter-annual velocity variations are greatest in the upper glacier, with higher velocities observed more frequently in recent decades. Future work should aim to elucidate the causes of this pattern, considering the overall rising air temperature trend in the western Himalaya. ARTICLE HISTORY
... As a significant agent of landscape changes, glacier movement or stagnation is of notable interest to the geomorphologist (Watson and Quincey 2015). The implication of glacier movement includes reconstructing the past glacier extent (Benn and Evans 2010), assessing glacial lake formation due to recession (Shugar et al. 2020) and associated hazards (Westoby et al. 2014), and examining glacier morphology and dynamics due to surge events (Quincey et al. 2011;Bhambri et al. 2020) or stagnation and surface lowering of debris-covered glaciers (Thompson et al. 2016;Steiner et al. 2021). Velocity assessment of debris-covered glaciers is important due to their complex response to climate change (Watson and Quincey 2015;Miles et al. 2018). ...
... Velocity assessment of debris-covered glaciers is important due to their complex response to climate change (Watson and Quincey 2015;Miles et al. 2018). Debris-covered glaciers that are stagnating and losing mass primarily through surface lowering can be more prone to the development of supraglacial ponds, which can accelerate mass loss (Thompson et al. 2016;Miles et al. 2020;Steiner et al. 2021). Therefore, it can be assumed that the surface velocity has a significant impact on the health and sustenance of a glacier. ...
Article
Full-text available
Spatiotemporal surface velocity measurements of the alpine valley type debris-covered Miyar Glacier of the Chandrabhaga (Chenab) basin, western Himalaya, were assessed based on the cross-correlation of Landsat images spanning nearly three decades (1992-2019). Long-term (1950-2015) temperature and precipitation trends were evaluated using Asian Precipitation Highly Resolved Observational Data Integration Towards Evaluation (APHRODITE) datasets. The mean velocity (1992-2019) of the Miyar Glacier is ∼29 m/yr, with spatial patterns revealing that the debris-covered tongue is nearly stagnant (∼5 m/yr) compared to the debris-free up-glacier zone (∼35 m/yr). The transition zone from clean to debris-covered ice in the mid-ablation area shows the highest long-term mean velocities of ∼60 m/yr during the observation period, likely resulting from a steep surface gradient and greater ice thickness than the other regions of this glacier. The slow-moving and nearly stagnant debris-covered area reveals the highest amount of surface lowering due to the expansion of supraglacial ponds. Miyar Glacier experiences summer speed-up of ∼67-80% in seasonal velocity compared to winter, interpreted as a result from enhanced basal sliding during summer months due to warmer temperatures inputting more meltwater into the subsurface drainage system. Inter-annual velocity variations are greatest in the upper glacier, with higher velocities observed more frequently in recent decades. Future work should aim to elucidate the causes of this pattern, considering the overall rising air temperature trend in the western Himalaya.
... Sakai et al., 1998Sakai et al., , 2000Buri et al., 2016;Miles et al., 2016;Nicholson et al., 2018), that have been shown to contribute disproportionately to glacier-wide ablation (e.g. Sakai et al., 1998;Immerzeel et al., 2014;Juen et al., 2014;Thompson et al., 2016). ...
... Declining surface gradient and mean ice thickness during negative mass balance phases both act to reduce the driving stress and ice flow within the debris-covered tongue, typically causing progressive stagnation (e.g. Bolch et al., 2008b;Quincey et al., 2009;Thompson et al., 2016). Some debris-covered glaciers form large, impounding lateroterminal moraine complexes, as are prevalent in the eastern Nepalese Himalaya (e.g. ...
Article
Full-text available
Ongoing changes in mountain glaciers affect local water resources, hazard potential and global sea level. An increasing proportion of remaining mountain glaciers are affected by the presence of a surface cover of rock debris, and the response of these debris-covered glaciers to climate forcing is different to that of glaciers without a debris cover. Here we take a back-to-basics look at the fundamental terms that control the processes of debris evolution at the glacier surface, to illustrate how the trajectory of debris cover development is partially decoupled from prevailing climate conditions, and that the development of a debris cover over time should prevent the glacier from achieving steady state. We discuss the approaches and limitations of how this has been treated in existing modeling efforts and propose that “surrogate world” numerical representations of debris-covered glaciers would facilitate the development of well-validated parameterizations of surface debris cover that can be used in regional and global glacier models. Finally, we highlight some key research targets that would need to be addressed in order to enable a full representation of debris-covered glacier system response to climate forcing.
