Tellus (2004), 56A, 501–513 Copyright C© Blackwell Munksgaard, 2004
Printed in UK. All rights reserved T E L L U S
An examination of the precipitation delivery
mechanisms for Dolleman Island, eastern
Antarctic Peninsula
By ANDREW RUSSELL 1∗, GLENN R. McGREGOR 1 and GARETH J . MARSHALL 2,
1University of Birmingham, School of Geography, Earth and Environmental Sciences, Edgbaston, Birmingham,
B15 2TT; 2British Antarctic Survey, Natural Environment Research Council, High Cross, Madingley Road,
Cambridge, CB3 0ET
(Manuscript received 15 December 2003; in final form 24 May 2004)
ABSTRACT
The variability of size and source of significant precipitation events were studied at an Antarctic ice core drilling site:
Dolleman Island (DI), located on the eastern coast of the Antarctic Peninsula. Significant precipitation events that occur
at DI were temporally located in the European Centre for Medium-Range Weather Forecasting (ECMWF) reanalysis
data set, ERA-40. The annual and summer precipitation totals from ERA-40 at DI both show significant increases over
the reanalysis period. Three-dimensional backwards air parcel trajectories were then run for 5 d using the ECMWF
ERA-15 wind fields. Cluster analyses were performed on two sets of these backwards trajectories: all days in the
range 1979–1992 (the climatological time-scale) and a subset of days when a significant precipitation event occurred.
The principal air mass sources and delivery mechanisms were found to be the Weddell Sea via lee cyclogenesis, the
South Atlantic when there was a weak circumpolar trough (CPT) and the South Pacific when the CPT was deep. The
occurrence of precipitation bearing air masses arriving via a strong CPT was found to have a significant correlation
with the southern annular mode (SAM); however, the arrival of air masses from the same region over the climatological
time-scale showed no such correlation. Despite the dominance in both groups of back trajectories of the westerly
circulation around Antarctica, some other key patterns were identified. Most notably there was a higher frequency of
lee cyclogenesis events in the significant precipitation trajectories compared to the climatological time-scale. There was
also a tendency for precipitation trajectories to come from more northerly latitudes, mostly from 50–70◦S. The El Nin˜o
Southern Oscillation (ENSO) was found to have a strong influence on the mechanism by which the precipitation was
delivered; the frequency of occurrence of precipitation from the east (west) of DI increased during El Nin˜o (La Nin˜a)
events.
1. Introduction
Over the last 50 yr, meteorological observations from the west-
ern Antarctic Peninsula demonstrate that the region has un-
dergone one of the largest temperature changes reported any-
where on the globe. For example, since 1951 the annual mean
temperature at Faraday station (Fig. 1) has risen by ∼3◦C
(see http://www.antarctica.ac.uk/met/gjma/ for Antarctic surface
temperature data), while Vaughan et al. (2001) have shown
significant positive annual temperature trends at other western
Antarctic Peninsula stations. On the north-eastern side of the
peninsula, smaller increases in annual mean temperature over a
50-yr period have been observed at Esperanza, but perhaps more
∗Corresponding author.
e-mail: axr141@bham.ac.uk
importantly, the summer mean temperature increase for this loca-
tion is approximately twice the magnitude of the summer warm-
ing seen on the western side of the peninsula (Skvarca et al.,
1998). Such dramatic summer warming has occurred in paral-
lel with the demise of the Larsen Ice Shelf (Doake et al., 1998;
Rott et al., 2002), which may have led to an acceleration and
consequent decline in mass of the glaciers of western Antarctica
(De Angelis and Skvarca, 2003). In order to place recent ob-
served increases in temperature in a regional and global palaeo-
climatic context, proxy records of temperature are extracted from
ice cores (Folland et al., 2001). This method is of particular use
in the data sparse Antarctic and there have been six drilling sites
studied on the relatively easily accessible peninsula region (Peel,
1992; Mulvaney et al., 2002). This reconstruction of past climate
is made possible due to the effect that temperature has upon
the isotopic balance in the moisture that arrives at Antarctica
Tellus 56A (2004), 5 501

502 A. RUSSELL ET AL.
Fig 1. Map of the Antarctic Peninsula showing the location of places
referred to in the text.
as precipitation (Legrand and Mayewski, 1997). However, the
source of, and the atmospheric route taken by, this moisture will
also influence the chemical composition and, more importantly,
the isotopic ratios recorded in these ice cores. Delaygue et al.
(2000a) have used GCM data to study these effects on central
Antarctic precipitation and concluded that the effects of precip-
itation seasonality and the cooling of moisture sources have a
limited and opposite effect and are, therefore, thought to not
influence the validity of the ‘isotopic thermometer’ for central
Antarctica. However, the mechanisms by which precipitation is
formed and delivered to the Antarctic coast and, in particular,
the Antarctic Peninsula are quite different from central Antarc-
tica because of the proximity to the circumpolar trough (CPT)
and the synoptic systems therein (King and Turner, 1997; Turner
et al., 1998; Simmonds et al., 2003). This results in increased pre-
cipitation quantity and variability at coastal and peninsula sites
but does not have a particularly large effect on the Antarctic
interior. These factors may have implications for the interpre-
tation of the regional proxy climatic records contained within
coastal and peninsula ice cores. For example, Schlosser (1999)
suggested that variability in the position of the CPT had a sig-
nificant effect on the Neumayer δ18O record; this station is in
Dronning Maud Land (DML), located to the east of the Wed-
dell Sea. For the period 1982–1991 the Neumayer δ18O record
implied that the temperature varied in the order of 5◦C when
the observed variability from the station itself for the same pe-
riod was less than 1◦C. In addition, Noone et al. (1999) studied
the signals from atmospheric circulation that could be detected
in the accumulation of an Antarctic ice core, again from DML.
They concluded that the glaciological record is connected to
global atmospheric circulation characteristics and could, thus,
be used as a tool in reconstructing past atmospheric circulation
changes.
With this point in mind, a variety of methods used to iden-
tify the source of Antarctic precipitation will now be discussed.
Ciais et al. (1995) concluded from a study of snow deuterium
excess data from a South Pole snow pit that the dominant mois-
ture source for that precipitation was 20–40◦S. This result is
in partial disagreement with the low-resolution GCM-based (8◦
× 10◦) findings of Delaygue et al. (2000b), who reported that
moisture from 30–60◦S plays a significant role in the precip-
itation climatology of Antarctica as a whole. Bromwich and
Weaver (1983) investigated the δ18O record for 1974 from the
coastal station Syowa (69◦S, 39◦E) and inferred that the mois-
ture source was around 55–58◦S. Further, Reijmer et al. (2002)
analysed air parcel back trajectories (BTs) for a number of ice
core sites from eastern Antarctica and the western Antarctic ice
sheet. They found that the main moisture sources for all sites
were located in the latitude range of 50–60◦S from the near-
est ocean to the west of the site in question. The above studies
of precipitation source for continental Antarctica tend to have
found the source to be further north than Turner et al. (1995),
who examined the precipitation sources for Rothera station, lo-
cated on the western Antarctic Peninsula (Fig. 1). Using mete-
orological observations and satellite imagery for a 1-yr period,
Turner et al. (1995) concluded that half of the cyclones that de-
liver precipitation to Rothera develop south of the 60◦S parallel.
