Introduction
In the light of future climate changes, it is essential to fully understand
past temperature fluctuations . Ice cores from the
Antarctic ice sheet have been used successfully for several decades in
palaeoclimatology to provide high-resolution records of past changes in snow
and ice isotopic composition, a key tool to reconstruct past local
temperature changes . A linear relationship
between the annual mean air temperature and the stable water isotope ratios
(D / H, 18O/16O) of the snow (and thus ice) at the drilling site has
been found . The ratio of water stable isotopes is usually given with respect to the Vienna Standard Mean Ocean
Water (VSMOW) in the δ-notation:
δ18O=18O16Osample18O16OSMOW-1‰.
However, has developed a theoretical framework showing
that the isotopic composition of precipitation does not only depend on
condensation temperature (Tc), but also on the distillation occurring
along the air mass trajectory, from the first evaporation from an oceanic
source to the final deposition at the ice core drilling site. The source
region of precipitation plays an important role, as it determines the
meteorological conditions during the first evaporation and the length and
position of the moisture transport path ;
the latter being determined by the large-scale synoptic situation.
found distinct differences in isotope ratios of fresh
snow samples at Neumayer Station, East Antarctica, depending on the moisture
origin. found a strong spatial variation of the
local isotope-temperature slope. They suggested that one reason for this is
the spatial variability in moisture origin.
Quantitative temperature reconstructions from ice core records have been
developed using combined water stable isotopes (D and 18O) in order to
assess past changes in moisture source characteristics (e.g. sea surface temperature,
SST; relative humidity, RH; and wind speed) from the second-order isotopic parameter deuterium excess
(d=δD -8×δ18O, ) and improve the Antarctic
temperature estimate . However, very few studies
have been performed on present-day Antarctic precipitation and the
relationships between its isotopic composition and moisture source and
atmospheric transport characteristics. For a better understanding of the
proxy data from ice cores, it is necessary to analyse the processes
determining the isotopic composition of the snow at the drilling locations
for recent time periods for which both meteorological and stable isotope
data are available.
The origin of precipitation is determined by the synoptic weather situations
producing snowfall. Snowfall associated with synoptic precipitation events is
mostly associated with warm events, which may be recorded in water stable
isotopes . While in the coastal areas precipitation is
caused by frontal activity in the circumpolar trough, different precipitation
mechanisms are at play on the high Antarctic Plateau. Generally, two main
types of precipitation can be distinguished here. The first is that of
diamond dust or clear-sky precipitation, which consists of very fine ice
needles and is formed by radiative cooling of a nearly saturated air mass.
The second is one in which distinctly higher amounts of snowfall occur due to
amplified Rossby waves that lead to advection of relatively warm and moist
air to the interior of the continent. This occurs several times per year.
When moist oceanic air is flowing southward, it is lifted by the steep slopes
of the Antarctic continent, thereby adiabatically cooled. When saturation is
reached, precipitation starts to fall and the absolute humidity of the air
decreases rapidly towards the plateau. It has been estimated that on the high
Antarctic Plateau approximately 25–60 % of the annual
precipitation originates from a few synoptic precipitation events
. The remaining
precipitation results from diamond dust. The spatial distribution of the
precipitation is highly influenced by the topography .
The present study aims at an improved understanding of the atmospheric
influences on the isotopic composition of Antarctic precipitation. Of
particular interest is the influence of the position of the moisture source
region and the conditions in this area and along the moisture transport
paths.
To reach this goal, precipitation conditions at the East Antarctic deep ice
core drilling site Dome Fuji were investigated using the unique data set of
daily precipitation and stable isotope measurements by .
The atmospheric flow conditions and synoptic situations causing precipitation
events at Dome Fuji were analysed with the help of Antarctic Mesoscale Prediction
System (AMPS) archive data. Moisture source areas for the precipitation
events were estimated with the combined information of back-trajectory
calculations and synoptic weather charts.
Furthermore, the ability of a simple Rayleigh-type isotopic model and an
isotopically enhanced general circulation model (GCM) to reproduce the
measured isotopic composition of precipitation was investigated. The
conditions at the estimated source regions were used as initial conditions
for the Rayleigh-type isotope model. Finally, the influence of the
precipitation origin and the general atmospheric flow on the stable isotopes
and, in particular, on the deuterium excess, were investigated using the
combined information from synoptic analysis, trajectory calculations,
observational data, and isotopic modelling.
Previous work
Synoptic analyses
In the past, several studies were conducted to investigate the synoptic
situations causing high-precipitation events on the plateau of Dronning Maud
Land (DML)
.
One of the first to study the intrusion of synoptic weather systems into the
interior of Antarctica was . He found that in December
1978 an intensifying ridge over DML and cyclogenesis in the Weddell Sea
caused record-high temperatures in the interior of Antarctica.
used reanalysis data from ECMWF to analyse
high-precipitation events in DML. They found that the most common weather
situation causing high-precipitation events develops through amplified long
waves with a cyclone in the Weddell Sea and an Atlantic block forcing
enhanced northerly advection. investigated synoptic
situations causing high precipitation rates at Kohnen using AMPS archive
data. Kohnen lies north-west of Dome Fuji and is around 1000 m lower in
altitude (Fig. ). Therefore, it is more influenced by the
coastal climate than Dome Fuji. distinguished
between five weather situations that cause high precipitation events: a
strong deep cyclone over/north of Kohnen, a blocking high east of Kohnen
Station with north-westerly flow, a blocking high east of Kohnen Station with
north-easterly flow, a (weak) ridge above Kohnen Station and an upper-air low
south of Kohnen.
At Dome Fuji, no investigation about the frequency of different weather
situations has been carried out so far. However, in several case studies,
blocking events that caused large amounts of precipitation were analysed
. The
investigated precipitation events were associated with a strong increase of
wind speed, a breakdown of the surface inversion, and a large temperature
increase.