... Our finding indicates that the mean mass balance between 2000 and 2020 for New Zealand glaciers that have ice-marginal lakes may be more negative than that of glaciers without lakes (Fig. 6B), which agrees with previous analyses of glaciers changes over the last few decades in Alaska (Field et al., 2021) and the Himalaya (e.g. Basnett et al., 2013;Thompson et al., 2016;King et al., 2018;Tsutaki et al., 2019). To a certain extent, more negative mass balance for glaciers with ice-marginal lakes compared to landterminating glaciers can also be attributed by the general characteristics of such glaciers with wide, gently-sloping tongues which is an indication for long response times and, hence, a larger imbalance with current climate (Field et al., 2021). ...
Article
Full-text available
Global glacier mass loss is causing expansion of proglacial landscapes and producing meltwater that can become impounded as lakes within natural topographic depressions or ‘overdeepenings’. It is important to understand the evolution of these proglacial landscapes for water resources, natural hazards and ecosystem services. In this study we (i) overview contemporary loss of glacier ice across the Southern Alps of New Zealand, (ii) analyse ice-marginal lake development since the 1980s, (iii) utilise modelled glacier ice thickness to suggest the position and size of future lakes, and (iv) employ a large-scale glacier evolution model to suggest the timing of future lake formation and future lake expansion rate. In recent decades, hundreds of Southern Alps glaciers have been lost and those remaining have fragmented both by separation of tributaries and by detachment of ablation zones. Glaciers with ice-contact margins in proglacial lakes (n > 0.1 km² = 20 in 2020) have experienced the greatest terminus retreat and typically twice as negative mass balance compared to similar-sized land-terminating glaciers. Our analysis indicates a positive relationship between mean glacier mass balance and rate of lake growth (R² = 0.34) and also with length of an ice-contact lake boundary (R² = 0.44). We project sustained and relatively homogenous glacier volume loss for east-draining basins but in contrast a heterogeneous pattern of volume loss for west-draining basins. Our model results show that ice-marginal lakes will increase in combined size by ~150% towards 2050 and then decrease to 2100 as glaciers disconnect from them. Overall, our findings should inform (i) glacier evolution models into which ice-marginal lake effects need incorporating, (ii) studies of rapid landscape evolution and especially of meltwater and sediment delivery, and (iii) considerations of future meltwater supply and water quality.
... We observed a pronounced change in the mass loss in the snout region (section 4.3), which may be attributed to certain localized geomorphic features such as dense networking of ice cliffs and the development of glacial lakes (proglacial and supraglacial ponds) in this zone (Fig. 1e, f). These features are known to amplify the local ablation rates (Thompson et al., 2016;Steiner et al., 2019) and is evidenced from the subsidence of the frontal portion (Fig. 5 c1, c2, c3; Garg et al., 2019) of the Kangriz glacier. The glacier is also losing mass through ice calving, observed frequently in the left snout portion Fig. 1e; Supplementary Fig. S3). ...
Article
Glacier specific studies, in a relatively unexplored terrain of Ladakh, hold immense importance tocomprehend not only the glacier response but also its synchronicity with the general regional trend. Accordingly, in this study, the Kangriz glacier in the Suru sub-basin, western Himalaya, has been taken up for multiparametric (area, terminal retreat, debriscover, snow line altitude, surface dynamics) assessment for the period 1971–2018. Results reveal an overall shrinkage of 3.3 ± 1.6%, with an expansion in the supraglacial debris cover by 45% (1971–2018). Concomitantly, the glacier surface velocity has reduced by 10.85 ± 5.68 ma⁻¹ (35%), from 31.2 ± 5.8 ma⁻¹ (1993/94) to 20.3 ± 1.7 ma⁻¹ (2017/18), with mass wastage of −0.52 ± 0.19 m w.e.a⁻¹ during 2000–17. The notable glacier degeneration is synchronous with regional warming (Tmax increase by 7%, Tmin increase by 43%, Tavg increase by 64%) and a decrease in precipitation by 3% (significant at α <0.05). Besides, frontal dynamics have changed recently, with an enhanced intensity of terminal retreat (2016–2018: 57 ± 13(Stdev) ma⁻¹). The overall glacier status suggests a degenerative pattern of the glacier, which is in sync with the other western Himalayan glaciers. In view of the recently amplified ice-calving events and rapid mass loss observed in the snout region, the frontal glacier morphology may change drastically in the coming years .
... However, because of bad weather conditions and wind, the prior lines were deviated in real-time surveying. Bathymetric models were created within ArcGIS 10.3 using natural neighbor interpolation (Haritashya et al. 2018;Thompson et al. 2016), and lake volume was calculated. Bathymetric modeling requires an outline with zero depth. ...
