Acceleration of ice loss across the Himalayas over the past 40 years

Abstract

Himalayan glaciers supply meltwater to densely populated catchments in South Asia, and regional observations of glacier change over multiple decades are needed to understand climate drivers and assess resulting impacts on glacier-fed rivers. Here, we quantify changes in ice thickness during the intervals 1975–2000 and 2000–2016 across the Himalayas, using a set of digital elevation models derived from cold war–era spy satellite film and modern stereo satellite imagery. We observe consistent ice loss along the entire 2000-km transect for both intervals and find a doubling of the average loss rate during 2000–2016 [−0.43 ± 0.14 m w.e. year ⁻¹ (meters of water equivalent per year)] compared to 1975–2000 (−0.22 ± 0.13 m w.e. year ⁻¹ ). The similar magnitude and acceleration of ice loss across the Himalayas suggests a regionally coherent climate forcing, consistent with atmospheric warming and associated energy fluxes as the dominant drivers of glacier change.

Full text

CLIMATOLOGY Copyright © 2019
The Authors, some
rights reserved;
exclusive licensee
American Association
for the Advancement
of Science. No claim to
original U.S. Government
Works. Distributed
under a Creative
Commons Attribution
NonCommercial
License 4.0 (CC BY-NC).
Acceleration of ice loss across the Himalayas over the
past 40 years
J. M. Maurer
1,2
*, J. M. Schaefer
1,2
, S. Rupper
3
, A. Corley
1
Himalayan glaciers supply meltwater to densely populated catchments in South Asia, and regional observations
of glacier change over multiple decades are needed to understand climate drivers and assess resulting impacts
on glacier-fed rivers. Here, we quantify changes in ice thickness during the intervals 1975–2000 and 2000–2016
across the Himalayas, using a set of digital elevation models derived from cold war–era spy satellite film and
modern stereo satellite imagery. We observe consistent ice loss along the entire 2000-km transect for both
intervals and find a doubling of the average loss rate during 2000–2016 [−0.43 ± 0.14 m w.e. year
−1
(meters
of water equivalent per year)] compared to 1975–2000 (−0.22 ± 0.13 m w.e. year
−1
). The similar magnitude and
acceleration of ice loss across the Himalayas suggests a regionally coherent climate forcing, consistent with
atmospheric warming and associated energy fluxes as the dominant drivers of glacier change.
INTRODUCTION
The Intergovernmental Panel on Climate Change 5th Assessment
Report estimates that mass loss from glaciers contributed more to
sea-level rise than the ice sheets during 1993–2010 (0.86 mm year
−1
versus
0.60 mm year
−1
, respectively), yet uncertainties for the glacier con-
tribution are three times greater (1). Glaciers also contribute locally
to water resources in many regions and serve as hydrological buffers
vital for ecology, agriculture, and hydropower, particularly in High
Mountain Asia (HMA), which includes all mountain ranges surround-
ing the Tibetan Plateau (2,3). Shrinking Himalayan glaciers pose
challenges to societies and policy-makers regarding issues such as
changing glacier melt contributions to seasonal runoff, especially in
climatically drier western regions (3), and increasing risk of outburst
floods due to expansion of unstable proglacial lakes (4). Yet, substantial
gaps in knowledge persist regarding rates of ice loss, hydrological re-
sponses, and associated climate drivers in HMA (2).
Mountain glaciers are known to respond dynamically to a variety
of drivers on different time scales, with faster response times than the
large ice sheets (5,6). In the Himalayas, in situ studies document sig-
nificant interannual variability of mass balances (7–9) and relatively
slower melt rates on debris-covered glacier tongues over interannual
time scales (10,11). Yet, the overall effects of surface debris cover are
uncertain, as many satellite observations suggest similar ice losses
relative to clean-ice glaciers over similar or longer periods (12,13).
Because of the complex monsoon climate in the Himalayas, dominant
causes of recent glacier changes remain controversial, although atmo-
spheric warming, the albedo effectdue to deposition of anthropogenic
black carbon (BC) on snow and ice, and precipitation changes have
been suggested as important drivers (14–16).
Model projections of future Himalayan ice loss and resulting impacts
require robust observations of glacier response to past and ongoing cli-
mate change. Recent satellite remote sensing studies have made substan-
tial advances with improved spatial coverage and resolution to quantify
ice mass changes during 2000–2016 (12,17,18), and former records
extending back to the 1970s have been presented for several areas using
declassified spy satellite imagery (13,19–22). These long-term records
are especially critical for extracting robust mass balance signals from
the noise of interannual variability (6). Many studies also report the
highly heterogeneous behavior of glaciers in localized regions, with
some glaciers exhibiting faster rates of ice loss during the 21st century
(20,22). Independent analyses document simultaneously increasing
atmospheric temperatures at high-elevation stations in HMA (23–26).
To robustly quantify the regional sensitivity of these glaciers to climate
change, a reliable Himalaya-wide record of ice loss extending back several
decades is needed.
Here, we provide an internally consistent dataset of glacier mass
change across the Himalayan range over approximately the past 40 years.
We use recent advances in digital elevation model (DEM) extraction
methods from declassified KH-9 Hexagon film (27)andASTERstereo
imagery to quantify ice loss trends for 650 of the largest glaciers during
1975–2000 and 2000–2016. All aspects of the analysis presented here only
use glaciers with data available during both time intervals unless specified
otherwise. We investigate glaciers along a 2000-km transect from Spiti
Lahaul to Bhutan (75°E to 93°E), which includes glaciers that accumu-
late snow primarily during winter (western Himalayas) and during the
summer monsoon (eastern Himalayas), but excludes complications of
surging glaciers in the Karakoram and Kunlun regions where many
glaciers appear to be anomalously stable or advancing (2). Our compi-
lation includes glaciers comprising approximately 34% of the total gla-
cierized area in the region, which represents roughly 55% of the total
ice volume based on recent ice thickness estimates (15,28). This diverse
dataset adequately captures the statistical distribution of large (>3 km
2
)
glaciers, thus providing the first spatially robust analysis of glacier change
spanning four decades in the Himalayas. We extract DEMs from declas-
sified KH-9 Hexagon images for the 650 glaciers, compile a corresponding
set of modern ASTER DEMs, fit a robust linear regression through
every 30-m pixel of the time series of elevations, sum the resulting
elevation changes for each glacier, divide by the corresponding areas,
and translate the volume changes to mass using a density conversion
factor of 850 ± 60 kg m
−3
(see Materials and Methods).
RESULTS
Glacier mass changes
Our results indicate that glaciers across the Himalayas experienced
significant ice loss over the past 40 years, with the average rate of ice
loss twice as rapid in the 21st century compared to the end of the 20th
1
Lamont-Doherty Earth Observatory, Columbia University, Palisades, NY, USA.
2
De-
partment of Earth and Environmental Sciences, Columbia University, New York, NY,
USA.
3
Department of Geography, University of Utah, Salt Lake City, UT, USA.
*Corresponding author. Email: jmaurer@ldeo.columbia.edu
SCIENCE ADVANCES |RESEARCH ARTICLE
Maurer et al., Sci. Adv. 2019; 5: eaav7266 19 June 2019 1of12
on June 20, 2019http://advances.sciencemag.org/Downloaded from

century (Fig. 1). We calculate a regional average geodetic mass balance
of −0.43 ± 0.14 m w.e. year
−1
(meters of water equivalent per year)
during 2000–2016, compared to −0.22 ± 0.13 m w.e. year
−1
during
1975–2000 (−0.31 ± 0.13 m w.e. year
−1
for the full 1975–2016 interval)
(see Materials and Methods). A 30-glacier moving average shows a
quasi-consistent trend across the 2000-km longitudinal transect
during both time intervals (Fig. 1), and subregions have similar means
and distributions of glacier mass balance. Some central catchments
Fig. 1. Map of glacier locations and geodetic mass balances for the 650 glaciers. Circle sizes are proportional to glacier areas, and colors delineate clean-ice,
debris-covered, and lake-terminating categories. Insets indicate ice loss, quantified as geodetic mass balances (m w.e. year
−1
) plotted for individual glaciers along a
longitudinal transect during 1975–2000 and 2000–2016. Both inset plots are horizontally aligned with the map view. Gray error bars are 1suncertainty, and the yellow
trend is the (area-weighted) moving-window mean, using a window size of 30 glaciers.
SCIENCE ADVANCES |RESEARCH ARTICLE
Maurer et al., Sci. Adv. 2019; 5: eaav7266 19 June 2019 2of12
on June 20, 2019http://advances.sciencemag.org/Downloaded from