Few studies have performed direct precipitation measurements in Antarctica.
already sampled fresh snow for isotope analysis in 1964/65
at the South Pole, without measuring precipitation amounts.
Additionally, they used radiosonde data to determine the lifting condensation
level to be able to relate the temperature at this level to the stable
isotope ratios of precipitation. At Dome C, daily precipitation samples have
been analysed since 2006 . The measurements include the
precipitation amount, analysis of water stable isotopes, and crystal structure
analysis.
Origin of precipitation
Since the origin of precipitation largely influences its stable isotope
ratios, it has been studied by various authors. The first studies combined
surface snow isotopic data, and theoretical Rayleigh distillation
calculations, and concluded that the origin of inland Antarctic precipitation
was mainly situated in the subtropics . However, more recent studies with trajectory models have found
source regions much further south .
determined an average source region between 50
and 60∘ S for precipitation at Antarctic ice core drilling sites.
used a trajectory model and ERA-40 reanalysis data to
investigate the transport towards Dome Fuji for the year 1997. They
distinguished between clear and snowy conditions using observations of
snowfall and clouds. The backward trajectories were calculated for five days
and arrived at 500 hPa. They found that during snowy conditions in
winter the travel time of five days was sufficient to identify the evaporation
area at the sea surface. In summer, the trajectories were often positioned
entirely above the continent and at greater height, thus not yielding
information about moisture origin. However, the authors did not distinguish
between clear-sky precipitation and synoptic precipitation, and the cloud
observations had large uncertainties. studied the
seasonality of moisture sources for Antarctic precipitation using Lagrangian
moisture source diagnostic. They traced water vapour transport
backwards in time for 20 days to estimate the precipitation origin for Antarctica and
found a source region for Dome Fuji at a mean latitude of 44∘ S.
They state that their results are consistent with findings from GCM with
tagged tracers. However, it is not trivial to understand the dynamics of the
calculated transport. The shorter trajectories in the present study were
cross-checked with the synoptic analysis.
AMPS topography of Antarctica and location of selected stations.
Stable isotopes
Commonly, the second-order isotopic parameter deuterium excess is used to
extract information about the moisture source region for precipitation in ice
cores . The deuterium excess
reflects the different behaviour of HD16O and H218O during
fractionation. On average, the impact of the equilibrium fractionation on
global precipitation is eight times higher for HD16O than for H218O.
The kinetic fractionation is nearly equal for both isotopes. The deuterium
excess is thus a way to quantify the kinetic fractionation
. Kinetic fractionation during evaporation is dependent on
the meteorological conditions at the time: SST, RH, and wind speed
. measured the isotopic composition of
moist air at 15 m altitude in the Southern Ocean from 30
to 65∘ S. They found an anticorrelation of d with RH at the
measurement height and a positive correlation with SST. They showed that RH
explains variations of d on short timescales.
recently performed in situ measurements of d in the marine boundary layer of
the North Atlantic. They could explain 84 % of the d variance by RH.
This relationship was dependent on SST. However, they could not find any
correlation of d with wind speed.
In palaeoclimatology, the deuterium excess from ice cores was used to extract
information about the conditions at the moisture source region
. Note that without additional information one
cannot distinguish between a change of the meteorological conditions at a
given moisture source and a change in the position of the moisture source. It
is also assumed that the signal of the initial deuterium excess is preserved
in final precipitation. However, during the moisture transport, additional
kinetic fractionation occurs in the clouds, mainly during deposition on ice
crystals. In mixed clouds the Bergeron–Findeisen effect induces kinetic
fractionation. It is caused by the difference in the saturation vapour
pressure over water and ice which leads to a net evaporation of liquid
droplets and a net deposition of vapour on ice crystals .
Apart from non-equilibrium processes, d is also influenced by the final
temperature during deposition: since the equilibrium fractionation
coefficients are temperature dependent, the slope between δD and
δ18O is smaller than eight for low temperatures. This leads to an
increase of d with decreasing temperature. Also the role of kinetic
fractionation at supersaturation on ice crystals could increase the
anticorrelation of d with temperature . The
relative importance of the processes determining d in Antarctic snow is not
understood sufficiently.
To reconstruct information about the temperature both at the deposition site
and at the source region, simple Rayleigh-type models are used. These
models consider the fractionation in an isolated air parcel using only
moisture source and condensation conditions as input. They do not include
dynamic processes or turbulent mixing. In a pure Rayleigh model it is assumed
that the condensate in the cloud is immediately removed by precipitation
after formation e.g.. In other simple models
e.g., the fraction of the condensate that is
precipitated can be chosen. A frequently used model is the mixed cloud
isotope model (MCIM) . It calculates the isotopic
fractionation of an isolated air parcel on the cooling path from the first
evaporation to the final deposition. It considers equilibrium and kinetic
fractionation processes. Several studies have tried to test the performance
of the model to reproduce the present-day isotopic composition of snow from
measurements . To include atmospheric
dynamics and find appropriate source conditions, some studies have combined
Rayleigh-type models with trajectory models. Trajectory models use
three-dimensional atmospheric fields to trace an air parcel back from the
precipitation site to the evaporation area. For example,
used MCIM and a trajectory model to investigate the
relationship between the seasonal cycle of deuterium excess and the
precipitation origin with snow samples from Neumayer Station. The model could
reproduce the annual cycle of δ18O and δD but the amplitude
was underestimated. They found on average a lower d for trajectories
originating from lower latitudes than for trajectories originating from the
Antarctic continent. combined a 5-day backward
trajectory model and MCIM to investigate isotope records from snow pits from
four locations in western DML. Instead of source conditions at the sea
surface, they used the monthly mean vapour isotopic composition from ECHAM4
at the start height of the trajectories above the ocean. The model was able
to simulate the seasonal cycle of the measured values, but at the low
temperatures in the Antarctic interior, the isotopic depletion was
underestimated.
