The South Shetland Islands are located at the northern tip of the Antarctic
Peninsula (AP). This region was subject to strong warming trends in the
atmospheric surface layer. Surface air temperature increased about
Antarctic peripheral glaciers and ice caps cover an area of
Trends in surface air temperature in the AP have been analysed and discussed
in several studies obtained from meteorological observations of either manned
or automatic weather stations (AWSs)
As a consequence of the observed warming, striking glaciological changes have
happened along the whole length of AP's western and eastern coasts.
Studies along the AP show basal and surficial enhanced melting on ice shelves
accompanied by subsequent collapse
Systematic glaciological field studies are very scarce on both sides of the AP and in especially there are none that try to sufficiently capture inter- and intra-annual variability. In this paper, we present a 6-year record of continuous glaciological observations obtained on very high temporal resolution to resolve winter melt periods and properly define the start of glacial accumulation and ablation periods. The time series is analysed with regard to climatic drivers and glacial melt is estimated to provide the necessary boundary conditions for interdisciplinary studies on the ongoing changes in the biota and species composition of the coastal waters in the South Shetland region. The following sections will describe the meteorological and glaciological data time series, evaluate calibration and validation of the glaciological model, and discuss surface mass balance and simulated glacial discharge with regard to seasonal and interannual variability. Finally, results of equilibrium line altitude (ELA) and accumulation area ratio (AAR) are used to assess future glacier extent.
The main scientific objective of this paper is to investigate the impact of the concurrent climatic conditions on meltwater discharge to the coastal environments and the glacier ELA, which sets the boundary conditions for the observed environmental change.
Map of research area on King George Island (South Shetland Islands,
northern Antarctic Peninsula) including the locations of our own installations
and external data time series. Potter creek basins of Potter North and Potter
South with drainage channels and mass balance stake locations along two
transects, PG0
King George Island (KGI) is the largest of the South Shetland Islands,
located 130
In this study, a physical-process-based, fully distributed energy balance model is applied to estimate glacier melt and glacier runoff into the Potter cove. The temporal resolution is defined by the resolution of the meteorological input data, which were sampled to hourly values. The spatial resolution is defined by the spatial input grids of the digital terrain model (DTM), the surface facies, aspect and slope, etc. First, the meteorological data sets, quality control and gap filling are discussed. Second, the glaciological measurements and post-processing are reviewed. Third, the glaciological model is described, and the necessary input grids are defined.
AWS installed on the Fourcade Glacier with view to the Potter Cove and Three Brothers Hill. The photo was taken during winter on 30 May 2012.
An AWS was installed in November 2010 at 62
AWS installed on the Fourcade Glacier with view to the Potter Cove and Three Brothers Hill. The photo was taken on 4 March 2012 and shows the AWS during the ablation period when pyroclastic material resurface due to melting of the winter snow layer.
The Argentinean Carlini Station (formerly Jubany Station) is located at
62
The Russian Bellingshausen Station is located on Fildes Peninsula of KGI at
The 5-year meteorological data time series for the AWS location on the glacier of the Potter Peninsula was screened for flawed and unrealistic values caused either by instrument malfunctioning or by environmental impacts such as hoar frost. During wintertime, power outages are more likely to occur and maintenance is often prevented by unfavourable weather conditions. Gap-filling routines were implemented and are discussed in the following.
Methods using the monthly mean diurnal cycle to fill data gaps were rejected,
because (a) data gaps in wintertime are by far more frequent and also the
time period of missing data longer, and (b) the diurnal variability is
dominated by the seasonal course. For each sensor, the data were checked for
malfunctioning or other environmental impact, but also for statistical
properties according to
The AWS air temperature measurements were screened for spikes (every value
outside the 6-fold standard deviation of the long-term average) and air
temperature readings below
Apart from a crossing of a synoptic frontal system, sudden spikes to
low values in the barometric pressure are usually associated
with problems and sudden drops in the power supply. All values below
930
The 2 m wind velocity was screened for sensor malfunctioning. The
cleaned 5-year statistics show a linear relationship between the horizontal
wind speeds measured at Carlini Station and the AWS on the glacier of
The search of a similar relationship for the relative
humidity yielded a linear regression between the time series at
Carlini Station and at the AWS with a very poor correlation coefficient
of
Ice temperature measurements were taken from the lowest
sensor level, originally installed at 10
Meteorological time series (gap-filled) aggregated to hourly resolution at the AWS site on the Fourcade Glacier, King George Island, during the time period November 2010–November 2015.
