JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 111, D15109, doi:10.1029/2005JD006524, 2006
Modeling the isotopic composition of Antarctic snow
using backward trajectories: Simulation of snow pit
M. M. Helsen,1 R. S. W. van de Wal,1 M. R. van den Broeke,1 V. Masson-Delmotte,2
H. A. J. Meijer,3 M. P. Scheele,4 and M. Werner5
Received 26 July 2005; revised 13 February 2006; accepted 7 April 2006; published 11 August 2006.
[1] The quantitative interpretation of isotope records (d18O, dD, and d excess) in ice cores
can benefit from a comparison of observed meteorology with associated isotope variability.
For this reason we studied four isotope records from snow pits in western Dronning Maud
Land (DML), Antarctica, covering the period 1998–2001. Timing and magnitude of
snowfall events on these locations were monitored using sonic height rangers. For the
distinguished snowfall events we evaluated the isotopic composition of the moisture during
transport by combining backward trajectory calculations with isotopic modeling, using a
Rayleigh-type distillation model (MCIM). The initial isotope ratio of the moisture was
determined from monthly mean isotope fields from a general circulation model (ECHAM4).
The trajectory analysis showed that the southern Atlantic Ocean is the major moisture
source for precipitation in DML. Modeling results along the trajectories revealed that most of
the isotopic depletion occurred during the last day of the transport. Finally, a diffusion
model was applied to describe the diffusion in the firn layer such that the modeled isotopes
could be compared with the observed isotope records. The resulting modeled isotope profiles
were mostly in good agreement with the observed seasonal variability in the snow.
However, at low temperatures (especially on the Antarctic interior), magnitude of the total
distillation was underestimated. Regarding the d excess parameter, our results show a large
influence of advection height on the final value of d excess in precipitation. This in turn
points to the importance of the vertical structure of d excess over the oceanic source region,
which obscures the classical interpretation of this parameter in terms of temperature and
relative humidity in the moisture source region.
Citation: Helsen, M. M., R. S. W. van de Wal, M. R. van den Broeke, V. Masson-Delmotte, H. A. J. Meijer, M. P. Scheele, and
M. Werner (2006), Modeling the isotopic composition of Antarctic snow using backward trajectories: Simulation of snow pit records,
J. Geophys. Res., 111, D15109, doi:10.1029/2005JD006524.
1. Introduction
[2] Ice cores from polar regions contain a wealth of
paleoclimatic information [e.g., GRIP Members, 1993; Petit
et al., 1999; North Greenland Ice Core Project Members,
2004; EPICA Community Members, 2004]. The most important parameter from these ice cores used as a proxy for
temperature (T) changes is the isotopic composition of water
(d18O and/or dD, usually expressed in per mill as the
deviation from the Vienna Standard Mean Ocean Water,
VSMOW). The strong empirical relationship between mean
Institute for Marine and Atmospheric Research Utrecht, Utrecht
University, Utrecht, Netherlands.
Laboratoire des Sciences du Climat et de l’Environnement, UMR
CEA/CNRS 1572, L’Orme des Merisiers CEA, Saclay, Gif sur Yvette,
Centre for Isotope Research, Groningen, Netherlands.
Royal Netherlands Meteorological Institute, De Bilt, Netherlands.
Max-Planck-Institute for Biogeochemistry, Jena, Germany.
Copyright 2006 by the American Geophysical Union.
annual T and the isotopic composition d of snow is the basis
of the use of stable water isotopes as a climate proxy.
Furthermore, the second-order parameter deuterium excess
(d = dD 8*d18O, hereafter called d excess) contains
additional information about nonequilibrium fractionation
of moisture, as was first synthesized by Dansgaard [1964].
As such, d excess is often used to infer climatic information
from both the source region of the moisture and of the
precipitation site [e.g., Stenni et al., 2001; Cuffey and
Vimeux, 2001; Masson-Delmotte et al., 2003].
[3] Although water isotope ratios are regarded as a powerful tool to investigate paleoclimate [e.g., Jouzel et al., 1997,
2003], some controversy exists to what extent this proxy can
be used as a T indicator, since the d-T relation varies in space
and time [e.g., Robin, 1983; International Atomic Energy
Agency, 1992; Cuffey et al., 1995; Johnsen et al., 1995; Jouzel
et al., 1997, 2003; Masson-Delmotte et al., 2003; Landais et
al., 2004]. Moreover, in low accumulation areas, the shortterm correlation (i.e., over several years) between d and local
T is not strong enough to translate isotope records with
confidence into T changes [e.g., Helsen et al., 2005].
1 of 19
Figure 1. Schematic outline of the approach.
[4] Apart from local T, numerous other factors influence
the isotopic composition of precipitation, such as condensation temperature [e.g., Dansgaard, 1964; Aldaz and
Deutsch, 1967; Peel et al., 1988], changing conditions in
the water vapor source area [Boyle, 1997; Cuffey and
Vimeux, 2001], microphysical processes in clouds during
snow formation [Fisher, 1991; Ciais and Jouzel, 1994],
changes in magnitude of the ratio between advective and
turbulent transport [Kavanaugh and Cuffey, 2003], changes
in strength of the inversion layer [Van Lipzig et al., 2002],
seasonality in precipitation [Werner et al., 2000].
[5] To provide the d-T relationship a better physical basis,
attempts have been made to model the isotopic fractionation
of atmospheric water, either using Lagrangian parcel models
based on Rayleigh distillation or using general circulation
models (GCMs) equipped with water isotope tracers. The
GCM approach is very suitable for the validation of the d-T
relationship since it offers the possibility to take into
account all the relevant processes involved in the determination of the d value. Moreover, with increasing resolution
GCMs are increasingly capable to reproduce global and
regional patterns of isotope variability [e.g., Joussaume et
al., 1984; Hoffmann et al., 1998; Werner and Heimann,
2002; Noone and Simmonds, 2002b; Vuille et al., 2003].
However, the complexity of GCMs reduces their usefulness
in terms of interpretation. Besides that, GCMs in climate
mode generate their own internal variability of e.g., timing
of accumulation, which hampers direct comparison with
observed isotope records.
[6] On the other hand, Lagrangian parcel models describe
isotopic fractionation of moisture in an isolated air parcel,
from a single moisture source toward a precipitation site. In
these models, details of cloud processes are often parameterized, which enables them to account for the bulk effect of
cloud processes on kinetic isotopic fractionation [Jouzel and
Merlivat, 1984; Ciais and Jouzel, 1994]. Because of their
relative simplicity, these models are suitable to study the
influences of different processes along transport. Moreover,
the influence of nonequilibrium fractionation (both in the
moisture source region and along transport) on the d excess
is largely based on results from this type of Lagrangian
parcel models [Merlivat and Jouzel, 1979; Johnsen et al.,
1989; Ciais et al., 1995; Vimeux et al., 1999]. However,
these models are often poorly constrained by meteorological
data. In an attempt to improve this, Helsen et al. [2004]
combined back trajectory calculations with isotopic modeling using a Lagrangian parcel model, for a major snowfall
event in Dronning Maud Land (DML), Antarctica. They
revealed a general inconsistency in assumptions in this type
of model, regarding the (noncontinuous) saturation of air
masses that transport moisture to the polar regions. Furthermore, they showed that the moisture overlying the South
Atlantic Ocean was initially more depleted than expected
from local evaporative conditions.
[7] In the present study we aim to increase our knowledge
about the controls on polar isotopic variability, by extending
the approach of Helsen et al. [2004], combining isotopic
modeling with meteorological data, to be able to compare
modeling results with observed isotope records. Figure 1
gives a general outline of the followed approach. We use
4 years of automatic weather station (AWS) data to infer the
accumulation history of four different sites in western DML.
After establishing the accumulation history (section 2), we
use back trajectory calculations to trace the transport
history of the moisture (section 3). Then, we combine this
meteorological data with isotope distillation modeling,
using a modified Rayleigh-type fractionation model, which
is able to deal with the occurrence of undersaturation and
evaporative recharge (section 4). This gives us modeled
isotopic values of the detected snowfall events, which still
cannot readily be compared with the observed isotope
records from the snow pits, since postdepositional diffusion
has smoothed the original values of the individual events.
Therefore we simulate the diffusion process (Appendix A),
and the resulting modeled isotope records are compared
with observations in section 5. In section 6 a discussion
follows, and conclusions are drawn in section 7.
2. Study Area and Accumulation History
[8] The mass balance of western DML has been monitored for several years using AWSs [Reijmer and Van den
Broeke, 2003]. For the present study, we use results from
four AWSs situated along a transect connecting the coastal
area (AWS 4), with the high Antarctic plateau (AWS 9)
(Figure 2). The escarpment region (AWS 5 and 6) forms the
transition between these two areas. The accumulation is
relatively low in this area, and ranges from 74 kg m2 yr1
on the Antarctic plateau (AWS 9) to 393 kg m2 yr1 on the
ice shelf (AWS 4). Table 1 gives a summary of the
topographic and meteorological characteristics of these
[9] The AWSs are equipped with a sonic height ranger
(SHR) (Campbell SR50) that measured the local (change in)
surface height. This instrument enables us to retrieve the
exact timing of snowfall events [Helsen et al., 2005]. The
SHR record of the surface height change is plotted in
Figure 3a. The accumulation is discontinuous; sudden
increases of the snow surface are followed by longer periods
of slowly decreasing or constant surface height. The sudden
increases of the surface are most likely the result of
snowfall. However, a net deposition of drifting snow cannot
be ruled out. The surface slowly subsided during periods
without accumulation, as a result of settling of the snow and
2 of 19
hampers an event-based comparison of observed and modeled isotope records (see section 5).
[11] Figure 3b displays an overview of the timing and
magnitude of the preserved snowfall events. The intermittent character of the accumulation is apparent, resulting in
only few large snowfall events that account for the bulk of
the total accumulation. This has resulted in the exclusion of
entire seasons from the isotope records, which can hamper
the interpretation of isotopic variability on a seasonal scale.