Data
Precipitation and stable isotopes
were the first to perform direct precipitation
measurements and sampling for isotopic measurements in central Antarctica.
Daily snow samples were collected in plastic containers, placed on the
station roof. The observation height of 4 m minimized the measurement
error due to drifting snow. The precipitation amount and the stable isotope
ratios of the samples were measured from 3 February 2003 to 20 January 2004.
The samples were analysed in Japan with a mass spectrometer using the
CO2-H2O / H2-H2O equilibrium method. The accuracy of the measurements of
δ18O is 0.05 ‰ and of δD
0.5 ‰. The precision of d is 0.64 ‰. The advantage of direct precipitation
measurements compared to accumulation measurements is that alterations
through wind scouring and sublimation after the snowfall are reduced to a
minimum. Similarly, the stable isotope ratios of the precipitation samples
are not affected by post-depositional processes, such as exchange with the
atmosphere at the snow–air interface or diffusion within the snowpack/ice.
AWS data
The Antarctic Meteorological Research Center (AMRC) and Automatic Weather
Station (AWS) Program are sister projects of the University of
Wisconsin–Madison funded under the United States Antarctic Program (USAP).
These projects focus on data for Antarctic research support, providing
real-time and archived weather observations and satellite measurements as
well as supporting a network of automatic weather stations across Antarctica.
The current AWS at Dome Fuji, which measures the standard meteorological
variables of air temperature, pressure, wind speed, wind direction, and
humidity, was set up by the AMRC in February 1997. Data can be obtained from
the AMRC website (http://amrc.ssec.wisc.edu).
AMPS archive data
For the analysis of the weather situations and trajectory calculation, the
Antarctic Mesoscale Prediction System (AMPS) archive data
was used. AMPS has been created to provide numerical weather prediction
guidance for the forecasters of the USAP, but has also widely been used for
scientific studies . For
the 2003–2004 period analysed here, AMPS employed Polar MM5, a mesoscale
atmospheric model run at high resolution and adapted to the special
conditions of the polar environment. It is a version of the fifth-generation
Pennsylvania State University/NCAR Mesoscale Model, with modifications made
to better represent conditions in polar regions . These
include an improved representation of sea ice with fractional sea ice coverage
in grid cells and using the latent heat of sublimation for
calculating heat fluxes over ice surfaces. Other modified areas were revised
cloud–radiation interactions, ice phase microphysics, and an improved
treatment of heat transfer through snow and ice surfaces
. compared AMPS precipitation in
western DML to the surface mass balance derived from glaciological data for
the years 2001–2006. They found similar spatial patterns in both data sets
related to topography and prevailing wind systems. However, they noted that
on the high Antarctic Plateau, precipitation might be underestimated by the
model because the MM5 does not include the formation of diamond dust.
AMPS was run with a set of nested domains of varying extents and grid sizes.
In this study, the gridded model output used was that from the outermost
domain with a grid spacing of 90 km and its nest with a grid of
30 km.
Methods
Definition of precipitation events
To quantify the frequency of precipitation events and to compare measurements
and model results, a threshold was defined to distinguish precipitation
events from diamond dust. In most cases, the amount of precipitation is
considerably higher for dynamically caused events than for diamond dust
events. However, sometimes it is difficult to distinguish between a small
amount of synoptic precipitation and diamond dust .
Furthermore, the errors in both measured and modelled precipitation for
specific events can be large. In previous studies, various ways have been
found to determine a threshold for precipitation events, e.g.
2 mmw.e. at 75∘ S and 0∘ E for precipitation
events in DML , a threshold with which 50 % of the
total precipitation is counted as snowfall or a deviation
from the mean of one standard deviation . A reasonable way
to define precipitation events is to consider percentiles
. This accounts for the fact that precipitation does
not have a Gaussian distribution. Here, from the frequency distribution, the
90 % percentile was chosen as the threshold. For this definition it
was necessary that all precipitation events for which a synoptic origin was
visible on the weather charts, and which were accompanied by increased
temperature and wind speeds, exceeded the threshold.
Trajectory calculation
To estimate the source region of precipitation events, three-dimensional
backwards trajectories were calculated using the graphics software RIP
(read/interpolate/plot) and AMPS archive data. The
three-dimensional displacement of an air parcel is calculated using the
following iterative scheme and an iteration step Δt of 600 s:
Xn+1=X0+Δt2vX0,t+vXn,t+Δt.
Xn+1 is the (n+1)th iterative approximation of the position vector at
the time t+Δt. It is calculated from the position vector of the air
parcel at time t(X0), the wind vector at the position X0 and time
t (v(X0,t)) and the wind vector at the position of the previous
iteration and time t+Δt (v(Xn,t+Δt)).
Isotope modelling
Generally, two types of isotopic models are distinguished: Rayleigh-type
models (see above) and isotopic GCMs (e.g. ECHAM5-wiso,
), which include an explicit representation of water stable isotopes into a three dimensional atmospheric model. Equilibrium and
kinetic fractionation processes are calculated for each phase change
.
included kinetic fractionation processes in mixed clouds in
a Rayleigh-type model developed by , resulting in the
so-called mixed cloud isotope model (MCIM).
In contrast to the original Rayleigh-type model, an adjustable fraction of
the condensate is kept in the cloud. In a range of temperatures which is also
adjustable, the cloud can hold both liquid droplets and ice crystals. In this
temperature range, additional kinetic fractionation processes occur due to
the Bergeron–Findeisen process. The initial isotopic composition after
evaporation can be derived either from the assumption that evaporation and
precipitation are balanced on a global scale (closure equation), or it can be
taken from an isotopic GCM at the required height. found a
systematic bias when using the closure equation on a regional scale. But the
advantage of the closure equation over the GCM approach is that the
relationship between the source conditions and the calculated or observed
isotopes can be investigated directly. Therefore, this approach was chosen
here. As input, the closure equation uses SST, the mean sea level pressure,
RH calculated at SST, and wind speed at the evaporation area. The maximum
wind speed allowed in MCIM is 10 ms-1. In 30 % of cases
the wind speed in the estimated source regions was higher and had to be
adjusted. The initial conditions for the closure equation were taken from
ECMWF-Interim Reanalysis (ERA-Interim) at the estimated source regions of the
precipitation events. The required variables are available from ECMWF
(http://apps.ecmwf.int/data sets/data/interim-full-daily/). The arrival
conditions were taken from AMPS at the arrival height of the trajectories.
Furthermore, simulations with the isotopic GCM ECHAM5-wiso
were used. This model enhances the atmospheric GCM ECHAM5 by stable water
isotope diagnostics. A similar approach was already performed with the
predecessors ECHAM3 and ECHAM4 . In
ECHAM5, additionally to H2O16, water containing O18 and D has
been implemented in the water cycle. For each phase change, fractionation
processes are considered. compared the output of
ECHAM5-wiso with varying horizontal and vertical resolution to observational
data. They found a good agreement of the simulations with observational data
on a global scale. A higher horizontal and vertical resolution clearly
improved the results. For the Antarctic continent, they detected a warm bias
of the surface temperatures from ECHAM5. Therefore, the model precipitation
was less depleted of heavy isotopes than the observations there. In this
study, an ECHAM5-wiso simulation with a grid size of
1.125∘ × 1.1215∘ and 31 vertical levels for the
point of Dome F were used. The simulation was performed for the period 1979
to 2013, with prescribed values of ocean surface conditions (SST, sea ice
concentration), insolation, and atmospheric greenhouse gas concentrations.
Monthly SST and sea ice concentration were derived from the ERA-Interim data
set, and the dynamic–thermodynamic state of the atmosphere was nudged to
ERA-Interim data as well.
In this study, we will compare the Dome F precipitation isotopic
measurements
with two different types of model outputs. First, we will use atmospheric
back-trajectories combined with MCIM, a theoretical distillation model, in which a single moisture source is assumed. Second, we will use the outputs of
ECHAM5-wiso, a GCM equipped with water stable isotopes and nudged to
atmospheric reanalyses. The advantage of the first approach is to make best
use of regional atmospheric circulation data, albeit with simplifications for
isotopic processes. The advantage of the second approach is the physically
consistent framework of the global atmospheric model, albeit with the usual
caveats of model resolution and physical biases.
Results
Precipitation and temperature
The total precipitation in the measurement period from 3 February 2003 to
20 January 2004 was 27.5 mmw.e. The mean daily precipitation was
0.08 mmw.e. On only 24 days no precipitation was observed.
Figure a shows the time series of the direct precipitation
measurements. It shows several high precipitation events on single days or on
a few consecutive days and only low amounts of diamond dust during the rest
of the time. The highest observed daily precipitation is 2.1 mmw.e.
on 15 December 2003.
In Fig. a the daily AMPS precipitation is shown by the red dots.
The total modelled precipitation in the measurement period is
16.0 mm, considerably less than the measured precipitation. For many
days with large measured precipitation amounts, AMPS forecasts increased
precipitation on the same day or shifted by one day. However, the predicted
precipitation amount is mostly too low compared to the observations.
(a) Measurements and AMPS simulations of daily
precipitation in mm w.e. Green dashed line: 90 % percentiles from
measurements. Orange dashed line: 90 % percentile from AMPS.
(b) Daily measurements of temperature, (c) daily
measurements of δ18O, and (d) daily measurements of
deuterium excess. The open and filled circles in panels (b–d) mark
the days where the precipitation exceeded the 90 and 95 %
percentiles respectively. The 90 % percentile is defined as the
threshold for precipitation events.
(a) Histogram of the precipitation measurements in mm w.e.
with a distance of 0.015 mmw.e. between the bins; the two highest
values (1.4 and 2.1 mm w.e) are not shown. (b) Histogram of the
precipitation simulations from AMPS with the same distance of
0.015 mm between the bins.
The mean daily temperature ranges between -77.8 and -25.3 ∘C.
The average for the entire period is -54.7 ∘C. The measurements
show a high variability (Fig. b). Especially in winter and spring
time, several short-term temperature increases of more than 20 ∘C
were observed. These rapid temperature rises are mostly associated with an
increased amount of precipitation.
(a) Daily measurements of δ18O plotted against
2 m temperature; the measurements during precipitation events are
marked by red crosses. Linear fits for all measurements and for the days with
a measured precipitation event are represented by a green and a red line.
(b) Daily measurements of deuterium excess plotted against
δ18O.
Figure displays the frequency distribution of the measured and
modelled precipitation. On 194 out of 337 days, precipitation amounts less
than 0.03 mmw.e. were observed. The frequency of the values
decreases rapidly with increasing precipitation. The histogram of the model
precipitation differs from that of the measurements mainly in a higher
frequency of values between 0 and 0.015 mm and no model values for
precipitation amount to over 0.6 mm. The 90 % percentile,
which is the threshold for precipitation events, is marked in
Figs. a and . It has a value of 0.16 mmw.e.
for the measurements and 0.12 mm for AMPS precipitation. Following this definition,
34 days with event-type precipitation were detected in the observations and 31 days
in the AMPS simulations. Taking into account that one precipitation event can have a
duration of several days, this corresponds to 21 observed precipitation events and 19
events simulated by AMPS. With this threshold, 60 % of the measured and
43 % of the modelled amount of precipitation is caused by
precipitation events, while 57 % of the 21 measured precipitation events
can be assigned to a modelled precipitation event.
Stable isotopes
Figure c shows the time series of the daily measured
δ18O in precipitation. As expected from isotopic distillation, the
course of δ18O mostly follows the temperature development
throughout the year. The mean δ18O is -61.3 ‰ with a
minimum of -81.9 ‰ in August and a maximum of -33.0 ‰
in December. The annual cycle of d (Fig. d) is anticorrelated
to δ18O and temperature with a maximum in winter and low values in
summer. The deuterium excess varies between -52.6 and 66.9 ‰ and
has a mean value of 17.4 ‰ in the measurement period.
In Fig. , the days with an observed precipitation event are marked
by black circles. Additionally, to distinguish between larger and smaller
precipitation events, days where the precipitation is higher than the
95 % percentile are marked with black dots. δ18O often has
a local maximum during precipitation events. It mostly corresponds to a local
temperature maximum at the same time. On average, δ18O is
4 ‰ higher on days with a precipitation event than for days with
diamond dust. The behaviour of deuterium excess is only uniform in winter,
where from June to August a local minimum is observed during most precipitation events.
Example of an amplified high-pressure ridge:
(a) 500 hPa geopotential height from AMPS, 1 August 2003
12:00 UTC, (b) 6 h precipitation in mm from
AMPS, 1 August 2003 06:00 UTC.
Figure a displays the scatter plot of δ18O vs.
temperature. The green crosses represent all daily measurements throughout
the measurement period. The dark green line is the associated linear fit with
δ18O =0.76±0.02×T-19.1±1.3 as regression
equation and a correlation coefficient R of 0.88. The δ18O–T
slope is close to the average Rayleigh distillation slope of 0.8. The red
crosses show the measurements from days with precipitation events. The
correlation coefficient of 0.93 for these days is higher than for the whole
data set. The corresponding linear equation is δ18O =0.69±0.05×T-23.2±2.6, plotted in Fig. a as a dark red
line. It has a slightly lower slope than the regression for the whole period.
The relationship between deuterium excess and δ18O is plotted in
Fig. b. The anticorrelation of the two parameters is clearly
visible. The slope between d and δ18O increases for lower
δ18O values, which correspond to lower temperatures. For
precipitation events, the range of deuterium excess is smaller than for all
measurements.
Analysis of synoptic situations causing precipitation events
The synoptic situations of the precipitation events were divided into five
categories using a manual classification scheme based on the 500 hPa
geopotential height fields from the AMPS archive. For each category, one
example is shown.
Amplified high-pressure ridge
In this situation, Dome Fuji is situated underneath an extended high-pressure
ridge with strongly amplified planetary waves north-west of Dome F. An example
event occurred from 1 to 2 August 2003. Figure a shows the
500 hPa geopotential height for 1 August 12:00 UTC. A trough
in the Atlantic Ocean with large meridional extent leads to strong advection
from the southern midlatitudes (north of 50∘ S) towards DML. East
of the trough, an extended ridge expands towards the south over the entire
western half of East Antarctica. The strong northerly flow leads to the
advection of relatively warm and moist oceanic air masses onto the Antarctic
Plateau. During the ascent of the air onto the plateau, it is cooled
adiabatically, and orographic precipitation forms. The largest precipitation
amounts are found north of Dome Fuji on the slope of the plateau, as can be
seen in Fig. b. When the air-mass reaches the high plateau, most
of the moisture has already fallen out and the amount of precipitation
considerably decreases towards Dome F.
Example of a weak high-pressure ridge: (a) 500 hPa
geopotential height from AMPS, 13 August 2003 12:00 UTC,
(b) 6 h precipitation in mm from AMPS, 14 August 2003
00:00 UTC.
Example of a blocking high: (a) 500 hPa
geopotential height from AMPS, 3 November 2003 00:00 UTC,
(b) 6 h precipitation in mm from AMPS, 3 November
2003 00:00 UTC.
Weak high-pressure ridge
The second situation is similar to the previous case, but with less strongly
amplified waves. A weak high-pressure ridge stretches over Dome F. The
moisture origin of the precipitation thus lies south of 50∘ S. On
14 August 2003, an event like this occurred. The 500 hPa geopotential
height for 13 August shows – as in the previous case – a ridge progressing
southwards on the Antarctic Plateau and causing northerly advection from the
coast towards the interior of the continent (Fig. a). In
contrast to the previous case, the waves are not as strongly amplified. The
source region of the precipitation in this case is the Southern Ocean.
Blocking high
In this scenario, a stable high-pressure system remains
above Dome F for several days. It thus induces stronger advection of relatively warm and moist
air than in the situations described before. An intense blocking situation
occurred from 31 October to 6 November 2003. Figure a shows the
situation on 3 November, when the most precipitation was measured. A
stationary ridge stretches over large parts of East Antarctica towards the
South Pole. Between the ridge and the trough west of it a strong northerly
advection from the midlatitudes takes place. In the following days, an
upper-level high is cut off and stays at this position until 6 November. The
advance of the ridge is accompanied by a strong warm air intrusion onto the
continent. From 30 October to 2 November, the upper-air temperature increases
by 18 ∘C above Dome Fuji. The warm air mass has a relatively high
absolute humidity. This causes high precipitation on 3 November on the slope
of the Antarctic Plateau that also reaches Dome F (Fig. b).
Example for the event Southerly Flow; (a) 500 hPa
geopotential height from AMPS, 1 October 2003 00:00 UTC,
(b) 6 h precipitation in mm from AMPS, 1 October 2003
00:00 UTC.
Southerly flow
In this situation, Dome Fuji is situated between an upper-level low to the
east and a ridge to the west. West of Dome Fuji, the air is advected from the
north, then moves anticyclonically around the ridge, thus reaching Dome Fuji
from a southerly direction. Because of the long transport path over the
continent, this weather situation does not always cause precipitation at
Dome F. Although it is a rather infrequent weather condition, it is counted
as a separate category because of its distinctive characteristics. An example
occurs on 1 October 2003 (Fig. a). In this case, the forecast and
the observed precipitation amount do not match well. AMPS simulates
precipitation only in the western part of the continent where the terrain
rises slowly to the plateau (Fig. b). The measurements, however,
show a precipitation event of 0.24 mmw.e. on 1 October. It is
possible that the amount of transported moisture was underestimated by the model.
Previous precipitation event
The last type of precipitation event differs from the other categories in
that the synoptic situation responsible for transporting the moisture onto
the Antarctic Plateau has changed by the day of the high precipitation.
Relatively moist and warm air from an event that took place on a previous day
in the vicinity is advected towards Dome F. When the air starts to cool over
the continent, new precipitation forms. At the time when precipitation is
observed at Dome F, no distinct synoptic situation causing it is visible any
more. However moisture and temperature are often still elevated. If the lower
layer of the air contains a high amount of moisture and is cooled to
dew point, hoar frost can also develop, which is then counted as precipitation.
Large amounts of precipitation can result from these mechanisms. AMPS often
underestimated the precipitation amount during this kind of event.
The event of 14–16 December 2003 provides an example. On 14 December, a
cut-off high is situated over Dome F (Fig. a). It was cut off from
a ridge stretching from the north-east onto the Antarctic Plateau. Dome F is
positioned west of the centre of the cut-off high in a north-easterly flow.
Extremely high snowfall was observed with 2.1 mmw.e. on 15 December
and 1.4 mmw.e. on 16 December – the highest values observed in the
whole measurement period. AMPS simulates patches of precipitation around
Dome F (Fig. b). The precipitation is probably caused by local
orographic lifting and radiative cooling.
Example for the event type Previous Precipitation Event;
(a) 500 hPa geopotential height from AMPS, 14 December 2003
00:00 UTC, (b) 6 h precipitation in mm from
AMPS, 14 December 2003 00:00 UTC.
Frequency distribution of the synoptic patterns for days with
precipitation events in % for (a) the measured
precipitation events and (b) the precipitation events from AMPS.
Frequency distribution of the weather situations
The frequency distribution of the described synoptic patterns for days with a
measured precipitation event is shown in Fig. a. Note that due to
the brevity of the measuring period, our results have no climatological
representativity. On approximately half of the days, precipitation is caused
by an upper-level ridge with either weakly or strongly amplified waves. The
“previous precipitation event” classification is found on approximately one
quarter of the days. Both the “blocking” and “southerly flow” patterns
caused precipitation only once each in the measurement period. For the model
precipitation events (Fig. b) the fractions of the “previous
precipitation event” and the “shallow ridge” situations are smaller than
for the observed events whereas the fractions of “blocking” and “southerly
flow” are larger than for the measured events.
Moisture source regions
In Fig. , the horizontal and vertical course of the trajectories
for the precipitation events are shown for the three arrival levels: 600,
500, and 300 hPa. Ideally, the arrival level should be around the
lifting condensation level determined by radiosonde data. Unfortunately, no
upper-air measurements are available for the study period. Thus the
calculations were done for the mentioned standard levels. The trajectories
with 600 hPa arrival level were found not to be representative for
the general flow because the surface pressure exceeds 600 hPa
33 % of the time in the considered period. During the rest of the
time, they are strongly influenced by the near surface and thus often stay
above the continent for the entire five days of the calculation. For the
300 hPa arrival level, the back-trajectories often end far above the
sea surface. Furthermore, the absolute humidity is extremely low at this
level (on average 0.025 gkg-1 compared to
0.21 gkg-1 at 500 hPa). In contrast to that, many
trajectories that arrive at 500 hPa reach a level close to the
surface after five days. Furthermore, of the three compared levels, the
500 hPa level has on average the highest temperature. Therefore it is
assumed that the flow around this level represents the main moisture
transport. Similarly, the temperature at the 500 hPa level was
assumed to be representative of Tc for isotopic calculations,
which is an even stronger simplification than the choice of the main
transport level.
The end point of the 5-day back-trajectory was not automatically assumed to
be the moisture source area: each trajectory was cross-checked with the
synoptic analysis, and the moisture source was estimated using the combined
information of trajectory and general atmospheric flow conditions. The
height of the trajectory end point, which was not always at the ocean
surface, was also taken into account. Pathway and transport time were adjusted
according to the available information. The source regions were defined by
latitude and longitude ranges. Figure shows the resulting
estimated source regions. The higher the assumed condensation level is, the
further the spread of the source regions is, due to the increase of wind
speed with altitude.
Horizontal course of the available trajectories for days with either
measured or modelled precipitation events with arrival times at 00:00 and
12:00 UTC arriving at (a) 600 hPa,
(b) 500 hPa and (c) 300 hPa. The colours
show the pressure level in hPa. The red dot marks the location of Dome F.
Estimated source regions of the modelled and measured precipitation
events for the available trajectories arriving at
(a) 600 hPa, (b) 500 hPa, and
(c) 300 hPa. The stronger the colour, the more frequently
the respective source area was found.
Isotopic modelling
In this section, the skill of MCIM and ECHAM5-wiso in reproducing the
measured isotopic composition at Dome F during precipitation events is
tested. The model parameters of MCIM were tuned to the measured isotopic
composition of precipitation. For this, the evaporation and deposition
conditions determined from the 500 hPa arrival level were used. The
result is shown in Fig. together with the measurements from the
fresh snow samples. The parameters could not be adjusted to reproduce the
measurements perfectly. MCIM simulates on average not enough fractionation
along the transport path. The annual cycle of δ18O is reproduced
correctly, but it has a smaller amplitude than the measurements and a
systematic offset of on average 11.6 ‰. It is largest in winter
(Fig. a). The annual cycle of deuterium excess is reproduced well
by MCIM as well (Fig. b), but it has on average a negative offset
of 7.7 ‰. The root mean square of the deviation from observations is
13.5 ‰ for d and 12.3 ‰ for δ18O. Deriving the
initial isotopic composition of the vapour from the ECHAM5-wiso
three-dimensional vapour isotopic composition field at the end point of the
back-trajectory did not improve the agreement of model and observations.
The output of ECHAM5-wiso for Dome F is shown in Fig. with green
circles. While both the modelled δ18O and d generally have a
smaller offset from the measurements in ECHAM5-wiso than in the MCIM
simulations (3.1 ‰ for δ18O and 3.8 ‰ for d),
the annual cycle is captured better by ECHAM5-wiso only for δ18O.
ECHAM5-wiso shows no clear annual cycle of deuterium excess. This corresponds
to a relatively high root mean square of the deviation from the measurements
for d of 10.4 ‰, whereas it is only 5.6 ‰ for
δ18O. Figure a shows the modelled δ18O plotted
against the measurements. δ18O from ECHAM5-wiso correlates with a
correlation coefficient of R=0.85 with the measurements, MCIM with 0.64.
The correlation of d with the measurements is higher for MCIM (R=0.44)
than for ECHAM5-wiso (R=0.21).
There is no evidence that the reason for the inability of ECHAM5-wiso to
reproduce the annual cycle of d is due to a bias in the modelled local
meteorological conditions at Dome F. The 2 m temperature during
precipitation events is on average only 0.7 ∘C higher than the
observations and the timing of the precipitation events is reproduced
reasonably well by ECHAM5-wiso.
Isotopic composition of precipitation for the days with a measured
or modelled precipitation event (excluding days where no trajectories were
available) used from measurements, MCIM (with arrival level at
500 hPa) and ECHAM5-wiso: (a) δ18O and
(b) deuterium excess.
Influence of the weather situation and the precipitation origin on the deuterium excess
Because the weather situation has an impact on the precipitation origin, it
also influences the isotopic composition of the precipitation at Dome Fuji.
To investigate this, the measurements of δ18O and d during
different types of weather situations are compared. Of special interest is
the comparison of the measurements for amplified and shallow ridges. They
represent clearly different origins of precipitation, with source regions
further north for amplified ridges. The deuterium excess is supposed to
contain information about the source conditions. According to the previous
studies, from the higher source temperatures associated with amplified
ridges, a higher d than for a more southern source region would be expected.
Isotopic composition of precipitation from MCIM with arrival level
at 500 hPa (red circles) and ECHAM5-wiso for the same days (green
circles) plotted against the measurements. The black line marks the line of
equal values: (a) δ18O and (b) deuterium excess.
Average values (with standard deviations in brackets) of the
estimated source latitude and observations at Dome F of 2 m
temperature, the estimated SST, observed precipitation, deuterium excess, and
δ18O in precipitation for two different types of precipitation
events. N is the number of days on which the event occurs.
N
Lat (∘)
T2m (∘C)
SST (∘C)
Prec. (mmw.e.)
δ18O (‰)
d (‰)
Ampified ridge
8
-45 (7.4)
-44.3 (10.2)
8.0 (5.4)
0.54 (0.24)
-53.7 (8.3)
9.5 (10.4)
Shallow ridge
10
-57 (4.7)
-56.2 (9.4)
0.4 (1.8)
0.42 (0.30)
-62.9 (6.9)
21.0 (8.2)
Table shows the meteorological and isotopic measurements averaged
for days with a precipitation event resulting from a shallow ridge and from
an amplified ridge. For an amplified ridge the 2 m temperature at
Dome F is on average 12 ∘C higher and the estimated SST at the
moisture source 7.6 ∘C higher than for a shallow ridge. The reason
for the higher 2 m temperature and thus Tc for amplified
ridges is the advection of warmer air masses from lower latitudes. In spite
of the more northern moisture source, the δ18O for an amplified
ridge is on average 9.2 ‰ higher than for a shallow ridge because
the higher Tc for amplified ridges reduces the temperature
difference between moisture source area and final deposition site and thus
causes less fractionation along the transport path. Accordingly, the
deuterium excess is not, as expected, higher for an amplified ridge, but on
average 11.5 ‰ lower than for the shallow ridge situation since the
influence of the higher Tc on d is stronger than the influence
of the higher SST. This is due to the temperature dependence of the ratio
between the equilibrium fractionation coefficients of HD16O and
H218O.
Figure shows the measured d plotted against SST and RH in the
estimated source regions of the precipitation events. Contrary to the
literature, neither for SST nor RH can a relationship with d be detected. The correlation coefficient R between d and SST is only
-0.13 and between d and RH it is -0.14.
Discussion and conclusion
The influence of the synoptic situation and the precipitation origin on the
isotopic composition of precipitation at Dome Fuji has been investigated
using a 1-year set of observations. The synoptic situations causing
precipitation were analysed with help of AMPS archive data. For the
measurement period, 21 synoptically induced precipitation events were
identified that caused 60 % of the total annual precipitation,
whereas the remaining 40 % mainly stemmed from diamond dust. This
breakdown depends on the definition of “synoptic precipitation” for
which a 90 % percentile threshold was used here. Better results
might be achieved by crystal analysis of the precipitation samples in order
to distinguish between diamond dust, snowfall, and hoar frost, as has been
done for Dome C by .