Article
Full-text available
Shrinkage of some of the glaciers has direct impacts on the formation and expansion of glacial lakes. Sudden glacial lake outburst floods (GLOFs) are a major threat to lives and livelihoods downstream as they can cause catastrophic damage. In this study, we present the dynamics of the Lower-Barun glacier and glacial lakes and their GLOF susceptibility. We used multi temporal Landsat and Sentinel satellite imagery and extracted the lake outlines using the Normalized Difference Water Index (NDWI) with manual post-correction while the glacier outline was digitized manually. Multi-criteria decision-based method was used to assess the GLOF susceptibility. For the estimation of peak discharge and failure time, an empirical model developed by Froelich (1995) was used. The surface area of the Lower-Barun glacial lake was increased by 86% in the last 40 yrs (from 1979 to 2018), with a mean increase of 0.0432 km2/yr. The shrinkage in the glacier area is around 0.49 km2/yr and has shrunk by 8% in the last four decades. The retreat of the Lower-Barun glacier was 0.20% per year in the last four decades. The susceptibility index was 0.94, which suggests that the lake is very highly susceptible to the GLOF. The peak discharge of 5768 m3/s is produced when the breach depth is 20 m and the entire water volume is released. Likewise, in the case of 15 m breach depth, the peak discharge of 4038 m3/s is formed. Breach depth scenario of 10 m, peak discharge of 2442 m3/s is produced and in case of breach depth of 5 m produces the peak discharge of 1034 m3/s. If GLOF occurs, it can exert disastrous impacts on the livelihood and infrastructure in the downstream. So, it is necessary to examine such lakes regularly and mitigation measures to lower the GLOF susceptibility should be emphasized.
Article
This study explores the dynamics of the Satopanth (SPG) and Bhagirathi-Kharak (BKG) glaciers of the Mana basin in the Central Himalayan region of India under the scenario of climate change. The spatio-temporal changes in the glacier area, volume, and shift in snout positions during the period from 1968 to 2017 were assessed based on the CORONA photograph and satellite data. The study also intended to identify the common drifts in glacier areas for observing temporal and seasonal (during the winter-spring-summer transition) changes in the glacial landscapes and ably evaluate glacier velocity and ice-marginal lake stage. The ice velocity was found to vary from 0.117 md⁻¹ to 0.165 md⁻¹ for January to October months. It was found to have increased during the winter-spring-summer transition periods used to estimate the ice flow intensity and its phase information regular basis. Besides this, the linear trends of the MOD11A2 land surface temperature product were obtained and temperature variability has been analyzed from 2001 to 2020. The results show significant warming over the basin along with a significant negative trend in snow cover area during different seasons. The derived information is quite useful in analyzing the climate change footprints such as snowline movement, snow volume changes, and glacier mass balance-related studies.
Article
Full-text available
Impacts of climate change on snow cover, permafrost and glaciers at varied elevations are wide ranging and well-known. Melting water from glaciers forms glacial lakes; where, under the changing climate and accelerating glacial retreats have been a prime cause of glacial lake growth and outbursts. Indian Himalayan Region has also witnessed those impacts on glacial lakes with several evidences of glacial lake outbursts in recent years affecting people and ecosystems. Thus, to provide an understanding of geoclimatic environment and glacial lakes of the region, this article has attempted to look into the spatio-temporal changes of climatic conditions and physical landscape in Chamoli district of Uttarakhand in Indian Himalayan Region. In this study, mapping of glacial lakes in the district has been done to look into the aspects of associated hazards. The assessment has been done using historical trend analysis of climatic parameters, land surface temperature and physical landscape change in the study area. Seismic activities of the region have also been taken into account to understand the vulnerability of glacial lakes. The results indicate dynamic environmental conditions with visible physical landscape change; where, increase in built-up area by 541.57 Km², decrease in dense forest and fallow land by 76.96 Km² and 364.09Km² respectively over a period of two decades (2000–2020) have been observed. Minimum temperature has been gradually increasing by 0.68 °C per decade. The geospatial analysis provides identification of ∼500 glacial lakes, where 40.92% are situated at 5000–6000 m elevation. It is understood that glacial lakes of the study area are prone to produce disasters in the region because of changing climate and active seismicity, and lakes at higher elevation are highly susceptible to GLOF (Glacial Lakes Outburst Flood). Moreover, prevalence of large number of supraglacial lakes in the study region which are dynamic in nature, makes the area more vulnerable to hazards. The very old supraglacial lakes in high altitude region need attention and detailed study in terms of their changing characteristics and stability. Anthropogenic interventions in the area are contributing towards higher intensity of impacts on human and resources. It is suggestive that further development activities should consider the potential hazards and vulnerable zones in study area for disaster risk reduction.