At AWS 4, 5 and 6 the accumulation is highest in autumn
and winter (with 65 –75% of annual accumulation). AWS 9
shows a more evenly distribution of accumulation through
the year [Reijmer and Van den Broeke, 2003]. The relation
between local T and the isotope records is further discussed
by Helsen et al. [2005]. In the present study, we follow an
alternative approach, by explicitly modeling the isotopic
composition of each accumulation event.
3. Trajectories
Figure 2. Map of Western Dronning Maud Land, with the
locations of the automatic weather stations.
sublimation. Sudden decreases of the surface height are
most likely caused by wind erosion.
[10] From these records, accumulation events that are
preserved in the snow are distinguished following the
procedure described in Helsen et al. [2005]. The accumulation history of each site is hereby resolved in 12-hour
intervals. Because of compaction of the underlying snow,
the distance between a certain snow layer and the SHR
increases slowly through time. We corrected for this effect
using measured density profiles, to give a best estimate of
the depth of each layer at the end of the monitoring period.
This was done to facilitate a comparison with measured
isotope records (Table 2, see section 5). However, due to
unknown changes in the density profiles through time, the
final depth of each snowfall event is only an estimate, which
[12] In order to model the isotopic composition of snowfall, information is needed about the transport history of the
moisture. We obtained this information by using a trajectory
model, that calculates 5-day backward trajectories for air
parcels arriving at the locations of the AWSs. This trajectory
model was developed by the Royal Netherlands Meteorological Institute (KNMI [Scheele et al., 1996]), and it
computes the large-scale three-dimensional displacement
of an air parcel during a time step D t, using an iterative
Xnþ1 ¼ X0 þ
½vðX0 ; t Þ þ vðXn ; t þ Dt Þ
where X0 is the position vector of the air parcel at time t, Xn
is the nth iterative approximation of the position vector at
t + Dt, and v(X, t) is the three-dimensional wind vector at
position X and time t. The iteration time step Dt is 10 min.
The iteration stops when the horizontal distance between Xn
and Xn+1 is less than 300 m, and the relative vertical
(pressure) difference (pn+1 pn)/pn+1 is less than 104. For
comparison, the mean horizontal displacement in 10 min is
typically 4 – 5 km.
[13] As an input for the trajectory model we used the
European Centre for Medium-range Weather Forecasts
(ECMWF) reanalysis (ERA-40) data set. We used the
Table 1. AWS Topographic and Climate Characteristics, 1998 – 2001a
Start of observation
End of observation
Elevation, m above sea level
Surface slope, m km1
SSMB,b kg m2 yr1
Snow density, kg m3
Temperature, K
Relative humidity, %
Specific humidity, g kg1
10 m wind speed, m s1
22 Dec 1997
21 Dec 2001
7245.20S, 1529.90W
3 Feb 1998
2 Feb 2001
7306.30S, 1309.90W
15 Jan 1998
14 Jan 2002
7428.90S, 1131.00W
1 Jan 1998
31 Dec 2001
7500.20S, 000.40E
From Van den Broeke et al. [2004].
SSMB, specific surface mass balance.
3 of 19
Figure 3. (a) Surface height and (b) identified accumulation events from the AWSs over the period 1998 –2001.
Because of a malfunctioning SHR, the accumulation history
of AWS 4 could only be established until early 2001, with a
data gap in winter 2000.
analyzed fields, to obtain the most accurate estimate of the
‘‘true’’ state of the atmosphere. The spectral resolution of
this archive is T159 (corresponding to 125 km horizontal
spatial resolution and 60 vertical levels), but as input for the
trajectory model these data are gridded on a constant 1.0
resolution in the horizontal plane, 60 vertical levels, and
6-hourly data. Interpolation in space and time was therefore
necessary. The spatial interpolation of the trajectory model
is bilinear in the horizontal, and linear with log(p) in the
vertical. The interpolation in time is quadratic, but for
temperature and moisture this produced unrealistic variations, which made us apply a linear interpolation for these
[14] Uncertainties in the resulting trajectories can be
considerable. The choice of trajectory type, interpolation
schemes and spatial resolution of the wind fields introduces
an uncertainty in the order of 1000 km after 5-day backward
calculation [Stohl et al., 1995]. In reality, the error can be
even larger, due to the presence of convective systems (e.g.,
fronts, convective storms) or the vicinity of the earth
surface, which can cause the parcel to lose its identity.
These errors are difficult to quantify and not included in the
uncertainty estimate, so computed trajectories must be
interpreted with care.
[15] Apart from the above mentioned difficulties, an
additional problem that can influence the isotope modeling
is that a trajectory model only traces the advective pathway
of an air parcel. This is not necessarily equal to the pathway
of the moisture itself, since turbulent moisture fluxes are
neglected. Nevertheless, we apply the trajectory method
because we consider it the best available estimate of
transport histories for individual snowfall events.
[16] For the Antarctic region this trajectory model has
previously been used to define moisture source regions for
several deep drilling sites [Reijmer et al., 2002], whereas
Helsen et al. [2004] used the trajectory model in combination with an isotope model to investigate the isotopic
distillation of Antarctic moisture along transport.
[17] To capture the best estimate of the transport history
of the moisture that brings snowfall to the study sites, the
trajectories for air parcels should ideally be calculated from
arrival locations at the exact time and location of the snow
formation above the accumulation sites. For the timing of
the snowfall events, we used the accumulation records from
the SHRs (Figure 3). These records consist of accumulation
amounts in 12-hour intervals, which are bounded by 0600
and 1800 UT; if accumulation occurred in such an interval,
we defined the moment of snow fall either at 0000 or
1200 UT. To determine the vertical (pressure) level of snow
formation, vertical profiles of cloud water content (CWC)
from ERA-40 were considered. We defined the height of
snow formation as the level with maximum CWC. If
successive 12-hour intervals showed accumulation, we
defined the interval with the highest maximum CWC as
the interval with most snow fall, and in addition, this
moment is regarded to be the representative timing of snow
formation of this event. Then, 5-day backward trajectories
were calculated from five pressure levels (50, 25, 0, 25
and 50 hPa) centered around this pressure level with
maximum CWC.
[18] Figure 4 shows the weighted mean of the trajectories
calculated for all snowfall events at the AWSs. The magnitude of each accumulation event detected by the SHR
(Figure 3b) is used as a weighting factor (wi). Using wi,
means are calculated by averaging the positions of the
trajectories per time step over a sphere, resulting in a weighted
mean pathway of the moisture. The main picture for all four
AWSs shows a cyclonic transport path over the southern
Atlantic Ocean, which is in agreement with earlier studies of
moisture transport to DML [Noone et al., 1999; Reijmer et al.,
2002]. To indicate the variability of the air parcel positions,
covariance ellipses are plotted around the mean locations
Table 2. Characteristics of Samples Taken From Snow Pits in the Field Season 2001 – 2002
Sampling date
Depth of the pit, cm
d sampling interval, cm
r sampling interval, cm
15 m from AWS 4
25 Dec 2001
2 m from AWS 5
17 Dec 2001
under SHR
14 Jan 2002
under SHR
17 Jan 2002
4 of 19
Figure 4. Weighted mean 5-day backward trajectories (thick lines) and covariance ellipses for days
when snowfall occurred at (a) AWS 4, (b) AWS 5, (c) AWS 6, and (d) AWS 9. Covariance ellipses are
associated with air parcel locations at 5 days (light gray), 3 days (medium gray), and 1 day (dark gray)
before arrival at the study sites. Weighting is performed using the magnitude of snowfall events from
Figure 3.
at 1, 3, and 5 days before arrival (Figure 4). The size, shape,
and angle of these ellipses are based on the weighted
variances (s2x and s2y ) and covariances (sxy) of the position
vector components (x and y refer to distances on a sphere in
km in longitudinal and latitudinal direction, respectively), in
analogy with Kottmeier and Fay [1998],
wi ðxi
sxy ¼
wi ðyi
wi ðxi xÞðyi yÞ
1 i
using sx and sy as lengths for the major and minor axis of
the ellipse, respectively. The orientation of major axis of the
ellipse is determined by sxy/s2x .
[19] As expected, the trajectories converge from a large
variability at 5 days before arrival toward a confined area
1 day before they reach their final destination. Although the
pathways for the four study sites look quite similar, there is
a major difference in the air parcels’ position at 5 days
before arrival. While the trajectories for the site on the ice
shelf (AWS 4, Figure 4a) are located at 61S, at 5 days
before arrival, the trajectories for the Antarctic plateau
(AWS 9, Figure 4d) show a position at 54S at 5 days
before arrival.
[20] Generally, the air parcel positions are more northward for destinations higher on the Antarctic plateau. This
pattern reflects an increase in advection speed. One should
keep in mind that the trajectories shown here are weighted
with accumulation at the arrival location. The larger advection speed of the trajectories for the plateau compared to the
coastal sites can be attributed to the scarcity of cyclonic
activity (and associated snowfall) in the Antarctic interior.
Only intense cyclones penetrate onto the Antarctic plateau
and bring snowfall to this area. Hence a large snowfall event
will be associated with strong advection, which explains the
more distant origin. This is in agreement with findings of
Noone and Simmonds [2002a], who showed that an air
parcel requires a critical amount of kinetic energy to reach
the high Antarctic plateau. The difference in transport paths
can have implications for the source area of the moisture,
and its isotopic composition, especially regarding the d
excess parameter (discussed in section 6).
[21] The difference in intensity of advection is further
explained by Figure 5a, which shows the pressure level of
the air parcel during transport toward the arrival sites. The
5 of 19
Figure 5. Weighed mean values of (a) air pressure,
(b) temperature, and (c) specific and (d) relative humidity
along trajectories for the four study sites.
higher altitude of AWS 9 results in a higher starting level,
and this higher transport level is maintained during all 5 days
before arrival. The longer advection path of the trajectories
arriving at AWS 9 is obviously also associated with the
higher transport path.