Measured deuterium excess plotted against areal mean of
(a) SST and (b) RH calculated at SST from ERA-interim in
the estimated source regions for days with trajectories available.
Whereas precipitation in the coastal areas is related to frontal activity in
the circumpolar trough, synoptic precipitation events at Dome F were found to
be mostly induced by amplified planetary waves with an upper-level ridge
stretching onto the Antarctic Plateau, usually associated with a strong
northerly advection, which transports moisture towards Dome F.
The synoptic analysis was combined with 5-day back-trajectory calculations in
order to estimate possible moisture origins for the precipitation events. A
mean moisture source centred at approximately 55∘ S was determined.
Conditions at the thus-defined moisture source areas were used as input for a
simple isotopic model often used in temperature reconstructions from ice
cores, the MCIM, in order to simulate the stable isotope ratios of the
precipitation samples.
Both measured δ18O and deuterium excess of fresh snow samples
showed a clear annual cycle with a maximum of δ18O in summer and an
opposite cycle with a maximum in winter for the deuterium excess. Since
synoptic precipitation events are mostly accompanied by advection of
relatively warm air masses, Tc is significantly higher during these events
than on average. δ18O usually has a local maximum during
precipitation events that is associated with a local temperature maximum. A
corresponding local minimum in deuterium excess is found only in the winter
months from June to August. MCIM was able to represent the annual cycle of
the stable isotope ratios and of deuterium excess fairly well. The amount of
isotopic fractionation, however, was on average underestimated by the model.
While the output of an isotopically enhanced GCM (ECHAM5-wiso) for Dome F was
on average closer to the measured isotopic composition of the snow samples,
this model could not reproduce the annual cycle of deuterium excess.
In spite of the abundance of data, MCIM could not be tuned to simulate
sufficient isotopic fractionation along the moisture transport path. This
issue has also appeared in previous studies on the Antarctic plateau using
this model . MCIM is a semi-empirical model that
was developed in the 1980s and includes strong simplifications like the
assumption of an isolated air parcel and a single moisture source. Apart from
the simplifications of MCIM, possible reasons for this disagreement are
errors in the initial water vapour isotope ratio and the location and
meteorological characteristics of the moisture source region. Furthermore,
while the assumption that the flow at the 500 hPa level is
representative for the main moisture transport to Dome F is plausible, it is
a strong simplification to use the model temperature at this level as
condensation temperature. An error in the modelled temperature might add to
the error in the height of the condensation level. While it would be
desirable to have radiosonde data to determine the lifting condensation
level, for the measuring period no upper-air data are available for Dome F.
The correlation coefficient between the measured δ18O and
2 m temperature is higher for precipitation events than for the
entire measurement period (that includes diamond dust) because the transport
paths are mostly more clearly defined for synoptic precipitation events than
for diamond dust precipitation. Also, the surface temperature inversion is
usually weakened or removed during event-type precipitation by high wind
speeds and back radiation from clouds. This decreases the difference between
Tc and T2m and could cause a more direct relation
of the two temperatures and thus improve the correlation of 2 m
temperature and δ18O.
In spite of the given uncertainties in the moisture source estimation, we
conclude that the results of the present study do not support the common
assumption that the deuterium excess of the
initial water vapour is preserved along the transport path until the final
deposition at the ice core drilling site. We do not find evidence of a
relationship between moisture source conditions and deuterium excess for the
analysed precipitation events at Dome Fuji.
The second significant impact of the results of the present study on ice core
interpretation is that, contrary to the assumption used for decades in ice
core studies and references therein, a more
northern moisture source does not necessarily mean a larger temperature difference
between source area and deposition site and thus precipitation that is more
depleted in heavy isotopes and has a higher deuterium excess. The warm air
advection associated with the event-type precipitation discussed in this and
some previous studies increases the condensation temperature considerably,
thus decreasing the temperature difference between the moisture source area
and the deposition site, which means less fractionation than for colder
conditions.
It is not clear to what extent the results from precipitation events in a
1-year period can be transferred to longer time scales. However, the
physical mechanisms at play stay the same and it is important to understand
how changes in the general atmospheric circulation (e.g. during
glacial-interglacial transitions) change moisture sources and transport paths
for Antarctic precipitation and thus the stable isotope ratios measured in
the ice cores. This study therefore stresses the importance of Antarctic
precipitation monitoring data sets for longer time periods. They also offer
the advantage of not being affected by post-deposition processes associated
with snow metamorphism, therefore providing the best tool to assess the
meteorological controls on snowfall isotopic composition.
The fact that ECHAM5-wiso does not correctly simulate the variability of d
further challenges the use of GCMs equipped with isotopes for interpretation
of d for past periods.
Recently, it became possible to measure δ17O in addition to δ18O and
δD. The 17Oexcess is then calculated
from δ18O and δ17O. This variable is believed to preserve
information on RH at the source region and to be independent of SST
. Thus, it should be possible to disentangle the
information about temperature and relative humidity at the moisture source
area. However, emphasized that, like deuterium
excess, 17Oexcess is also highly influenced by kinetic
fractionation in supersaturation conditions. Thus, it may also be sensitive
to temperature, especially for low temperatures at the condensation site. For
a more exact quantitative climatic interpretation of stable isotope records
from ice cores, further detailed process studies with present-day data
combined with modelling on larger timescales are necessary to investigate
both the atmospheric and the post-depositional influences on stable isotope
ratios in Antarctic snow.
Data availability
The ERA-Interim data are available at
http://www.ecmwf.int/en/elibrary/8174-era-interim-archive-version-20.
For access a registration is necessary. AMPS archive data are available at
http://polarmet.osu.edu/AMPS/amps_mm5.html. The Automatic weather
station data can be accessed at
ftp://amrc.ssec.wisc.edu/pub/aws/antrdr/.