Thesis
Full-text available
Cold polar and mountainous periglacial environments are characterised by highly dynamic ground surfaces that move under the action of frost and gravity, and contribute to shaping the landscape. The movement rates and directions are spatially and temporally variable, depending on the involved periglacial processes and their environmental controlling factors. Spaceborne Synthetic Aperture Radar (SAR) has revolutionised the investigation of the ground surface in polar and mountainous regions, due to its ability to image large and remote areas independently of light and meteorological conditions. By comparing images taken at different times, the SAR Interferometry (InSAR) technique can remotely detect ground surface displacements at centimetre to millimetre accuracy. This thesis exploits the InSAR technology to analyse the spatial distribution and temporal variability of the ground surface displacements in periglacial environments. The spatio-temporal displacement patterns are documented on a regional scale, allowing for the study of the kinematic signatures of various frost- and gravity-driven processes. The research takes advantage of the Copernicus Sentinel-1 SAR mission that provides open access images with unprecedented spatial coverage and at a weekly temporal resolution. Sentinel-1 InSAR is complemented with results based on the high spatial resolution TerraSAR-X images. The advantages and limitations of InSAR in the scope of periglacial research are discussed based on case studies in central and western Spitsbergen (Svalbard) and in Troms and Finnmark (Northern Norway). In Papers I–V, we demonstrate the ability of InSAR to document the kinematic properties of the periglacial ground dynamics, characterised by displacement rates ranging from a millimetre to a metre over a season or a year. InSAR allows for investigating the displacement progression caused by the ground freeze–thaw cycles in permafrost lowlands and the gradual downslope creep of periglacial landforms in mountainous environments. InSAR observations are compared with in-situ measurements and geomorphological mapping, and coupled with statistical and physical modelling. This integration contributes to a better understanding of the factors controlling the spatio-temporal patterns of the ground movement. This research suggests novel ways to develop dedicated InSAR products relevant for the assessment of geohazards and the systematic observation of ground dynamics in the context of climate change. The results show the value of combining satellite InSAR with complementary remote sensing techniques to document fast-moving landforms and provide decadal time series. Finally, this dissertation outlines perspectives for furthering the work in the scope of InSAR applied to periglacial research.
Article
Full-text available
In this note, the total dissipative melting in temperate glaciers is studied. The analysis is based on the notion that the dissipation is determined by the loss of potential energy due to the downward motion of mass (ice, snow, meltwater and rain). A mathematical formulation of the dissipation is developed and applied to a simple glacier geometry. In the next step, meltwater production resulting from enhanced ice motion during a glacier surge is calculated. The amount of melt energy available follows directly from the lowering of the centre of gravity of the glacier. To illustrate the concept, schematic calculations are presented for a number of glaciers with different geometric characteristics. Typical dissipative melt rates, expressed as water-layer depth averaged over the glacier, range from a few centimetres per year for smaller glaciers to half a metre per year for Franz Josef Glacier, one of the most active glaciers in the world (in terms of mass turnover). The total generation of meltwater during a surge is typically half a metre. For Variegated Glacier a value of 70 cm is found, for Kongsvegen 20 cm. These values refer to water layer depth averaged over the entire glacier. The melt extit{rate} depends on the duration of the surge. It is generally an order of magnitude greater than water production by `normal' dissipation. On the other hand, the additional basal melt rate during a surge is comparable in magnitude with the water input from meltwater and precipitation. This suggests that enhanced melting during a surge does not grossly change the total water budget of a glacier. Basal water generated by enhanced sliding is an important ingredient in many theories of glacier surges. It provides a positive feedback mechanism that actually makes the surge happen. The results found here suggest that this can only work if water generated by enhanced sliding accumulates in a part of the glacier base where surface meltwater and rain have no or very limited access. This finding seems compatible with the fact that, on many glaciers, surges are initiated in the lower accumulation zone.
