Ice-sheet losses track high-end sea-level rise
Observed ice-sheet losses track the upper range of the IPCC Fifth Assessment Report sea-level predictions,
recently driven by ice dynamics in Antarctica and surface melting in Greenland. Ice-sheet models must account for
short-term variability in the atmosphere, oceans and climate to accurately predict sea-level rise.
Thomas Slater, Anna E. Hogg and Ruth Mottram
The Antarctic and Greenland
ice-sheets contain enough water
to raise global sea levels by 58 m
(ref. 1) and 7 m (ref. 2), respectively. As
the largest source of potential sea-level
rise (SLR)3, modest losses from these ice
sheets will increase coastal flooding4 and
affect oceans through freshwater input5.
Accurately forecasting SLR improves flood
risk assessment and adaptation. Since
the satellite record began in the 1990s,
Antarctica and Greenland together have
raised global sea levels by 17.8 mm, and
the volume of ice lost has increased over
time1,2. Of this, 7.2 mm originate from
Antarctica where ocean-driven melting
and ice-shelf collapse have accelerated ice
flow1; the remaining 10.6 mm come from
Greenland, which, despite holding less ice,
accounts for 60% of the recent ice-sheet
contribution as oceanic and atmospheric
warming have increased ice discharge and
surface meltwater runoff2. We compare
observations of Antarctic1 and Greenland
mass change2 to IPCC Fifth Assessment
Report (AR5) SLR projections3 during
their 10-year overlap, and we assess model
skill in predicting ice dynamic and surface
Observed and predicted mass change
Projecting the ice-sheet contribution
remains one of the most uncertain
components of the global sea-level budget3.
Progressive development of ice-sheet models
has improved their skill6 and will continue to
as descriptions of ice-sheet flow and climate
system interactions advance7. In AR5, the
ice-sheet contribution by 2100 is forecast
from process-based models simulating
changes in ice flow and surface mass balance
(SMB) in response to climate warming3.
Driven by the century-scale increase in
temperature forced by representative
concentration pathways (RCPs), global
mean SLR estimates range from 280–980
mm by 2100 (Fig. 1). Of this, the ice-sheet
contribution constitutes 4–420 mm (ref. 3).
The spread of these scenarios is uncertain,
scenario-dependent and increases rapidly
after 2030 (Fig. 1).
During 2007–2017, satellite observations
show total ice-sheet losses increased the
global sea level by 12.3 ± 2.3 mm and track
closest to the AR5 upper range (13.7–14.1
mm for all emissions pathways) (Fig. 1).
Despite a reduction in ice-sheet losses during
2013–2017 — when atmospheric circulation
above Greenland promoted cooler summer
conditions and heavy winter snowfall2 — the
observed average SLR rate (1.23 ± 0.24 mm
per year) is 45% above central predictions
(0.85 ± 0.07 mm per year) and closest to
the upper range (1.39 ± 0.14 mm per year)
(Fig. 2). These upper estimates predict an
additional 145–230 mm (179 mm mean) of
SLR from the ice sheets above the central
predictions by 2100. SLR of 150 mm will
double storm-related flooding frequency
across the west coasts of North America and
Europe and in many of the world’s largest
coastal cities4. Ice-sheet losses at the upper
end of AR5 predictions would expose 44–66
million people to annual coastal flooding
worldwide8. SLR in excess of 1 m could
require US$71 billion of annual investment
in mitigation and adaptation strategies9.
Separating ice-sheet processes
The ice-sheet response to climate forcing
comes from the SMB (net balance between
accumulation and ablation processes)
Antarctica and Greenland
1995 2000 2005 2010 2015 2020 2025 2030 2035 2040
Sea-level contribution (mm)
Fig. 1 | Observed and predicted sea-level contribution from Antarctic and Greenland ice-sheet mass
change. The Antarctic and Greenland ice-sheet contribution to global sea level according to IMBIE1,2
(black) reconciled satellite observations and AR53 projections between 1992–2040 (left) and 2040–
2100 (right). For each AR5 emission scenario, the upper (maroon), mid (orange) and lower (yellow)
estimates are taken from the 95th percentile, median and 5th percentile values of the ensemble range,
respectively3. Within the upper, mid and lower sets, AR5 pathways are represented by darker lines in
order of increasing emissions: RCP 2.6, RCP 4.5, RCP 6.0, SRES A1B and RCP 8.5. Shaded areas represent
the spread of AR5 scenarios and the 1σ estimated error on the observations. The dashed vertical lines
indicate the period during which the satellite observations and AR5 projections overlap (2007–2017).
AR5 projections have been offset to equal the satellite record value at their start date (2007).
