TCThe CryosphereTCThe Cryosphere1994-0424Copernicus PublicationsGöttingen, Germany10.5194/tc-10-2361-2016A daily, 1 km resolution data set of downscaled Greenland ice sheet surface mass balance (1958–2015)NoëlBriceb.p.y.noel@uu.nlvan de BergWillem Janhttps://orcid.org/0000-0002-8232-2040MachguthHorsthttps://orcid.org/0000-0001-5924-0998LhermitteStefhttps://orcid.org/0000-0002-1622-0177HowatIanFettweisXavierhttps://orcid.org/0000-0002-4140-3813van den BroekeMichiel R.https://orcid.org/0000-0003-4662-7565Institute for Marine and Atmospheric research Utrecht, Utrecht University, Utrecht, the NetherlandsDepartment of Geography, University of Zurich, Zurich, SwitzerlandDepartment of Geosciences, University of Fribourg, Fribourg, SwitzerlandGeological Survey of Denmark and Greenland GEUS, Copenhagen, DenmarkDepartment of Geoscience & Remote Sensing, Delft University of Technology, Delft, the NetherlandsByrd Polar Research Center and School of Earth Sciences, Ohio State University, Columbus, USADepartment of Geography, University of Liège, Liège, BelgiumBrice Noël (b.p.y.noel@uu.nl)13October2016105236123776June20169June201623August201613September2016This work is licensed under a Creative Commons Attribution 3.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by/3.0/This article is available from https://tc.copernicus.org/articles/10/2361/2016/tc-10-2361-2016.htmlThe full text article is available as a PDF file from https://tc.copernicus.org/articles/10/2361/2016/tc-10-2361-2016.pdf
This study presents a data set of daily, 1 km resolution
Greenland ice sheet (GrIS) surface mass balance (SMB) covering the period
1958–2015. Applying corrections for elevation, bare ice albedo and
accumulation bias, the high-resolution product is statistically downscaled
from the native daily output of the polar regional climate model RACMO2.3 at
11 km. The data set includes all individual SMB components projected
to a down-sampled version of the Greenland Ice Mapping Project (GIMP) digital
elevation model and ice mask. The 1 km mask better resolves narrow ablation
zones, valley glaciers, fjords and disconnected ice caps. Relative to the
11 km product, the more detailed representation of isolated glaciated areas
leads to increased precipitation over the southeastern GrIS. In addition, the
downscaled product shows a significant increase in runoff owing to better
resolved low-lying marginal glaciated regions. The combined corrections for
elevation and bare ice albedo markedly improve model agreement with a newly
compiled data set of ablation measurements.
Introduction
During the last 2 decades, the Greenland ice sheet (GrIS)
has experienced significant mass loss as a result of increased meltwater runoff
and sustained high solid ice discharge from marine-terminating outlet
glaciers .
To fill spatial and temporal gaps in the scarce in situ observations,
regional climate models (RCMs) are often used to produce maps of the GrIS
surface mass balance (SMB;
).
RCMs explicitly calculate the individual SMB components ,
i.e. precipitation, runoff and sublimation, over the entire ice sheet
(Fig. ) at high spatial and temporal resolution and over extended
periods. However, the current spatial resolution of RCMs, typically
5–20 km, remains too coarse to accurately resolve glaciated areas in
topographically complex regions such as small isolated ice caps and marginal
outlet glaciers flowing into narrow fjords. In these regions, the relatively
coarse elevation and land ice masks used in RCMs might result in runoff
underestimation , hampering realistic regional SMB
estimates. Performing higher resolution simulations to address these issues
would require a substantial computational effort and is thus restricted to
case studies of small regions and relatively short time periods.
Annual mean SMB modelled by RACMO2.3 at 11 km over the GrIS and
surrounding ice caps for the period 1958–2015. This figure also depicts the
location of 213 ablation measuring sites (yellow dots) and 182 accumulation
sites (white dots) used for downscaled SMB evaluation as well as the four
GrIS marginal regions (blue boxes), discussed in Sect. 5. Letters refer to
the different transects shown in Fig. 9.
As an alternative, statistical downscaling can be applied to RCM output.
Previously, this method has been applied to the GrIS using global reanalysis
and climate data .
downscaled near-surface temperature and precipitation from three different RCMs
(11–25 km spatial resolution) to force a glacier mass balance model on a
250 m grid derived from the Greenland Ice Mapping Project (GIMP) digital
elevation model (DEM) , accurately resolving local glaciers
and ice caps of Greenland. Vertical gradients of climate parameters were
iteratively calibrated to enable the mass balance model to generate a
realistic melt distribution for the period 1980–2010, but the very high
resolution restricted the analysis to a few regions.
statistically downscaled GrIS SMB by interpolating each component of the
Modèle Atmosphérique Régional (MAR) from the original 25 km grid
to a 15 km resolution. This method used local daily vertical gradients,
except for precipitation, to correct for elevation differences between MAR
and a down-sampled version of the 5 km DEM from . The
elevation correction significantly reduced SMB biases. However, a resolution
of 15 km remains insufficient to resolve the rugged topography at the ice
sheet margins; to address this issue, near-kilometre resolution is necessary.
Here, we present a new data set of daily, 1 km resolution GrIS SMB
components (precipitation, melt, runoff, refreezing, sublimation and
snowdrift erosion) covering the period 1958–2015. The SMB product is
statistically downscaled from data of the Regional Atmospheric Climate Model
version 2.3 (RACMO2.3) at 11 km (Fig. ), using an
elevation-dependent technique based on the elevation and ice mask from the
GIMP DEM , down-sampled to 1 km. The following section
briefly describes RACMO2.3, the GIMP DEM, observational data sets and MODIS
bare ice albedo product used to evaluate and correct the downscaled data set.
The downscaling algorithm is explained in Sect. 3. Downscaled SMB is
evaluated using ablation and accumulation measurements in Sect. 4. Section 5
discusses the downscaling results for four different regions and for the
entire ice sheet. The added value, limitations and uncertainties of the
downscaling method are argued in Sect. 6, followed by conclusions in Sect. 7.
Model and dataThe regional climate model RACMO2
A detailed description of the Regional Atmospheric Climate
Model (RACMO2) is presented
in . RACMO2 incorporates the atmospheric dynamics and
physics modules from the High Resolution Limited Area Model (HIRLAM) and the
European Centre for Medium-range Weather Forecasts Integrated Forecast System
(ECMWF-IFS, ). The polar version of RACMO2 is developed by
the Institute for Marine and Atmospheric Research (IMAU), Utrecht University,
and is especially adapted for use over ice sheets and other glaciated
regions. Polar RACMO2 is interactively coupled to a multi-layer snow module,
accounting for firn densification, meltwater percolation, refreezing and
runoff ; it also incorporates an albedo scheme with
prognostic snow grain size and a drifting snow module,
simulating snow erosion and the drifting snow contribution to sublimation
. Recently, RACMO2.1 has been updated to RACMO2.3 as
discussed in and . Model evaluation against
SMB measurements, collected in the accumulation and ablation zones of the
GrIS, showed generally improved agreement . The native 11 km
climate run is forced at the lateral boundaries by ERA-40 (1958–1978,
) and ERA-Interim (1979–2015, )
reanalyses and uses the 5 km DEM and ice mask from .
Elevation and ice mask (orange, shown in inset) as prescribed in
RACMO2.3 at 11 km (left) and derived from the GIMP DEM down-sampled to 1 km
(right) over central east Greenland (blue box 1 in Fig. 1).