The cloud coverage data were taken from the observations (cloud coverage
The four-component radiation sensor is prone to icing or
riming due to advection of warm, moist air masses into the
region. This affects especially the upward looking sensors for long-
and shortwave radiation. As criteria for detection of sensor riming,
The optical air mass number is given for
The downward component of the longwave radiation budget at the surface is
calculated using the emissivity of the atmosphere (
Accumulation measurements based on sonic ranging are available
during November 2010 to May 2011 and March 2012 to November 2012. Outside
these periods, accumulation was reconstructed using the readings of the
Bellingshausen 6-hourly precipitation data. These were resampled to hourly
data time series by linear interpolation, smoothed and normalized with the
total daily sum. The mass balance stake (MBS) data at the AWS location were then
used as an envelop for the daily sums of the resulting accumulation time
series. Figure
Two transects of MBSs were installed from the
top of the Warszawa Ice Dome down to the border of the glaciers Fourcade and
Polar Club to serve for calibration and validation of modelling efforts (see
Fig.
Locations of the mass balance stake (MBS) at the AWS and the
transects (PG0
MBS transects installed on the Fourcade Glacier during the ablation
period on 28 March 2012
Displayed is the time series of snow density observations on the
Fourcade Glacier, Potter Peninsula, during the time period June 2012 to
February 2016
Snow density samples were taken with an aluminum snow density cutter
(SnowHydro, Fairbanks, Alaska) with a defined volume of 0.001
Starting in June 2012 until February 2016, regular snow density values were
taken together with the MBS transect measurements.
A seasonality in the snow density time series is evident and presented in
Fig.
These findings are supported by the snow depth measurements
(
The snow density and snow depth time series were used to
compute the cumulative mass balance (CMB) in metres of water equivalent
(m w.e.) from MBS observations along the mass balance transects shown in
Figs.
Shown is the cumulative mass balance in m w.e. (water equivalent)
measured at the calibration mass balance stake transect PG0
Shown is the cumulative mass balance in m w.e. (water equivalent)
measured at the validation mass balance stake transect PG1
The resulting MBS time series graphs is differentiated by the transect
ID PG0 and PG1. The gradual shift from ablation at
the lowest MBS PG09 to accumulation at the highest MBS PG04 is clearly
visible. The PG1 transect, however, does not follow
this behaviour: there is considerably more accumulation at MBS PG19
than at the higher elevation PG17. Also within the accumulation zone,
the expected increase of accumulation with elevation does not apply:
the cumulative accumulation at MBS PG14 is considerably lower than at
MBS PG15. This can be explained by the different degree of exposure to
weather. KGI is prone to transient low-pressure systems connected to
storm events with high wind speeds from the northwest and precipitation
mostly in the form of rain, but also to katabatic winds due to
influence of the Antarctic high-pressure systems with high wind speeds
from the southeast
The very high temporal resolution of mass balance observation is
unique for the AP region and was chosen not only to resolve the high seasonal
and interannual variability of the onset of accumulation and ablation
period but also to capture winter melt periods to estimate glacier
meltwater runoff also during wintertime. The time series analysed
here encompass 6 years, but are still ongoing. Figures
The glacier surface mass balance model (GMM) by
To estimate the complete input of glacial and snow meltwater into the
Potter Cove, the GMM was run in catchment configuration. The model
area encompasses glacial and periglacial areas that are part of the
Fourcade Glacier catchment area draining into Potter Cove. The
boundary of the Fourcade Glacier is taken from
The GMM was applied to two different catchment areas: (1) the hydrological
catchment area draining into the Potter meltwater and discharge creeks and
(2) the Potter Cove catchment, i.e. the Fourcade Glacier, defined as the part
of the Warszawa Icefield that drains into Potter Cove. The Potter Cove
catchment has an area of 25.1
For the catchment boundary refinements, flow directions were calculated on
the basis of the data from our own differential GPS measurements on the MBS
transects taken at the glacial surface and then were interpolated to form
a topography of the surface. For the austral summer 2010–2011, the drainage
basins were estimated to encompass glacier elevations between
80 and 450
The GMM run period was chosen for the 5-year period 22 November 2010 to
21 November 2015. The glaciological observations encompass the time period
November 2010–May 2016, but the high-frequency data acquisition does not
start before February 2012. For calibration of the GMM, the mass balance
observations at the transect with transect ID (TID) PG0 were used (see
Fig.
Cumulative mass balance (CMB) from simulations with the glacier melt
model (modelled CMB) and observation (observed CMB) on the Warszawa Icefield
(transect ID PG0
Cumulative mass balance (CMB) from simulations with the glacier melt
model (modelled CMB) and observation (observed CMB) on the Warszawa Icefield
(transect ID PG1
Figures
Figures
SY 2014 contained the coldest summer in the GMM run period,
with the highest amount of precipitation as well. In the beginning of
the ablation period, air temperatures were below freezing and snow
fall occurred together with high wind speeds. This leads to an
underestimation in the simulated CMB since the GMM
does not account for snow drift. Then there is a period when model and
observations are in very good agreement. During the fall of SY 2014,
there were frequent rain events. Since the GMM was run in catchment
configuration, meaning that the model area contained also the
periglacial parts of the hydrological catchment, the snow module
Apart from these two climatic boundary conditions, the model results and the glaciological observations are in good agreement, and a drift or disagreement over the 5-year GMM run period cannot be seen in the data.
Definition of beginning and end of the glaciological summer deducted
from observations of local minima and maxima of the accumulation/ablation
time series at the mass balance transects (PG0
The GMM outputs for the first glaciological year 2010 (meaning
2010/11) were not considered in the further analysis due to the model
spin-up for the first few months. Results for the stake locations
close to the glacier border, i.e. PG09 and PG19, were excluded in the
graphs, as already discussed (Sect.
Specific summer, winter and net mass balance (
Specific summer, winter and net mass balance (
The glaciological year 2011/12 contained a very cold and dry winter in
2011. The summer 2012 showed an exceptionally high net radiation
balance amounting to 156
Time series of meltwater discharge from GMM run November 2010 to
November 2015 for the Fourcade Glacier, catchment of the Potter Cove,
separated into the different source areas (AID) of snow, firn, ice and rock
terrain (
The GMM calculates the discharge on an hourly basis according to the
temporal resolution of the meteorological input time series. The GMM
configuration allows for computation of partly glaciated catchment
areas as is the case for the Potter Cove catchment. The GMM
differentiates between the source areas of the meltwater discharge:
Seasonal meltwater discharge from GMM run for the Fourcade Glacier, a hydrological catchment of the Potter Cove (DJF: austral summer December–February; MAM: austral fall March–May; JJA: austral winter June–August; SON: austral spring September–November).
Linearly relating the time series of simulated discharge to the
positive degree day (PDD) time series derived from the Carlini air
temperature series
Equilibrium line altitude calculated from observations at transects
PG0
The observed trend in SAM and surface air temperature especially during
wintertime
Equilibrium line altitude estimates from this study (PG) and former
glaciological studies during the last decades. ELA estimates taken from
literature encompass: BD is Bellingshausen Dome, KGI
(
Glacier extent for equilibrium with actual climatic boundary
conditions and an ELA of 260
Time series of accumulation and ablation show a very high intra- and
interannual variability that concurs with the climatological
variability that is reported by
The ELA is defined as a set of points on a glacier surface where the climatic
mass balance is zero
Figure
Until recently, the remote sensing data from synthetic aperture radar
(SAR) measurements only allowed for analysis during the summer months
due to seasonal manning of the Chilean base O'Higgins responsible for
the download of data from the overpassing satellite, thus only
resulting in the estimation of the transient snow line, in this case
the firn line. Hence, the ELA would be systematically underestimated
during a negative surface mass balance year. Additionally, SAR data
were not corrected for incidental angles, thus not allowing for
differentiation of ablation patterns from superimposed ice. The ELA
analysis here is based on ground-based glaciological studies only, and
remote sensing studies have been disregarded to avoid methodological
bias in the ELA analysis. Multi-temporal SAR data analysis carried out
by
Figure
The expected ratio of accumulation area to the total glacierized area
for a glacier in equilibrium with its climate (
One of the most intuitive parameters to describe the equilibrium state
of a glacier is the AAR, meaning the state
and health of a glacier. Particularly, it is an indicator for expected
future retreat or growth of a glacier until it gets into equilibrium
with concurrent climatic conditions. It thus relates via the ELA with
the specific mass balance and changes of total glacier area. The more
negative the mass balance, the higher the elevation line of ELA and
the smaller the AAR. Once the long-term ELA reaches values higher than
the maximum altitude of the glacier dome, there is no more
accumulation area and the glacier will disappear sooner or
later.
This behaviour also depends on the bedrock topography, which remains
largely unknown for the Warszawa Icefield. If underneath the glacier
is mountainous terrain, then the stratigraphy of glacier ice mass
would persist longer due to higher elevation than if it would all be
ice mass underneath. Glaciers respond to climatic boundary conditions
with a time lag of a few decades to centuries until they are in
equilibrium with the climatic conditions
A tipping point will be reached for the glacier when the ELA rises
above the highest point of the glacier which implies that the
accumulation area goes to zero, i.e. an
The high variability in melt conditions and in especially missing
winter accumulation is found to enhance glacial discharge. Positive
air temperatures and thus non-frozen moraine landscape surface also
promote the intake of sediment load into the meltwater streams that
is then introduced into the coastal waters. Assuming the most extreme
scenario of glacial retreat according to the ELA–AAR analysis
(Fig.
All R codes are available on request from Ulrike Falk.
Supplementary data are available at
UF was the PI of the glaciological and climatological work package within the IMCOAST project and also led the glaciological modelling work on KGI within the IMCONet project. DL was responsible for the quality assessment and control of mass balance data time series, the establishment of final glaciological protocols and all communication with the overwinterers. All final analysis and post-processing of climatological and glaciological data time series was performed by UF. ASB investigated the hydrology of Potter Cove and refined the catchment definition grids for the glaciological surface mass balance model. Glaciological modelling work was carried out by UF, with ASB contributing to the calibration and validation procedures. The manuscript was mainly written by UF. DL contributed to the sections concerning mass balance stake observations and analysis, as well as glacier mass balance. ASB wrote parts for the hydrological catchment definition and contributed to the sections on glacier discharge and meltwater analysis.
There are no potential conflicts of interest regarding financial, political or other matters.
We would like to thank the Alfred Wegener Institute (AWI) from Germany and
the Instituto Antártico Argentino – Dirección Nacional del Antártico
(IAA-DNA) from Argentina for their support in Antarctica. A special
acknowledgement goes to Hernán Sala for assistance in the field and help
with logistics and to the overwintering scientists at Carlini Station and
Dallmann Laboratory: Daniel Viqueira, Juan Piscicelli, Facundo Alvarez,
Francisco Ferrer, Pablo Saibene, Martín Gingins and Julia Luna without whom
this work would not have been possible. During the period 2010–2015, the
overwintering crews from the Ejército Argentino at Carlini Station
continuously supported our scientific tasks and we would like to specifically
mention SP Norberto Leonardo Galván. We also than Regine Hoch and the
University of Fairbanks in Alaska for the introduction to the glacier melt
model. All graphs in this paper were produced using the R programming
language