Peel and Mulvaney (1992) are the only workers to consider the
source of precipitation for the eastern Antarctic Peninsula. They
studied the chemical composition of an ice core from Dolle-
man Island (DI), eastern Antarctic Peninsula, and concluded
that the marginal ice zone of the Weddell Sea has become an
important source of moisture for this region over recent times.
However, their results could be questioned with respect to one of
their assumptions, i.e. that an increase in non-sea-salt sulphate
(nssSO2−4 ) concentration is an indication of increased open sea.
It has since been argued that nssSO2−4 in this region is largely
derived from ‘frost flowers’ that form on new sea ice (Rankin
et al., 2002; Wolff et al., 2003). If this is correct, it implies that
the assumption of Peel and Mulvaney (1992) would need to be
reversed.
Given this debate concerning the origin of the moisture that
results in Antarctic precipitation and the relative lack of work
examining the eastern Antarctic Peninsula, the purpose of this
study is to examine the precipitation delivery mechanisms for
DI (70.3◦S, 60.5◦W) on the eastern side of the Antarctic Penin-
sula (Fig. 1). This was achieved by undertaking an air parcel
backwards trajectory (BT) analysis of the precipitation record
contained within a reanalysis data set and an ice core drilled on
the island in 1993.
Tellus 56A (2004), 5

PRECIPITATION DELIVERY MECHANISMS FOR DOLLEMAN ISLAND 503
In Section 2 of this paper we briefly describe the data and
methods employed in this study. As background, the precipita-
tion regime of DI will be discussed in Section 3. In Section 4
the results of the BT analysis are presented. The implications of
the analysis results will be discussed and conclusions drawn in
Section 5.
2. Data and methodology
An environment to circulation approach (Yarnal et al., 2001)
was used here to elucidate the large-scale precipitation delivery
mechanisms to DI. First, ‘significant’ precipitation events (i.e.
the largest precipitation events) were identified at the daily time-
scale. Secondly, 5-d BT analyses for the dates associated with
these events were performed in order to identify the source re-
gions and trajectories of the air masses bearing the precipitation.
A comparative analysis between the BT patterns for the signif-
icant precipitation events and the climatological pattern for the
study was then made. Finally, the annual accumulation for DI, as
recorded in an ice core drilled in 1993 and forecast by reanalysis
data, was analysed in relation to the interannual variability of the
BT patterns.
In this study European Centre for Medium-Range Weather
Forecasting (ECMWF) 44-yr reanalysis data (ERA-40) were
utilized to examine the general precipitation regime at DI for
the period 1958–2001. However, within this period the years of
1979–1992 were focused upon as 1992 represents the last year
of the DI ice core based annual accumulation record. Moreover,
at the time of writing, BT analysis was only completed using the
older ECMWF reanalysis data set, ERA-15, which is only avail-
able from 1979. It is considered that there is no methodological
problem in using these two data sets in this way because there is
a very high degree of correlation between the pressure data fields
for the whole of the Southern Hemisphere between ERA-40 and
ERA-15. Figure 2 shows the case for mean sea level pressure
(MSLP). Therefore, the BTs derived from the two reanalyses
would also be correlated to a significant degree.
The precipitation field in ERA-40 was calculated as a daily
forecast. The most accurate method (i.e. the method that most
reduces the spin-up error) of calculating this daily forecast is
thought to be the sum of (i) the difference between the (T +
12) and (T + 24) forecasts, where T is 1200 h on the previous
day and (ii) the equivalent value calculated where T is 0000 h
on the day in question (Genthon and Krinner, 1998). It is also
assumed that the daily precipitation forecast total represents one
precipitation event.
With respect to verifying the ERA-40 data for the study pe-
riod at DI, we can examine the temperature and atmospheric
pressure data recorded by an automatic weather station (AWS)
that was located on DI between 1986 and 1988. Figure 3 shows
a plot of temperature for this period from both the AWS and the
ERA-40 data. The correlation coefficient between the AWS and
the ERA-40 data for temperature and surface pressure from the
Fig 2. A plot of the correlation coefficient between the monthly MSLP
for 1979–1992 from the ERA-15 and ERA-40 reanalyses for the
Southern Hemisphere. Shaded areas are where r > 0.95. The
correlations are significant at the 0.01 level for the whole hemisphere.
Fig 3. Plot of daily mean temperature from the DI AWS (solid line)
and the ERA-40 data (dashed line).
closest model grid cell are 0.93 and 0.94, respectively, both of
which are significant at the 0.01 level. These correlations imply
that the ERA-40 data can be used to portray the general clima-
tology at DI with reasonable confidence. The root-mean-squared
(RMS) errors between the AWS and ERA-40 data are 2.9◦C and
4.8 hPa. In order to calculate these RMS errors, the differences
in height of DI in reality and in the model were accounted for:
the temperature was corrected by using a factor that reflects the
sensitivity of temperature to altitude as described by Krinner and
Genthon (1999) and the pressure was corrected using the hydro-
static equation. The difference that remains is likely to be due,
in part, to the smoothing of the peninsula topography and the
partially related inaccuracies in the land/sea mask of the model.
Tellus 56A (2004), 5

504 A. RUSSELL ET AL.
However, the most important data field for this study is precip-
itation and it should be borne in mind that precipitation fields are
derived and also difficult to validate. Despite this, some workers
have assessed the quality of the ECMWF reanalysis precipitation
data on a continental scale for Antarctica. For example, Genthon
and Braun (1995) describe the mean annual Antarctic precipi-
tation from the older and shorter ECMWF reanalysis, ERA-15
as being ‘fairly well represented’. Further, Genthon and Cosme
(2003) show that ERA-40 continental scale Antarctic mean sur-
face mass balance is in ‘relative agreement’ with glaciological
measurements from Vaughan et al. (1999). Genthon (2003) also
believes ERA-40 precipitation and, especially, evaporation data
to be more realistic than ERA-15 due to fewer spin-up problems
in the newer model. Indeed, there is noticeable difference be-
tween the time series of precipitation from the two reanalyses
at DI. For these reasons, the ERA-40 precipitation data were as-
sumed to be more reliable and were used in this study despite
using BTs driven by the ERA-15 data.