[22] The temperature evolution along the trajectories is
shown in Figure 5b. During the major part of the transport
path, T remains constant, or increases slightly. Only during
the final stage of the transport (last day before arrival), T
decreases strongly. This is the case for all sites, but most
pronounced for AWS 9.
[23] The changes in specific humidity (q, Figure 5c)
during transport toward the arrival site are very similar to
the T pattern. Important to note here is that Figure 5 shows
average values; individual trajectories often show large
variation in height, T and q. Increases of q in the individual
trajectories (not shown) occur either between 5 and 1 day
before arrival, (implying moisture uptake), or the moisture
has already entered the air parcel more than 5 days before
arrival. The strong decrease of T in the final stage of
transport causes a similar drop in q. This decrease of T
and q during the approach toward the Antarctic continent is
of primary importance for the final isotopic composition of
the moisture (see section 4).
[24] Using the variation of q along the trajectories, we can
find the region of maximum moisture uptake for the four
study locations, by weighting the location of moisture
uptake with the amount of increase of q, multiplied by the
weighting factor for each trajectory (as in the procedure for
the calculation of the weighed mean trajectories). This gives
us mean values and covariance ellipses of the moisture
uptake locations, which can be interpreted as the moisture
source regions for the four study sites (Figure 6). The
dominant moisture source regions are located over the
South Atlantic Ocean, which is in line with earlier findings
of, e.g., Reijmer et al. [2002]. It should be kept in mind that
these source areas are determined using only moisture
increases during the last 5 days of transport. The percentages in the ellipses in Figure 6 indicate the portion of the
total moisture that has entered the air parcel during these last
5 days of transport. The remaining part of the moisture was
already present in the air parcel at 5 days before arrival,
which hampers an estimation of the source area of this
[25] The difference in trajectory length that was noticed in
Figure 4 has a consequence for the moisture source regions,
which is clearly visible in Figure 6. Snow that is transported
inland onto the Antarctic plateau has a more distant moisture source than coastal precipitation: the moisture source
for AWS 4 is located between 45S and 70S, while the
moisture source for AWS 9 is between 40S and 65S.
This is in agreement with theoretical considerations [Robin
and Johnsen, 1983] and different GCM studies [Werner et
al., 2001; Delaygue et al., 2000] that also point toward an
increasing contribution of low-latitude moisture toward the
Antarctic interior.
[26] Figure 5d presents values of relative humidity (RH).
The value of RH is indicative for the presence of saturated
conditions, which are crucial for the occurrence of condensation and associated isotopic fractionation. We note fairly
high values of RH for the trajectories of AWS 4 and 5, while
the trajectories for AWS 6 and 9 show a lower value over
the first part of transport, slowly increasing toward saturated
values. This indicates that air parcels bringing snowfall to
the coastal area of DML tend to experience more frequent
saturation during the last 5 days of transport than those that
travel to the more elevated interior. The latter air parcels
only show saturated conditions during the final phase of
transport (1 day). These differences in RH can have
consequences for the isotopic distillation process, since
we do not expect condensation (nor isotopic fractionation)
during undersaturated conditions. This matter will be
addressed in more detail in section 4.
4. Isotopic Distillation
[27] To simulate the isotopic fractionation process of
moisture in air parcels, we use the Mixed Cloud Isotopic
6 of 19
Figure 6. Covariance ellipses representing the location of moisture uptake along trajectories for
(a) AWS 4, (b) AWS 5, (c) AWS 6, and (d) AWS 9. The percentages in the ellipses show the portion of
the total moisture that has entered the trajectory during the last 5 days of transport.
Model (MCIM [Ciais and Jouzel, 1994]), which describes
isotopic changes in an isolated Lagrangian air parcel. In
principle, this model is derived from a Rayleigh distillation
model as described by Dansgaard [1964]. However,
improvements on the Rayleigh distillation theory (regarding
the kinetic fractionation effect) have been incorporated both
in the source area [Merlivat and Jouzel, 1979] and in the
description of snow formation at low temperatures [Jouzel
and Merlivat, 1984]. Furthermore, Ciais and Jouzel [1994]
introduced a description of the processes occurring in mixed
clouds, allowing the interaction of vapor, liquid and ice.
These improvements have resulted in the MCIM, which has
succeeded in reproducing the main characteristics of stable
isotope variability in the middle and high latitudes [e.g.,
Jouzel et al., 1997]. In addition, the MCIM has been used to
support the interpretation of the seasonal cycle at Law
Dome, Antarctica [Delmotte et al., 2000; Masson-Delmotte
et al., 2003].
[28] Difficulties arise when a simple isotope distillation
model like the MCIM is combined with a realistic meteorological transport history of an air parcel [Helsen et al.,
2004]. One of those problems concerns the definition of the
isotopic composition of the initial vapor, certainly when
dealing with air parcels at varying height. Furthermore, air
parcels do not experience continuous saturation during
transport, which is an assumption in simple isotope models.
Therefore the isotope model has been modified, which is
addressed hereafter.
4.1. Initial Conditions
[29] The initial isotopic composition of vapor (dv0) in a
Lagrangian air parcel is not easily estimated because the
vapor and the ocean surface are not in isotopic equilibrium.
More specifically, the value of dv0 differs from the isotopic
content of evaporating water (de) [Craig and Gordon,
1965]. However, many Rayleigh-type distillation models
assume dv0 and de to have the same value, using global-scale
balance considerations between evaporation and precipitation. This enables the calculation of dv0 from conditions at
the ocean surface [Merlivat and Jouzel, 1979] and this
approach is known as the global-scale closure equation.
However, Jouzel and Koster [1996] pointed out that,
although evaporation and precipitation are in balance on a
global scale, this might not be justified at the regional scale.
Therefore they recommend to use GCM-generated values
for the isotopic composition of the vapor (dv), from grid
boxes of the surface layer overlying the ocean (as by, e.g.,
Delmotte et al. [2000]).
[30] Moving toward a trajectory approach, the value of
dv0 cannot directly be taken from these GCM surface layer
fields, since air masses are typically located 1 – 3 km above
the ocean surface when the calculations start (Figure 5a).
This is well above the marine atmospheric boundary layer,
and the moisture in these air masses is expected to be more
depleted compared to the moisture in the surface layer, due
to earlier condensation cycles. Helsen et al. [2004] showed
that moisture approaching the Antarctic continent has in-
7 of 19
Figure 7. Mean monthly d18Ov values of water vapor in (left) January and (right) July from ECHAM4.
(a) and (b) The d18Ov distribution of the level closest to the surface and (c) and (d) vertical cross sections
along the 0E/W meridian. The grey area in the lower left corner represents the Antarctic continent.
deed a more depleted dv value than can be expected from
evaporation from the local ocean surface.
[31] Therefore we make use of three dimensional GCMgenerated isotope fields, from which the value of dv0 is
taken as a starting point for the isotope model. A 20-year
present-day climate run is performed with the Hamburg
atmospheric climate model ECHAM4 [Roeckner et al.,
1996], which includes tracers for the isotopic species d18O
and dD [Hoffmann et al., 1998; Werner and Heimann,
2002]. The spectral resolution is T106 (1.1 latitude 1.1 longitude) with 19 vertical levels. This run provides
monthly mean values of the isotopic composition of water
vapor for d18O and dD. These fields can be regarded as an
isotope climatology. With the present resolution, ECHAM4
has proven to yield a realistic spatial distribution of isotope
variability [Vuille et al., 2003].
[32] To sketch the character of the isotopic distribution in
these fields, Figures 7a and 7b show values of d18Ov of the
near-surface level in January and July, respectively. Over
the ocean, d18Ov values remain relatively constant, whereas
strongly depleted values are visible over the Antarctic
continent. Very low values (<80%) are found over central
Antarctica during winter, but whether these low values also
occur in reality remains uncertain, since this would require
strong, continuous condensation near the surface.
[33] Figures 7c and 7d show vertical slices from the
ECHAM4 isotope fields, along the 0 E/W meridian. The
large vertical gradient in d18Ov is apparent, and indicates
that the vertical position of the starting point of the
trajectory will be of great influence for the initial isotopic
value of the moisture. Furthermore, the moisture just above
the ocean surface is more depleted in July than in January,
which is probably due to a stronger near-surface temperature gradient or caused by an increased sea ice extent in
winter, that prohibits isotopic equilibration of the moisture
with the ocean water.
[34] From the cross section in July it becomes clear that
the anomalously low values over the Antarctic continent
only show up in a shallow layer covering the surface. For
our purpose, the ECHAM4 isotope fields are only used to
define the initial value of dv0 (mostly over the ocean), and
therefore these possibly spurious parts of the model data do
not affect our isotopic modeling results.
[35] Figure 8 shows d excess values of vapor in the
ECHAM4 climatology. For the winter situation (Figures 8b
and 8d) very high values are found over the Antarctic
continent. These high vapor d excess values also produce
anomalous high d excess values in central Antarctic precipitation, which are not in agreement with observed d excess
patterns. This points to problems in the parameterization of
8 of 19
Figure 8. Same as Figure 7, but for mean monthly d excess values of water vapor in (left) January and
(right) July from ECHAM4. (a) and (b) The d excess distribution of the level closest to the surface and
(c) and (d) vertical cross sections along the 0E/W meridian. The grey area in the lower left corner
represents the Antarctic continent.
kinetic fractionation effects occurring during snow formation in ECHAM4 [Werner et al., 2001]. Nevertheless, as
explained above, these data are not expected to affect our
results, since it is the isotopic composition at lower latitudes
that determines the initial isotopic value of the moisture in
the air parcels (Figure 4).
[36] However, the d excess variability over the ocean is
important for our purpose, since it determines the initial
d excess signature of the moisture in the air parcels. Over
the ocean surface, ECHAM4 produces a gradient in
d excess values of the vapor, with increasing values with
altitude. Since the initial altitude of the trajectories (5 days
before arrival) can vary considerably, the vertical gradient of
d excess obviously is of large influence on our modeling
results. This matter will be addressed in more detail in
section 6.