GIMP DEM
To downscale RACMO2.3 output, we use the ice mask and topography from the
GIMP DEM, described in , and currently considered to be one of
the most complete ice masks for Greenland . A 1 km ice mask
and DEM are obtained by averaging the original 90 m GIMP grid cells in each
1 km pixel covering Greenland. A 1 km resolution is deemed an acceptable
trade-off between improved resolution, i.e. a 121-fold improvement compared
to the 11 km grid, and manageable data handling given the daily time
resolution, time span (1958–2015) and the number of SMB components. As an
example, Fig. shows the topography and ice mask from RACMO2.3 at
11 km in central east Greenland (left in Fig. ; blue box 1 in Fig. ) and the GIMP DEM at 1 km
(right). The latter better resolves small scale landforms such as narrow
fjords and calving glacier tongues. Integrated over the contiguous GrIS, the
ice-covered area of 1.69×106km2 for the 1 km grid
represents a 0.5 % decrease relative to the 11 km mask. For our SMB
calculations, we only consider grounded ice, i.e. we discarded floating ice
pixels using a 1 km version of the 90 m grounded ice mask used in
.
Ablation and accumulation measurements
To evaluate the daily downscaled SMB product, we use 1155 SMB measurements
collected in the GrIS ablation (1073) and accumulation (182) zones. The
ablation data set was compiled as part of the Programme
for Monitoring of the Greenland Ice Sheet (PROMICE) and includes
stake and automatic weather station measurements retrieved from 213 sites (yellow dots in
Fig. ). Accumulation observations were derived from 182 sites
including snow pits and firn cores as well as
airborne radar measurements (white dots in Fig. ).
We exclusively selected data that temporally overlap with RACMO2.3
simulations (1958–2015; 205 sites discarded). We rejected observations from
sites with a > 100 m height bias relative to the representative elevation
of the 1 km GIMP topography (one site discarded).
To compare modelled and downscaled SMB with observations, different selection
approaches were applied in the ablation and accumulation zones, as described
in . In the accumulation zone, we select the closest grid cell
on the 11 and 1 km grids to represent modelled and downscaled SMB,
respectively. In the ablation zone, an altitude correction is applied by
selecting the grid cell with the smallest elevation bias among the closest
pixel and its eight adjacent neighbours.
MODIS bare ice albedo
A 1 km version of the 500 m MODerate-resolution Imaging Spectroradiometer
(MODIS) 16-day Albedo product (MCD43A3) is used to retrieve estimates of bare
ice albedo in the GrIS ablation zone. Bare ice albedo is estimated as the
average of the 5 % lowest surface albedo measurements for the period
2000–2015. A similar ice albedo product is used in RACMO2.3 based on MODIS
observations between 2001 and 2010 . In RACMO2.3, bare ice
albedo ranges from 0.3, i.e. dark bare ice exposed in the low ablation zone,
to 0.55 under persistent snow cover in the GrIS accumulation zone. Bare ice
albedo of glaciated pixels with no valid MODIS estimate is set to 0.47.
Methods
The daily, 1 km SMB product consists of statistically downscaled output from
a previously conducted RACMO2.3 simulation at 11 km, covering the period
1958–2015. RACMO2.3 settings and lateral forcing are described in
. The downscaling algorithm corrects the interpolated SMB
components using their local regression to elevation. Figure shows
the spatial correlation of individual SMB components with elevation on the
11 km RACMO2.3 grid. The spatial correlation is calculated for each grid box
using eight adjacent ice-covered pixels.
The elevation correction is exclusively applied to the SMB components, which
show a significant and spatially homogeneous correlation with elevation, i.e.
melt, runoff and sublimation (Fig. ). These SMB components decrease
with decreasing air temperature, represented by a negative correlation with
elevation (Fig. b, d, e). Although precipitation negatively
correlates with elevation over most of the ice sheet, the correlation remains
small and highly heterogeneous at the margins (Fig. a). Snowdrift
erosion exhibits a noisy correlation pattern. Therefore, daily precipitation
and snowdrift erosion are bilinearly interpolated to the 1 km ice mask
without elevation corrections. Refreezing exhibits a marked bimodal
correlation pattern (not shown), gradually increasing with height in the
ablation zone, where pore space is more abundant, and decreasing towards the
ice sheet interior due to limited meltwater supply. For this reason, and in
order to have a consistent liquid water balance, daily refreezing is
calculated as a residual:
RF=RA+ME-RU,
where RF is the residual refreezing, RA is rainfall, ME is surface melt, and
RU is meltwater runoff.
Daily SMB values are obtained by summing the individually downscaled
components:
SMB=Ptot-RU-SU-ER,
where Ptot is total precipitation (liquid and solid), RU is
meltwater runoff, SU is total sublimation (from surface and drifting snow)
and ER is drifting snow erosion.
Correlation to elevation of annual mean (a) total
precipitation (solid and liquid), (b) runoff, (c) SMB,
(d) sublimation, (e) melt and (f) drifting snow
erosion modelled by RACMO2.3 and calculated on the 11 km grid for the period
1958–2015.
Elevation-dependent downscaling procedure: b11km
and a11km are respectively the daily local estimates of the SMB
components regression to elevation and the SMB components value at mean sea
level obtained on the RACMO2.3 grid at 11 km. The red line corresponds to
the regression (b11km) calculated using the current grid cell
(blue dot) and the adjacent ones (red dots). The dashed green line applies
the regression slope to the current grid cell to estimate
a11km.
Elevation-dependent downscaling
The downscaling algorithm interpolates daily SMB components to the 1 km
topography and ice mask in three successive steps (Fig. ).
First, the local dependence on elevation is calculated on the original
RACMO2.3 11 km grid. Regression parameters are computed on a daily basis and
are, therefore, only valid for that specific day. A local regression slope,
b11 km (mmWEm-1, Fig. ), is calculated for
each ice-covered RACMO2.3 grid point using the maximum number of points
available; i.e. we use a total of six to nine ice-covered grid cells, the
current one and the minimum five to maximum eight adjacent pixels. This
minimum number is chosen after testing the downscaling sensitivity to the
number of regression cells used, as discussed in Sect. 3.2. An approximation
of the SMB components at mean sea level, a11 km (mmWE,
Fig. ), is then obtained using b11 km and the current
pixel. The regression is applied to the current grid cell to prevent local
estimates of a11 km to significantly differ from the original
RACMO2.3 value. Local regression parameters for melt and runoff are only
computed for pixels experiencing ablation. Moreover, erroneous positive
regression slopes, i.e. increasing melt rates with altitude, are discarded
until the following stage.
Next, valid estimates of b11 km and a11 km are
extrapolated iteratively on the 11 km grid to fully cover the more extensive
1 km ice mask. To that end, daily regression parameters are extrapolated
outwards of the 11 km ice mask by averaging b11 km from at least
three ice-covered pixels from the eight cells surrounding the current one.
Finally, the extrapolated fields of b11 km and a11 km
are bilinearly interpolated to the 1 km ice mask, providing estimates of
b1 km and a1 km. The downscaled SMB components
(Xv0.2), i.e. runoff, melt and sublimation, are then computed as a
linear function of the high-resolution topography as
Xv0.2=a1 km+b1 km×elevation1 km.
The downscaled data set that is based on the above elevation-dependent
technique is hereafter referred to as version v0.2.