It should also be noted at this stage that the reliability of major
data fields in the ERA-40 reanalysis (MSLP, 2-m temperature and
500-hPa geopotential height) have been seriously questioned in
the southern high latitudes. Bromwich and Fogt (2004) have
highlighted some very low and even some negative correlations
between Antarctic station and ERA-40 data for the period before
the introduction of satellite observations (∼1972), especially in
winter. This is followed by a rapid increase in ERA-40 skill up to
the dawn of the satellite era. In summer, though, the high-latitude
data were found to be of a comparable quality as those from
the mid-latitude Southern Hemisphere continents. Despite the
largest of these problems being discovered in eastern Antarctica,
there are still significant problems with the ERA-40 data over the
Antarctic Peninsula. The implications of the work of Bromwich
and Fogt (2004) are that any ERA-40 data used before the early
1970s must be seriously scrutinized in this study.
Although it is impossible to verify the daily DI precipitation
from ERA-40, a comparison can be made between the annual ac-
cumulation record from ERA-40 (precipitation minus evapora-
tion or P−E) and the surface mass balance (SMB) data contained
in the DI ice core. First, though, given that Pasteur and Mulvaney
(2000) show that there is only ‘occasional slight melting’ in the
DI core, and that the summer melt period for the DI area, as
determined by Torinesi et al. (2003) from microwave sensors,
is relatively short, it is noted here that the SMB is assumed to
be equal to accumulation for DI in this paper. This provides an
opportunity for verification of accumulation at the annual time-
scale. In order to perform such a comparison, the ERA-40 annual
P−E totals were calculated over the 12-month period of August
to July, rather than the calendar year, as this former period co-
incides with the ice core’s annual cycle of isotopic stratigraphy
that was used to divide the ice core into annual layers (Peel et al.,
1988). Overall, the annual ERA-40 and ice core accumulation
demonstrate a weak (r = 0.29) and statistically insignificant as-
sociation over the full overlap period of 1958–1992 (Fig. 4). The
Fig 4. Comparison of the ice core accumulation (dashed line) and the
ERA-40 P−E (solid line).
serious problems associated with the ERA-40 data before the in-
troduction of satellite observations (Bromwich and Fogt, 2004)
undoubtedly had an impact here. Despite this, examination of
Fig. 4 reveals that for the period covered by the air parcel tra-
jectory study (1979–1992) the general upward trend in ERA-40
and ice core annual accumulation is similar. Furthermore, there
is a higher degree of statistical association between the ERA-40
P−E and ice core accumulation for this latter period (r = 0.59,
p = 0.05).
As noted earlier, BT analyses were conducted for days with
significant precipitation and all days over the period 1979–1992
(the climatological time-scale). Figure 5 shows the frequency of
ERA-40 precipitation events by size. It is clear that ERA-40 pre-
cipitation at DI is dominated by small events and this is possibly
Fig 5. Frequency of precipitation events from the ERA-40 data set for
1979–1992. After the first bar, each bar represents the number of events
in the 31 1-mm precipitation bins as a percentage of all the days in the
1979–92 range. The first bar represents days that receive no
precipitation. The horizontal arrow highlights the range of the largest
events that contribute 50% of the total precipitation at DI.
Tellus 56A (2004), 5

PRECIPITATION DELIVERY MECHANISMS FOR DOLLEMAN ISLAND 505
an unlikely characteristic. Although there is no precipitation data
from the DI AWS to confirm this, it has been shown by Reijmer
et al. (2002) for an Antarctic site in DML that the dominance of
small precipitation events in the reanalysis is unrealistic; when
compared to sonic altimeter measurements from their site it was
shown that only the larger events in the reanalysis data occurred
in reality. However, the DML site investigated by Reijmer et al.
(2002) is at a much greater altitude than DI and is some dis-
tance away on the opposite side of the Weddell Sea; therefore,
the two sites may display significantly different precipitation
characteristics. Indeed, Turner et al. (1997) used synoptic ob-
servations to describe a very different situation on the western
Antarctic Peninsula: they report that the mean number of precipi-
tation events at Rothera (Fig. 1) for the period 1956–1994 is 463
per year. These studies of DML and western Antarctic Penin-
sula precipitation represent the nearest and most relevant works
on precipitation regime in western Antarctica. However, Fig. 6
shows that for the ERA-40 data there is a relatively low correla-
tion coefficient between the DI precipitation regime and the areas
studied by Reijmer et al. (2002) and Turner et al. (1997). Taking
the differing precipitation regimes of these two sites into account,
this paper is, as a result, concerned with the investigation of air
mass trajectories that are associated with both the climatologi-
cal time-scale and the days that received a significant amount of
precipitation in the reanalysis. Given this, a significant precipi-
tation event at DI was defined in this study as the largest events
in the data set, namely, days that receive more than 3.57 mm of
precipitation (Fig. 5). This threshold was chosen as, collectively,
precipitation events larger than 3.57 mm d−1 are responsible for
Fig 6. A plot of the correlation coefficient between the annual mean
ERA-40 accumulation (P−E) at DI and at all other ERA grid cells in
the Antarctic region for 1979–1992. The correlations are significant at
the 0.01 level for all of the peninsula and Weddell Sea area but are less
significant throughout the rest of the plot. Dashed lines represent r < 0
and shaded areas where r > 0.5.
50% of the total ERA-40 precipitation at DI. This threshold was
broken on 654 d, which represents 12.8% of the study period
1979–1992.
The BT analysis was run using the British Atmospheric Data
Centre (http://www.badc.rl.ac.uk) trajectory model. This model
derives three-dimensional air parcel paths from a set of wind
components (u,v, omega) held on a 2.5◦ ×2.5◦ latitude/longitude
grid and used ECMWF ERA-15 six-hourly pressure data. Again,
it was deemed to be acceptable to use the ERA-15 data for the
BTs as there was found to be a relatively high degree of correla-
tion between the temporal occurrence of significant precipitation
events in ERA-40 and ERA-15 (Spearman’s rank correlation of
r = 0.75, p = 0.01). Therefore, the ERA-15 trajectories were
very likely to capture the characteristics of the ERA-40 precipi-
tation events. The BTs were initiated at the 850-hPa level starting
at 1200 h from 70.3◦S, 60.5◦W and output data at six-hourly in-
tervals for 5 d. The initial level of 850 hPa was chosen as this
exceeds the elevation of DI (398 m) but is still low enough to
be directly influenced by the smoothed Antarctic Peninsula to-
pography in the model (Turner et al., 1999). As a wide variety
of BTs were found to exist for the climatological time-scale and
for the days comprising significant precipitation events, individ-
ual BTs were grouped using hierarchical cluster analysis, based
on Ward’s solution, in order to identify the major BT patterns.
The number of BT groups retained for analysis was determined
by inspection of the cluster analysis dendrogram and by con-
sideration of the rate of decay of an agglomeration coefficient
that portrays the rate at which large numbers of small groups
are aggregated into a smaller number of large groups. The resul-
tant BT pattern groups for the significant precipitation events
were subsequently compared with the climatological pattern
(all days).