deviate somewhat from the Himalaya-wide mean during 2000–2016
(byapproximately0.1to0.2mw.e.year
−1
) in the Uttarakhand (~79.0°
to 80.0°E), the Gandaki catchment (~83.5° to 84.5°E), and the Karnali
catchment (~81° to 83°E), which has fewer larger glaciers and relatively
incomplete data coverage. Similar to previous in situ and satellite-
based studies (18,29), we observe considerable variation among in-
dividual glacier mass balances, with area-weighted SDs of 0.1 and
0.2 m w.e. year
−1
during each respective interval for the 650 glaciers. This
variability most likely reflects different glacier characteristics such as sizes
of accumulation zones relative to ablation zones, topographic shading,
and amounts of debris cover. Yet, we find that, in our survey (using a
rough average of 60 glaciers per 7000-km
2
subregion), local variations
tend to average out and mean values are similar across most catchments.
Contrasting distributions of glacier mass balances are evident
when comparing between time intervals. The 1975–2000 distribution
has a negative tail extending to −0.6 m w.e. year
−1
, while the 2000–2016
distribution is more negative, extending to −1.1 m w.e. year
−1
(Fig. 2A).
Duringthemorerecentinterval,glaciersarelosingicetwiceasfaston
average (Fig. 2B), though this varies somewhat between subregions. For
example, we find that the average rate of ice loss has increased by a
factor of 3 in the Spiti Lahaul region, and by a factor of 1.4 in West
Nepal. We also compile altitudinal distributions of ice thickness change
for the glaciers and create a Himalaya-wide average thickness change
profile versus elevation (Fig. 2, C and D). These distributed thinning
profiles are a function of both thinning by mass loss and of dynamic
thinning due to ice flow. We find that the 2000–2016 thinning rate
(m year
−1
) profile is considerably steeper, which is likely caused by a
combination of faster mass loss and widespread slowing of ice velocities
during the 21st century (2,30).
We multiply geodetic mass balances by the full glacierized area in
the Himalayas between 75° and 93° longitude to estimate region-wide
ice mass changes of −7.5 ± 2.3 Gt year
−1
during 2000–2016, compared
to −3.9 ± 2.2 Gt year
−1
during 1975–2000 (−5.2 ± 2.2 Gt year
−1
during
the full 1975–2016 interval). Recent models using Shuttle Radar To-
pography Mission (SRTM) elevation data for ice thickness inversion
estimate the total glacial ice mass in our region of study to be approx-
imately 700 Gt in the year 2000 (see Materials and Methods) (15,28). If
this estimate is accurate, our observed annual mass losses suggest that
of the total ice mass present in 1975, about 87% remained in 2000 and
72% remained in 2016.
Comparison of clean-ice, debris-covered, and
lake-terminating glaciers
We study mass changes for different glacier types by separating gla-
ciers into clean-ice (<33% area covered by debris), debris-covered
(≥33% area covered by debris), and lake-terminating categories based
on a Landsat band ratio threshold and manual delineation of pro-
glacial lakes (see Materials and Methods). All three categories have
undergone a similar acceleration of ice loss (Table 1), and debris-
covered glaciers exhibit similar and often more negative geodetic mass
balances compared to clean-ice glaciers over the past 40 years (Fig. 3).
Altitudinal distributions indicate slower thinning for lower-elevation
regions of debris-covered glaciers (glacier tongues where debris is most
concentrated) relative to clean-ice glaciers, but comparatively faster
thinning in mid- to upper elevations (Fig. 4). Lake-terminating glaciers
concentrated in the eastern Himalayas exhibit the most negative mass
balances due to thermal undercutting and calving (31), though they
only comprise around 5 to 6% of the estimated total Himalaya-wide
mass loss during both intervals.
Fig. 2. Co mparison of ice losses betwee n 1975–2000 and 2000–2016 for
the 650 glaciers. (A) Histograms of individual glacier geodetic mass balances
(m w.e. year
−1
) during 1975–2000 (mean = −0.21, SD = 0.15) and 2000–2016
(mean = −0.41, SD = 0.24). Shaded regions behind the histograms are fitted normal
distributions. (B) Result of dividing the modern (2000–2016) mass balances by the
historical (1975–2000) mass balances for each glacier, showing the resulting distri-
bution of the mass balance change (ratio) between the two intervals (mean = 2.01,
SD = 1.36). In this case, the shaded region is a fitted kernel distribution. (C)Altitu-
dinal distributions of ice thickness change (m year
−1
) separated into 50-m elevation
bins during the two intervals. (D) Normalized altitudinal distributions of ice thickness
change. Normalized elevations are defined as (z−z
2.5
)/(z
97.5
−z
2.5
), where zis eleva-
tion and subscripts indicate elevation percentiles. This scales all glaciers by their ele-
vation range (i.e., after scaling, glacier termini = 0 and headwalls = 1), allowing for
more consistent comparison of ice thickness changes across glaciers with different
elevation ranges. Note the abrupt inflection point in the 2000–2016 profile at ~0.1;
this is likely due to retreating glacier termini. Shaded regions in the altitudinal dis-
tributions indicate the SEM estimated as sz=ffiffiffiffiffi
nz
p,wheres
z
is the SD of the thinning
rate for each 50-m elevation bin and n
z
is the number of independent measurements
when accounting for spatial autocorrelation (see Materials and Methods).
SCIENCE ADVANCES |RESEARCH ARTICLE
Maurer et al., Sci. Adv. 2019; 5: eaav7266 19 June 2019 3of12
on June 20, 2019http://advances.sciencemag.org/Downloaded from

Approximation of required temperature change
As a first approximation of the consistency between observed glacier
mass balances and available temperature records, we estimate the
energy required to melt the observed ice losses and conservatively
estimate the atmospheric temperature change that would supply this
energy via longwave radiation to the glaciers, using a simple energy
balance approach (Materials and Methods). We propagate significant
uncertainties associated with input from global climate reanalysis
data, scaling of temperatures from coarse reanalysis grids to specific
glacier elevations, and averaging of climate data over the glacierized
region. Results suggest that the observed acceleration of ice loss can
Table 1. Himalaya-wide geodetic mass balances (m w.e. year
−1
).
1975–2000 2000–2016 1975–2016
All glaciers −0.22 ± 0.13 −0.43 ± 0.14 −0.31 ± 0.13
Clean-ice −0.19 ± 0.07 −0.38 ± 0.08 −0.27 ± 0.07
Debris-covered −0.24 ± 0.06 −0.44 ± 0.08 −0.32 ± 0.06
Lake-terminating −0.33 ± 0.07 −0.56 ± 0.08 −0.40 ± 0.07
Fig. 3. Comparison between clean-ice (<33% debris-covered area) and debris-covered (≥33% debris-covered area) glaciers for seven subregions. Circle sizes
are proportional to glacier areas, colors delineate clean-ice versus debris-covered categories, and boxplots indicate geodetic mass balance (m w.e. year
−1
). Box center marks
(red lines) are medians; box bottom and top edges indicate the 25th and 75th percentiles, respectively; whiskers indicate q
75
+1.5⋅(q
75
−q
25
)andq
25
−1.5 ⋅(q
75
−q
25
),
where subscripts indicate percentiles and “+”symbols are outliers.
SCIENCE ADVANCES |RESEARCH ARTICLE
Maurer et al., Sci. Adv. 2019; 5: eaav7266 19 June 2019 4of12
on June 20, 2019http://advances.sciencemag.org/Downloaded from