Sensitivity experiment
Figure shows the difference between 11 km and downscaled GrIS
integrated daily runoff in summer 2011. Each line represents a different
minimum number of grid cells, ranging from three to nine, used to estimate the local
regression of runoff with elevation (b11 km; Fig. ). The
results are moderately sensitive to the number of regression points used
except for the nine cells setting, which systematically underestimates runoff at
the beginning and the end of the melt season as it discards all low-lying
glaciated pixels at the edge of the GrIS that experience early melt and the
largest values of runoff. The standard deviation between the different
settings (∼ 0.2 Gtday-1) is smaller than the difference
between 11 and 1 km runoff (∼ 0.6 Gtday-1). The more
regression points used, the smoother the runoff to elevation gradient field
becomes, lowering the downscaled runoff and bringing it closer to the 11 km
model output. Conversely, a small number of regression points can lead to
spuriously large local gradients. To prevent the downscaling algorithm from
substantially converging to, or diverging away from, 11 km RACMO2.3 output,
we adopted a setting of a minimum of six regression points, which is closest to the
average value of the different experiments (±0.1 Gtday-1).
Summer 2011 time series of daily, ice sheet-integrated runoff
difference (Gtday-1) between the downscaled product at 1 km,
using a minimum threshold of three–nine regression points (legend), and the
RACMO2.3 model at 11 km.
Comparison of SMB measurements collected at 213 sites with
(a) modelled SMB from RACMO2.3 at 11 km; (b) downscaled
SMB at 1 km (v0.2) and (c) corrected downscaled SMB at 1 km
(v1.0). The red stars correspond to PROMICE station QAS_L located in south
Greenland (61.03∘ N, 46.85∘ W; 310 m a.s.l.). The red
dashed line represents the regression including all measurements using a
perpendicular fit.
Melt and runoff adjustments
RACMO2.3 uses a prescribed bare ice albedo field, typically ranging from 0.30
in the low ablation zone to 0.55 under persistent snow cover. It is based on
the lowest 5 % of MODIS surface albedo values averaged for the period
2001–2010 . A comparison with a similar 1 km MODIS product
averaged for 2000–2015, ranging from 0.15 to 0.55, shows a systematic
overestimation of ice albedo at 11 km, especially for low-lying marginal
glacier tongues (Fig. i). This causes melt energy to be
underestimated during the melt season. To correct for this, downscaled melt
and runoff are adjusted by estimating the missing amount of ice melt
(MEadd) resulting from underestimated absorption of downward
shortwave radiation (SWd). In addition, as RACMO2.3 calculates
radiative fluxes on a horizontal plane, the direct fraction of
SWd is corrected for the slope and orientation of each 1 km
glaciated grid cell, as described in . For simplicity, we
assume SWd to be equally partitioned between diffuse and direct
radiation, and that the sun is exactly in the south at noon. This assumption
is purely pragmatic; on the basis of data availability, it could be further
refined in future versions of the downscaling procedure. Figures b
and show that ablation underestimation in v0.2 is restricted to
the low ablation zone (SMB <-4 mWE), where bare ice is exposed
for long episodes in summer. Therefore, the following corrections are only
applied to the ablation zone on days of melting bare ice when both surface
runoff and melt are non-zero in the downscaled product v0.2:
MEadd=Δα×0.5(SWd, 1 kmLf+ξSWd, 1 kmLf),
where MEadd (mmWEday-1) is the additional amount of ice
melt calculated at 1 km; Δα (–) is the difference between the
averaged bare ice albedo retrieved from the set of regression cells used to
downscale runoff at 11 km and the MODIS albedo product at 1 km;
SWd, 1 km is the modelled daily cumulated downward shortwave
radiation bilinearly interpolated to 1 km; Lf is the latent heat
of fusion (3.337×105Jkg-1); and ξ (–) is the
correction factor for a tilted plane (Fig. ), applied to the direct
component of downward shortwave radiation:
ξ=cos(ζ*)cos(ζ),ζ*=sin(ζ)cos(a)cos(σ)cos(Θ)+sin(ζ)sin(σ)sin(Θ)+cos(ζ)cos(σ),ζ=acos(sin(ϕ)sin(δ)+cos(H)cos(ϕ)cos(δ)),
where ζ* is the solar angle of incidence for a tilted plane, ζ
is the solar zenith angle, a is the azimuth of the tilted plane, σ
is the local surface slope, Θ is the orientation, ϕ is the
latitude, δ is the solar declination and H is the hour angle set to
0 at noon (Fig. ). All angles are expressed in radians.
Comparison of accumulation observations collected at 182 sites with
modelled SMB from RACMO2.3 at 11 km (red) and downscaled SMB v0.2 at 1 km
(blue) in mWEyr-1. Note that bias correction has not yet been
applied.
Scheme of a tilted plane as described in the GIMP DEM at 1 km.
SWd is the downward shortwave radiation, ζ* is the solar
angle of incidence for a tilted plane, ζ is the solar zenith angle, a
is the azimuth of the tilted plane, σ is the local surface slope and
Θ is the orientation. n and z are respectively the
vector normal to the tilted plane and the local vertical axis.
The bare ice albedo bias correction aims at minimizing the misfit between
downscaled SMB v0.2 and in situ measurements (Fig. b) by estimating
the missing runoff in the low ablation zone. Additional runoff
RUadd is calculated by applying a daily specific fraction Γ
(–) to MEadd, estimating the melt contribution to surface
runoff. Γ is defined as the ratio between daily downscaled runoff and
melt in v0.2 estimated using elevation dependence only:
RUadd=Γ×MEadd.
Assuming that the residual misfit between reconstructed and observed SMB
(ΔSMB, Fig. b) for the different ablation sites can be
ascribed to underestimated runoff in the low ablation zone of the GrIS,
RUadd is then scaled by a factor fscale (–), obtained by
computing a least-squares fit, minimizing the difference between ΔSMB
and RUadd using all ablation measurements:
ΔSMB=fscale×RUadd,fscale=∑ΔSMB×RUadd∑RUadd2,
The least-squares fit yields a value of fscale=1.176 for the GrIS.
This means that RUadd, i.e. accounting for elevation and bare ice
albedo corrections, has yet to be increased by ∼ 18 % to optimize
the agreement between downscaled and in situ SMB (Fig. c). The fact
that fscale> 1 strongly suggests that additional processes might
play a role in enhancing surface ablation, e.g. underestimation of modelled
sensible heat flux from warm air advection along the GrIS periphery
and uncertainties in cloud representation
. However, as the statistical downscaling approach is not
designed to correct for these physical processes, we adopted the empirical
approach presented above. The adjusted amount of runoff (RUv1.0) is
obtained by adding the missing runoff to the downscaled runoff
(RUv0.2):
RUv1.0=RUv0.2+fscale×RUadd.
The corrected melt (MEv1.0) is obtained in a similar fashion, and
refreezing (RFv1.0) is estimated as a residual between adjusted
melt, runoff and rainfall:
MEv1.0=MEv0.2+MEadd,RFv1.0=RA+MEv1.0-RUv1.0.
The downscaled SMB data set resulting from the combined elevation correction
and runoff adjustment is referred to as version v1.0 in the following
sections.
Annual mean observed (red dots) and downscaled (blue dots, v1.0) SMB
for eight selected transects in the GrIS ablation zone (mWEyr-1).
Name and locations of these transects (Fig. ) are listed at the
bottom of each graph. Graphs also list the number of sites used for each
transect, linear SMB-to-elevation regression retrieved from observations and
downscaled (v1.0) data in mmWEyr-1m-1, the RMSE and the mean
bias.
Evaluation of daily downscaled SMB
Figure evaluates the original RACMO2.3 SMB at 11 km (a), the
1 km raw downscaled SMB version v0.2 (b) and the 1 km corrected downscaled
SMB version v1.0 (c) (mWEyr-1) with 1073 observations from 213
ablation sites (yellow dots in Fig. ). The observational period was
matched with the modelled and downscaled SMB using the exact number of days.