3. Precipitation regime
The precipitation regime of the Antarctic Peninsula is largely
defined by its position and topography. The western side ex-
periences a relatively warm, maritime climate as a result of its
protrusion into lower latitudes and, thus, into the prevalent west-
erly, circumpolar atmospheric circulation. The elevation of the
peninsula (typically 2 km) and the steep orography results in
the triggering of precipitation on the western side. The elevation
also has the effect of shielding the eastern side from most of the
weather systems approaching the peninsula from the west. This
leads to very different temperatures and precipitation regimes on
the two sides of the peninsula, and Fig. 6 illustrates this latter
point. This figure shows that the ERA-40 precipitation at DI has
a higher correlation with that over the Weddell Sea than with the
western Antarctic Peninsula and a negative correlation further
to the west over the Amundsen and Bellingshausen Seas (ABS).
Such climate contrasts are also seen in the annual cycle of pre-
cipitation across the peninsula. Figure 7 shows that there is a
summer peak in the ERA-40 precipitation at DI whereas there
Tellus 56A (2004), 5

506 A. RUSSELL ET AL.
Fig 7. The annual cycle of precipitation at DI from the ERA-40
reanalysis. The error bars represent standard deviation and the range
for each month over the 14-yr study period.
is a winter peak at most western Antarctic Peninsula locations
(King and Turner, 1997).
Regression analysis of the seasonal precipitation from ERA-
40 against time (independent variable) has revealed that only
summer has a statistically significant positive trend at DI
(r 2 = 0.45, p = 0.01). This positive trend in summer precip-
itation (Fig. 8) is chiefly responsible for the positive trend seen
in the ERA-40 annual accumulation totals (Fig. 4). This has
occurred as a result of an increase in the size of the summer
precipitation events in the reanalysis data. This is supported by
the fact that there is no significant trend or interannual varia-
tion in the mean annual significant precipitation event size or
frequency per year in the ERA-40 precipitation (not shown).
However, the trend to higher summer (DJF) precipitation totals
seen in the reanalysis appears not to be associated with summer
warming, as summer temperatures for this area from ERA-40
(Fig. 8) do not demonstrate a comparable trend. It must be borne
in mind that despite the major shortcomings in these data be-
fore ∼1972 occurring in non-summer months (Bromwich and
Fogt, 2004), these problems may still impact the summer data
discussed above. Interestingly, the absence of marked summer
warming contrasts with the trends found within the observational
record for Esperanza, which is also located in eastern Antarctic
Peninsula (Skvarca et al., 1998). This difference is probably re-
lated to the fact that Esperanza is much further north than DI. Es-
peranza is, however, the only eastern Antarctic Peninsula station
to have a temperature record of reasonable quality and length.
As noted in Section 2, both ERA-40 and the ice core demon-
strate an increase in annual accumulation at DI over the period
1979–1992 (Fig. 4). Of note is the considerable level of interan-
nual variability imposed on the positive trend of accumulation,
particularly evident in the ice core data. Assuming all other things
being equal, it may be speculated that the annual accumulation
is sensitive to the interannual variability of air mass trajecto-
Fig 8. Mean daily precipitation for the summer (DJF) at DI from
ERA-40 (top) and mean daily temperature for the four seasons at DI
from the ERA-40 reanalysis (bottom).
ries such that high accumulation years will possess basic BT
pattern differences compared to low accumulation years. This
contention is investigated in the next section.
4. Back trajectory analysis
BT cluster analysis for all days over the period 1979–92 revealed
eight predominant air mass trajectory groups (Fig. 9). These will
be referred to as the climatological or BTcl patterns. Although
there are eight climatological patterns, it is likely that BTcl6–8
are a subgroup of BTcl5 as they essentially possess the same tra-
jectory characteristics apart from an origin further to the west of
BTcl5. For the subgroup of BT patterns associated with the 654
significant precipitation (BTsp) events, five major and one minor
air mass trajectory pattern were revealed by the cluster analysis
(Fig. 9). The minor BTsp pattern (BTsp6) is undoubtedly a hybrid
of BTsp5. Table 1 shows the number of days belonging to each of
the climatological and significant precipitation BT patterns and
these are categorized further into BTs that originate from the east
or west of DI. Climatologically, DI experiences a passage of air
masses from the east for ∼40% of the time. Similarly, ∼39% of
the significant precipitation days are associated with air masses
moving in from the east over the Weddell Sea. However, when
Tellus 56A (2004), 5

PRECIPITATION DELIVERY MECHANISMS FOR DOLLEMAN ISLAND 507
Fig 9. Means of the 5-d backwards trajectories that fell into each
cluster found in the cluster analysis performed on the climatology
trajectories (top) and the 654 significant precipitation trajectories
(bottom).
the relative occurrence of specific BT patterns that are similar in
the BTsp and BTcl plots (Fig. 9) are compared, a greater propor-
tion of significant precipitation days are associated with BTsp2,
BTsp3 and, to a lesser extent, BTsp4 compared to climatology
(Table 1). Conversely BTsp1, BTsp5 and BTsp6 are relatively
unimportant in terms of significant precipitation days compared
to their climatological occurrence (Table 1). The climatological
importance of westerly atmospheric flow is demonstrated by the
fact that, respectively, six and three of the climatological and
significant precipitation BT patterns are clearly associated with
the circumpolar westerlies and thus the CPT.
The mean trajectories seen in the two sets of BT patterns are
remarkably similar; in fact, the only climatological BT pattern
that does not have a significant precipitation equivalent is BTcl3
Table 1. The number of days that fell into the clusters of significant
precipitation days (BTsp) and the climatological time-scale (BTcl)
Cluster 1 2 3 4 5 6 7 8
From east From west
BTsp 51 211 188 147 52 4
7.8% 32.3% 28.7% 22.5% 8.1% 0.6%
40.1% 59.9%
BTcl 792 1091 823 962 1038 38 44 58
16.3% 22.5% 17.0% 19.9% 21.4% 0.8% 0.9% 1.2%
38.8% 61.2%
(Fig. 9). Climatologically, BTcl3 is of almost equal importance
to the other major BT patterns that originate to the west of DI
(BTcl4 and BTcl5). All these westerly trajectories travel over
the peninsula to reach DI and are associated with higher pre-
cipitation amounts on the western side of the peninsula than the
eastern side due to orographic effects. Under such conditions DI
would effectively lie in a precipitation shadow and thus rarely
receive significant precipitation from these trajectories unless
the air mass had acquired a large amount of moisture. BTcl3
displays a very short trajectory that will induce a much slower
rate of moisture advection when compared to the longer westerly
trajectories. Therefore, an equivalent of BTcl3 is absent from the
significant precipitation delivery mechanisms.
Despite the overall similarity of the climatological and signif-
icant precipitation BTs, some noteworthy differences do exist.