be explained by an average temperature ranging from 0.4° to 1.4°C
warmer during2000–2016, relative to the 1975–2000 average. This ap-
proximately agrees with the magnitude of warming observed by me-
teorological stations located throughout HMA, which have recorded
air temperatures around 1°C warmer on average during 2000–2016,
relative to 1975–2000 (Fig. 5). More comprehensive climate observa-
tions and models will be essential for further investigation, but these
simple energy constraints suggest that the acceleration of mass loss in
the Himalayas is consistent with warming temperatures recorded by
meteorological stations in the region.
DISCUSSION
Implications for dominant drivers of glacier change in
the Himalayas
The parsing of Himalayan glacier energy budgets is not a straight-
forward task owing to the scarcity of meteorological data, in combi-
nation with the complex climate and topography of the region (2).
Furthermore, the Himalayas border hot spots of high anthropogenic
BC emissions, which may affect glaciers by direct heating of the at-
mosphere and decreasing albedo of ice and snow after deposition (14).
While improved analyses combining observations and high-resolution
atmospheric and glacier energy balance models will be required to
quantify these effects, the pattern of ice loss we observe has important
implications regarding dominant climate influences on Himalayan gla-
cier mass balances. Our results suggest that any drivers of glacier change
must explain the region-wide consistency, the doubling of the average
rate of ice loss in the 21st century compared to 1975–2000, and the ob-
servation that clean-ice, debris-covered, and lake-terminating glaciers
have all experienced a similar acceleration of mass loss.
Some studies have suggested a weakening of the summer monsoon
and reduced precipitation as primary reasons for negative glacier mass
balances, particularly in the Everest region (16). While decreasing ac-
cumulation rates may account for a significant portion of the mass
balance signal for some glaciers, an extreme Himalaya-wide decrease
in precipitation would be required to explain the extensive ice losses
we observe, especially given that monsoon-dominated glaciers with
high accumulation rates are knownto be much more sensitive to tem-
perature than accumulation changes (5,32). Regional studies of pre-
cipitation trends in the Himalayas do not suggest a widespread decrease
in precipitation over the past four decades (Supplementary Materials).
A uniform BC albedoforcing across the Himalayas is another possible
explanation, although BC concentrations measured in snow and ice in
the Himalayas have been found to be spatially heterogeneous (14,33),
and high-resolution atmospheric models also show large spatial vari-
ability of deposited BC originating from localized emissions in regions
of complex terrain (14,34). Future analyses focused on quantifying
the spatial patterns of BC deposition will reveal further insights, yet
given the rather homogeneous pattern of mass loss we observe across
the 2000-km Himalayan transect, a strong, spatially heterogeneous
mechanism seems improbable as a dominant driver of glacier ice loss
in the region.
Debris-covered glaciers
Similar thinning rates of debris-covered (thermally insulated) gla-
ciers relative to clean-ice glaciers have been observed by previous studies
Fig. 4. Altitudinal distributions of ice thickness change (m year
−1
) for the 650 glaciers. Glaciers are separated by time interval (top) and category (<33% versus
≥33% debris-covered area) (bottom). (A) Altitudinal distributions of ice thickness change for clean-ice glaciers during 1975–2000 and 2000–2016. The yaxes are
normalized elevation as in Fig. 2. (B) Same as (A), but for debris-covered glaciers. (C) Altitudinal distributions of ice thickness change during 1975–2000 for clean-ice and
debris-covered glaciers. (D) Same as (C), but for 2000–2016. (E) Altitudinal distributions of glacierized area for both glacier categories. Elevational extent of debris cover
varies widely between individual glaciers, but is mostly concentrated in lower ablation zones. The clean-ice category includes 478 glaciers and the debris-covered
category includes 124 glaciers.
SCIENCE ADVANCES |RESEARCH ARTICLE
Maurer et al., Sci. Adv. 2019; 5: eaav7266 19 June 2019 5of12
on June 20, 2019http://advances.sciencemag.org/Downloaded from

and have been not only ascribed to surface melt ponds and associated
ice cliffs acting as localized hot spots to concentrate melting but also
attributed to declining ice flux causing dynamic thinning and stagnation
of debris-covered glacier tongues (2).Whilewecannotyetdirectlyde-
convolve relative contributions from these processes, we find that aver-
age thinning rates for debris-covered glaciers are slower than clean-ice
glaciers at low elevations (glacier tongues where debris is most concen-
trated), which agrees with reduced melt rates from field studies. In turn,
debris-covered glaciers tend to have comparatively faster thinning at
mid-range elevations, where debris cover is sparser and also where
the majority of total glacierized area resides (Fig. 4). Models of debris-
covered glacier processes suggest that this pattern of thinning may
cause a reduction in down-glacier surface gradient, which, in turn, re-
duces driving stress and ice flux and explains why debris-covered ab-
lation zones become stagnant (35). We also find that clean-ice glaciers
exhibit a much more pronounced steepening of the thinning profile
Fig. 5. Compilation of previously published instrumental temperature records in HMA. (A) Regional temperature anomalies, relative to the 1980–2009 mean
temperatures for each record. The yellow trend (23) from the quality-controlled and homogenized climate datasets LSAT-V1.1 and CGP1.0 recently developed by the
China Meteorological Administration (CMA), using approximately 94 meteorological stations located throughout the Hindu Kush Himalayan region. The orange trend
(44) is from a similar CMA dataset derived from 81 stations more concentrated on the eastern Tibetan Plateau. The blue trend (24) is from three decades of temperature
data from 13 mountain stations located on the southern slopes of the central Himalayas. The black trend is the 5-year moving mean. (B) Temperature anomalies from
high-elevation stations at the Chhota Shigri glacier terminus (25); Dingri station in the Everest region (26); average from the Kanzalwan, Drass, and Patseo stations (45);
and average of 16 stations above 4000 m elevation on the Tibetan Plateau and eastern Himalayas (46). Here, temperature anomalies are relative to the mean of each
record. The gray trend line is the 5-year moving mean. (C) Difference in mean temperature (°C) between the two intervals, i.e., the mean of the 2000–2016 interval
relative to the mean of the 1975–2000 interval.
SCIENCE ADVANCES |RESEARCH ARTICLE
Maurer et al., Sci. Adv. 2019; 5: eaav7266 19 June 2019 6of12
on June 20, 2019http://advances.sciencemag.org/Downloaded from

over time, compared to debris-covered glaciers. It may be that both
glacier types experience a uniform thinning phase followed by a changing
terminus flux and retreat phase, but the clean-ice glaciers are in a later
phase of response to recent climate change (36).
Comparison with previous studies in the Himalayas
Tocompareourresultswithpreviousremotesensingstudiesthatderive
mass changes from various sensors (including Hexagon, SRTM, SPOT5,
ICESat, and ASTER), we restrict our results to the approximate geo-
graphical regions covered by each corresponding study (12,13,17–22)
and then compute area-weighted average geodetic mass balances. In
addition, we compare individual glacier mass balances for the Everest
and Langtang Himal regions, where mass changes were previously es-
timated using declassified Corona and Hexagon imagery (13,19,20).
Despite factors such as variable spatial resolutions, distinct void-filling
methods, heterogeneous spatial and temporal coverages, and different
definitions of glacier boundaries, we find that our average mass balances
generally agree with previous analyses and overlap within uncertainties
(table S1). However, because of the significant variability of individual
glacier mass changes within subregions, our results also highlight the
importance of sampling a large number of glaciers to obtain a robust
average trend for any given area.
Comparison with benchmark mid-latitude glaciers and
global average
To gain perspective on mass losses from these low-latitude glaciers
in the monsoonal Himalayas, we compare our results with benchmark
mid-latitude glaciers in the European Alps, as well as with a global av-
erage mass balance trend (fig. S1) (37). The Alps contain the most
detailed long-term glaciological and high-elevation meteorological
records on Earth, and the climatic sensitivity and behavior of these
European glaciers are well understood compared to glaciers in HMA.
Air temperatures in the Alps show an abrupt warming trend beginning
in the mid-1980s, and Alpine mass balance records display a concurrent
acceleration of ice loss, with a continual strongly negative mass balance
after that time. Himalayan weather station data indicate a more gradual
warming trend, with the strongest warming beginning in the mid-1990s
(fig. S1, A and B). We find that mass balance in the Himalayas is less
negative compared to the Alps and the global average, despite close
proximity to a known hot spot of increasing BC emissions with rapid
growth and accompanying combustion of fossil fuels and biomass in
South Asia (38). The concurrent acceleration of ice loss observed in both
the Himalayas and Europe over the past 40 years coincides with a dis-
tinct warming trend beginning in the latter part of the 20th century,
followed by the consistently warmest temperatures through the 21st
century in both regions.
Conclusion
Our analysis robustly quantifies four decades of ice loss for 650 of the
largest glaciers across a 2000-km transect in the Himalayas. We find
similar mass loss rates across subregions and a doubling of the average
rate of loss during 2000–2016 relative to the 1975–2000 interval. This
is consistent with the available multidecade weather station records
scattered throughout HMA, which indicate quasi-steady mean annual
air temperatures through the 1960s to the 1980s with a prominent
warming trend beginning in the mid-1990s and continuing into the
21st century (23–26). We suggest that degree-day and energy balance
models focused on accurately quantifying glacier responses to air tem-
perature changes (including energy fluxes and associated feedbacks)
will provide the most robust estimates of glacier response to future
climate scenarios in the Himalayas.
MATERIALS AND METHODS
Hexagon
U.S. intelligence agencies used KH-9 Hexagon military satellites for re-
connaissance from 1973 to 1980. A telescopic camera system acquired
thousands of photographs worldwide, after which film recovery cap-
sules were ejected from the satellites and parachuted back to Earth over
the Pacific Ocean. With a ground resolution ranging from 6 to 9 m,
single scenes from the mapping camera cover an area of approximately
30,000 km
2
with overlap of 55 to 70%, allowing for stereo photogram-
metric processing of large regions. Images were scanned by the U.S.
GeologicalSurvey (USGS) at a resolution of 7 mm and downloaded via
the Earth Explorer user interface (earthexplorer.usgs.gov). Digital
elevation models were extracted using the Hexagon Imagery Automated
Pipeline methodology, which is coded in MATLAB and uses the
OpenCV library for Oriented FAST and Rotated BRIEF (ORB) feature
matching, uncalibrated stereo rectification, and semiglobal block
matching algorithms (27). The majority of the KH-9 images here were
acquired within a 3-year interval (1973–1976), and we processed a total
of 42 images to provide sufficient spatial coverage (fig. S2).
ASTER
The ASTER instrument was launched as part of a cooperative effort
between NASA and Japan’s Ministry of Economy, Trade and Industry
in 1999. Its nadir and backward-viewing telescopes provide stereo-
scopic capability at 15-m ground resolution, and a single DEM covers
approximately 3600 km
2
. Approximately 26,000 ASTER DEMs were
downloaded via the METI AIST Data Archive System (MADAS) satellite
data retrieval system (gbank.gsj.jp/madas), a portal maintained by the
Japanese National Institute of Advanced Industrial Science and Technol-
ogy and the Geological Survey of Japan. To use all cloud-free pixels (in-
cluding images with a high percentage of cloud cover), no cloud selection
criteria were applied when downloading the images. We used the Data1.
l3a.demzs geotiff product, which has a spatial resolution of 30 m. After
downloading, the DEMs were subjected to a cleanup process: For a
given scene, any saturated pixels (i.e., equal to 0 or 255) in the nadir
band 3 (0.76 to 0.86 mm) image were masked in the DEM. Next, the
SRTM dataset was used to remove any DEM values with an absolute
elevation difference larger than 150 m (relative to SRTM), which
effectively eliminated the majority of errors caused by clouds. While
more sophisticated cloud masking procedures are possible, the ASTER
shortwave infrared detectors failed in April 2008, making cloud detec-
tion after this time impossible using standard methods. We examined
existing cloud masks derived using Moderate Resolution Imaging
Spectroradiometer images as another option (tonolab.cis.ibaraki.ac.
jp/ASTER/cloud/). However, these are not optimized for snow-covered
regions and often misclassify glacierpixelsasclouds.Instead,ourlarge
collection of multitemporal ASTER scenes, the SRTM difference
threshold, and our robust linear trend fitting algorithm [see description
of Random Sample Consensus (RANSAC) in the “Trend fitting of
multitemporal DEM stacks”section] effectively excluded any remain-
ing erroneous cloud elevations after the initial threshold. As a final
step, all ASTER DEMs were coregistered to the SRTM using a stan-
dard elevation–aspect optimization procedure (39). We did not apply
fifth-order polynomial correction procedures to the ASTER DEMs for
satellite “jitter”effects and curvature bias as done in some previous
SCIENCE ADVANCES |RESEARCH ARTICLE
Maurer et al., Sci. Adv. 2019; 5: eaav7266 19 June 2019 7of12
on June 20, 2019http://advances.sciencemag.org/Downloaded from