Each blue star corresponds to the cumulative SMB for a duration ranging from
10 days to a full hydrological year. The downscaled SMB v0.2 agrees better
with observations compared to the RACMO2.3 output at 11 km
(Fig. a, b): we find a significant decrease of the root mean square
error (RMSE) (190 mmWE or -16 %) and a smaller bias
(100 mmWE or -21 %). The deviation from unity of the regression
slope decreases from 0.28 to 0.21 (-25 %), and the variance explained
increases from 47 to 61 %. When applying the bare ice albedo and local
orientation corrections, we find further significant improvements relative to
version v0.2 (Fig. c), with now 78 % of the variance explained
and a significant decrease in RMSE (270 mmWE or -27 %) and bias
(310 mmWE or -84 %). Red stars represent data from PROMICE
station QAS_L (61.03∘ N, 46.85∘ W, 310 m a.s.l.;
yellow dot in Fig. a), situated in an extremely narrow ablation
zone (∼ 10 km) at the southwestern tip of Greenland. Here,
modelled ablation gradients at 11 km are strongly underestimated in
RACMO2.3 and are only marginally better resolved at 1 km. At this
site, the additional corrections are especially important to obtain agreement
with observations.
Figure compares annual mean observed and downscaled SMB (v1.0)
along eight different SMB transects. There is good agreement for most
transects, except for Helheim Glacier (66.41∘ N,
-38.34∘ W). The downscaled product fails at reproducing the
quasi-constant ablation rate (∼-1 mWe) characterizing the
Helheim transect. The reason for this low SMB gradient is not clear at
present; it may be due to uncertainties in individual observation covering
relatively short periods, i.e. 1 or 2 months, which are only limited to the
melt season (July–August). Another possible explanation is that Helheim
Glacier experiences large and variable winter accumulation at low elevations,
potentially caused by drifting snow transport, limiting summer ablation. In
addition, Nioghalvjerds-fjorden and Storstrømmen transects
(Fig. a, b) also show significant remaining biases between in situ
and downscaled SMB at elevations lower than 200 m. We hypothesize
that these SMB measurements are located on floating glacier tongues with melt
ponds, resulting in very low satellite albedo, while stake measurements are
performed between ponds on brighter surfaces. As a result, the bare ice
albedo correction could be overestimated.
Centre east: (a) ice sheet mask in RACMO2.3 at 11 km (red)
and in the down-sampled GIMP DEM at 1 km (orange) (blue box 1 in Fig. 1),
and the mask of disconnected glaciers and ice caps at 1 km (blue); average
(1958–2015) annual mean (b) total precipitation,
(c) runoff and (d) SMB (mmWEyr-1) modelled by
RACMO2.3 at 11 km; (e) elevation bias (m) between 1 and 11 km
resolutions. Panels (f), (g) and (h) represent
annual mean total precipitation, runoff and SMB downscaled to 1 km using
elevation dependence only (v0.2). Panel (i) shows the bare ice
albedo bias between MODIS measurements at 1 km (2000–2015) and RACMO2.3 at
11 km (2001–2010). Panels (j), (k) and (l) are
similar to (f), (g) and (h) but incorporate the
bare ice albedo and precipitation corrections (v1.0).
In the accumulation zone, a small improvement is also found in v0.2 compared
to RACMO2.3 (Fig. ), but accumulation remains underestimated. The
SMB bias and RMSE are reduced by 0.7 mmWE (-2 %) and
1.8 mmWE (-3 %), whereas the regression slope and variance
explained remain unchanged. In the accumulation zone, SMB is mostly driven by
precipitation which is bilinearly interpolated to 1 km without
elevation correction. In addition, changes in sublimation are small due to
the relatively homogeneous topography of the ice sheet interior, limiting SMB
changes through downscaling. Despite significant improvements in the cloud
scheme of RACMO2.3 , clouds become saturated and start to
produce precipitation at elevations that are too low, resulting in
overestimated precipitation at the margins, e.g. southeastern Greenland,
while the ice sheet interior experiences too dry conditions. This
precipitation bias is currently being investigated, and we aim to resolve it
in the upcoming version RACMO2.4. To overcome the systematic negative SMB
bias of RACMO2.3 in the GrIS accumulation zone
(-37.5 mmWEyr-1, Fig. ), the daily total
precipitation v0.2 is adjusted to correct for underestimation in the ice
sheet accumulation zone (SMB > 0 mmWEyr-1):
PRv1.0=PRv0.2+PRv0.2PRv0.2a×σSMB,
where PRv1.0 is the daily adjusted total precipitation v1.0,
PRv0.2 is the daily bilinearly interpolated total precipitation
v0.2, PRv0.2a is the annual cumulative bilinearly
interpolated total precipitation v0.2 and σSMB is the
accumulation zone SMB bias in the downscaled product v1.0.
The final SMBv1.0 product is reconstructed as
SMBv1.0=PRv1.0-RUv1.0-SU-ER.
High-resolution SMB patterns: case studies
Table lists annual mean modelled and downscaled SMB components
(Gtyr-1) integrated over four different regions (blue boxes in
Fig. ) as well as over the entire GrIS. These regions were selected
for their specific climates, rough topography and narrow glaciated features,
which were not well resolved at 11 km. Figures ,
, and show the ice sheet mask for the
selected regions at 11 km (red cells) and 1 km (orange cells)
as well as peripheral glaciers and ice caps at 1 km (blue cells), the
elevation bias between the 1 and 11 km DEMs and the bare ice albedo bias
between the 1 km MODIS product and RACMO2.3 at 11 km; the latter
figures moreover show the main SMB components at both resolutions for the two
downscaled products (v0.2 and v1.0). In the following sections, we discuss
the impact of downscaling on regional SMB. Here, SMB components are
exclusively integrated over the contiguous GrIS; the SMB of detached ice caps
will be discussed in a forthcoming paper.
Table listing (top) the annual mean integrated SMB components
(Gtyr-1) covering the period 1958–2015 over four different
regions, centre east (69.6–74.3∘ N; 21–31∘ W; blue box 1
in Fig. 1), centre west (69.3–72.5∘ N; 49–57∘ W; blue
box 2), south (59.5–63.3∘ N; 41–51∘ W; blue box 3) and
north (80.5–83∘ N; 42–62∘ W; blue box 4), and for the
entire GrIS at both resolutions as well as the difference between 1 and
11 km; (bottom) the same is denoted for the ice-covered area
(km2).
1958–2015RegionsCentre east Centre west South North GrIS ResolutionUnit11 km1 kmΔ11 km1 kmΔ11 km1 kmΔ11 km1 kmΔ11 km1 kmΔSMBv0.2Gtyr-15.0-0.3-5.3-4.6-5.4-0.844.347.63.3-2.6-0.32.3349.3351.32.0Runoffv0.2Gtyr-116.123.87.718.319.20.942.444.62.28.95.8-3.1284.1297.713.6Precipv0.2Gtyr-122.625.22.615.015.20.291.497.25.86.96.1-0.8675.4692.016.6SMBv1.0Gtyr-15.0-11.6-16.6-4.6-6.7-2.144.337.3-7.0-2.6-1.31.3349.3338.2-11.1Runoffv1.0Gtyr-116.136.720.618.321.12.842.457.515.18.97.7-1.2284.1367.082.9Precipv1.0Gtyr-122.626.84.215.015.80.891.499.88.46.97.00.1675.4748.272.8SublimationGtyr-12.12.10.01.61.60.04.44.70.30.80.7-0.141.341.90.6Snow driftGtyr-1-0.5-0.40.1-0.3-0.20.10.20.30.1-0.1-0.10.00.71.10.4Ice area104km25.96.00.12.72.7-0.027.78.20.53.53.1-0.4170.3169.4-0.9
Centre west: same as Fig. 10 but for central west Greenland (blue
box 2 in Fig. 1).