The most striking of these is the generally higher curvature of
the BTsp. This most likely reflects the high relative vorticity
of the flows within synoptic-scale systems that are associated
with significant precipitation at DI. BTsp2 is a good example
of this, as its trajectory mimics the highly curved cyclonic path
followed by air masses when cyclogenesis occurs in the west-
ern region of the Weddell Sea. Furthermore, BTsp4 and BTsp5,
and BTsp1 and BTsp2 have origins further west and east, respec-
tively, than their climatological equivalents. This may indicate
that weather systems that bring significant precipitation to DI
along such trajectories travel at much greater speeds than would
be otherwise expected based on climatology. Another major dif-
ference is the smaller proportion of precipitation trajectories that
come directly over the peninsula. This is an indication of mois-
ture depleted air masses reaching DI via this route and its absence
from the BTsp patterns that emerge is encouraging given the fact
that the peninsula topography is smoothed in the model. This
smoothing could have resulted in an artificially increased num-
ber of trajectories taking this route to DI than would happen in
reality, but this appears to have not had a significant impact on
these trajectories, which start from the 850-hPa level.
In addition to some basic differences in trajectory curvature
and speed, subtle contrasts also exist between the climatological
position of air mass origin and that for significant precipitation
days (Fig. 10). Air mass origin is defined here as the BT posi-
tion 4 d before arrival at DI. This 4-d figure was chosen as a
Tellus 56A (2004), 5

508 A. RUSSELL ET AL.
Fig 10. Plots to show the longitudinal (top) and latitudinal (bottom)
position of the precipitation trajectories 4 d before they arrive at DI.
The vertical dashed line shows the longitude/latitude of DI.
result of the work of Reijmer and van den Broeke (2001). They
used BTs and associated moisture data from the ERA-15 data
to conclude that most of the moisture uptake for DML precip-
itation in 1998 occurred 3–5 d before the event. Therefore, it
was assumed that this was also the likely source of most of the
DI precipitation. In terms of longitude (Fig. 10), air mass tra-
jectories associated with significant precipitation demonstrate
an occurrence peak at approximately 2◦ east of DI. This re-
flects the importance of significant precipitation trajectories as-
sociated with lee cyclogenesis east of the Antarctic Peninsula
(Turner et al., 1998). Climatologically, a greater proportion of
air masses originate further west compared to those associated
with significant precipitation days. This may be explained by the
dominance and high variability of synoptic-scale activity in the
CPT over the ABS (Connolley, 1997). Compared to longitudinal
differences, the latitudinal contrast in air mass origin (Fig. 10) is
far greater between the climatology and significant precipitation
time-scales. For air masses associated with significant precipi-
tation, the point of origin is approximately 5◦ further north than
the climatological position. This not only indicates a more rapid
poleward movement of systems associated with significant pre-
cipitation at DI but possible origin of these over warmer ocean
surfaces or a smaller percentage of time spent over sea ice, both
of which bear implications for vertical and, ultimately, horizon-
tal moisture transport over DI. It is also evident from Fig. 10
that the dominant latitudinal position of precipitation bearing air
masses for DI 4 d before the event is between 50 and 70◦S.
In order to understand the relationship between the nature
of the large-scale pressure field and the BTs associated with
significant precipitation, composite MSLP anomaly plots were
constructed from the ERA-40 data for the six significant pre-
cipitation BTs (Fig. 11). Pressure anomalies were calculated
thus: the MSLP for the day of precipitation minus the monthly
mean MSLP for the period 1979–1992 for the month that the
significant precipitation day occurred. Associated with BTsp1
(Fig. 11a) is a positive anomaly around most of Antarctica as-
sociated with a weak CPT. The strongest negative anomaly and
the most significant feature of BTsp1 is over the Weddell Sea,
indicative of a low-pressure system in this region. This would
imply that under these synoptic-scale conditions precipitation
would be delivered to DI by tracking from the South Atlantic
and subsequently over the Weddell Sea. The trajectory taken by
such systems is similar to the climatological path of mesocy-
clones that form in the eastern Weddell Sea (King and Turner,
1997). However, it is doubtful that the resolution of the ERA-40
grid will pick out any but the largest of these features so the pre-
cipitation is likely to be due to synoptic, or multiple mesoscale
(Carleton, 1995), weather systems that advect moisture over DI
from a north-easterly direction. The pressure anomaly pattern
for BTsp2 (Fig. 11b) demonstrates a slightly deeper CPT than
that of BTsp1. However, the largest negative and most signifi-
cant anomaly is over the Bellingshausen Sea. This anomalously
low pressure also extends over the peninsula into the Weddell
Sea. This pattern most likely represents lee cyclogenesis over
the Weddell Sea. The mean trajectory of air masses associated
with this pressure configuration would appear to support such a
contention. For BTsp3 (Fig. 11c), a CPT of comparable strength
and character to that of BTsp2 is evident. The greatest difference
between BTsp2 and BTsp3 is that the low over the Bellingshausen
Sea associated with BTsp3 does not extend as far into the Wed-
dell Sea region as that seen for BTsp2. As such, the trajectories
associated with BTsp3 originate over the Bellingshausen Sea and
track over the northern tip of the peninsula. They subsequently
move around the small area of low pressure over the Weddell Sea
and move towards DI. This BT pattern therefore exhibits some
characteristics of lee cyclogenesis but also of weather systems
moving towards DI via a weak westerly circumpolar circulation.
To some extent, BTsp1–3 are all associated with a high/low pres-
sure couplet across the peninsula. Such a pressure configuration
resembles that observed around the Antarctic Peninsula during
El Nin˜o – i.e. El Nin˜o Southern Oscillation (ENSO) warm –
events (Turner, 2004). In relation to this, the number of BTsp1–3
events per year demonstrates interannual variability worthy of
note (Fig. 12). BTsp2 appears to oscillate with a wavelength of
about 4 yr with frequency peaks at 1982–1983, 1987 and 1991,
Tellus 56A (2004), 5

PRECIPITATION DELIVERY MECHANISMS FOR DOLLEMAN ISLAND 509
Fig 11. The MSLP anomalies of the days
that fell into each cluster of significant
precipitation events. They are presented
separately for BTsp1–6 in Figs. 11a–f,
respectively. Dashed contours show negative
anomalies and the shaded areas are
significant (i.e. those areas that fall outside
the mean ±2 standard deviations). The thick
dashed line in each plot is the mean
backwards trajectory for the members of
each cluster (see Fig. 9, bottom).
which all coincide with El Nin˜o events from the Southern Oscil-
lation Index of Stocker et al. (2001). Although possessing much
less pronounced interannual variation, the frequency of BTsp1
and BTsp3 also appears to peak around the time of El Nin˜o events.
These observations are in line with those made above concerning
the nature of El Nin˜o related pressure anomalies.
The pressure anomaly patterns for BTsp4, BTsp5 and BTsp6
(Figs. 11d–f) are characterized by an increasingly more deep
CPT in the ABS region for each case. This also indicates in-
creasingly strong circumpolar westerly air flows. These westerly
trajectories lie over the South Pacific and originate at greater dis-
tances from DI in each successive case. In contrast to the situa-
tion for BTsp1–3, the pattern of anomalously low pressure over
the ABS bears some similarity to that observed over the west-
ern Antarctic region during La Nin˜a (i.e. ENSO cold) events
(Turner, 2004). Such events appear to have an impact on the
joint frequency of BTsp4, BTsp5 and BTsp6 (Fig. 12) as marked
increases in the frequency of the BTs originating over the ABS
occurred in 1985 and 1989, coincident with pronounced La Nin˜a
events. Conversely, frequency troughs occurred in 1983, 1987
and 1990–1992 when El Nin˜o conditions dominated the Pacific
Basin.