[37] It should be noted that the use of climatological mean
ECHAM4 isotope values as initial isotopic values can
introduce a significant error in our final modeling results.
Isotopic values in single storm events can show large
deviations from the mean state (in the order of 5 – 10%
[Hoffmann et al., 1998]). Especially in regions where
cyclonic disturbances occur only occasionally, the isotopic
composition during intense storms might deviate substan-
tially from the mean state. However, while there exist no
reanalyses of atmospheric isotope fields, climatological
values seem to be the best available estimate for defining
the initial isotopic composition.
4.2. Isotopic Recharge
[38] Apart from the above mentioned difficulties in determining the initial isotopic composition, realistic transport
histories often show subsequent increases in q before the
final rain-out toward the precipitation site. The trajectories
show that such increases in q occur over the southern
Atlantic Ocean (Figure 6). These increases in q are associated with isotopic recharge, i.e., the uptake of moisture with
a different isotopic composition. Since the MCIM addresses
the isotopic changes in an isolated air parcel, any increase of
q that exceeds the total amount of moisture in the air parcel
will result in model failure. To enable realistic description of
the isotopic evolution of air parcels, the MCIM has been
adapted to account for such increases of q, assuming that
this moisture is mixed into the air parcel from ambient air.
[39] Similarly as for the initial isotopic composition of the
vapor (dv0), we use the ECHAM4 isotope fields to determine the isotopic composition of the ambient water vapor
(dv,ECHAM). As soon as all available cloud water and/or ice
is reevaporated, the additional increase of q is ‘‘extracted’’
9 of 19
Figure 9. Output from the adapted MCIM showing the
effect of isotopic recharge with a constant isotopic
composition of the ambient moisture. (a) Imposed T (solid
line) and q (dashed line) history as a function of time and
(b) isotopic composition of the ambient moisture (dotted
line) and the resulting isotopic changes in vapor (dashed
line) and precipitation (solid line).
This is not in line with realistic moisture transport, as was
earlier noted by Helsen et al. [2004].
[42] In the ECMWF model (and many other GCMs),
cloud water is assumed to form when RH exceeds a
specified threshold [Tiedtke, 1993; Jakob, 1999]. In analogy
with these models, we included a parameter RHthreshold =
80% as an indicative value for the presence of clouds. The
RH value obtained from the trajectories can be considered
representative for grid average conditions but not necessarily on cloud microscale level. This in an important difference, since the MCIM calculates kinetic fractionation
effects as a function of the amount of supersaturation (i.e.,
RH > 100%). To overcome this discrepancy, we follow the
original MCIM in defining its a supersaturation value as a
function of T (using the parameterization of Petit et al.,
1991), to ensure realistic modeling results for d excess,
while we follow our trajectory data as an indicator for the
presence of clouds.
[43] The implication of the introduction of RHthreshold is
schematically explained in Figure 10. We impose a schematic T, p and q history on an air parcel (Figure 10a).
Initially, the air parcel is well below saturation (RH = 60%).
As T decreases and q is kept constant, RH increases. As
soon as RH > RHthreshold, condensation is assumed to occur,
and, although the RH value according to the imposed
humidity history are still below 100%, on a cloud microscale level (super-) saturated conditions are defined as a
function of T and p following the original tuning of the
MCIM described by Ciais and Jouzel [1994]. Isotopic
from ambient air, and the resulting dv(t) value is calculated
by mixing dv,ECHAM(t) with dv(t Dt). Using this approach,
a strong increase of q will result in a value of dv close to the
value of dv,ECHAM at that particular location, which is
schematically illustrated in Figure 9.
[40] Figure 9 is a schematic history of T and q, in which
an isotopic recharge occurs between 3 and 4 days before
arrival. In this case we kept the isotopic composition of the
ambient moisture constant (d18Ov,ECHAM = 20%, dotted
line in Figure 9b). Between 4 and 5 days before arrival,
fractionation occurs in the same way as in the original
MCIM. Then, between 3 and 4 days before arrival, T (solid
line in Figure 9a) and q (dashed line in Figure 9a) in the air
parcel increase, and the isotopic composition of the vapor
(d18Ov, dashed line in Figure 9b) increases as well. The
value of d18Ov approaches d18Ov,ECHAM but will never
entirely reach its value due to the presence of the depleted
vapor at 4 days before arrival. After this recharge, a drop in
T causes a fractionation comparable to the original model.
The resulting final precipitation is slightly more depleted
compared to the case without an earlier condensation cycle.
4.3. Fractionation Threshold
[41] As demonstrated in section 3 (in particular for AWS
6 and 9), on average, the trajectories only show saturated
conditions during the final stage of transport, and lower
values of RH during earlier stages of the transport. It is not
expected that moisture in an undersaturated air mass experiences isotopic fractionation. However, simple Rayleightype isotope models like the MCIM calculate the isotopic
distillation as a function of only T and p, assuming constant
saturation from the source regions to the deposition site.
Figure 10. Output from of the adapted MCIM showing
the effect of a threshold value of RH for isotopic
fractionation. (a) Schematic moisture changes along trajectory: specific (solid line) and relative (dashed line) humidity
as a function of T and (b) isotopic changes of the moisture
as a function of T. The dashed line represents vapor in the
adapted model, the solid line the resulting precipitation. The
dotted line is the vapor if the original MCIM is used, and
the dash-dotted line is the accompanying precipitation.
10 of 19
Figure 11. Mean isotopic distillation histories along trajectories for (a, b) AWS 4, (c, d) AWS 5, (e, f)
AWS 6, and (g, h) AWS 9. Weighted mean fractionation history (left) for d18O and (right) for the
d excess. Minimum and maximum values of the final isotopic value of the snow are plotted at the end
of each distillation history, to indicate the variability.
fractionation starts (Figure 10b), and from then on the same
gradient is followed as would occur using the original
MCIM. Clearly, the impact of this threshold is that the
new modeled isotopic fractionation entirely depends on the
difference between TRH=80% and Tfinal. In the original
MCIM, the air parcel was lifted until saturation was
reached, and then followed until the final point of the
trajectory. Hence the difference between the source area
(at the level of saturation) and the arrival location was the
main controlling parameter. The quantitative impact of
RHthreshold on the final results is different for each trajectory,
depending on the mutual behavior of T and RH. However,
this matter will be addressed in section 6.
4.4. Fractionation History
[44] Weighted mean isotope distillation histories are
shown in Figure 11 for the trajectories calculated for the
four study sites. As for the trajectory results presented in
Figure 5, weighting for the averaging are obtained from
accumulation amounts (Figure 3).
[45] The mean isotopic change along transport of both
d18O (Figure 11, left) and d excess (Figure 11, right) are
shown for both vapor (dashed lines) and precipitation (solid
lines). Also shown are the accompanying values of the
monthly mean isotope values of the vapor from the
ECHAM4 fields, at the location of the air parcels (dotted
lines). Minimum and maximum values of the final precipitation are also indicated with the gray bars.
[46] The general picture from the fractionation histories in
Figure 11 is that the modeled d18O value of the moisture is
relatively constant during most of the transport, whereas
during the last day of transport the major part of the
fractionation takes place. This is very well explained by
the changes in T and q (Figure 5) during the last day of
transport. The isotopic amplitude is only minor at AWS 5,
since most precipitation occurred in spring or autumn
[Helsen et al., 2005], while AWS 9 shows the largest
[47] It is interesting to compare modeled d18Ov and
ambient d18Ov,ECHAM from the ECHAM4 fields. Because
of the definition of the initial state, the values of
d18Ov,ECHAM and d18Ov are equal when the simulation starts.
From that moment onward, d18Ov,ECHAM slowly increases,
while d18Ov hardly changes, due to (on average) a stable
pattern of T, p and q. However, during the last day of
transport, d18Ov,ECHAM plunges to more negative values
than d18Ov for AWS 6 and 9.
[48 ] This can be understood, since the values of
d18Ov,ECHAM are climatic mean values of a certain latitude
and altitude. On the other hand, the values of d18Ov are
only calculated for snowfall events. In Antarctica, snowfall is generally associated with advection of much
11 of 19
Figure 12. Seasonal mean d18Ov distillation paths for
AWS 6 as a function of T. Solid line represents summer
(DJF) distillation, and dashed line represents winter (JJA).
Numbers indicate day before arrival.
warmer air than average [e.g., Noone et al., 1999; Helsen
et al., 2005]. A comparison of T values from both our
trajectories and in the ECHAM4 fields (not shown)
indeed points to a difference in T. Therefore the lower
gradient of d18Ov compared to d18Ov,ECHAM is associated
with a smaller drop in T during accumulation, compared
to the mean temperature distribution over the Antarctic
region, and hence explains why d18Ov is higher then
[49] Focussing on the seasonal amplitude, Figure 12
shows the seasonally averaged distillation of d18Ov as a
function of T for trajectories to AWS 6. The drop in T
(which is strongest during the last day of transport) produces a strong distillation of the moisture. The average d/T
gradient increases from 0.6% K1 around the freezing point
to 1.0% K1 at 253 K, which is in agreement with previous
results obtained with the MCIM [Ciais and Jouzel, 1994].
Between 5 and 3 days before arrival, increases in T are
accompanied by isotopic recharge, which causes a somewhat irregular pattern of the average values of T and d18Ov.
[50] Interestingly, looking at the difference in T over
which condensation occurs (DT) over the seasonally averaged trajectories, Figure 12 shows that DTJJA is hardly
larger than DTDJF. Since DT determines the magnitude of
the isotopic distillation, not much difference would be
expected in the final isotopic composition, notwithstanding
that there is a seasonal difference in the d18O value of the
initial moisture in the air parcel. This implies that the
seasonal amplitude of d18O can only to a small extent be
explained by seasonal changes in DT. To a much larger
extent it is a reflection of a more regional signal in d18O of
atmospheric moisture.