Article
Full-text available
The recent evolution of Pamir-Karakoram-Himalaya (PKH) glaciers, widely acknowledged as valuable high-altitude as well as mid-latitude climatic indicators, remains poorly known. To estimate the region-wide glacier mass balance for 9 study sites spread from the Pamir to the Hengduan Shan (eastern Himalaya), we compared the 2000 Shuttle Radar Topography Mission (SRTM) digital elevation model (DEM) to recent (2008–2011) DEMs derived from SPOT5 stereo imagery. During the last decade, the region-wide glacier mass balances were contrasted with moderate mass losses in the eastern and central Himalaya (−0.22 ± 0.12 m w.e. yr−1 to −0.33 ± 0.14 m w.e. yr−1) and larger losses in the western Himalaya (−0.45 ± 0.13 m w.e. yr−1). Recently reported slight mass gain or balanced mass budget of glaciers in the central Karakoram is confirmed for a larger area (+0.10 ± 0.16 m w.e. yr−1) and also observed for glaciers in the western Pamir (+0.14 ± 0.13 m w.e. yr−1). Thus, the "Karakoram anomaly" should be renamed the "Pamir-Karakoram anomaly", at least for the last decade. The overall mass balance of PKH glaciers, −0.14 ± 0.08 m w.e. yr−1, is two to three times less negative than the global average for glaciers distinct from the Greenland and Antarctic ice sheets. Together with recent studies using ICESat and GRACE data, DEM differencing confirms a contrasted pattern of glacier mass change in the PKH during the first decade of the 21st century.
Article
Full-text available
The ablation amount for entire ice cliffs reaches about 2017( of that at the whole debris-covered area, the ablation rate at the ice cliff mainly depends on shortwave radiation, which differs widely with the orientation of all ice cliff. Therefore, the distribution of ice cliffs ill relation to their orientation was observed oil a debris-covered glacier. The south-facing, cliffs were small in area because they have low slope angles and tended to be covered with debris. The north-facing cliffs. oil the other hand, were large in the studied area and maintain a slope angle larger than the repose angle of debris. They, therefore, are stable. Longwave radiation from the debris surface opposite the ice cliffs was larger on the lower portion of ice cliffs than on the upper portion in every azimuth. This difference in longwave radiation maintained a steep slope angle on ice cliff. Shortwave radiation was stronger at the upper portion of ice cliffs than at tire lower portion due to tire local shading effect, causing gentle sloping of ice clift's. This was especially pronounced at cliffs facing to south. Therefore, the dependency of the ice cliff angle in orientation call be explained by tire difference in local radiation between the upper and lower portion of the ice cliff.
Article
Full-text available
There are many supraglacial ponds on debris-covered glaciers in the Nepal Himalayas. The heat absorbed at the surface of a pond was estimated from heat budget observations on the Lirung Glacier in Langtang Valley, Nepal. The results indicated an average heat absorption of 170 W m-2 during the summer monsoon season. This rate is about 7 times the average for the whole debris-covered zone. Analysis of the heat budget for a pond suggests that at least half of the heat absorbed at a pond surface is released with the water outflow from the pond, indicating that the water warmed in the pond enlarges the englacial conduit that drains water from the pond and produces internal ablation. Furthermore, the roof of the conduit could collapse, leading to the formation of ice cliffs and new ponds, which would accelerate the ablation of the debris-covered glacier.
Article
Full-text available
We evaluated elevation changes at four sites on debris-covered ablation area of Khumbu Glacier, Nepal Himalaya, since 1978. In 2004, we carried out a ground survey by differential GPS in the upper- and lowermost areas of the ablation area. The amount of surface lowering was calculated by comparing digital elevation models (DEMs) with 30-m grid size, as generated from survey data corrected in 1978, 1995, and in the present study. Because we could not access the middle parts of the debris-covered area due to surface roughness, for this area we used an ASTER-DEM calibrated by the ground survey data. The amount of surface lowering during the period 1978-2004 was insignificant near the terminus. A remarkable acceleration of surface lowering was found in the middle part of the debris-covered ablation area, where the glacier surface is highly undulating. In the uppermost area, surface lowering has continued at a steady rate. Surface flow speeds have decreased since 1956, revealing that the recent decrease in ice flux from the upper accumulation area would have accelerated the rate of surface lowering of the debris-covered area of Khumbu Glacier during the period 1995-2004.
Article
Repeat photographs and field survey reveal the mechanism of short-term ice-cliff evolution at Maud Glacier, a temperate lake-calving glacier in New Zealand. Calving is cyclic, each cycle involving four stages: (1) waterline melting and collapse of the roof of a sub-horizontal notch at the cliff foot; (2) calving of ice flakes from the cliff face leading to a growing overhang from the waterline upwards, and crack propagation from the glacier surface; (3) large but infrequent calving of slabs in response to the developing overhang, returning the cliff to an “initial” vertical profile; (4) rare subaqueous calving ofa submerged ice foot. Results indicate that the rate-controlling process is the speed of waterline melting, and that calving rate is independent of water depth (at least at time-scales of weeks to months). Slowly calving lake-terminating glaciers have mass balances more negative than land-terminating glaciers, but nevertheless advance and retreat in response to mass-balance driven changes in ice velocity.