One objective of this study was to establish whether the inter-
annual variability of air mass trajectories influences the annual
Tellus 56A (2004), 5

510 A. RUSSELL ET AL.
Fig 12. Plots to show the number of trajectories that fell into each of
the six BTsp clusters for each year (top) and the sum of the annual
number of members of BTsp4, BTsp5 and BTsp6 (bottom).
accumulation at DI. Although correlation analysis revealed no
statistical association between the frequencies of individual BTs
associated with significant precipitation events and the ERA-40
or ice core based accumulation, significant relationships were
found between DI accumulation and the climatological occur-
rence of BT frequencies. For example, of the climatological BT
patterns, DI accumulation appears to be most sensitive to the
occurrence BTcl1 and BTcl4. In the former case, the associa-
tion is positive (r = 0.56, p = 0.04) such that on an annual
basis a greater frequency of air masses from the Weddell Sea
results in higher accumulation at DI, as recorded in the ice core
accumulation. In contrast, a greater occurrence of air masses
originating over the ABS is associated with low annual ERA-40
accumulation (r = − 0.51, p =0.05). The physical explanation
for such associations is clearly seen in the MSLP anomalies as-
sociated with the significant precipitation equivalent of BTcl1
and BTcl4 (Figs. 11a and 11d). The occurrence of BTsp1/BTcl1
is associated with negative pressure anomalies over the Weddell
Sea and, thus, cyclogenesis. At the same time, positive pres-
sure anomalies and anomalous subsiding southerly flows are ob-
tained over the ABS (Fig. 11a), the consequence of which will
be a weakened CPT and few eastward travelling weather sys-
tems originating from this region. Conversely, a high frequency
of BTsp4/BTcl4 is associated with anomalous low pressure over
the ABS, indicative of an active CPT. Simultaneously, over the
Weddell Sea, a strong blocking situation occurs which will pre-
vent the movement over DI of any moisture bearing systems
originating west of the peninsula, the consequence of which will
be low annual accumulation. This implies that annual accumula-
tion is more a product of the daily variation of air mass delivery
and not just the air masses associated with large precipitation
events.
Given the association between synoptic activity over the Wed-
dell Sea and ABS, of interest is the trend in the location of air
mass origin as it relates to accumulation at DI. As noted earlier,
statistically significant positive trends are evident for both the
ERA-40 and the ice core accumulation for 1979–1992. These are
matched by a statistically significant decline in the occurrence of
BTcl4 over the same period. Based on the pressure anomaly pat-
terns evident in Figs. 11a and 11d, this would imply a sustained
weakening of synoptic activity over the ABS and thus enhanced
cyclonic and/or reduced blocking activity over the Weddell Sea.
The outcome of this would be increasing annual accumulation
totals as shown for the study period.
5. Discussion and conclusions
The principal goal of this study was to shed light on the pre-
cipitation delivery mechanisms for DI. In order to achieve this,
a BT analysis of ERA-40 significant precipitation events (i.e.
the largest events that deliver 50% of the total precipitation) was
undertaken. BT patterns for such events were subsequently com-
pared with the climatological pattern in order to establish if dif-
ferences in BT pattern characteristics exist between climatology
and significant precipitation days. Study results have revealed
that, at the general level, there appears to be little difference in
BT patterns between climatology and significant precipitation
events, with both showing air masses arriving from the South
Atlantic, South Pacific and the Weddell Sea. However, subtle
differences are apparent. Compared to the climatological pattern
of air mass trajectories arriving over DI, significant precipitation
events appear to be associated with faster moving air masses and,
therefore, possibly greater rates of moisture advection over DI.
Furthermore, these air masses take a much less direct route to DI
as they follow a highly curved trajectory. This implies that the
flows that deliver significant precipitation amounts to DI pos-
sess high relative vorticity, and also that they do not traverse the
highest regions of the Antarctic Peninsula. A further noteworthy
characteristic is the importance of easterly air mass trajectories
for significant precipitation events, implying that the Weddell
Sea region is an important source of moisture for accumulation
over DI. This is an important finding given the dominance of
the circumpolar westerlies over the wider study area (Reijmer
et al., 2002). Further, it is proposed that the dominant latitudinal
Tellus 56A (2004), 5

PRECIPITATION DELIVERY MECHANISMS FOR DOLLEMAN ISLAND 511
source of precipitation bearing air masses for DI is between the
latitudes of 50–70◦S.
Regarding the implications of the precipitation delivery mech-
anisms that have been discovered here, it has been proposed
that western Antarctic precipitation will harbour a particularly
strong ENSO signature because of its location relative to the Pa-
cific Ocean (Turner, 2004). Several studies have reported these
ENSO links with Antarctica in certain precipitation data sets:
Cullather et al. (1996) in the ECMWF operational analyses (or
EOP); Noone et al. (1999) in ERA-15 and ice core accumu-
lation; Bromwich et al. (2000) in EOP and ERA-15; Marshall
(2000) in ERA-15; Genthon et al. (2003) in the National Center
for Environmental Prediction–National Center for Atmospheric
Research (NCEP–NCAR) reanalysis, ERA-15, GCM and satel-
lite data; Genthon and Cosme (2003) in ERA-40; and Turner
(2004) summarizes the signals observed in ice cores. In this
paper, a signature of ENSO is highlighted in the precipitation
delivery mechanisms for DI via the frequency of occurrence of
precipitation events from the east (west) of DI during El Nin˜o
(La Nin˜a) events.
Given the nature of the pressure anomaly fields associated
with easterly air masses arriving at DI, it would appear that lee
cyclogenesis plays an important role in precipitation delivery to
DI. This corroborates the satellite image based findings of Turner
et al. (1998) who concluded that the frequent occurrence of low-
pressure systems developing in the lee of the peninsula possessed
the potential to deliver significant precipitation amounts to the
study area. Although significant precipitation delivery from the
east via lee cyclogenesis is important for the precipitation cli-
matology at DI, equally important are air masses that have orig-
inated to the west of the peninsula. Accordingly, activity within
the CPT also appears to play a role in precipitation delivery
to DI as approximately 60% of significant precipitation events
have air mass trajectories that can be traced back to the ABS
(Table 1).