[51] We summarized the relative contribution of DT to the
total isotopic distillation in Table 3. The isotopic composition of the moisture that enters the air parcel appears to
make a large difference for AWS 4 and 6, whereas DT can
explain the bulk of the seasonal isotopic cycle for AWS 5
and 9. The large isotopic differences in this initial moisture
along trajectories to AWS 4 and 6 are not due to large shifts
in source areas, but can partly be attributed to seasonal
changes in the isotopic composition of vapor around Antarctica and also to differences in initial height of the air
parcels (which can be of large influence as explained in
sections 4.1 and 4.2).
[52] The changes in d excess along the trajectories are
shown in Figure 11 (right). In the first phase of transport,
when the air parcels are located over the ocean, a gradual
decrease can be observed in the d excess values of both
the moisture in our modeled air parcels as in the
ECHAM4 fields. In the last part of the transport, when
T strongly drops when the parcel approaches the Antarctic
continent, our modeled d excess values of the vapor do not
react strongly, whereas the ECHAM4 fields show strongly
increasing d excess values. The modeled d excess values
in the precipitation along the trajectories show a decrease
during the T drop in the final stage of transport.
[53] This drop of d excess in precipitation during the final
stage of transport of the air parcels appears to be in conflict
with the general pattern of increasing d excess values with
decreasing T over Antarctica, which is present in both
observations [Petit et al., 1991; Dahe et al., 1994] and in
model results [Hoffmann et al., 1998]. If we focus on the
general behavior of d excess in precipitation as a function of
T in our model (Figure 13, which shows results of a model
run using a linear decreasing T), it appears that d excess
slowly decreases down to T 250 K, and subsequently
increases at lower T. Furthermore, changes in p also play a
role in the exact shape of the d excess curve. On average,
our trajectories did not encounter low enough T to reach the
increasing branch of the d excess curve. This is the reason
that the d excess does not increase (yet) in the direction
of the Antarctic interior. We thus expect higher values of
d excess for trajectories toward more remote regions in
Table 3. Characteristics of Seasonally Averaged Distillation Pathsa
d18Op, %
d18ODT, %
d18Op, %
d18ODT, %
Dd18Otot, %
Dd18ODT, %
D d18Odv,ini, %
The d18Op denotes final isotopic composition of the snowfall, DT is the difference in T over which distillation has occurred, d18ODT is the amount of
distillation explained by DT. The DJF-JJA columns give the total seasonal isotopic difference in the snowfall, the amount of this difference explained by DT,
and the (remaining) difference explained by the initial isotopic composition of the moisture, respectively.
12 of 19
Figure 13. Modeled d excess of vapor (dashed line) and
precipitation (solid line) as a function of temperature along
an idealized trajectory. The d excess values of vapor from
ECHAM are plotted in gray against temperature in ECHAM
for all air parcel positions for AWS 9.
[54] An indication of higher d excess values with low T
can be found in the variability in d excess (shown as the
grey bars in Figure 11). Much more deviations from the
mean value on the positive side of the d excess range occur
compared to deviations on the negative side. This is most
evident for AWS 9 and the explanation for this is that events
leading to such high d excess values are associated with
lower T than average. Since these cold accumulation events
are typically associated with only a small amount of
snowfall, the weighted mean d excess is much lower than
would be expected from the minimum and maximum value.
[55] We can also compare our modeled values of d excess
in the vapor phase with ECHAM4. In Figure 13, mean
monthly d excess values of the vapor in ECHAM4 are
plotted in grey as a function of mean monthly temperatures
in ECHAM4 for pathways toward AWS 9. For the sake of
clarity, modeled d excess is plotted against trajectory T only
for an idealized temperature history. This comparison shows
that the two parameters generally behave in a comparable
way. Only at very low T, the d excess in ECHAM4 increases
more strongly than our modeled d excess. These results
point out that the apparent large difference in d excess in
ECHAM4 and in our modeling results in Figure 11 is
caused by T differences between the ECHAM4 climatology
(all conditions) on the one hand and the trajectories for
snowfall events on the other hand, during which warm
conditions prevail. The different behavior of d excess at
very low T can be attributed to a slightly lower degree
of supersaturation (Si) in ECHAM [Hoffmann et al.,
1998] (parameterized as a function of T as follows: Si =
1 0.003 T ) compared to Si applied here (Si = 1.02 0.0038T ). Schmidt et al. [2005] showed that higher values
of Si (such as used here) yields a more realistic d excess
pattern over Antarctica compared to the parameterization
generally used in GCMs.
5. Simulated Snow Pit Records
[56] To enable a comparison of modeled isotopic composition of precipitation with that measured in snow pits, the
diffusion in the firn is taken into account (see Appendix A).
By modeling both isotopic distillation during atmospheric
transport and postdepositional diffusion occurring in the firn
layer, we have obtained simulated isotope records that can
be compared with the isotope records, as sampled in the
Antarctic field season 2001– 2002 (Table 2). A comparison
of modeled and observed records can reveal the validity of
our modeling approach.
[57] Both modeled (solid lines) and observed records
(dashed lines) are shown in Figure 14. Furthermore, we
summarized our results in Table 4 in terms of mean values
and standard deviations (which can be interpreted as an
indicator for the seasonal amplitude). Table 4 also shows a
column with isotope values from ECHAM4. These values
result from an experiment using monthly mean isotope
values in precipitation from ECHAM4 (not shown in
Figure 14), instead of our model experiment using backward trajectories.
5.1. The D18O Records
[58] The results for the d18O records differ from site to
site. For the coastal site AWS 4 (Figure 14a) it was not
possible to reveal the entire accumulation record, due to
failure of the SHR, which prohibits a reconstruction of the
upper 1.5 m. The remaining part of the reconstructed record
resembles the observations reasonably, although the winter
values are not as depleted as the observations. This results
in a less negative mean value compared to the observations
(Table 4).
[59] Modeled and observed records are out of phase over
a large part of the record of AWS 5 (between 0.5 m and
1.5 m; Figure 14c). The most probable reason for this is
that for AWS 4 and 5, the snow pit samples were extracted
somewhat away from the AWSs (Table 2). This implies that
the SHR might have recorded a different accumulation
history than was sampled in the snow pit, since spatial
accumulation variability can be substantial at very short
distances due to sastrugi formation [Isaksson et al., 1996;
Frezzotti et al., 2002; King et al., 2004]. Nevertheless, both
mean value and seasonal amplitude of the record at AWS 5
are well reconstructed.
[60] The snow pit samples at AWS 6 and 9 were taken
exactly under the SHR (Table 2), which should lead to a
much better correspondence between modeled and observed
isotope profiles. For AWS 6, this is indeed the case: the
shape of the modeled d18O record at AWS 6 (Figure 14e)
agrees well with the measured record. However, over the
entire period, modeled d18O is shifted compared to the
observed record, i.e., fractionation along the trajectories is
underestimated by 5.3% on average (Table 4). This underestimation of the fractionation is larger in winter, since the
modeled winter minima are less pronounced than the
observed winter minima.
[61] Agreement is not so good for AWS 9. The mean
modeled d18O is not depleted enough (difference is 8.2% on
average), and although the samples were taken directly
under the SHR, the location of the seasonal maxima do
not coincide with the observed record. However, the magnitude of the seasonal amplitude does agree with the
observations. Possible reasons for the underestimation of
the isotopic distillation will be addressed in section 6.
[62] If we compare our model results to the records
obtained using monthly mean values from the ECHAM4
13 of 19
Figure 14. Modeled and observed isotope records at the four sites. Dashed lines represent the
observations, open circles are the modeled undiffused values and the solid lines show the modeled
isotope records after diffusion.
climatology (Table 4) it appears that our model results
match the mean observed records better for AWS 4 and 5,
while the ECHAM records are closer to the observed mean
values for AWS 6 and 9. However, looking at the seasonal
amplitude, we see that our approach is much more successful in reproducing the magnitude of the seasonal cycle
compared to ECHAM4, which may be due to the use of
monthly mean isotopic values from ECHAM4, instead of
event-based values.
[63] A robust measure for the comparison between model
results and observations is of course the correlation coefficient r. However, for a sound estimation of r, it is important
that modeled and observed isotope values correspond to the
same event. Because of uncertainties concerning the estimation of the depth of each event (see section 2), and taken
into account that the snow pits at AWS 4 and 5 are not
sampled directly under the SHR, a comparison of the
records only resulted in an acceptable correlation between
modeled and observed isotope records for AWS 6 (r =
0.86). At AWS 9, a combination of the low accumulation
(only 74 kg m2 yr1) and the (unknown) influence of
redistribution of snow by snow drift is probably the reason
for the low resemblance between the modeled and observed
record. In general, the error introduced by the use of
monthly mean ECHAM4 isotope fields as initial isotopic
values may be the reason of the discrepancy with the
observed isotope variability.
5.2. Deuterium Excess
[64] The precision of the mass spectrometer used to
determine the isotopic composition of the samples is in
the order of 0.1% in d18O and 2.0% in dD. Because of the
Table 4. Statistical Properties of Modeled and Observed Isotope
Observed, ECHAM, Modeled, Observed, ECHAM, Modeled,
14 of 19
d Excess
relatively high uncertainty in the resulting d excess parameter (2.2%), the raw d excess records were somewhat noisy.
We applied a three-point running average on d excess data
to smooth the records (dashed lines in Figure 14).
[65] In general, the modeled d excess variations are rather
close to the observed records. Only at AWS 5 a bias is
found from the observed record, but this can be explained
by the difference in sampled snow and monitored accumulation by the SHR, as mentioned above. Even at AWS 9,
where the modeled d18O record was not correctly reproduced, the d excess values are close to the observed record.
This means that dD deviates in a comparable way from the
observations as d18O. While for the mean d18O values our
approach agreed best with AWS 4 and 5, the mean d excess
is better reproduced for AWS 6 and 9. As for the d18O
records, the seasonal amplitude in d excess is successfully
reproduced (Table 4).