studies (18). We found that while these types of corrections may reduce
theoverallaverageelevationerrorinascene,thepolynomialfittingcan
be unwieldy and may introduce unwanted localized biases. By stacking
many ASTER DEMs (with 20.5 being the average number of observa-
tions per pixel stack during the ASTER trend fitting, see fig. S3E) and
using a robust fitting procedure, we found that biases do not correlate
across overlapping scenes, and thus tend to cancel out one another.
Furthermore, the elevation change results from this portion of our
study overlap within uncertainties with Brun et al.(18) (Supplementary
Materials) who did perform polynomial corrections. This suggests that
for a large-scale regional study using a high number of overlapping
ASTER scenes, the satellite jitter and curvature bias corrections have a
relatively minimal impact on the final results.
Glacier polygons
To delineate glaciers during all portions of the analysis, we used man-
ually refined versions of the Randolph Glacier Inventory (RGI 5.0) (40).
Starting with the original RGI dataset, we edited the glacier polygons
to reflect glacier areas during 1975, 2000, and 2016. For the 1975 edit,
we used a combination of Hexagon imagery, the Global Land Survey
(GLS) Landsat Multispectral Scanner mosaic (GLS1975), and glacier
thickness change maps derived from differencing the Hexagon and
modern ASTER DEMs, which are particularly useful for debris-covered
glacier termini that often have spectral characteristics indistinguishable
from surrounding terrain. Debris-covered areas for each glacier were
delineated using a Landsat DN TM4/TM5 band ratio with a threshold
of 2.0, and glaciers with ≥33% debris cover were assigned to the debris-
covered category. For the 2000 edit, we used the GLS2000 Landsat
Enhanced Thematic Mapper Plus mosaic, along with glacier thickness
change maps derived from differencing ASTER DEMs. For the 2016
edit, we used a custom mosaic of Landsat 8 imagery with acquisition
dates spanning 2014–2016. The individually edited glacier polygons
were used for all ice volume change and geodetic mass balance compu-
tations. The polygons were also used during alignment of the DEMs,
where the shapefiles were converted to raster masks with a dilation (mor-
phological operation) of 250 m on the binary rasters. This effectively
excluded unstable terrain surrounding the glaciers during the DEM
alignment process, such as steep avalanching slopes and unstable
moraines.
Trend fitting of multitemporal DEM stacks
Glacier polygons were processed individually—all DEMs from a given
time interval (1975–2000 or 2000–2016) that overlap a polygon were
selected and resampled to the same 30-m resolution using linear in-
terpolation. The overlapping DEMs were sampled with a 25% extension
around each glacier to include nearby stable terrain for alignment and
uncertainty analysis (fig. S4). After ensuring that there is no vertical bias,
the aligned DEMs were sorted in temporal order as a three-dimensional
matrix, and linear trends were fit to every pixel “stack”(i.e., along the
third dimension of the matrix) using the RANSAC method. During
each RANSAC iteration, a random set of two elevation pixels per stack
were selected. A linear trend was fit to these two values, and then all
remaining elevation pixels were compared to the trend. Any elevation
pixels within 15 m of the trend line were marked as inliers. This process
was repeated for 100 iterations, and the iteration with the greatest num-
ber of inliers was selected. A final linear fit was performed using all in-
liers from the best iteration, and this trend was used for each pixel
stack’s thickness change estimate. The thickness change maps were
subjected to outlier removal using thresholds for maximum slope, max-
imum thickness change, minimum count per pixel stack, minimum
timespan per pixel stack, maximum SD of inlier elevations per pixel
stack, and maximum gradient of the thickness change map (fig. S3).
In addition, the thickness change pixels were separated into 50-m ele-
vation bins, and pixels falling outside the 2 to 98% quantile range were
excluded. Any bins with less than 100 pixels were removed and then
interpolated using the two adjacent bins. Before computing ice volume
change for the glaciers, the final thickness change maps were visually
inspected, any remaining erroneous pixels (which occurred almost ex-
clusively in low-contrast, snow-covered accumulation zones) were man-
ually masked, and a 5 × 5 pixel median filter was applied. We did not
attempt to perform seasonality corrections, as no seasonal snowfall
records are available and because nearly all the Hexagon DEMs were
acquired during winter, thus minimizing any seasonality offsets be-
tween regions. For the 1975–2000 interval, we used the Hexagon DEMs
and sampled ASTER thickness change trends at the start of the year
2000. For regions with multiple overlapping Hexagon DEMs, we used
the same RANSAC method. During the 1975–2000 interval, only two
DEMs were available for most glaciers. In these cases, the RANSAC
iterations were unnecessary, and we simply differenced the two available
DEMs.WedidnotuseSRTMforanythicknesschangeestimates;thus,
no correction for radar penetration was necessary.
Mass changes
To compute (mean annual) ice volume changes for individual glaciers,
all thickness change pixels falling within a glacier polygon were
transformed to an appropriate projected WGS84 UTM coordinate
system (zones 43 to 46, depending on longitude of the glacier). Pixel
values (m year
−1
) were then multiplied by their corresponding areas
(pixel width × pixel height) and summed together. The resulting ice
volume changewas then divided by the average glacier area to obtain a
glacier thickness change. We used the average of the initial and final
glacier areas for a given time interval and excluded slopes greater than
45° to remove any cliffs and nunataks. Last, the glacier thickness change
was multiplied by an average ice-firn density (41)of850kgm
−3
and
then divided by the density of water (1000 kg m
−3
) to compute glacier
geodetic mass balance in m w.e. year
−1
. Because of cloud cover, shadows,
and low radiometric contrast, some glacier accumulation zones had
gaps in the DEMs and resulting thickness change maps. This is partic-
ularly evident in the Hexagon DEMs for the Spiti Lahaul region owing
to extensive cloud cover. To fill these gaps, we tested two different void-
filling methods for comparison. In the first method, we defined a cir-
cular search area with a radius of 50 km around the center of a given
glacier. All thickness change pixels from glaciers in this surrounding
area were binned (into 50-m elevation bins, and following the same
outlier-removal procedure given in the preceding section), and any
missing data in the glacier were set to this “regional bin”mean val-
ue at the corresponding elevation. In the second method, we filled
data gaps using an interpolation procedure, where voids in an individ-
ual glacier were linearly interpolated using bin values at upper and lower
elevations relative to the missing data (those belonging to the same gla-
cier), and assumed zero change at the highest elevation bin (headwall).
Both methods yielded similar results (table S1). In addition, no obvious
trends were apparent when geodetic mass balance was plotted versus
percent data coverage or glacier size (fig. S5). While smaller glaciers ex-
hibited more scatter, the average mass balance was similar for all glacier
sizes. These observations indicate that our representative sample of gla-
ciers, while biased toward larger glaciers, adequately captures the statis-
tical distribution of glacier mass balances in the Himalayas.
SCIENCE ADVANCES |RESEARCH ARTICLE
Maurer et al., Sci. Adv. 2019; 5: eaav7266 19 June 2019 8of12
on June 20, 2019http://advances.sciencemag.org/Downloaded from