In relation to the importance of the ABS for the precipitation
regime at DI is the intriguing finding that there appears to be a
trend to an increasing number of significant precipitation events
over DI that have origins over the ABS. However, at the clima-
tological scale, the number of air masses arriving at DI from the
general region of the ABS appears to be decreasing. This ap-
parently contradictory finding may indicate that although fewer
air masses may arrive at DI from the west, the ones that do are
delivering greater amounts of precipitation than otherwise might
be expected. Such a trend may be associated with a deepening
of the CPT, which, in turn, is influenced by the SAM. The fact
that the SAM has demonstrated a positive trend over recent years
(Marshall, 2003), and that this trend explains approximately 20%
of the variation (r = 0.45, p = 0.05) in the joint frequency of
trajectories originating over the ABS associated with significant
precipitation at DI (BTsp4–6), indicates a possible linkage be-
tween CPT dynamics and DI accumulation. However, such a
contention will have to await an analysis of the trend of the ac-
tual precipitation amounts associated with air masses originating
over the ABS. Such an analysis would also assist with an inter-
pretation of the interannual variability of accumulation at DI as
recorded by ERA-40 and the DI ice core, as it has been suggested
in this study that the frequency of significant precipitation events
alone cannot explain such variability.
The SAM also appears to have had an impact on the annual
precipitation totals at DI. The positive trend in annual precipi-
tation totals in the ERA-40 data is a result of an increase in the
size of the summer precipitation events; it is also during summer
that the greatest positive trends in the SAM have been observed
(Marshall, 2003). This connection deserves more investigation
in the future via a seasonal analysis of the BT data.
A major assumption of this study is that ERA-40 and the DI
ice core offer a realistic picture of the accumulation regime at
DI. For the study period of 1979–1992, the estimates of annual
accumulation from ERA-40 and the ice core are quite well cor-
related; however, for the duration of ERA-40 this is not so and is
of concern. This is either indicative of the unreliability of ERA-
40 for making estimates of accumulation before 1979 related to
more fundamental problems in the reanalysis (Bromwich and
Fogt, 2004) and/or the non-representativeness of the accumula-
tion record contained within the single DI ice core. In the case
of ERA-40, topographic and sea/land classification inaccuracies
in the reanalysis model, as well as the fact that the ERA-40 fore-
casts represent a whole grid cell, may all contribute to unreliable
estimates of DI accumulation. These factors are over and above
the problems already discussed that are attributed to the reanal-
ysis before the introduction of satellite data. Possible problems
with the ice core record also need to be acknowledged. These
include the method by which the accumulation record has been
partitioned into annual totals (based on the troughs in the ice core
isotope stratigraphy being assigned as the beginning of August),
in situ snow/ice melt and accumulation that occurs as a result of
local microclimatic effects.
Despite the above concerns, study results have shown that the
climatological occurrences of BT patterns are able to account for
a significant proportion of the variation in annual accumulation
at DI as represented by ERA-40 and the DI ice core. This is
an interesting result with respect to the lack of understanding
of the precipitation regime in this area of the eastern Antarctic
Peninsula. It also implies that it would be of interest for future
work of this sort, in this region, to investigate the impact of
precipitation events using a lower threshold than the significant
precipitation event used in this paper.
In summary, the key centres of action of precipitation delivery
for DI appear to be the Weddell Sea region to the east and the
ABS to the west of DI. Although this study has shown that an
association between BT pattern frequency and DI accumulation
is plausible on physical grounds, further testing of this hypoth-
esis will have to await an analysis of BT–accumulation associa-
tions for the full period of ERA-40 data that is suitably reliable.
Restriction of the BT analysis to the years of overlap between
Tellus 56A (2004), 5

512 A. RUSSELL ET AL.
the ERA-15 data and the ice core has also restricted conclusions
being drawn regarding sympathetic frequency trends of accumu-
lation and BT patterns. For example, that the weakly positive,
but statistically insignificant, trend in the frequency of BTs with
origins over the ABS associated with significant precipitation at
DI may be part of a longer-term climatic oscillation.
6. Acknowledgments
We would like to thank the British Atmospheric Data Centre
for their help in providing the backwards air trajectory data and
Rob Mulvaney of the British Antarctic Survey for making the
Dolleman Island ice core data available. The useful comments
of the two reviewers of this work also need to be acknowledged.
References
Bromwich, D. H. and Fogt, R. L. 2004. Strong trends in the skill of the
ERA-40 and NCEP/NCAR reanalyses in the high and middle latitudes
of the Southern Hemisphere, 1958–2001. J. Climate in press
Bromwich, D. H. and Weaver, C. J. 1983. Latitudinal displacement from
main moisture source controls δ18O of snow in coastal Antarctica.
Nature 301, 145–147.
Bromwich, D. H., Rogers, A. N., Ka˚llberg, P., Cullather, R. I., White,
J. W. C. et al. 2000. ECMWF analyses and reanalyses depiction of
ENSO signal in Antarctic precipitation. J. Clim. 13, 1406–1420.
Carleton, A. M. 1995. On the interpretation and classification of
mesoscale cyclones from satellite IR imagery. Int. J. Rem. Sens. 16,
2457–2485.
Ciais, P., White, J. W. C., Jouzel, J. and Petit, J. R. 1995. The origin
of present-day Antarctic precipitation from surface snow deuterium
excess data. J. Geophys. Res. 100(D9), 18 917–18 927.
Connolley, W. M. 1997. Variability in annual mean circulation in the
southern high latitudes. Climate Dyn. 13, 745–756.
Cullather, R. I., Bromwich, D. H.and van Woert, M. L. 1996. Interan-
nual variations in Antarctic precipitation related to El Nin˜o southern
oscillation. J. Geophys. Res. 101(D14), 19 109–19 118.
De Angelis, H. and Skvarca, P. 2003. Glacier surge after Ice Shelf col-
lapse. Science 299, 1560–1562.
Delaygue, G., Jouzel, J., Masson, V., Koster, R. D. and Bard, E. 2000a.
Validity of the isotopic thermometer in central Antarctica: limited im-
pact of glacial precipitation seasonality and moisture origin. Geophys.
Res. Lett. 27, 2677–2860.
Delaygue, G., Masson, V., Jouzel, J., Koster, R. D. and Healy, R. J.
2000b. The origin of Antarctic precipitation: a modelling approach.
Tellus 52B, 19–36.
Doake, C. S. M., Corr, H. F. J., Rott, H., Skvarca, P. and Young, N. W.
1998. Breakup and conditions for stability of the northern Larsen Ice
Shelf. Nature 391, 778–780.
Folland, C. K., Karl, T. R., Christy, J. R., Clarke, R. A., Gruza, G. V. et al.
2001. Observed climate variability and change. In: Climate Change
2001: The Scientific Basis. Contribution of Working Group I to the
Third Assessment Report of the Intergovernmental Panel on Climate
Change, (eds. J. T. Houghton, Y. Ding, D. J. Griggs, M. Noguer,
P. J. van der Linden et al.). Cambridge University Press, Cambridge,
99–182.
Genthon, C. 2003. Climate and surface mass balance of the polar ice
sheets in ERA40/ERA15. ECMWF Reanalysis Project Report Series
3, 299–316.
Genthon, C. and Braun, A. 1995. ECMWF analyses and predictions of
the surface climate of Greenland and Antarctica. J. Climate 8, 2324–
2332.