NATURE CLIMATE CHANGE | www.nature.com/natureclimatechange
and the dynamic response to changes in
ice flow, calving of icebergs and melting
at the ice–ocean interface. AR5 provides
separate projections for these components
(Fig. 2)3. AR5 SMB simulations were
based on a regional climate model (RCM)
ensemble, extended with temperature-based
polynomials driven by surface air
temperatures from general circulation
models (GCMs)3. Ice dynamic contributions
were derived from studies carried out using
ice-sheet models forced by, but not coupled
to, atmospheric and oceanic model outputs.
In this way, the atmosphere and ocean can
impact the ice sheet but not vice versa. In
2013, when AR5 was released, few models
were available to simulate the complex
calving processes and ice dynamical
contributions to SLR. Instead, ice dynamics
were projected using parameterizations
for calving at selected outlet glaciers and
scaled based on the published range of SLR3.
Process-based models considered in AR5
have generally produced lower estimates
of SLR than semi-empirical models based
on palaeoclimate reconstructions10. As SLR
from SMB and dynamic components of
ice-sheet mass balance differ substantially in
Antarctica and Greenland, we consider their
We compare the observed1,2 and
modelled3 ice dynamical and SMB
contributions during the overlap period
(Fig. 2). During 2007–2017, Antarctic
ice dynamics contributed 4.6 ± 2.3 mm
(Supplementary Fig. 1) to global sea level,
at the same average rate projected by the
AR5 mid-level scenario (0.47 ± 0.05 mm
per year) (Fig. 2). We note, however, a large
spread between AR5 Antarctic ice dynamic
projections, which range from 3–34 mm
by 2040, and predict a negative sea-level
contribution in the lower scenarios from
2030 (Supplementary Fig. 1). Despite all
scenarios predicting Antarctic mass gains
from increasing snowfall, the continent’s
estimated SMB (0.05 ± 0.13 mm per year)
has reduced slightly and is closest to the
upper range (–0.02 ± 0.04 mm per year).
In Greenland, dynamic ice losses estimated
from satellite observations during 2007–
2017 (0.26 ± 0.13 mm per year) track the
lower range of predictions (0.22 ± 0.04 mm
per year). However, these AR5 projections
were based on kinematic scaling and do not
explicitly simulate ice flow3. Surface mass
losses in Greenland raised global sea levels
by an estimated 4.6 ± 1.8 mm during 2007–
2017 at an average rate of 0.46 ± 0.23 mm
per year, 28% higher than the upper range of
scenarios (0.36 ± 0.06 mm per year).
High interannual variability in the
observed mass change — notably for the
Antarctic dynamic (0.46 ± 0.16 mm per
year) and Greenland surface (0.46 ± 0.23
mm per year) components (Fig. 2) — is
not reproduced in AR5 and may not
represent the longer-term mass imbalance.
For Greenland in particular, changes in
atmospheric circulation-induced11 extreme
melting12 and substantial variability in
meltwater runoff are not captured in
AR5 predictions2, which are forced by
annual temperature changes and do not
reproduce the persistence in the North
Sea-level contribution (mm yr–1)
Fig. 2 | Observed and predicted annual rates of sea-level rise from Antarctic and Greenland ice-sheet mass change and their individual ice dynamic and
surface mass components. Average annual rates of sea-level rise and their standard deviations from IMBIE1,2 (black) and AR5 (ref. 3) projections during
2007–2017, including upper (95th percentile, maroon), mid (median, orange) and lower (5th percentile, yellow) estimates. Results are partitioned into the
surface and ice dynamic mass change, along with the combined sea-level contribution from both ice sheets.
NATURE CLIMATE CHANGE | www.nature.com/natureclimatechange
Atlantic driving these short-term weather
events. In addition, clouds modulate13
surface melting, and climate model biases
in clouds and their formation processes
may be partly responsible for both over-
and under-estimating surface melt. Future
studies would benefit from a comparison
over the full 25-year observational record,
during which satellites provide continuous
and complete coverage over both ice sheets,
to better contextualize variability within the
Advances in ice-sheet modelling are
expected through experiments such as the
Ice-sheet Model Intercomparison project
for CMIP6 (ISMIP6)6, which will deliver
process-based projections from standalone
ice-sheet models forced by output from
coupled atmosphere–ocean GCMs in time
for AR6 in 2022. These efforts will improve
predictions of the ice dynamical response,
particularly in Antarctica where the spread
among AR5 scenarios is large, through
advanced representations of ice–ocean
interactions which extrapolate GCM ocean
forcing into ice-shelf cavities7. Modelling of
surface processes is also improved by using
RCMs to increase the spatial resolution of
atmospheric GCM forcing and capture SMB
variations found in steep topography at
Challenges remain in modelling
ice-sheet dynamic and SMB processes.