To calculate regional geodetic mass balances, we separated glaciers
into four subregions (Spiti Lahaul, West Nepal, East Nepal, and Bhutan)
as defined by Brun et al.(18). We then calculated the average mass bal-
ance for each of these four subregions, weighted by individual glacier
areas. Last, we calculated a final average mass balance for the Himalayas,
weighted by the total glacierized area (from the RGI 5.0 database) in
each of the four subregions, between 75° to 93° longitude. Because of
the relatively homogeneous mass balance distribution, we found that
this approach resulted in similar values (well within the uncertainties)
compared to simply calculating the area-weighted average mass bal-
ance of the 650 measured glaciers. To obtain the total mass changes in
Gt year
−1
, we multiplied each subregion mass balance by its total gla-
cierized area and then summed the results from all subregions to get
Himalaya-wide totals of −3.9 Gt year
−1
for 1975–2000 and −7.5 Gt
year
−1
for 2000–2016. To calculate contributions to sea-level rise, we
used a global ocean surface area of 361.9 × 10
6
km
2
(fig. S4G).
To estimate the total ice mass present in our region of study, we
used ice thickness estimates from Kraaijenbrink et al.(15), who used
the Glacier bed Topography version 2 model to invert for ice thickness
(28) with input from the SRTM DEM (acquired in February of 2000).
The ice thickness estimates from (15) did not include glaciers smaller
than 0.4 km
2
, and to estimate the additional mass contribution from
these smallest glaciers (along with any other glaciers that are missing
thickness estimates), we fit a second-order polynomial to the loga-
rithm of glacier volumes versus the logarithm of glacier areas and eval-
uated this fit equation for any glaciers without volume data (fig. S6).
We then converted glacier volume to mass using a density value of
850 kg m
−3
. Over our region of study, the ice volumes from the thick-
ness data amounted to 649 Gt, with an additional contribution of 35 Gt
from the fitting procedure, for a total of 684 Gt.
Uncertainty assessment
We quantified statistical uncertainty for individual glaciers using an
iterative random sampling approach. For a given glacier, the SD of
elevation changes from the surrounding stable terrain (s
z
) was first
calculated. For any missing thickness change pixels within the glacier
polygon, we also included an extrapolation uncertainty s
e
. This accounts
for additional error in regions with incomplete data, i.e., those glacier
regions filled by extrapolating thickness changes from surrounding
glaciers, or linear interpolation assuming zero change at the headwall,
as described in the previous section. We found that in the Himalaya-
wide altitudinal distributions, the maximum SD of thickness change in
any 50-m elevation bin above 5000 m is 0.56 m year
−1
. Nearly all re-
gions with incomplete data coverage are abovethis elevation, resulting
from poor radiometric contrast for snow-covered glacier accumula-
tion zones. We thus conservatively set s
e
equal to 0.6 m year
−1
.We
then combined both sources of error to get s
p
for every individual
thickness change pixel
sp¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
s2
zþs2
e
qð1Þ
To account for spatial autocorrelation, we started with a normally
distributed random error field (with a mean of 0 and an SD of 1) the
same size as the thickness change map and then filtered it using an
n-by-nmoving window average to add spatial correlation, where
nis defined as the spatial correlation range divided by the spatial
resolution of the thickness change map. We used 500 m for the spatial
correlation range, which is a conservative value based on semivariogram
analysis in mountainous regions from previous studies (18,21,42). The
resulting artificial error field E
n
(now with spatial correlation) is scaled
by the s
p
values and added to the thickness change map DHas follows,
where (x,y) are pixel coordinates
DHEðx;yÞ¼DHðx;yÞþEnðx;yÞ⋅spðx;yÞ
snð2Þ
If thickness change data exist at a given pixel location (x,y)onthe
glacier, s
n
istheSDofthesetofallE
n
values where data exist (i.e., where
s
e
is equal to zero). Conversely, if thickness change data do not exist at a
given pixel location (x,y)ontheglacier,s
n
istheSDofthesetofallE
n
values where data do not exist (i.e., where s
e
is equal to 0.6 m year
−1
). In
this way, the second term of Eq. 2 assigns larger uncertainties to regions
with incomplete data. Last, all glacier thickness change pixels in DH
E
were summed together to compute a volume change with the intro-
duced error. This procedure was repeated for 100 iterations, and the
volume change uncertainty s
DV
was computed as the SD of the resulting
distribution (fig. S4). For region-widevolumechangeestimates,wecon-
servatively assumed total correlation between glaciers and computed
region-wide uncertainty as follows, where gis the total number of gla-
ciers (~17,400)
sDVregion ¼∑
g
1
sDVð3Þ
For glaciers where thickness change data are not available, a measure
of uncertainty is still required to factor into the final regional uncertainty
estimate. For these glaciers, we estimated s
DV
as (42)
sDV¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
s2
zregion⋅Acor
5⋅A
rð4Þ
Acor ¼p⋅L2ð5Þ
In this case, s
zregion
is the region-wide SD of elevation change over
stable terrain (0.42 m year
−1
)(fig.S7),A
cor
is the correlation area, Lis the
correlation range (500 m), and Ais the glacier area. Last, all s
DV
and
s
DV regjon
estimates were combined with an area uncertainty (43)of
10% and a density uncertainty (41)of60kgm
−3
using standard uncor-
related error propagation.
Sensitivity of region-wide glacier mass change estimates
We further tested the sensitivity of our region-wide estimates to poten-
tial biases, including (i) the exclusion of small glaciers, (ii) incomplete
data coverage for many glacier accumulation zones during 1975–2000,
and (iii) void-filling technique. First, we note that our geodetic mass
balance analysis only includes glaciers larger than 3 km
2
.Thisisbecause
mass balance uncertainties increase with decreasing glacier size, and we
find that uncertainties often exceed the magnitude of mass changes for
glaciers smaller than ~3 km
2
. To test whether the neglected small gla-
ciers appreciably affect the result, we also computed mass balances
using all available glaciers (i.e., all glaciers with ≥33% data coverage,
including those smaller than 3 km
2
). We find that including the full set
of smaller glaciers changes the region-wide geodetic mass balance
estimates by a maximum of 0.04 m w.e. year
−1
(fig. S4G). Next, we note
SCIENCE ADVANCES |RESEARCH ARTICLE
Maurer et al., Sci. Adv. 2019; 5: eaav7266 19 June 2019 9of12
on June 20, 2019http://advances.sciencemag.org/Downloaded from