[66] A comparison with the d excess values resulting
from the ECHAM4 climatology reveals a large deviation
from the observed d excess value at AWS 9. Clearly,
ECHAM4 produces too high d excess values at low temperatures, which is not in line with our observations. Generally,
d excess records are deconvolved into a local forcing (site
temperature) and a forcing due to changes in the moisture
source area (sea surface temperature, RH, and wind speed)
[e.g., Stenni et al., 2001; Masson-Delmotte et al., 2004]. In
section 6 we will address to what extent d excess is
determined by the source area, compared to changes during
6. Discussion
[67] There is a tendency in our results obtained using the
MCIM to underestimate isotopic distillation at lower site
temperatures. This is most clearly shown at AWS 9
(Figure 14, where the difference between the mean modeled
and observed d18O value at this site is 8.2% (Table 4). To a
lesser extent, the observed d18O record at AWS 6 also shows
more depleted values than the modeled isotope record,
especially in winter. There are several possible explanations
for this underestimation.
[68] First, we imposed a RHthreshold value of 80% as an
indicator for condensation to occur. From Figure 5 it
appears that especially the trajectories to AWS 6 and 9
show lower RH values along a major part of the transport.
To investigate the influence of RHthreshold, we repeated the
MCIM runs for AWS 9 with RHthreshold = 60%. The
resulting mean modeled d18O value decreased only 2.1%
to 37.5%. This is only a minor improvement, especially
since it is not expected that condensation will occur in
reality when RH is only 60%. From Figure 5d it appears
that RH generally is higher than 80% during the last day of
transport, when the major drop in T occurs. This explains
the relatively small influence of the threshold value of RH
on the total depletion of the moisture. These results point
toward another reason for the large difference between
model results and observations at AWS 9.
[69] Second, the MCIM assumes that part of the water
that forms during condensation is kept in the cloud. This
slightly reduces the isotopic distillation compared to Rayleigh fractionation (i.e., all the condensed water is immediately removed from the cloud). Using pure Rayleigh
fractionation, the final d18Op value for AWS 9 decreases
by only 1.9%. This indicates that the choice of the distillation-type cannot explain the difference in observed and
modeled d18O values either.
[70] Third, the quality of the ERA-40 data set as an input
for the trajectory model can be a reason for the underestimation of the fractionation. ERA-40 is badly constrained by
observations in a data-sparse region like the high Antarctic
plateau. An underestimation of the T difference between
TRH=80% and Tc,final in ERA-40 could well explain the lack
of sufficient fractionation toward high elevation sites like
AWS 9. However, it is not possible to test this hypothesis,
since no independent information of TRH=80% is available.
We can only compare the final condensation temperature in
the trajectories (Tc,final) with results from the regional
climate model RACMO2/ANT [Reijmer et al., 2005]. This
model is forced by ERA-40 at its boundaries, but it
simulates the Antarctic climate at a much higher resolution
(55 km). We defined condensation temperature in the
regional climate model (Tc,RACMO) as the temperature with
maximum CWC above the location of the AWSs [Helsen et
al., 2005]. Since RACMO2/ANT has proven to yield more
realistic results for the Antarctic region than ERA-40
[Reijmer et al., 2005], we expected to find differences
between Tc,final and Tc,RACMO, but no substantial differences
were revealed. If Tc,RACMO is a good representation of the
real Tc over AWS 9, these findings reject our suggestion that
the quality of ERA-40 is the reason for the underestimation
of the distillation at low temperatures.
[71] There are some other possible explanations for the
underestimation of the depletion, which are more difficult to
verify. For instance, isotopic equilibration of snow and firn
with vapor from the boundary layer [Waddington et al.,
2002] or from deeper firn layers [Motoyama et al., 2005]
has not been accounted for in the diffusion model. Furthermore, the contribution of clear-sky precipitation to the total
accumulation is unknown, but it is believed to be an
important contribution to the total accumulation in the
Antarctic interior [Bromwich, 1988]. This type of precipitation has not been accounted for in our approach, since
trajectories are only calculated when condensed moisture is
detected above the surface. Clear-sky precipitation is likely
more depleted than cyclonic precipitation because it is
formed from residual vapor from previous storms, that
condensates at lower temperatures due to radiative cooling
of saturated air. On the other hand, as snowfall events at
AWS 9 are coupled to more intense cyclones compared to
more coastal sites, the isotopic composition of the moisture
within these cyclones may deviate more from the mean state
as within low-intense cyclones. Together with the fewer
number of large events in the Antarctic interior, this may
prohibit the use of climatologically monthly mean
ECHAM4 isotope fields as a starting value for the trajectory
study for this location. Finally, the influence of turbulent
mixing of moisture in the air masses with more depleted
ambient moisture has not been taken into account, since the
calculated trajectories just follow the air parcels and not
necessarily the moisture within them. This can be an
important additional source of more depleted moisture.
[72] There is a similarity between modeled and observed
d excess values (Table 4 and Figure 14), and this offers the
possibility to study the forcing behind the d excess signal.
15 of 19
We separated the modeled d excess results for AWS 6, into
two groups of high and low d excess values, using the local
mean value of 1.3% as separation. Weighted mean trajectories were calculated for both groups. It appears that all
relevant parameters (e.g., T, source area) are practically
equal for both groups. There was only a difference in
height: the trajectories resulting in high d excess values
were advected toward their arrival location at a higher
altitude compared to the low d excess trajectories. This is
illustrated in Figure 15b, in which the dotted line indicates
the mean pressure of the trajectories with a high d excess
value, and the dashed line represents the mean pressure of
the trajectories resulting in low d excess values. The
implication of this difference in height can be seen in
Figure 15a, which shows a contour plot of d excessv,ECHAM
values, along a section following the meridian over AWS 6.
A strong vertical gradient can be recognized. The effect of
this vertical gradient on our modeling results is obvious:
trajectories starting near the ocean surface start with vapor
with a much lower initial d excess value (from the
ECHAM4 fields) than trajectories that originate from a
height of several km. For the two groups considered, the
difference in d excess due to the initial value is 3.2%, which
is 60% of the final difference in d excess (5.3%).
[73] This remaining part of the final difference in d excess
between the two groups of trajectories (40%) can be
attributed to differences in kinetic fractionation along the
transport path, especially during the last day of transport.
The solid line in Figure 15c indicates Dd excess, and it
illustrates that for the two groups considered, the behavior
of d excess is rather similar along the major part of the
transport. The final drop of d excess during the last day of
transport (previously seen in Figure 11) is slightly stronger
for the trajectories with low d excess values (dashed line),
which explains the remaining 2.1% difference in the mean
value of d excess.
[74] This analysis shows that the ECHAM4 fields influence the d excess modeling results to a large extent.
Because of a general lack of measurements of atmospheric
d excess values, it remains unclear whether GCMs like
ECHAM4 produce realistic patterns of atmospheric d excess. A comparison of different GCM-generated d excess
fields (such as the Stable Water Isotope intercomparison
Group (SWING) (M. Werner, personal communication,
2005)) seems therefore valuable. Nevertheless, the similarity between modeled and observed d excess records from
this study is an indication that the vertical gradient in the d
excess values of the vapor in ECHAM4 is in line with
[75] The d excess parameter is often used to extract
(paleo-) climatic information of moisture source areas, i.e.,
RH and sea surface temperature (SST) [e.g., Johnsen et al.,
1989; Vimeux et al., 1999; Masson-Delmotte et al., 2003]. It
is clear that the relative imprint of the initial d excess value
compared to the final d excess value of the precipitation
differs depending on the distillation history. Unfortunately,
our approach does not provide information about evaporation conditions of the moisture. Consequently, we cannot
directly determine to which extent these conditions are
preserved along the trajectories. However, our results suggest a prominent influence of the vertical gradient in d
excess over the moisture source region. Noone and
Figure 15. (a) Mean d excess values of vapor for January
in ECHAM4, (b) mean pressure levels, and (c) modeled
d excess values of precipitation along trajectories calculated
for AWS 6. Dotted lines represent trajectories producing
high d excess values, dashed lines represents trajectories
producing low d excess values. Solid line indicates Dd
Simmonds [2004] pointed out that variations in sea ice
distribution can have a considerable effect on atmospheric
transport and ultimately on the d excess signal. In our
analysis however, no direct dependence of d excess on
sea ice fraction was identified along the trajectories.
However, it is likely that the vertical isotopic gradients
in d excess are influenced by the presence of sea ice,
through a reduced equilibration of atmospheric vapor with
evaporating water. Furthermore, a reduced sea ice cover will
reduce the amount of diabatic cooling over the pack ice,
which will result in a higher level of turbulent ascent of the
moisture [Noone and Simmonds, 2004]. Hence a strong
gradient of d excess (increasing d excess values with height)
will especially develop over sea ice, while open water will
suppress the development of such a gradient. This would
mean that the higher d excess values on a higher atmospheric level reflect a more distant (tropical?) moisture
source, while the low near-surface d excess values can be
attributed to evaporation at higher latitudes, which is in line
with the current interpretation of d excess in polar ice cores
[e.g., Vimeux et al., 1999]. However, it is questionable to
what extent RH or SST can leave an imprint in the vertical
gradient in d excess over the oceanic source area. Therefore
16 of 19
Figure A1. Evolution of the d18O record at AWS 6. Arrows indicate the vertical shifts of winter minima
in subsequent years.
more insights are needed in the dynamics behind the
vertical distribution of the initial d excess values of the
vapor over the ocean.
7. Conclusions
[76] Simple isotope distillation models have since long
been proved useful in terms of explaining observed spatial
gradients [e.g., Jouzel and Merlivat, 1984] and seasonal
cycles [e.g., Ciais et al., 1995]. In this paper we went
further in applying such a model on individual events.