Genthon, C. and Cosme, E. 2003. Intermittent signature of ENSO
in west-Antarcitc precipitation. Geophys. Res. Lett. 30, 2081
(doi:10.1029/2003GL018280).
Genthon, C. and Krinner, G. 1998. Convergence and disposal of energy
and moisture on the Antarctic polar cap from ECMWF reanalyses and
forecasts. J. Climate 11, 1703–1716.
Genthon, C., Krinner, G. and Sacchettini, M. 2003. Interannual Antarctic
tropospheric circulation and precipitation variability. Climate Dyn. 21,
289–307.
King, J. C. and Turner, J. 1997. Antarctic Meteorology and Climatology.
Cambridge University Press, Cambridge, 425 pp.
Krinner, G. and Genthon, C. 1999. Altitude dependence of the ice sheet
surface climate. Geophys. Res. Lett. 26, 2227–2230.
Legrand, M. and Mayewski, P. 1997. Glaciochemistry of polar ice cores:
a review. Rev. Geophys. 35, 219–243.
Marshall, G. J. 2000. An examination of the precipitation regime at
Thurston Island, Antarctica, from ECMWF reanalysis data. Int. J.
Climatol. 20, 255–277.
Marshall, G. J. 2003. Trends in the Southern Annular Mode from obser-
vations and analyses. J. Clim. 16, 4134–4143.
Mulvaney, R., Oerter, H., Peel, D. A., Graf, W., Arrowsmith, C. et al.
2002. 1000 year ice-core records from Berkner Island, Antarctica.
Ann. Glaciol. 35, 45–51.
Noone, D., Turner, J. and Mulvaney R. 1999. Atmospheric signals and
characteristics of accumulation in Dronning Maud Land, Antarctica.
J. Geophys. Res. 104(D16), 19 191–19 211.
Pasteur, E. C. and Mulvaney, R. 2000. Migration of methane sulphonate
in Antarctic firn and ice. J. Geophys. Res. 105(D9), 11 525–11 534.
Peel, D. A. 1992. Ice core evidence from the Antarctic Peninsula re-
gion. In: Climate since AD 1500. (eds. R. S. Bradley and P. D. Jones).
Routledge, London, 549–571.
Peel, D. A. and Mulvaney, R. 1992. Time-trends in the pattern of ocean-
atmosphere exchanges in an ice core from the Weddell Sea sector of
Antarctica. Tellus 44B, 430–442.
Peel, D. A., Mulvaney, R. and Davison, B. M. 1988. Stable-isotope/air-
temperature relationships in ice cores from Dolleman Island and the
Palmer Island plateau, Antarctic Peninsula. Ann. Glaciol. 10, 130–
136.
Rankin, A. M., Wolff, E. W. and Martin, S. 2002. Frost flowers: Im-
plications for tropospheric chemistry and ice core interpretation. J.
Geophys. Res. 107(D23), 4683 (doi:10.1029/2002JD002492).
Reijmer, C. H. and van den Broeke, M. R. 2001. Moisture source of
precipitation in Western Dronning Maud Land, Antarctica. Antarctic
Sci. 13, 210–220.
Reijmer, C. H., van den Broeke, M. R. and Scheele, M. P. 2002. Air
parcel trajectories and snowfall related to five deep drilling locations
in Antarctica based on the ERA-15 dataset. J. Climate 15, 1957–
1968.
Rott, H., Rack, W., Skvarca, P. and De Angelis, H. 2002. Northern Larsen
Ice Shelf, Antarctica: further retreat after collapse. Ann. Glaciol. 34,
277–282.
Tellus 56A (2004), 5

PRECIPITATION DELIVERY MECHANISMS FOR DOLLEMAN ISLAND 513
Schlosser, E. 1999. Effects of seasonal variability of accumulation on
yearly mean δ18O values in Antarctic snow. J. Glaciol. 45, 463–468.
Simmonds, I., Keay, K. and Lim, E.-P. 2003. Synoptic activity in the
seas around Antarctica. Mon. Wea. Rev. 131, 272–288.
Skvarca, P., Rack, W., Rott, H. and Dona´ngelo, T. I. Y. 1998. Evidence
of recent climatic warming on the eastern Antarctic Peninsula. Ann.
Glaciol. 27, 628–632.
Stocker, T. F., Clarke, G. K. C., Le Treut, H., Lindzen, R. S., Meleshko,
V. P. et al. 2001. Physical climate processes and feedbacks. In: Climate
Change 2001: The Scientific Basis. Contribution of Working Group
I to the Third Assessment Report of the Intergovernmental Panel on
Climate Change. (eds. J. T. Houghton, Y. Ding, D. J. Griggs, M.
Noguer, P. J. van der Linden et al.). Cambridge University Press,
Cambridge, 417–470.
Torinesi, O., Fily, M. and Genthon, C. 2003. Variability and trends of
the summer melt period of Antarctic ice margins since 1980 from
microwave sensors. J. Climate 16, 1047–1060.
Turner, J. 2004. The El Nin˜o Southern Oscillation and the Antarctic. Int.
J. Climatol. 24, 1–31.
Turner, J., Lachlan-Cope, T. A., Thomas, J. P. and Colwell, S. R. 1995.
The synoptic origins of precipitation over the Antarctic Peninsula.
Antarctic Sci. 7, 327–337.
Turner, J., Colwell, S. R. and Harangozo, S. A. 1997. Variability of pre-
cipitation over the coastal western Antarctic Peninsula from synoptic
observations. J. Geophys. Res. 102, D12, 13 999–14 007.
Turner, J., Marshall G. J. and Lachlan-Cope, T. A. 1998. Analysis
of synoptic-scale low pressure systems within the Antarctic Penin-
sula sector of the circumpolar trough. Int. J. Climatol. 18, 253–
280.
Turner, J., Connolley, W. M., Leonard, S., Marshall, G. J. and Vaughan,
D. G. 1999. Spatial and temporal variability of net snow accumula-
tion over the Antarctic from ECMWF reanalysis project data. Int. J.
Climatol. 19, 697–724.
Vaughan, D. G., Bamber, J. L., Giovinetto, M., Russell, J. and Cooper,
A. P. R. 1999. Reassessment of net surface mass balance in Antarctica.
J. Climate 12, 933–946.
Vaughan, D. G., Marshall, G. J., Connolley, W. M., King, J. C. and
Mulvaney, R. 2001. Devil in the detail. Science 293, 1777–1779.
Wolff, E. W., Rankin, A. M. and Ro¨thlisberger, R. 2003. An ice core
indicator of Antarctic sea ice production? Geophys. Res. Lett. 30,
2158 (doi:10.1029/2003GL018454).
Yarnal, B., Comrie, A. C., Frakes, B. and Brown, D. P. 2001. Devel-
opments and prospects in synoptic climatology. Int. J. Climatol. 21,
1923–1950.
Tellus 56A (2004), 5