Descriptions of ice–ocean interactions are
hindered by coarse GCM resolution, and
potential feedbacks in ocean circulation
due to freshwater input are not accounted
for6. Dynamic ice loss is driven by marine
melt and iceberg calving; improved
representations of these processes in
ice-sheet models, and dense time series of
outlet glacier observations, will improve
understanding. Surface forcing for ISMIP6
experiments is provided as annual averages,
and establishing the effects of shorter-term
atmospheric variability and circulation
changes on ice-sheet SMB requires further
work. The quality of SMB forcing is also
affected by inadequacies in GCM output —
for example, in accurate representations of
clouds and surface albedo. Such challenges
can be partly addressed with two-way
coupling of Antarctic and Greenland
ice-sheet models to the atmosphere–ocean
system. However, this remains a significant
undertaking: differing spatial and temporal
resolutions required by model components
must be negotiated, and improving related
parameterizations is essential.
Ice-sheet observational and modelling
communities must also continue to
collaborate. For example, regional case
studies of extreme events driven by
short-term variability can improve our
understanding of ice-sheet processes.
Partitioning ice-sheet projections into
SMB and ice dynamics in AR6, as in AR5,
will allow these processes to be further
understood and evaluated separately. Recent
experiments have assessed the ability of
models to reproduce historical change5,14,15,
increasing confidence in sea-level
projections and gauging the likelihood
of extreme SLR from marine ice-sheet
and ice-cliff instabilities. Reducing
uncertainty in observational datasets
through collaborative processes such as
IMBIE, and generating new datasets (for
example, of SMB and ice-shelf melt rates),
will help reduce present-day biases in
ice-sheet models. Used together, ice-sheet
observations and models will continue to
inform scientific debate and climate policy
for decades to come. ❐
Thomas Slater 1 ✉ , Anna E. Hogg2 and
Ruth Mottram 3
1Centre for Polar Observation and Modelling, School
of Earth and Environment, University of Leeds,
Leeds, UK. 2School of Earth and Environment,
University of Leeds, Leeds, UK. 3Danish
Meteorological Institute, Copenhagen, Denmark.
Published: xx xx xxxx
1. e IMBIE Team. Nature 558, 219–222 (2018).
2. e IMBIE Team. Nature 579, 233–239 (2020).
3. IPCC Climate Change 2013: e Physical Science Basis (eds
Stocker, T. F. et al.) (Cambridge Univ. Press, 2013).
4. Vitousek, S. et al. Sci. Rep. 7, 1399 (2017).
5. Golledge, N. R. et al. Nature 566, 65–72 (2019).
6. Nowicki, S. M. J. et al. Geosci. Model Dev. 9, 4521–4545 (2016).
7. Pattyn, F. et al. Nat. Clim. Change 8, 1053–1061 (2018).
8. Kulp, S. A. & Strauss, B. H. Nat. Commun. 10, 4844 (2019).
9. Hinkel, J. et al. Proc. Natl Acad. Sci. USA 111, 3292–3297 (2014).
10. Garner, A. J. et al. Earth’s Future 6, 1603–1615 (2018).
11. Bevis, M. et al. Proc. Natl Acad. Sci. USA 116, 1934–1939 (2019).
12. Nghiem, S. V. et al. Geophys. Res. Lett. 39, L20502 (2012).
13. Hofer, S., Tedstone, A. J., Fettweis, X. & Bamber, J. L. Nat. Clim.
Change 9, 523–528 (2019).
14. Ritz, C. et al. Nature 528, 115–118 (2015).
15. Edwards, T. L. et al. Nature 566, 58–64 (2019).
This work is an outcome of the Ice-sheet Mass Balance
Inter-Comparison Exercise (IMBIE) supported by the
ESA Climate Change Initiative and the NASA Cryosphere
Program. T.S. was funded by the NERC Centre for
Polar Observation and Modelling through a Natural
Environment Research Council (cpom300001) grant,
and A.E.H. was funded by a NERC Fellowship (NE/
R012407/1). R.M. acknowledges the support of the
ESA CCI+ for Greenland ice-sheet under ESA-ESRIN
contract number 4000104815/11/I-NB and the Danish
State through the National Centre for Climate Research
(NCKF). We thank the European Space Agency, National
Aeronautics Space Administration and the German
Aerospace Centre for provision of satellite data, without
which this study would not have been possible. We also
thank A. Shepherd for leading IMBIE, which produced
the reconciled observations of ice-sheet mass change, and
for useful discussions during the course of this study. The
satellite data used here are freely available at http://imbie.
org/data-downloads/, and IPCC sea-level rise projections
can be downloaded from http://www.climatechange2013.
The authors declare no competing interests.
is available for this paper at https://doi.org/10.1038/
NATURE CLIMATE CHANGE | www.nature.com/natureclimatechange