South: same as Fig. 10 but for south Greenland (blue box 3 in
Fig. 1). The yellow dot in (a) locates station
QAS_L.
North: same as Fig. 10 but for north Greenland (blue box 4 in
Fig. 1). The green line in (a) shows the grounded ice mask at
1 km. The yellow dot in (a) locates the Petermann glacier
site settled on a floating ice tongue.
Central east Greenland
Central east Greenland (blue box 1 in Fig. ) is characterized by a
large body of interconnected valley glaciers, mostly terminating in narrow
glacial fjords. Figure a, e and i underline the inability of the
11 km mask to properly represent many glaciated areas, local topography or
bare ice albedo. In the 1 km mask, the ice-covered area increases
(∼ 2 %), while the elevation bias can locally exceed 500 m
over glacial valleys and small-scale promontories (Table ,
Fig. e); the average elevation bias is 80 m. These
differences affect SMB in two ways. First, precipitation increases by
2.6 Gtyr-1 (12 %) in v0.2 (Table ,
Fig. b, f), exclusively caused by the expansion of glaciated area
(no elevation correction is applied). Another 1.6 Gtyr-1
(6 %) of precipitation is added in v1.0 (Fig. j) to compensate
for the systematic negative SMB bias in the GrIS accumulation zone, as
discussed in Sect. 4. For both downscaling versions, changes in runoff mirror
the elevation change between the two resolutions (Fig. e),
highlighting the high sensitivity of runoff to elevation. In version v0.2,
integrated runoff increases by 7.7 Gtyr-1 (48 %)
(Fig. c, g). Furthermore, Fig. i reveals a systematic
overestimation of bare ice albedo at 11 km. Correcting for this
further increases runoff over the glaciers tongues (Fig. k),
accounting for ∼ 13 Gtyr-1 (55 %) of additional runoff
with respect to v0.2 (Table ). Negligible changes in sublimation
and drifting snow are found (Table ). As a consequence,
integrated SMB on the 1 km mask decreases by 5.3 Gtyr-1 in
version v0.2 (Fig. d, h) and by 16.6 Gtyr-1 in
version v1.0 (Fig. l). This analysis for central east Greenland
demonstrates the importance of accurately reproducing small-scale topography
and ice albedo to realistically capture local SMB variations.
Central west Greenland
The 11 km resolution DEM provides a reasonable representation of the wide,
gently sloping western ablation zone of the GrIS, where most glaciers are
land-terminating. The northern part of the selected area includes several
marine-terminating glaciers which are better represented at 1 km
(Fig. d, h).
Owing to negligible difference in glaciated area, precipitation remains
almost unchanged for the two resolutions and versions
(∼ 15 Gtyr-1). In both downscaled versions, enhanced
runoff is mostly obtained over narrow, low-lying glaciers tongues and
detached ice caps (Fig. c, g, k) where most of the elevation and
ice albedo biases are found (Fig. e, i). On the ice sheet, the
elevation correction increased runoff by about 1 Gtyr-1
(5 %) (Fig. g), while an additional
∼ 2 Gtyr-1 (10 %) (Fig. k) can be ascribed
to the ice albedo correction (Table ).
South Greenland
Southeast Greenland (blue box 3 in Fig. ) is a rugged region
(Fig. e), characterized by multiple topographically forced
precipitation maxima (Fig. b, f) and narrow marginal ablation
zones (Fig. c, g, k). Similar to central east Greenland, the
larger glaciated area (+6.5 %, Fig. a) at 1 km enhances
integrated precipitation by ∼ 6 Gtyr-1 (+7 %) in
v0.2 and 8.4 Gtyr-1 (+9 %) in v1.0. Increased runoff
(2.2 Gtyr-1 or 5 % in version v0.2) at the southern margins
can be ascribed to additional melt production over the better resolved narrow
ablation zones (Fig. g, k) combined with a moderate mean elevation
difference (∼ 17 m) between both resolutions. In v0.2, the ice
mask expansion explains most of the integrated SMB changes, leading to an
overall mass gain of 3.3 Gtyr-1.
Figure b reveals considerable ablation underestimation in south
Greenland, expressed as a systematic SMB bias of 2–4 mWe relative to
measurements collected at PROMICE station QAS_L (red dots in
Fig. a). The main reason for this underestimation is that SMB at
this location is characterized by a rare combination of high snowfall and
strong summer melt.
The remaining ablation underestimation in v0.2 can be partly ascribed to an
overestimated bare ice albedo (0.47) prescribed in RACMO2.3 ;
observed albedo at QAS_L frequently falls to 0.2 during the melt season
. As a result, the additional bare ice albedo correction
significantly improves runoff at station QAS_L (Fig. c).
Integrated over region 3, runoff increases by another
∼ 13 Gtyr-1 (29 %) relative to v0.2
(Fig. k). The increased marginal mass loss leads to the expansion
of the southern ablation zone towards higher elevations
(Fig. k, l), in line with local observations (Fig. c).
North Greenland
In north Greenland (blue box 4 in Fig. ), the climate is dry, and
most glaciers are marine-terminating. The ice sheet surface is relatively
smooth and homogeneous. The wide ablation zone is reasonably well captured at
11 km, leading to a modest deviation in elevation (∼ 43 m)
(Fig. e). However, the ice-covered area decreases by
∼ 11 % between both resolutions as the 11 km grid contained
erroneous floating glacier tongues (Fig. a). The ice area
reduction at 1 km affects precipitation (-0.8 Gtyr-1 or
-12 %) (Fig. b, f) and runoff (-3.1 Gtyr-1 or
-35 %) (Fig. c, g), resulting in a small SMB increase
(2.3 Gtyr-1) in version v0.2 (Fig. d, h). Large bare
ice albedo discrepancies can be found on five major glaciers
(Fig. i) where runoff increases substantially
(∼ 2 Gtyr-1 or 34 %) in version v1.0, further
decreasing the integrated SMB by 1.0 Gtyr-1 compared to v0.2
(Fig. h, l).
Greenland ice sheet
Although similar in area, the 1 km ice sheet mask better resolves peripheral
glaciers at the GrIS margins than RACMO2.3 at 11 km. GrIS-integrated
precipitation increases by 16.6 Gtyr-1 (+2 %) in v0.2,
most of which can be ascribed to ice area expansion in the east
(2.6 Gtyr-1) and south of Greenland (5.8 Gtyr-1),
where precipitation is large. An additional 56.2 Gtyr-1
(+8 %) is obtained in v1.0 when correcting for the accumulation zone
SMB bias. The smooth topography of the ice sheet interior results in a small
elevation difference of 4 m between both resolutions. Significant
elevation biases are mostly restricted to peripheral glaciers and narrow
ablation zones at the GrIS margins. As a result, runoff increases by
13.6 Gtyr-1 (+5 %) in version v0.2. Accounting for the
bare ice albedo bias in RACMO2.3 further increases runoff by
69.3 Gtyr-1 in version v1.0, leading to a much improved
agreement with ablation measurements. Of our selected areas, central east and
south Greenland contribute 25 and 18 % to the total runoff increase in the downscaled
product v1.0 owing to the many low-lying glaciers tongues that can only be
resolved at 1 km. Due to their smoother topography, north and centre
west Greenland contribute much less to the runoff change (∼ 3 and
1 %, respectively). Integrated over the contiguous ice sheet, SMB is not
significantly affected by the elevation dependence for which enhanced
precipitation (16.6 Gtyr-1) yearly balances the moderate
increase in runoff (13.6 Gtyr-1). In contrast, the bare ice
albedo and precipitation corrections substantially increase marginal runoff
(82.9 Gtyr-1) and accumulation (72.8 Gtyr-1),
resulting in a decrease of SMB of -11.1 Gtyr-1 (-3 %)
relative to the 11 km product.