[77] We used a combination of back trajectory calculations and isotopic modeling to simulate the isotopic composition of four snow pits in Dronning Maud Land,
Antarctica. The trajectories indicate a moisture source area
in the southern Atlantic Ocean. The major part of the
isotopic fractionation occurs during the last day of transport,
when the air masses experience cooling due to the orographic lift over the Antarctic continent. The strength of this
final T drop is of primary importance to both the d18O and
the d excess value of the final precipitation. However, the
seasonal amplitude can only partly be explained by the
magnitude of this T drop. Seasonal variations of the isotopic
composition of the initial moisture are also important for the
final isotopic composition of the snow.
[78] Although we are aware that the use of climatological
isotope values (as modeled by the ECHAM4 GCM) as the
initial state is a possible major source of error, the resemblance between modeled and observed isotope variations is
fairly good, from which we conclude that this approach is
able to realistically simulate the isotopic fractionation.
However, for the high Antarctic plateau the isotopic distillation is underestimated. The reasons for this underestimation remain unclear. A forthcoming study will further
address the spatial and temporal patterns of both d18O and
d excess in Antarctic precipitation.
Appendix A
[79] Isotopic diffusion in firn is taken into account by
using the firn diffusion model as described by Johnsen et al.
[2000]. This model is able to realistically describe the
effective smoothing of isotope records through time. For
a full explanation of the theory we refer to their work,
here we briefly explain how this model was included in
our analyses.
[80] Considering a coordinate system with a vertical z axis
and an origin at the surface, the isotopic composition di of
nondeforming firn changes over time t according to
@di mes wai 1
RT ai t rf rice
@ 2 di
where m is the molar weight of water, es is the saturation
vapor pressure over ice, wai is the diffusivity of the isotopic
species i in open air, R is the universal gas constant, T
temperature, ai is the ice-vapor equilibrium fractionation
factor for the isotopic species i, t is the tortuosity factor, rf
is the density of the firn, and rice is the density of ice. Using
equation (A1), the diffusion rate is especially sensitive for T
and rf. We obtained subsurface values of T from energy
balance calculations by Van den Broeke et al. [2005].
During sampling, rf has been measured. We used a fitted
power law function for rf at each site, and assumed that this
relation is constant over time. For all other parameters in
equation (A1) we used expressions as suggested by Johnsen
et al. [2000].
[81] To minimize the effects of boundary conditions, we
added one artificial sinusoidal seasonal cycle and started the
diffusion model at 1 January 1998. Each day when accumulation occurred according to the SHR record, this snow is
added on top of the diffusion domain, and the origin of the
z axis is shifted to the surface. We account for the
densification of the snow by calculating new density values
for the modeled domain each time a shift of the z axis
occurs (i.e., after each accumulation event).
[82] In Figure A1, modeled growth of the d18O record at
AWS 6 is shown. The d18O record is plotted for each
modeled year, starting with the prescribed sinusoidal seasonal cycle at 1 January 1998, followed by four subsequent
years of accumulation and diffusion, and ending on the day
17 of 19
when the snow pit samples were taken, early 2002. The
isotope record is initially irregular and often includes
several subseasonal variations. However, firn diffusion
quickly smoothes these features, and after some years
the only remaining signal is the annual cycle. Striking are
also the differences in decrease of seasonal amplitudes,
which are largely dependent on the second derivative
term in equation (A1): the diffusion process has a much
larger effect on the strongly negative d18O winter 2000
minimum, compared to the broader minimum of winter
1999 (Figure 16).
[83] Acknowledgments. This work is a contribution to the European
Project for Ice Coring in Antarctica (EPICA), a joint ESF (European Science
Foundation)/EC scientific programme, funded by the European Commission and by national contributions from Belgium, Denmark, France,
Germany, Italy, Netherlands, Norway, Sweden, Switzerland, and the
United Kingdom. This is EPICA publication 151. Financial support was
obtained from the Netherlands Organisation for Scientific Research
(NWO) by a grant of the Netherlands Antarctic Programme. We thank
three anonymous reviewers, whose comments have considerably improved
this manuscript.
Aldaz, L., and S. Deutsch (1967), On a relationship between air temperature
and oxygen isotope ratio of snow and firn in the South Pole region, Earth
Planet. Sci. Lett., 3, 267 – 274.
Boyle, E. A. (1997), Cool tropical temperatures shift the global d18O-T
relationship: An explanation for the ice core d18O – borehole thermometry
conflict?, Geophys. Res. Lett., 24(3), 273 – 276.
Bromwich, D. H. (1988), Snowfall in high southern latitudes, Rev. Geophys., 26(1), 149 – 168.
Ciais, P., and J. Jouzel (1994), Deuterium and oxygen 18 in precipitation:
Isotopic model, including mixed cloud preocesses, J. Geophys. Res.,
99(D8), 16,793 – 16,803.
Ciais, P., J. W. C. White, J. Jouzel, and J. R. Petit (1995), The origin of
present-day Antarctic precipitation from surface snow deuterium excess
data, J. Geophys. Res., 100(D9), 18,917 – 18,927.
Craig, H., and A. Gordon (1965), Deuterium and oxygen 18 variations in the
ocean and the marine atmosphere, in Stable Isotopes in Oceanic Studies and
Paleotemperatures, pp. 9 – 130, Lab. di Geol. Nucl., Pisa, Italy.
Cuffey, K. M., and F. Vimeux (2001), Covariation of carbon dioxide and
temperature from the Vostok ice core after deuterium-excess correction,
Nature, 412, 523 – 527.
Cuffey, K. M., G. D. Clow, R. B. Alley, M. Stuiver, E. D. Waddington, and
R. W. Saltus (1995), Large Arctic temperature change at the WisconsinHolocene glacial transition, Science, 270, 455 – 458.
Dahe, Q., J. R. Petit, J. Jouzel, and M. Stievenard (1994), Distribution of
stable isotopes in surface snow along the route of the 1990 International
Trans-Antarctic Expedition, J. Glaciol., 40(134), 107 – 118.
Dansgaard, W. (1964), Stable isotopes in precipitation, Tellus, 16, 436 –
Delaygue, G., V. Masson, J. Jouzel, R. D. Koster, and R. J. Healy (2000),
The origin of Antarctic precipitation: A modelling approach, Tellus, Ser.
B, 52, 19 – 36.
Delmotte, M., V. Masson, J. Jouzel, and V. Morgan (2000), A seasonal
deuterium excess signal at Law Dome, coastal eastern Antarctica: A
Southern Ocean signature, J. Geophys. Res, 105(D6), 7187 – 7197.
EPICA Community Members (2004), Eight glacial cycles from an Antarctic ice core, Nature, 429, 623 – 628.
Fisher, D. A. (1991), Remarks on the deuterium excess in precipitation in
cold regions, Tellus, Ser. B, 43, 401 – 407.
Frezzotti, M., S. Gandolfi, and S. Urbini (2002), Snow megadunes in
Antarctica: Sedimentary structure and genesis, J. Geophys. Res.,
107(D18), 4344, doi:10.1029/2001JD000673.
GRIP Members (1993), Climate instability during the last interglacial period recorded in the GRIP ice core, Nature, 364, 203 – 207.
Helsen, M. M., R. S. W. Van de Wal, M. R. Van den Broeke, E. R. T.
Kerstel, V. Masson-Delmotte, H. A. J. Meijer, C. H. Reijmer, and M. P.
Scheele (2004), Modelling the isotopic composition of snow using backward trajectories: A particular accumulation event in Dronning Maud
Land, Antarctica, Ann. Glaciol., 39, 293 – 299.
Helsen, M. M., R. S. W. Van de Wal, M. R. Van den Broeke, D. Van As,
H. A. J. Meijer, and C. H. Reijmer (2005), Oxygen isotope variability in
snow from western Dronning Maud Land, Antarctica and its relation to
temperature, Tellus, Ser. B, 57(5), 423 – 435.
Hoffmann, G., M. Werner, and M. Heimann (1998), Water isotope module
of the ECHAM atmospheric general circulation model: A study on time
scales from days to several years, J. Geophys. Res., 103(D14), 16,871 –
International Atomic Energy Agency (1992), Statistical treatment of data on
environmental isotopes in precipitation, IAEA Tech. Rep. Ser. 331, Vienna.
Isaksson, E., W. Karlén, N. Gundestrup, P. Mayewski, S. Whitlow, and
M. Twickler (1996), A century of accumulation and temperature changes
in Dronning Maud Land, Antarctica, J. Geophys. Res., 101(D3), 7085 –
Jakob, C. (1999), Cloud cover in the ECMWF Reanalysis, J. Clim., 12(4),
947 – 959.
Johnsen, S. J., W. Dansgaard, and J. W. C. White (1989), The origin of
Arctic precipitation under present and glacial conditions, Tellus, Ser. B,
41, 452 – 468.
Johnsen, S. J., D. Dahl-Jensen, W. Dansgaard, and N. Gundestrup (1995),
Greenland palaeotemperatures derived from GRIP bore hole temperature
and ice core isotope profiles, Tellus, Ser. B, 47, 624 – 629.
Johnsen, S. J., H. B. Clausen, K. M. Cuffey, G. Hoffmann, J. Schwander,
and T. Creyts (2000), Diffusion of stable isotopes in polar firn and ice:
The isotope effect in firn diffusion, in Physics of Ice Core Records, edited
by T. Hondoh, pp. 121 – 140, Hokkaido Univ. Press, Sapporo, Japan.
Joussaume, J., R. Sadourny, and J. Jouzel (1984), A general circulation
model of water isotope cycles in the atmosphere, Nature, 311, 24 – 29.
Jouzel, J., and R. D. Koster (1996), A reconsideration of the initial conditions used for stable water isotope models, J. Geophys. Res., 101(D17),
22,933 – 22,938.
Jouzel, J., and L. Merlivat (1984), Deuterium and oxygen 18 in precipitation: Modeling of the isotopic efects during snow formation, J. Geophys.
Res., 89(D7), 11,749 – 11,757.
Jouzel, J., et al. (1997), Validity of the temperature reconstruction from
water isotopes in ice cores, J. Geophys. Res., 102(C12), 26,471 – 26,487.