that the Hexagon DEMs in particular have poor data coverage over gla-
cier accumulation zones (figs. S8 and S9). However, the vast majority of
thinning occurs in glacier ablation zones, and the amount of thinning
decreases with elevation in a quasi-linear fashion, especially in mid- to
upper regions of the glaciers where data gaps are most common. Thus,
we hypothesize that we can extrapolateandinterpolatewithreasonable
confidence over accumulation areas. To test the robustness of this as-
sumption, we used the 2000–2016 glacier change data. The ASTER
data over this interval have superior radiometric contrast and ade-
quately capture elevation changetrends for most accumulation zones.
We first set all 2000–2016 thickness change pixels to be empty where
the 1975–2000 data are missing to simulate the same data gaps over
accumulation zones as in the 1975–2000 data. We then performed the
same geodetic mass balance calculations and found that the region-wide
geodetic mass balance only changes by 0.01 m w.e. year
−1
(fig. S4G,
comparing test 3 to test 1). Last, we performed two separate void-filling
methods for all tests (see the “Mass changes”section for descriptions of
void-filling methods) and observed a maximum change in geodetic
mass balance of 0.04 m w.e. year
−1
. Overall, the relatively small impact
of each test suggests that our results are robust to the exclusion of small
glaciers, incomplete data coverage over glacier accumulation zones, and
void-filling technique.
SUPPLEMENTARY MATERIALS
Supplementary material for this article is available at http://advances.sciencemag.org/cgi/
content/full/5/6/eaav7266/DC1
Fig. S1. Comparison of Himalayan temperature trends and regional mass balance with
benchmark mid-latitude glaciers and a global average trend.
Fig. S2. Coverage of glacierized area in the Himalayas.
Fig. S3. Trend fit examples for two large glaciers using ASTER DEMs during 2000–2016,
histograms of ASTER pixel counts and timespans per stack (glacier averages), and outlier
thresholds.
Fig. S4. Illustration of uncertainty estimation procedure for a single iteration/glacier and
Himalaya-wide sensitivity tests.
Fig. S5. Geodetic mass balances during 1975–2000 and 2000–2016 plotted against various
parameters.
Fig. S6. Log-log plot of glacier volumes versus areas used to estimate the total ice mass
present in our region of study.
Fig. S7. Analysis of elevation change for nonglacier pixels (stable terrain) during both intervals.
Fig. S8. Thickness change maps used in the analysis.
Fig. S9. Thickness change maps for the three remaining Himalayan regions.
Table S1. Geodetic mass balance comparisons with prior studies.
References (47–70)
REFERENCES AND NOTES
1. J. A. Church, P. U. Clark, A. Cazenave, J. M. Gregory, S. Jevrejeva, A. Levermann,
M. A. Merrifield, G. A. Milne, R. S. Nerem, P. D. Nunn, A. J. Payne, W. T. Pfeffer, D. Stammer,
A. S. Unnikrishnan, Climate change 2013: The physical science basis. Contribution of
Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate
Change, in Sea Level Change (2013), pp. 1137.
2. M.F.Azam,P.Wagnon,E.Berthier,C.Vincent,K.Fujita,J.S.Kargel,Reviewofthe
status and mass changes of Himalayan-Karakoram glaciers. J. Glaciol. 64,61–74
(2018).
3. A. F. Lutz, W. W. Immerzeel, A. B. Shrestha, M. F. P. Bierkens, Consistent increase in High
Asia's runoff due to increasing glacier melt and precipitation. Nat. Clim. Chang. 4,
587–592 (2014).
4. S. Harrison, J. S. Kargel, C. Huggel, J. Reynolds, D. H. Shugar, R. A. Betts, A. Emmer,
N. Glasser, U. K. Haritashya, J. Klimeš, L. Reinhardt, Y. Schaub, A. Wiltshire, D. Regmi,
V. Vilímek, Climate change and the global pattern of moraine-dammed glacial lake
outburst floods. Cryosphere 12, 1195–1209 (2018).
5. J. Oerlemans, J. P. F. Fortuin, Sensitivity of glaciers and small ice caps to greenhouse
warming. Science 258, 115–117 (1992).
6. G. H. Roe, M. B. Baker, F. Herla, Centennial glacier retreat as categorical evidence of
regional climate change. Nat. Geosci. 10,95–99 (2017).
7. P. Wagnon, C. Vincent, Y. Arnaud, E. Berthier, E. Vuillermoz, S. Gruber, M. Ménégoz,
A. Gilbert, M. Dumont, J. M. Shea, D. Stumm, B. K. Pokhrel, Seasonal and annual mass
balances of Mera and Pokalde glaciers (Nepal Himalaya) since 2007. The Cryosphere 7,
1769–1786 (2013).
8. C. Vincent, A. Ramanathan, P. Wagnon, D. P. Dobhal, A. Linda, E. Berthier, P. Sharma,
Y. Arnaud, M. F. Azam, P. G. Jose, J. Gardelle, C. Vincent, Balanced conditions or slight
mass gain of glaciers in the Lahaul and Spiti region (northern India, Himalaya) during the
nineties preceded recent mass loss. Cryosphere 7, 569–582 (2013).
9. M. Zemp, H. Frey, I. Gärtner-Roer, S. U. Nussbaumer, M. Hoelzle, F. Paul, W. Haeberli,
F. Denzinger, A. P. Ahlstrøm, B. Anderson, S. Bajracharya, C. Baroni, L. N. Braun,
B. E. Cáceres, G. Casassa, G. Cobos, L. R. Dávila, H. Delgado Granados, M. N. Demuth,
L. Espizua, A. Fischer, K. Fujita, B. Gadek, A. Ghazanfar, J. Ove Hagen, P. Holmlund,
N.Karimi,Z.Li,M.Pelto,P.Pitte,V.V.Popovnin,C.A.Portocarrero,R.Prinz,
C. V. Sangewar, I. Severskiy, O. Sigurđsson, A. Soruco, R. Usubaliev, C. Vincent,
Historically unprecedented global glacier decline in the early 21st century. J.Glaciol.
61,745–762 (2015).
10. B. Pratap, D. P. Dobhal, M. Mehta, R. Bhambri, Influence of debris cover and altitude on
glacier surface melting: A case study on Dokriani Glacier, central Himalaya, India.
Ann. Glaciol. 56,9–16 (2015).
11. C. Vincent, P. Wagnon, J. M. Shea, W. W. Immerzeel, P. Kraaijenbrink, D. Shrestha,
A. Soruco, Y. Arnaud, F. Brun, E. Berthier, S. F. Sherpa, Reduced melt on debris-covered
glaciers: Investigations from Changri Nup Glacier, Nepal. Cryosphere 10, 1845–1858
(2016).
12. J. Gardelle, E. Berthier, Y. Arnaud, A. Kääb, Corrigendum to “Region-wide glacier mass
balances over the Pamir-Karakoram-Himalaya during 1999–2011”.Cryosphere 7,
1885–1886 (2013).
13. F. Pellicciotti, C. Stephan, E. Miles, S. Herreid, W. W. Immerzeel, T. Bolch, Mass-balance
changes of the debris-covered glaciers in the Langtang Himal, Nepal, from 1974 to 1999.
J.Glaciol. 61, 373–386 (2015).
14. C. G. Gertler, S. P. Puppala, A. Panday, D. Stumm, J. Shea, Black carbon and the Himalayan
cryosphere: A review. Atmos. Environ. 125, 404–417 (2016).
15. P. D. A. Kraaijenbrink, M. F. P. Bierkens, A. F. Lutz, W. W. Immerzeel, Impact of a global
temperature rise of 1.5 degrees Celsius on Asia’s glaciers. Nature 549, 257–260 (2017).
16. F. Salerno, N. Guyennon, S. Thakuri, G. Viviano, E. Romano, E. Vuillermoz, P. Cristofanelli,
P. Stocchi, G. Agrillo, Y. Ma, G. Tartari, Weak precipitation, warm winters and springs
impact glaciers of south slopes of Mt. Everest (central Himalaya) in the last 2 decades
(1994–2013). Cryosphere 9, 1229–1247 (2015).
17. A. Kääb, D. Treichler, C. Nuth, E. Berthier, Brief communication: Contending estimates of
2003–2008 glacier mass balance over the Pamir–Karakoram–Himalaya. Cryosphere 9,
557–564 (2015).
18. F. Brun, E. Berthier, P. Wagnon, A. Kääb, D. Treichler, A spatially resolved estimate of
High Mountain Asia glacier mass balances from 2000 to 2016. Nat. Geosci. 10, 668–673
(2017).
19. T. Bolch, T. Pieczonka, D. I. Benn, Multi-decadal mass loss of glaciers in the Everest area
(Nepal Himalaya) derived from stereo imagery. Cryosphere 5, 349–358 (2011).
20. S. Ragettli, T. Bolch, F. Pellicciotti, Heterogeneous glacier thinning patterns over the last
40 years in Langtang Himal, Nepal. Cryosphere 10, 2075–2097 (2016).
21. J. M. Maurer, S. B. Rupper, J. M. Schaefer, Quantifying ice loss in the eastern Himalayas
since 1974 using declassified spy satellite imagery. Cryosphere 10, 2203–2215 (2016).
22. Y. Zhou, Z. Li, J. Li, R. Zhao, X. Ding, Glacier mass balance in the Qinghai–Tibet Plateau
and its surroundings from the mid-1970s to 2000 based on Hexagon KH-9 and SRTM
DEMs. Remote Sens. Environ. 210,96–112 (2018).
23. Y.-Y. Ren, G.-Y. Ren, X.-B. Sun, A. B. Shrestha, Q.-L. You, Y.-J. Zhan, R. Rajbhandari,
P.-F. Zhang, K.-M. Wen, Observed changes in surface air temperature and precipitation in
the Hindu Kush Himalayan region over the last 100-plus years. Adv. Clim. Change Res. 8,
148–156 (2017).
24. D. B. Kattel, T. Yao, Recent temperature trends at mountain stations on the southern
slope of the central Himalayas. J. Earth Syst. Sci. 122, 215–227 (2013).
25. A. P. Dimri, W. W. Immerzeel, N. Salzmann, R. J. Thayyen, Comparison of climatic
trends and variability among glacierized environments in the Western Himalayas.
Theor. Appl. Climatol. 134,155–163 (2017).
26. X. Yang, T. Zhang, D. Qin, S. Kang, X. Qin, Characteristics and changes in
air temperature and glacier’s response on the north slope of Mt. Qomolangma
(Mt. Everest). Arct. Antarct. Alp.Res. 43, 147 (2018).
27. J. Maurer, S. Rupper, Tapping into the Hexagon spy imagery database: A new automated
pipeline for geomorphic change detection. ISPRS J. Photogramm. Remote Sens. 108,
113–127 (2015).
28. H. Frey, H. Machguth, M. Huss, C. Huggel, S. Bajracharya, T. Bolch, A. Kulkarni,
A. Linsbauer, N. Salzmann, M. Stoffel, Estimating the volume of glaciers in the Himalayan–
Karakoram region using different methods. Cryosphere 8, 2313–2333 (2014).
29. S. F. Sherpa, P. Wagnon, F. Brun, E. Berthier, C. Vincent, Y. Lejeune, Y. Arnaud,
R. B. Kayastha, A. Sinisalo, Contrasted surface mass balances of debris-free glaciers
SCIENCE ADVANCES |RESEARCH ARTICLE
Maurer et al., Sci. Adv. 2019; 5: eaav7266 19 June 2019 10 of 12
on June 20, 2019http://advances.sciencemag.org/Downloaded from

observed between the southern and the inner parts of the Everest region (2007–15).
J. Glaciol. 63, 637–651 (2017).
30. A. Dehecq, N. Gourmelen, A. S. Gardner, F. Brun, D. Goldberg, P. W. Nienow, E. Berthier,
C. Vincent, P. Wagnon, E. Trouvé, Twenty-first century glacier slowdown driven by
mass loss in High Mountain Asia. Nat. Geosci. 12,22–27 (2019).
31. A. Sakai, K. Nishimura, T. Kadota, N. Takeuchi, Onset of calving at supraglacial lakes on
debris-covered glaciers of the Nepal Himalaya. J.Glaciol. 55, 909–917 (2009).
32. S. Rupper, G. Roe, Glacier changes and regional climate: A mass and energy balance
approach. J. Climate 21, 5384–5401 (2008).
33. S. Kaspari, T. H. Painter, M. Gysel, S. M. Skiles, M. Schwikowski, Seasonal and elevational
variations of black carbon and dust in snow and ice in the Solu-Khumbu, Nepal and
estimated radiative forcings. Atmos. Chem. Phys. 14, 8089–8103 (2014).
34. Y. Qian, W. I. Gustafson Jr., L. R. Leung, S. J. Ghan, Effects of soot-induced snow albedo
change on snowpack and hydrological cycle in western United States based on Weather
Research and Forecasting chemistry and regional climate simulations. J. Geophys.
Res. Atmos. 114, 10.1029/2008JD011 (2009).
35. D. I. Benn, T. Bolch, K. Hands, J. Gulley, A. Luckman, L. I. Nicholson, D. Quincey,
S. Thompson, R. Toumi, S. Wiseman, Response of debris-covered glaciers in the Mount
Everest region to recent warming, and implications for outburst flood hazards.
Earth Sci. Rev. 114, 156–174 (2012).
36. G. H. Roe, M. B. Baker, Glacier response to climate perturbations: An accurate linear
geometric model. J.Glaciol. 60, 670–684 (2014).
37. World Glacier Monitoring Service, Global Glacier Change Bulletin No. 2 (2014-2015).
M. Zemp, S. U. Nussbaumer, I. Gärtner-Roer, J. Huber, H. Machguth, F. Paul, and
M. Hoelzle, Eds. [ICSU(WDS)/IUGG(IACS)/UNEP/UNESCO/WMO, World Glacier Monitoring
Service, 2017], 244 pp.
38. T. C. Bond, S. J. Doherty, D. W. Fahey, P. M. Forster, T. Berntsen, B. J. DeAngelo,
M. G. Flanner, S. Ghan, B. Kärcher, D. Koch, S. Kinne, Y. Kondo, P. K. Quinn, M. C. Sarofim,
M. G. Schultz, M. Schulz, C. Venkataraman, H. Zhang, S. Zhang, N. Bellouin,
S. K. Guttikunda, P. K. Hopke, M. Z. Jacobson, J. W. Kaiser, Z. Klimont, U. Lohmann,
J. P. Schwarz, D. Shindell, T. Storelvmo, S. G. Warren, C. S. Zender, Bounding the role of
black carbon in the climate system: A scientific assessment. J. Geophys. Res. Atmos. 118,
5380–5552 (2013).
39. C. Nuth, A. Kääb, Co-registration and bias corrections of satellite elevation data sets for
quantifying glacier thickness change. Cryosphere 5, 271–290 (2011).
40. A. Arendt, A. Bliss, T. Bolch, J. G. Cogley, A. S. Gardner, J.-O. Hagen, R. Hock, M. Huss,
G. Kaser, C. Kienholz, W. T. Pfeffer, G. Moholdt, F. Paul, V. Radić, L. Andreassen,
S. Bajracharya, N. E. Barrand, M. Beedle, E. Berthier, R. Bhambri, I. Brown, E. Burgess,
D. Burgess, F. Cawkwell, T. Chinn, L. Copland, B. Davies, H. De Angelis, E. Dolgova,
L. Earl, K. Filbert, R. Forester, A. G. Fountain, H. Frey, B. Giffen, N. Glasser, W. Q. Guo,
S. Gurney, W. Hagg, D. Hall, U. K. Haritashya, G. Hartmann, C. Helm, S. Herreid, I. Howat,
G. Kapustin, T. Khromova, M. König, J. Kohler, D. Kriegel, S. Kutuzov, I. Lavrentiev,
R. LeBris, S. Y. Liu, J. Lund, W. Manley, R. Marti, C. Mayer, E. S. Miles, X. Li, B. Menounos,
A. Mercer, N. Mölg, P. Mool, G. Nosenko, A. Negrete, T. Nuimura, C. Nuth, R. Pettersson,
A. Racoviteanu, R. Ranzi, P. Rastner, F. Rau, B. Raup, J. Rich, H. Rott, A. Sakai, C. Schneider,
Y. Seliverstov, M. Sharp, O. Sigurðsson, C. Stokes, R. G. Way, R. Wheate, S. Winsvold,
G. Wolken, F. Wyatt, N. Zheltyhina. Randolph Glacier Inventory—A Dataset ofGlobal Glacier
Outlines: Version 5.0 (2015).
41. M. Huss, Density assumptions for converting geodetic glacier volume change to mass
change. Cryosphere 7, 877–887 (2013).
42. C. Rolstad, T. Haug, B. Denby, Spatially integrated geodetic glacier mass balance and
its uncertainty based on geostatistical analysis: Application to the western Svartisen ice
cap, Norway. J.Glaciol. 55, 666–680 (2009).
43. F. Paul, N. E. Barrand, S. Baumann, E. Berthier, T. Bolch, K. Casey, H. Frey, S. P. Joshi,
V. Konovalov, R. Le Bris, N. Mölg, G. Nosenko, C. Nuth, A. Pope, A. Racoviteanu, P. Rastner,
B. Raup, K. Scharrer, S. Steffen, S. Winsvold, On the accuracy of glacier outlines
derived from remote-sensing data. Ann. Glaciol. 54, 171–182 (2013).
44. Y. Xu, A. Knudby, H. C. Ho, Y. Shen, Y. Liu, Warming over the Tibetan Plateau in the last
55 years based on area-weighted average temperature. Regional Environ. Change 17,
2339–2347 (2017).
45. M. S. Shekhar, H. Chand, S. Kumar, K. Srinivasan, A. Ganju, Climate-change studies in the
western Himalaya. Ann. Glaciol. 51, 105–112 (2010).
46. L. Yan, X. Liu, Has climatic warming over the Tibetan Plateau paused or continued in
recent years. J. Earth Ocean Atmos. Sci. 1, 13 (2014).
47. A. Ohmura, Physical basis for the temperature-based melt-index method.
J. Appl. Meteorol. 40, 753–761 (2001).
48. G. Roe, Feedbacks, timescales, and seeing red. Annu. Rev. Earth Planet. Sci. 37,93–115
(2009).
49. D. P. Dee, S. M. Uppala, A. J. Simmons, P. Berrisford, P. Poli, S. Kobayashi, U. Andrae,
M. A. Balmaseda, G. Balsamo, P. Bauer, P. Bechtold, A. C. M. Beljaars, L. van de Berg,
J. Bidlot, N. Bormann, C. Delsol, R. Dragani, M. Fuentes, A. J. Geer, L. Haimberger,
S. B. Healy, H. Hersbach, E. V. Hólm, L. Isaksen, P. Kållberg, M. Köhler, M. Matricardi,
A. P. McNally, B. M. Monge-Sanz, J. J. Morcrette, B. K. Park, C. Peubey, P. de Rosnay,
C. Tavolato, J. N. Thépaut, F. Vitart, The ERA-Interim reanalysis: Configuration and
performance of the data assimilation system. Q. J. R. Meteorol. Soc. 137, 553–597 (2011).
50. E.Palazzi,J.vonHardenberg,A.Provenzale,PrecipitationintheHindu-Kush
Karakoram Himalaya: Observations and future scenarios. J. Geophys. Res. Atmos. 118,
85–100 (2013).
51. Y. Ageta, K. Higuchi, Estimation of mass balance components of a summer-accumulation
type glacier in the Nepal Himalaya. Geografiska Annaler. Series A. Physical Geography 66,
249–255 (2017).
52. W. W. Immerzeel, L. Petersen, S. Ragettli, F. Pellicciotti, The importance of observed
gradients of air temperature and precipitation for modeling runoff from a glacierized
watershed in the Nepalese Himalayas. Water Resour. Res. 50, 2212–2226 (2014).
53. A. B. Shrestha, C. P. Wake, J. E. Dibb, P. A. Mayewski, Precipitation fluctuations in the
Nepal Himalaya and its vicinity and relationship with some large scale climatological
parameters. Int. J. Climatol. 20, 317–327 (2000).
54. M. R. Bhutiyani, V. S. Kale, N. J. Pawar, Climate change and the precipitation variations in
the northwestern Himalaya: 1866–2006. Int. J. Climatol. 30, 535–548 (2010).
55. A. Sinha, G. Kathayat, H. Cheng, S. F. M. Breitenbach, M. Berkelhammer, M. Mudelsee,
J. Biswas, R. L. Edwards, Trends and oscillations in the Indian summer monsoon rainfall
over the last two millennia. Nat. Commun. 6, 6309 (2015).
56. V. Krishnamurthy, B. N. Goswami, Indian monsoon–ENSO relationship on interdecadal
timescale. J. Climate 13, 579–595 (2000).
57. M. A. Bollasina, Y. Ming, V. Ramaswamy, Anthropogenic aerosols and the weakening of
the South Asian summer monsoon. Science 334, 502–505 (2011).
58. S. Rupper, J. M. Schaefer, L. K. Burgener, L. S. Koenig, K. Tsering, E. R. Cook, Sensitivity
and response of Bhutanese glaciers to atmospheric warming. Geophys. Res. Lett. 39,
10.1029/2012GL053010 (2012).
59. N. Naito, A. Yutaka, N. Masayoshi, E. D. Waddington, C. F. Raymond, H. Conway, Response
sensitivities of a summer-accumulation type glacier to climate changes indicated with a
glacier fluctuation model. Bull. Glaciol. Res. 18,1–8 (2001).
60. J. M. Shea, W. W. Immerzeel, An assessment of basin-scale glaciological and hydrological
sensitivities in the Hindu Kush–Himalaya. Ann. Glaciol. 57, 308–318 (2016).
61. S. Vijay, M. Braun, Elevation change rates of glaciers in the Lahaul-Spiti (Western
Himalaya, India) during 2000–2012 and 2012–2013. Remote Sens. (Basel) 8, 1038 (2016).
62. K. Mukherjee, A. Bhattacharya, T. Pieczonka, S. Ghosh, T. Bolch, Glacier mass budget
and climate reanalysis data indicate a climatic shift around 2000 in Lahaul-Spiti, western
Himalaya. Clim. Change 148, 219–233 (2018).
63. T. Nuimura, K. Fujita, S. Yamaguchi, R. R. Sharma, Elevation changes of glaciers revealed
by multitemporal digital elevation models calibrated by GPS survey in the Khumbu
region, Nepal Himalaya, 1992-2008. J.Glaciol. 58, 648–656 (2012).
64. T. Bolch, A. Kulkarni, A. Kääb, C. Huggel, F. Paul, J. G. Cogley, H. Frey, J. S. Kargel, K. Fujita,
M. Scheel, S. Bajracharya, M. Stoffel, The state and fate of Himalayan glaciers. Science 336,
310–314 (2012).
65. A. Racoviteanu, Y. Arnaud, M. W. Williams, W. F. Manley, Spatial patterns in glacier
characteristics and area changes from 1962 to 2006 in the Kanchenjunga–Sikkim area,
eastern Himalaya. Cryosphere 9, 505–523 (2014).
66. S. R. Bajracharya, S. B. Maharjan, F. Shrestha, The status and decadal change of glaciers in
Bhutan from the 1980s to 2010 based on satellite data. Ann. Glaciol. 55, 159–166 (2014).
67. S. Ojha, K. Fujita, K. Asahi, A. Sakai, D. Lamsal, T. Nuimura, H. Nagai, Glacier area shrinkage
in eastern Nepal Himalaya since 1992 using high-resolution inventories from aerial
photographs and ALOS satellite images. J.Glaciol. 62, 512–524 (2016).
68. M. Begert, C. Frei, Long-term area-mean temperature series for Switzerland—Combining
homogenized station data and high resolution grid data. Int. J. Climatol. 38,
2792–2807 (2018).
69. M. Fischer, M. Huss, M. Hoelzle, Surface elevation and mass changes of all Swiss glaciers
1980–2010. Cryosphere 9,525–540 (2015).
70. F. Paul, W. Haeberli, Spatial variability of glacier elevation changes in the Swiss Alps
obtained from two digital elevation models. Geophys. Res. Lett. 35, 10.1029/
2008GL034718 (2008).
Acknowledgments: We thank C. Small and S. Hemming for valuable discussions on the
research and manuscript, and B. Raup for helping archive the data at NSIDC (National Snow
and Ice Data Center). Funding: J.M.M. was funded by a NASA Earth and Space Science
Fellowship (NNX16AO59H). J.M.S. was funded by NSF/EAR-1304351, the G. Unger Vetlesen
Foundation, and the Center for Climate and Life, Columbia University. S.R. was funded by
NASA 15-HMA15-0030. Author contributions: J.M.M., S.R., and J.M.S. framed the research
questions and designed the study. J.M.M. processed the DEMs, implemented the trend fitting
procedure, and performed the uncertainty assessment. A.C. and J.M.M. refined the glacier
polygons and calculated geodetic mass balances. S.R. performed the energy and
temperature change analysis. All authors interpreted the results. J.M.M. led the writing of the
manuscript with input and contributions from all coauthors. Competing interests: The
SCIENCE ADVANCES |RESEARCH ARTICLE
Maurer et al., Sci. Adv. 2019; 5: eaav7266 19 June 2019 11 of 12
on June 20, 2019http://advances.sciencemag.org/Downloaded from

authors declare that they have no competing interests. Data and materials availability: All
data needed to evaluate the conclusions in the paper are present in the paper and/or the
Supplementary Materials. Additional data related to this paper may be requested from
the authors. Glacier thickness change data from this study are archived at the NSIDC
(https://doi.org/10.5067/GGGSQ06ZR0R8). KH-9 Hexagon products can be downloaded from
USGS Earth Explorer (earthexplorer.usgs.gov), and the ASTER products used in this study are
available from the MADAS satellite data retrieval system (gbank.gsj.jp/madas).
Submitted 15 October 2018
Accepted 15 May 2019
Published 19 June 2019
10.1126/sciadv.aav7266
Citation: J. M. Maurer, J. M. Schaefer, S. Rupper, A. Corley, Acceleration of ice loss across the
Himalayas over the past 40 years. Sci. Adv. 5, eaav7266 (2019).
SCIENCE ADVANCES |RESEARCH ARTICLE
Maurer et al., Sci. Adv. 2019; 5: eaav7266 19 June 2019 12 of 12
on June 20, 2019http://advances.sciencemag.org/Downloaded from

Acceleration of ice loss across the Himalayas over the past 40 years
J. M. Maurer, J. M. Schaefer, S. Rupper and A. Corley
DOI: 10.1126/sciadv.aav7266
(6), eaav7266.5Sci Adv
ARTICLE TOOLS http://advances.sciencemag.org/content/5/6/eaav7266
MATERIALS
SUPPLEMENTARY http://advances.sciencemag.org/content/suppl/2019/06/17/5.6.eaav7266.DC1
REFERENCES http://advances.sciencemag.org/content/5/6/eaav7266#BIBL
This article cites 64 articles, 3 of which you can access for free
PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions
Terms of ServiceUse of this article is subject to the
registered trademark of AAAS. is aScience Advances Association for the Advancement of Science. No claim to original U.S. Government Works. The title
York Avenue NW, Washington, DC 20005. 2017 © The Authors, some rights reserved; exclusive licensee American
(ISSN 2375-2548) is published by the American Association for the Advancement of Science, 1200 NewScience Advances
on June 20, 2019http://advances.sciencemag.org/Downloaded from