Added value, limitations and uncertainties
The downscaled SMB v1.0 is the first data set to provide daily SMB estimates
for all outlet glaciers of the GrIS at a 1 km resolution for 58 years
(1958–2015). Relative to the original RACMO2.3 output, this data set
improves local SMB values (Fig. ) and produces more realistic SMB
patterns over rugged glaciated areas along the GrIS margins
(Figs. , , , ).
Figures and show that SMB v1.0 is an overall
improvement on the original RACMO2.3. To further investigate this,
Fig. shows the annual mean SMB RMSE (model vs. observations) of
the 11 km SMB field in RACMO2.3 (red), the downscaled product v0.2 (green)
and v1.0 (blue) as a function of observed SMB, binned in 0.5 mWE intervals.
In the ablation zone (SMB < 0), the SMB RMSE is reduced by 29–65 %
in v1.0 relative to the 11 km product, owing to the elevation correction in
v0.2 (9–23 %) and the additional albedo correction (20–42 %). In
the accumulation zone, the elevation dependence (9 %) and the
precipitation adjustment (19 %) also contribute to reduce the SMB RMSE by
28 % in v1.0. The largest RMSE reduction occurs in the lower GrIS
ablation zone, where improvements in topography and bare ice albedo in v1.0
are greatest.
Although significantly improved, the downscaled SMB v1.0 is likely to be
locally underestimated for four reasons: (a) the bare ice albedo correction
is evenly applied to both snow-covered and bare ice regions experiencing
surface melt and runoff, as no relevant proxy, reflecting day-to-day snow
coverage, could be derived from RACMO2.3. However, this issue should have a
limited effect on the magnitude of downscaled melt and runoff since the
albedo correction is most efficient in summer, when the snow cover of
low-lying glaciers has likely melted. (b) The MODIS ice albedo product at
1 km becomes less accurate at high latitudes, likely suffering from bare
soil contamination resulting from mixed reflectance signals recorded in both
the tundra- and ice-covered regions. Note that floating glacier tongues also
show surface albedo that is too low, e.g. Petermann glacier (yellow dot in
Fig. a), resulting from mixed signals from adjacent dark melt pond
and brighter dry ice. The resulting albedo underestimation over low-lying
floating tongues below 200 m leads to overestimated ablation
(∼ 0.2 mWEyr-1; Fig. a, b). (c) The average
1 km MODIS ice albedo product for 2000–2015 used in the melt correction
remains constant in time and might underestimate the bare ice albedo prior to
2000 as the period 2000–2015 encompasses multiple record high melt years.
(d) The MODIS Terra sensor has degraded . These limitations underline the high sensitivity of
the downscaled product to the input fields used to initialize the downscaling
procedure, i.e. RCM version used, the resulting modelled SMB components, bare
ice albedo records, ablation measurements, topography and ice mask. The
downscaled SMB v1.0 presents an estimated uncertainty of
∼ 12 Gtyr-1 in the GrIS ablation zone, which was estimated
by integrating the SMB bias in v1.0 (60 mmWE, Fig. c) over
the ablation zone of the contiguous ice sheet
(∼ 202 000 km2).
Annual mean modelled SMB RMSE (model vs. observations) of the 11 km
SMB field in RACMO2.3 (red dots), the downscaled SMB data set v0.2 (green
dots) and v1.0 (blue dots) as a function of observed SMB (395 observations).
Modelled SMB is grouped in 0.5 mWEyr-1 bins except for the first
bin, which ranges from -6.00 to -3.75 mWEyr-1. Numbers
indicate the number of observations used in each bin.
We anticipate that the new, 1 km Greenland SMB product is especially useful
for studies that address the mass balance of Greenland outlet glaciers that
are too steep and/or narrow to be properly resolved at the typical horizontal
resolution of regional climate models (∼ 5–20 km). Future
downscaled products can have even higher resolution (100 m) and will
be based on further improved RCM output fields of precipitation and melt.
Conclusions
The relatively coarse spatial resolution currently used in RCMs
remains insufficient to properly resolve small-scale variations in elevation
and ice cover at the ice sheet margins, significantly affecting the
calculation of melt and runoff. In the present study, we statistically
downscale individual SMB components from RACMO2.3 at 11 km to a 1 km ice
mask and topography derived from the GIMP DEM, using a daily specific
elevation dependence. Moreover, runoff and melt are corrected for biases in
bare ice albedo in RACMO2.3. Precipitation and snowdrift erosion are
bilinearly interpolated without applying an elevation correction. Total
precipitation is also adjusted to compensate for the dry accumulation bias of
RACMO2.3 in the ice sheet interior. Downscaled daily SMB is then retrieved
for the period 1958–2015 by summing daily downscaled precipitation, runoff,
sublimation and drifting snow erosion. An evaluation of the downscaled SMB
product against observations, collected both in the ablation and accumulation
zones of the GrIS, shows improved agreement. In the ablation zone, the
variance explained by the downscaled product v1.0 increased by 31 %
relative to the original RACMO2.3 11 km output, mainly through better
resolved narrow outlet glaciers at the GrIS margins.
Integrated over the GrIS, precipitation increased by 16.6 Gtyr-1
due to the larger glaciated area in south and east Greenland at 1 km; an
additional correction of 56.2 Gtyr-1 must account for the
accumulation bias in the ice sheet interior in RACMO2.3. Likewise, a
13.6 Gtyr-1 increase in runoff is attributed to elevation
corrections on the 1 km topography, and another 69.3 Gtyr-1
extra runoff can be ascribed to underestimated bare ice albedo over narrow
outlet glaciers at the GrIS margins. A small area in central east Greenland
alone, characterized by multiple narrow glacier tongues poorly resolved at
11 km, accounts for ∼ 25 % of the total additional runoff.
Data availability
The daily, 1 km SMB data set v1.0 presented in this study is freely available from the
authors without conditions.
Acknowledgements
Brice Noël, Willem Jan van de Berg and Michiel R. van den Broeke
acknowledge support from the Polar Programme of the Netherlands Organization
for Scientific Research (NWO/ALW) and the Netherlands Earth System Science
Centre (NESSC). Ian Howat and the GIMP project are supported by the U.S.
National Aeronautics and Space Administration (NASA). Horst Machguth
acknowledges support from the Programme for Monitoring of the Greenland Ice
Sheet (PROMICE), funded by the Danish Energy Agency's (DANCEA) program.
Edited by: L. Koenig
Reviewed by: two anonymous referees
ReferencesBales, R. C., McConnell, J. R., Mosley-Thompson, E., and Csatho, B.:
Accumulation over the Greenland ice sheet from historcal and recent records,
J. Geophys. Res., 106, 33813–33825, 10.1029/2001JD900153, 2001.Bales, R. C., Guo, Q., Shen, D., McConnell, J. R., Du, G., Burkhart, J. F.,
Spikes, V. B., Hanna, E., and Cappelen, J.: Annual accumulation for Greenland
updated using ice core data developed during 2000–2006 and analysis of daily
coastal meteorological data, J. Geophys. Res., 114, D06116,
10.1029/2008JD011208, 2009.Bamber, J. L., Ekholm, S., and Krabill, W. B.: A new, high-resolution digital
elevation model of Greenland fully validated with airborne laser altimeter
data, J. Geophys. Res., 106, 6733–6745, 10.1029/2000JB900365, 2001.Burgess, E. W., Forster, R. R., Box, J. E., Mosley-Thompson, E., Bromwich,
D. H., Bales, R. C., and Smith, L. C.: A spatially calibrated model of annual
accumulation rate on the Greenland Ice Sheet (1958–2007), J. Geophys. Res.,
115, F02004, 10.1029/2009JF001293, 2010.Dee, D. P., Uppala, S. M., Simmons, A. J., Berrisford, P., Poli, P.,
Kobayashi, S., Andrae, U., Balmaseda, M. A., Balsamo, G., Bauer, P.,
Bechtold, P., Beljaars, A. C. M., van de Berg, L., Bidlot, J., Bormann, N.,
Delsol, C., Dragani, R., Fuentes, M., Geer, A. J., Haimberger, L., Healy,
S. B., Hersbach, H., Hólm, E. V., Isaksen, L., Kållberg, P.,
Köhler, M., Matricardi, M., McNally, A. P., Monge-Sanz, B. M., Morcrette,
J.-J., Park, B.-K., Peubey, C., de Rosnay, P., Tavolato, C., Thépaut,
J.-N., and Vitart, F.: The ERA-Interim reanalysis: configuration and
performance of the data assimilation system, Q. J. Roy. Meteor. Soc., 137,
553–597, 10.1002/qj.828, 2011.Enderlin, E. M. and Howat, I. M.: Submarine melt rate estimates for floating
termini of Greenland outlet glaciers (2000–2010), J. Glaciol., 59, 67–75,
10.3189/2013JoG12J049, 2013.Enderlin, E. M., Howat, I. M., Jeong, S., Noh, M.-J., van Angelen, J. H., and
van den Broeke, M. R.: An improved mass budget for the Greenland ice sheet,
Geophys. Res. Lett., 43, 866–872, 10.1002/2013GL059010, 2014.Ettema, J., van den Broeke, M. R., van Meijgaard, E., van de Berg, W. J.,
Box, J. E., and Steffen, K.: Climate of the Greenland ice sheet using a
high-resolution climate model – Part 1: Evaluation, The Cryosphere, 4,
511–527, 10.5194/tc-4-511-2010, 2010a.Ettema, J., van den Broeke, M. R., van Meijgaard, E., and van de Berg, W. J.:
Climate of the Greenland ice sheet using a high-resolution climate model –
Part 2: Near-surface climate and energy balance, The Cryosphere, 4, 529–544,
10.5194/tc-4-529-2010, 2010b.Fausto, R. S., van As, D., Box, J. E., Colgan, W., Langen, P. L., and
Mottram, R. H.: The implication of nonradiative energy fluxes dominating
Greenland ice sheet exceptional ablation area surface melt in 2012, Geophys.
Res. Lett., 43, 1944–8007, 10.1002/2016GL067720, 2016.Fettweis, X.: Reconstruction of the 1979–2006 Greenland ice sheet surface
mass balance using the regional climate model MAR, The Cryosphere, 1, 21–40,
10.5194/tc-1-21-2007, 2007.Fettweis, X., Gallée, H., Lefebre, F., and van Ypersele, J.-P.: Greenland
surface mass balance simulated by a regional climate model and comparison
with satellite-derived data in 1990–1991, Clim. Dynam., 24, 623–640,
10.1007/s00382-005-0010-y, 2005.Fettweis, X., Tedesco, M., van den Broeke, M., and Ettema, J.: Melting trends
over the Greenland ice sheet (1958–2009) from spaceborne microwave data and
regional climate models, The Cryosphere, 5, 359–375,
10.5194/tc-5-359-2011, 2011.Franco, B., Fettweis, X., Lang, C., and Erpicum, M.: Impact of spatial
resolution on the modelling of the Greenland ice sheet surface mass balance
between 1990–2010, using the regional climate model MAR, The Cryosphere, 6,
695–711, 10.5194/tc-6-695-2012, 2012.Hanna, E., Huybrechts, P., Janssens, I., Cappelen, J., Steffen, K., and
Stephens, A.: Runoff and mass balance of the Greenland ice sheet: 1958–2003,
J. Geophys. Res., 110, D13108, 10.1029/2004JD005641, 2005.Hanna, E., Huybrechts, P., Steffen, K., Cappelen, J., Huff, R., Shuman, C.,
Irvine-Fynn, T., Wise, S., and Griffiths, M.: Increased Runoff from Melt from
the Greenland Ice Sheet: A Response to Global Warming, J. Climate, 21,
331–341, 10.1175/2007JCLI1964.1, 2008.Hanna, E., Huybrechts, P., Cappelen, J., Steffen, K., Bales, R. C., Burgess,
E., McConnell, J. R., Steffensen, J. P., den Broeke, M. V., Wake, L., Bigg,
G., Griffiths, M., and Savas, D.: Greenland Ice Sheet surface mass balance
1870 to 2010 based on Twentieth Century Reanalysis, and links with global
climate forcing, J. Geophys. Res., 116, D24121, 10.1029/2011JD016387,
2011.Howat, I. M., Negrete, A., and Smith, B. E.: The Greenland Ice Mapping
Project (GIMP) land classification and surface elevation data sets, The
Cryosphere, 8, 1509–1518, 10.5194/tc-8-1509-2014, 2014.Kuipers Munneke, P., van den Broeke, M. R., Lenaerts, J. T. M., Flanner,
M. G., Gardner, A. S., and van de Berg, W. J.: A new albedo parameterization
for use in climate models over the Antarctic ice sheet, J. Geophys. Res.,
116, D05114, 10.1029/2010JD015113, 2011.Lenaerts, J. T. M., van den Broeke, M. R., van Angelen, J. H., van Meijgaard,
E., and Déry, S. J.: Drifting snow climate of the Greenland ice sheet: a
study with a regional climate model, The Cryosphere, 6, 891–899,
10.5194/tc-6-891-2012, 2012.Lucas-Picher, P., Wulff-Nielsen, M., Christensen, J. H.,
Aðdalgeirsdóttir, G., Mottram, R., and Simonsen, S. B.: Very high
resolution regional climate model simulations over Greenland: Identifying
added value, J. Geophys. Res., 117, D02108, 10.1029/2011JD016267, 2012.Machguth, H., Rastner, P., Bolch, T., Mölg, N., Sørensen, L. S.,
Adalgeirsdottir, G., van Angelen, J. H., van den Broeke, M. R., and Fettweis,
X.: The future sea-level rise contribution of Greenland's glaciers and ice
caps, Environ. Res. Lett., 8, 025005, 10.1088/1748-9326/8/2/025005,
2013.Machguth, H., Thomsen, H., Weidick, A., Ahlstrøm, A. P., Abermann, J.,
Andersen, M. L., Andersen, S., Bjørk, A. A., Box, J. E., Braithwaite,
R. J., Bøggild, C. E., Citterio, M., Clement, P., Colgan, W., Fausto,
R. S., Gubler, K. G. S., Hasholt, B., Hynek, B., Knudsen, N., Larsen, S.,
Mernild, S., Oerlemans, J., Oerter, H., Olesen, O., Smeets, C., Steffen, K.,
Stober, M., Sugiyama, S., van As, D., van den Broeke, M., and van de Wal,
R. S.: Greenland surface mass balance observations from the ice sheet
ablation area and local glaciers, J. Glaciol., 27 pp.,
10.1017/jog.2016.75, online first, 2016.Noël, B., van de Berg, W. J., van Meijgaard, E., Kuipers Munneke, P.,
van de Wal, R. S. W., and van den Broeke, M. R.: Evaluation of the updated
regional climate model RACMO2.3: summer snowfall impact on the Greenland Ice
Sheet, The Cryosphere, 9, 1831–1844, 10.5194/tc-9-1831-2015, 2015.Overly, T. B., Hawley, R. L., Helm, V., Morris, E. M., and Chaudhary, R. N.:
Greenland annual accumulation along the EGIG line, 1959–2004, from ASIRAS
airborne radar and neutron-probe density measurements, The Cryosphere, 10,
1679–1694, 10.5194/tc-10-1679-2016, 2016.Polashenski, C. M., Dibb, J. E., Flanner, M. G., Chen, J. Y., Courville,
Z. R., Lai, A. M., Schauer, J. J., Shafer, M. M., and Bergin, M.: Neither
dust nor black carbon causing apparent albedo decline in Greenland's dry snow
zone: Implications for MODIS C5 surface reflectance, Geophys. Res. Lett., 42,
9319–9327, 10.1002/2015GL065912, 2015.Rastner, P., Bolch, T., Mölg, N., Machguth, H., Le Bris, R., and Paul,
F.: The first complete inventory of the local glaciers and ice caps on
Greenland, The Cryosphere, 6, 1483–1495, 10.5194/tc-6-1483-2012, 2012.Rignot, E., Box, J. E., Burgess, E., and Hanna, E.: Mass balance of the
Greenland ice sheet from 1958 to 2007, Geophys. Res. Lett., 35, L20502,
10.1029/2008GL035417, 2008.Rignot, E., Velicogna, I., van den Broeke, M. R., Monaghan, A., and Lenaerts,
J.: Acceleration of the contribution of the Greenland and Antarctic ice,
Geophys. Res. Lett., 38, L05503, 10.1029/2011GL046583, 2011.Sasgen, I., van den Broeke, M. R., Bamber, J. L., Rignot, E., Sandberg
Sørensenf, L., Wouters, B., Martinec, Z., Velicogna, I., and Simonsen,
S. B.: Timing and origin of recent regional icemass loss in Greenland, Earth
Planet. Sc. Lett., 333–334, 293–303, 10.1016/j.epsl.2012.03.033,
2012.Shepherd, A., Ivins, E. R., Geruo, A.,
Barletta, V. R., Bentley, M. J., Bettadpur, S., Briggs, K. H., Bromwich,
D. H., Forsberg, R., Galin, N., Horwath, M., Jacobs, S., Joughin, I., King,
M. A., Lenaerts, J. T. M., Li, J., Ligtenberg, S. R. M., Luckman, A.,
Luthcke, S. B., McMillan, M., Meister, R., Milne, G., Mouginot, J., Muir, A.,
Nicolas, J. P., Paden, J., Payne,A. J., Pritchard, H., Rignot, E., Rott, H.,
Sandberg Sørensen, L., Scambos, T. A., Scheuchl, B., Schrama, E. J. O.,
Smith, B., Sundal, A. V., van Angelen, J. H., van de Berg, W. J., van den
Broeke, M. R., Vaughan, D. G., Velicogna, I., Wahr, J., Whitehouse, P. L.,
Wingham, D. J., Yi, D., Young, D., and Zwally, H. J.: A Reconciled Estimate
of Ice-Sheet Mass Balance, Science, 338, 1183–1189,
10.1126/science.1228102, 2012.Stark, J. D., Office, E. M., Donlon, C. J., Martin, M. J., and McCulloch,
M. E.: OSTIA: An operational, high resolution, real time, global sea surface
temperature analysis system, in: OCEANS 2007 – Europe, 18–21 June 2007,
Aberdeen, Scotland, UK, IEEE, 10.1109/OCEANSE.2007.4302251, 1–4, 2007.
Unden, P., Rontu, L., Järvinen, H., Lynch, P., Calvo, J., Cats, G.,
Cuxart, J., Eerola, K., Fortelius, C., Garcia-Moya, J. A., Jones, C.,
Lenderlink, G., Mcdonald, A., Mcgrath, R., Navascues, B., Nielsen, N. W.,
Degaard, V., Rodriguez, E., Rummukainen, M., Sattler, K., Sass, B. H.,
Savijarvi, H., Schreur, B. W., Sigg, R., and The, H.: HIRLAM-5 Scientific
Documentation, Technical Report, 2002.
Uppala, S. M., KÅllberg, P. W., Simmons, A. J., Andrae, U., Da Costa
Bechtold, V., Fiorino, M., Gibson, J. K., Haseler, J., Hernandez, A., Kelly,
G. A., Li, X., Onogi, K., Saarinen, S., Sokka, N., Allan, R. P., Andersson,
E., Arpe, K., Balmaseda, M. A., Beljaars, A. C. M., Van De Berg, L., Bidlot,
J., Bormann, N., Caires, S., Chevallier, F., Dethof, A., Dragosavac, M.,
Fisher, M., Fuentes, M., Hagemann, S., Hólm, E., Hoskins, B. J., Isaksen,
L., Janssen, P. A. E. M., Jenne, R., Mcnally, A. P., Mahfouf, J.-F.,
Morcrette, J.-J., Rayner, N. A., Saunders, R. W., Simon, P., Ster, A.,
Trenberth, K. E., Untch, A., Vasiljevic, D., Viterbo, P., and Woollen, J.:
The ERA-40 re-analysis, Q. J. Roy. Meteor. Soc., 131, 2961–3012, 2005.Van Angelen, J. H., van den Broeke, M. R., Wouters, B., and Lenaerts,
J. T. M.: Contemporary (1969–2012) evolution of the climate and surface mass
balance of the Greenland ice sheet, Surv. Geophys., 35, 1155–1174,
10.1007/s10712-013-9261-z, 2013.Van As, D., Fausto, R. S., Ahlstrøm, A. P., Andersen, S. B., Andersen,
M. L., Citterio, M., Edelvang, K., Gravesen, P., Machguth, H., Nick, F. M.,
Nielsen, S., and Weidick, A.: Temperature and ablation records from the
Programme for Monitoring of the Greenland Ice Sheet (PROMICE), Geol. Surv.
Den. Greenl., 23, 73–76, available at: www.geus.dk/publications/bull,
last access: 27 September 2016, 2011.Van den Broeke, M. R., Smeets, P., and Ettema, J.: Surface layer climate and
turbulent exchange in the ablation zone of the west Greenland ice sheet, Int.
J. Climatol., 29, 2309–2323, 10.1002/joc.1815, 2009.Van Meijgaard, E., van Ulft, L. H., van de Berg, W. J., Bosveld, F. C.,
van den Hurk, B., Lenderink, G., and Siebesma, A. P.: Technical Report 302:
The KNMI regional atmospheric climate model RACMO version 2.1, Royal
Netherlands Meteorological Institute, De Bilt, 2008.
Van Tricht, K., Lhermitte, S., Lenaerts, J. T. M., Gorodetskaya, I. V.,
L'Ecuyer, T. S., Noël, B., van den Broeke, M. R., Turner, D. D., and
van Lipzig, N. P. M.: Clouds enhance Greenland ice sheet meltwater runoff,
Nature Communications, 7, 10266, 10.1038/ncomms10266, 2016.Van Wessem, J. M., Reijmer, C. H., Lenaerts, J. T. M., van de Berg, W. J.,
van den Broeke, M. R., and van Meijgaard, E.: Updated cloud physics in a
regional atmospheric climate model improves the modelled surface energy
balance of Antarctica, The Cryosphere, 8, 125–135,
10.5194/tc-8-125-2014, 2014.Weiser, U., Olefs, M., Schöner, W., Weyss, G., and Hynek, B.: Correction
of broadband snow albedo measurements affected by unknown slope and sensor
tilts, The Cryosphere, 10, 775–790, 10.5194/tc-10-775-2016, 2016.