Jouzel, J., F. Vimeux, N. Caillon, G. Delaygue, G. Hoffmann, V. MassonDelmotte, and F. Parrenin (2003), Magnitude of isotope/temperature scaling for interpretation of central antarctic ice cores, J. Geophys. Res.,
108(D12), 4361, doi:10.1029/2002JD002677.
Kavanaugh, J. L., and K. M. Cuffey (2003), Space and time variation of
d18O and dD in Antarctic precipitation revised, Global Biogeochem. Cycles, 17(1), 1017, doi:10.1029/2002GB001910.
King, J. C., P. S. Andersen, D. G. Vaughan, G. W. Mann, S. D. Mobbs, and
S. B. Vosper (2004), Wind-borne redistribution of snow across an Antarctic ice rise, J. Geophys. Res., 109, D11104, doi:10.1029/
Kottmeier, C., and B. Fay (1998), Trajectories in the antarctic lower troposphere, J. Geophys. Res., 103(D9), 10,947 – 10,959.
Landais, A., J. M. Barnola, V. Masson-Delmotte, J. Jouzel, J. Chappellaz,
N. Caillon, C. Huber, and S. J. Johnsen (2004), A continuous record of
temperature evolution over a sequence of Dansgaard-Oeschger events
during Marine Isotopic Stage 4 (76 to 62 kyr BP), Geophys. Res. Lett.,
31, L22211, doi:10.1029/2004GL021193.
Masson-Delmotte, V., M. Delmotte, V. Morgan, D. Etheridge, T. V.
Ommen, S. Tartarin, and G. Hoffmann (2003), Recent southern Indian
Ocean climate variability inferred from a Law Dome ice core: New insights for the interpretation of coastal Antarctic isotopic records, Clim.
Dyn., 21(2), 153 – 166.
Masson-Delmotte, V., B. Stenni, and J. Jouzel (2004), Common millennialscale variability of Antarctic and Southern Ocean temperatures during the
past 5000 years reconstructed from the EPICA Dome C ice core, Holocene, 14(2), 145 – 151.
Merlivat, L., and J. Jouzel (1979), Global climatic interpretation of the
deuterium-oxygen 18 relationship for precipitation, J. Geophys. Res.,
84(C8), 5029 – 5033.
Motoyama, H., N. Hirasawa, K. Satow, and O. Watanabe (2005), Seasonal
variations in oxygen isotope ratios of daily collected precipitation and
wind drift samples and in the final snow cover at Dome Fuji Station,
Antarctica, J. Geophys. Res., 110, D11106, doi:10.1029/2004JD004953.
Noone, D., and I. Simmonds (2002a), Annular variations in moisture transport mechanisms and the abundance of d18O in Antarctic snow, J. Geophys. Res., 107(D24), 4742, doi:10.1029/2002JD002262.
Noone, D., and I. Simmonds (2002b), Associations between d18O of water
and climate parameters in a simulation of atmospheric circulation for
1979 – 95, J. Clim., 15(22), 3150 – 3169.
Noone, D., and I. Simmonds (2004), Sea ice control of water isotope
transport to Antarctica and implications for ice core interpretation,
J. Geophys. Res., 109, D07105, doi:10.1029/2003JD004228.
Noone, D., J. Turner, and R. Mulvaney (1999), Atmospheric signals and
characteristics of accumulation on Dronning Maud Land, Antarctica,
J. Geophys. Res., 104(D16), 19,191 – 19,211.
18 of 19
North Greenland Ice Core Project Members (2004), High-resolution record
of Northern Hemisphere climate extending into the last interglacial
period, Nature, 431, 147 – 151.
Peel, D. A., R. Mulvaney, and B. R. Davison (1988), Stable-isotope/airtemperature relationships in ice cores from Dolleman Island and the
Palmer Land plateau, Antarctic Peninsula, Ann. Glaciol., 10, 130 – 136.
Petit, J. R., J. W. C. White, N. W. Young, J. Jouzel, and Y. Korotkevich
(1991), Deuterium excess in recent Antarctic snow, J. Geophys. Res.,
96(D3), 5113 – 5122.
Petit, J. R., et al. (1999), Climate and atmospheric history of the past
420,000 years from the Vostok ice core, Antarctica, Nature, 399, 429 –
Reijmer, C. H., and M. R. Van den Broeke (2003), Temporal and spatial
variability of the surface mass balance in Dronning Maud Land, Antarctica, as derived from automatic weather stations, J. Glaciol., 49(167),
512 – 520.
Reijmer, C. H., M. R. Van den Broeke, and M. P. Scheele (2002), Air parcel
trajectories and snowfall related to five deep drilling locations in Antarctica based on the ERA-15 dataset, J. Clim., 15, 1957 – 1968.
Reijmer, C. H., E. Van Meijgaard, and M. R. Van den Broeke (2005),
Evaluation of temperature and wind over Antarctica in a Regional Atmospheric Climate Model using one year of automatic weather station data
and upper air observations, J. Geophys. Res., 110, D04103, doi:10.1029/
Robin, G. (1983), The climatic record from ice cores, in The Climatic
Record in Polar Ice Sheets, edited by G. Robin, pp. 180 – 195, Cambridge
Univ. Press, New York.
Robin, G., and S. J. Johnsen (1983), Atmospheric processes, in The Climatic Record in Polar Ice Sheets, edited by G. Robin, pp. 47 – 52, Cambridge Univ. Press, New York.
Roeckner, E., K. Arpe, L. Bengtsson, M. Christoph, M. Claussen,
L. Dmenil, M. Esch, M. Giorgetta, U. Schlese, and U. Schulzweida
(1996), The atmospheric general circulation model ECHAM-4: Model
description and simulation of present-day climate, MPI Rep. 218, MaxPlanck-Inst. for Meteorol., Hamburg, Germany.
Scheele, M. P., P. C. Siegmund, and P. F. J. Van Velthoven (1996), Sensitivity of trajectories to data resolution and its dependence on the starting
point: In or outside a tropopause fold, Meteorol. Appl., 3, 267 – 273.
Schmidt, G. A., G. Hoffmann, D. T. Shindell, and Y. Hu (2005), Modeling
atmospheric stable water isotopes and the potential for constraining cloud
processes and stratosphere-troposphere water exchange, J. Geophys. Res.,
110, D21314, doi:10.1029/2005JD005790.
Stenni, B., V. M.-D. S. Johnsen, J. Jouzel, A. Longinelli, E. Monnin,
R. Röthlisberger, and E. Selmo (2001), An oceanic cold reversal during
the last deglaciation, Science, 293, 2074 – 2077.
Stohl, A., G. Wotawa, P. Seibert, and H. Kromp-Kolb (1995), Interpolation
errors in wind fields as a function of spatial and temporal resolution and
their impact of different types of kinematic trajectories, J. Appl.
Meteorol., 34(10), 2149 – 2165.
Tiedtke, M. (1993), Representation of clouds in large-scale models, Mon.
Weather Rev., 121(11), 3040 – 3061.
Van den Broeke, M., C. Reijmer, and R. Van de Wal (2004), Surface
radiation balance in Antarctica as measured with automatic weather stations, J. Geophys. Res., 109, D09103, doi:10.1029/2003JD004394.
Van den Broeke, M., C. Reijmer, D. Van As, R. Van de Wal, and
J. Oerlemans (2005), Seasonal cycles of Antarctic surface energy balance
from automatic weather stations, Ann. Glaciol., 41, 131 – 139.
Van Lipzig, N. P. M., E. Van Meijgaard, and J. Oerlemans (2002), The
effect of temporal variations in the surface mass balance and temperatureinversion strength on the interpretation of ice-core signals, J. Glaciol.,
48(163), 611 – 621.
Vimeux, F., V. Masson, J. Jouzel, and M. Stievenard (1999), Glacial-interglacial changes in ocean surface conditions in the Southern Hemisphere,
Nature, 398, 410 – 413.
Vuille, M., R. S. Bradley, M. Werner, R. Healy, and F. Keimig (2003),
Modelling d18O in precipitation over the tropical Americas: 1. Interannual
variability and climatic controls, J. Geophys. Res., 108(D6), 4174,
Waddington, E. D., E. J. Steig, and T. A. Neumann (2002), Using characteristic times to assess whether stable isotopes in polar snow can be
reversibly deposited, Ann. Glaciol., 35, 118 – 124.
Werner, M., and M. Heimann (2002), Modeling interannual variability of
water isotopes in Greenland and Antarctica, J. Geophys. Res., 107(D1),
4001, doi:10.1029/2001JD900253.
Werner, M., U. Mikolajewicz, M. Heimann, and G. Hoffmann (2000),
Borehole versus isotope temperatures on Greenland: Seasonality does
matter, Geophys. Res.Lett., 27(5), 723 – 726.
Werner, M., M. Heimann, and G. Hoffmann (2001), Isotopic composition
and origin of polar precipitation in present an glacial climate simulations,
Tellus, Ser. B, 53, 53 – 71.
M. M. Helsen, M. R. van den Broeke, and R. S. W. van de Wal, Institute
for Marine and Atmospheric Research, Utrecht University, P.O. Box
80.005, NL-3508 TA Utrecht, Netherlands. ([email protected])
V. Masson-Delmotte, Laboratoire des Sciences du Climat et de
l’Environnement (LSCE), UMR CEA/CNRS 1572, Bat 709, L’Orme des
Merisiers CEA, Saclay, F-91191 Gif sur Yvette Cedex, France.
H. A. J. Meijer, Centre for Isotope Research (CIO), Nijenborgh 4,
NL-9747, AG Groningen, Netherlands.
M. P. Scheele, Royal Netherlands Meteorological Institute, P.O. Box 201,
NL-3730 AE De Bilt, Netherlands.
M. Werner, Max-Planck-Institute for Biogeochemistry, P.O. Box 10 01
64, D-07701 Jena, Germany.
19 of 19
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF