Annals ofGlaciology 27 1998
© International Glaciological Society
Implications for the interpretation of ice-core isotope data
from analysis of modelled Antarctic precipitation
D.
NOO NE,
I.
SIMMON DS
School of Earth Sciences, University of Melbourne, Parkville, Victoria 3052, Australia
ABSTRACT. By co nsider a tion of m odel- ge ne ra ted a tmosph eric da ta, domin a nt
anom alies in th e synoptic circul ation patterns a re observed under conditions of high Antarctic precipitation. T hi s is associated with strong moisture advecti on of m a rine origin.
Examining precipi ta tion at individu al locations reveals a strong rel a ti onship between
local surface temperat ure a nd precipitation a mount. Days with > 5 mm of precipita tion
(which, on average, corresponds to about 8 % of days over Anta rctica) have surface temperatures that a re a round 10°C warmer tha n the mean. This bias suggest that abnormal
conditi ons a re captured in the ice-core record and that interpretation or reconstruction of
palaeotemperatures will succeed only under the possibly fl awed assumption th at similar
abnormal condi tions existed at the time of deposition. Although isotopic a na lysis of Anta rctic ice cores has been used successfull y in pa laeocl imate studi es, a compl ete understa nding of the underl ying processes affecting the deposition of the core rem ains to be
found. It is reaso ned that by obtaining such a n understanding, it m ay be possible to reconstruct the synoptic cond itions under which accumulation occurred.
INTRODUCTION
Th e work of Da nsgaard (1964) o n the behaviour of stable
water isotopes in precipitation has prompted m any studies
of pa laeocl im ates from ice-core data. It is genera lly accepted that the high correlati on between the annual D/H
or 18 0 / 6 0 ratios a nd mean ambient temperature can be
used directly to infer palaeoclim atic temperatures from
ice-core samples (e.g. J ohnsen and others, 1995; Jouzel a nd
others, 1996). J ones a nd others (1993) point out, however,
that la rge-scale temperature trend s cannot be extrapolated
much beyond the locality of the drill-site.
T his empirical approach reli es solely on the observationa l d ata rather tha n the physical processes involved. An
understa nding of the role of atmos pheric circulation in influencing the isotopic sig nat ure revealed in Anta rctic d rill
cores is crucia l to the interp retation of th e core d ata. Wi thout such knowledge, conclusions rega rding palaeoenvironments from a ny typ e of proxy data could be seriously
misleading (Taylor a nd others, 1995). T he key to understanding and inte rpreting the isotopic signa ls in ice cores over
time is two-fold. First, we must consider the a tmospheric
processes th at transport water from a region of evaporati on
to the place of precipitation, and secondly, we need to understa nd the processes governing the isotope concentration
of th e accumulated water. We wo uld expect, for example,
tha t days with precipitation wo uld have greater than average cloud cover and stronger therm a l ad vec tion from lower
latitudes. T hi s would influence the local surface temperature. Such co nditi ons are by defini tion anomalous a nd p erhaps not fully representative of mean temperature. Whil e
thi has been appreciated for some time in principle (e.g.
K ato, 1978; A rista rain a nd others, 1986), littl e a ttempt has
been m ade to explain or quantify the uncertainties to which
398
the interpretation of the isotope- temperature relationship is
subj ect.
To highlight the important as pects of interpretation of
the palaeoclimatic signa l at a given locality, we examine
here some of the processes invo lved in deposition of water
at the drill-site. To perform such a n ana lysis, we need highresoluti on (both spatial a nd temporal) data. Meas urements
of d aily precipitation and other atmospheric para meters a re
un avail able from the observational network. Therefore, to
perform this analysis, we must m a ke use of output from a
global climate model with a complex physical representation of the atmosphere. The implications that a comprehensive understanding of the synoptic processes influencing the
isoto pic signatu re has for existing pa laeotemperature reco nstructions, and for successfull y estimating the synoptic
vari ability of palaeoclimate systems from the ice-core data,
a re disc ussed.
OVERVIEW OF THE NUMERICAL MODEL
The Melbourne University general circulation model
(G CM ) is a non-linear, physically based, spectral, primitive
equation model and is discussed in detail elsewh ere
(Simm ond s, 1985; Simmond s a nd others, 1988; Simm onds
a nd Law, 1995). The spatial resolution is defined by truncation of the spectral series at wavenumber 21, roughly corresponding to a 5.6° x 3.3° grid. In the vertical, there are nine
d iscrete levels in the terrain following "sigma" coordinates,
with the lowest level about 75 m above the surface. For th e
m odel run at this resoluti on th ere are approximately 350
points ove r the Antarctic continent. Many physical processes not governed by the primitive equations are included
via appropriate para meterisations. These include the radiati on scheme of Fels and Schwar zkopf (1975) and Schwa rzkopf a nd Fels (1991), the convecti ve scheme suggested by
Noone and Simmonds: Precipitation analysis and interpretation qf ice-core data
Manabe and others (1965), a soil hydrology scheme following Deardorff's (1977) model and a parameterisation of heat
fluxes over sea ice applied by Simmonds and Budd (1990).
Clouds are included interactively at three levels (Argete
and Simmonds, 1996) to affect the radiation, and snow cover
is prescribed as the climatology from passive-microwave
observations by the Nimbus 7 satellite. The topography
originates from continental heights of Smith and others
(1966) and there is forcing at the lower boundary from the
sea-surface temperatures ofReynolds (1988).
Although modelling Antarctic moisture is a difficult
task in practice, Simmonds (1990) showed that the modelgenerated accumulation rates over the Antarctic are representative of the accumulation estimated from stake measurements in both amount and spatial distribution. As such,
it is thought that the Antarctic precipitation is modelled
with sufficient accuracy and is appropriate for use in this
study. The model results from the second decade of a 20 year
control integration are used in this study.
(a
.. . . ...
c;;;~
: ~
i
.
L_ t.,3,,~
..
:/
Cl
.". '
' 0
O~:~~;~j
SYNOPTIC CONDITIONS
Precipitation is obviously an integral part of ice-core analysis
through accumulation at a drill-site. To determine what
synoptic conditions lead to precipitation, we examine the
mean sea-level pressure (MSLP) anomalies for each day in
the model dataset. One calculates the anomalies by subtracting, from the given pattern, the monthly climatology temporally interpolated to the appropriate calendar day. A
subset of the anomalies is chosen and averaged for the days
on which more than a specified amount of precipitation fell
at a given grid location. The amount chosen here is
5 mm d- \ as it represents high-precipitation conditions and
occurs on roughly one in ten days at coastal locations. This
fraction varies depending on precisely which grid point is
considered. Further, even the most arid points (e.g. Vostok)
show a number of tota ls over 5 mm because model data are
used.
Figure 1 shows the averaged MSLP anomaly pattern for
the coastal point near Casey base (67.7" S, 112.5 E ) and the
inland Vostok base (77.6 ° S, 106.9° E ). Coastal sites (e.g.
Casey ) commonly show a local feature associated with the
approach of a "cyclonic system". This dipolar structure, seen
clearly in Figure la, suggests a northerly air stream which
would advect relatively warm, moist marine air over the site.
Upon cooling, this air would easi ly become saturated and
condense into the modelled precipitation and release latent
heat at the site. For inland cases (e.g. Fig. Ib ), such a dipolar
feature is not usually present, and indeed it is more common
for a single cyclonic feature to be present. This low-pressure
anomaly is probably associated with precipitation formed
by ascend ing air associated with surface convergence.
Kinematic trajectories were generated in three dimensions for the 6 days prior to arrival at the site under consideration, for each day on which more than 5mm of
precipitation occurred. Figure 2 shows the density of the
origin of air parcels 6 days prior to precipitation, together
with the calculated mean trajectory path. This trajectory
reveals that indeed the air is of marine origin for the coastal
site. The direction of approach deviates from that suggested
by the geostrophic assumption as the driving winds from
the GCM include the effects of katabatic flow. Few of the individual trajectories penetrate the region of the circumpolar
0
Fig. 1. MSLPanomaly on days with> 5 mm precipitationfor
( a) Casey and (b) Vastok locations. The contour interval is
0.5 hPa and the zero contour is shown in bold.
trough. The MSLP anomalies and the trajectory analysis
suggest that there are active links between the precipitation
and the local synoptic conditions on any given day,
Analysis of precipitation
We now turn to consider the precipitation over the Antarctic continent in more detail by examining the daily precipitation totals at the 350 model points located over Antarctica.
Although more precipitation events tended to occur in
winter months, individual precipitation events with high
totals occurred throughout the year. These data revealed
that about half of the days showed no or negligible precipitation, From the same 10 year model dataset, we also extract
the surface temperatures. For each model point, we calcu late the surface-temperature anomaly as the difference
between the daily temperature at the site and the climatolo399
Noone and Simmonds: Precipitation analysis and interpretation if ice-core data
test days are warmer than the average by up to 15°C, and
days with over 5 mm of precipitation (around 8% of days )
are over 10°C warmer. If the dataset is broken down seasonally (Fig. 3b), we Gnd that the trend is stronger in autumn
and winter (up to 20 ° and l8 °C, respectively) than in
summer and spring (5° and 12°C). The gradient of this trend
is also slightly larger if only coastal locations are considered.
A similar trend is found at other model temperature Gelds
although it is not as pronounced. For example, the lowest
atmospheric level in the model (sigma level 0.991) displays
warm tendencies up to 2°C.
During precipitation events warm air is advected from
the north, and the coastal regions of the continent are subject to greater cloud cover. Hence, it is not clear a priori
whether the warm surface conditions of high-rainfall days
are associated more closely with the extra southward sensible-heat transport or the additional trapping of longwave
25
20
15
........
Cb)
...........-.
20
2:'
'0"
E
10
•..•...
c:
5
.§
...;::l'"
Q)
15
0
j
-;;;
...
0.
Q)
c:;
Q)
...
0.
6
-5
10
~
-10<1.1
"...
~
- 15
,.'
x
;::::s
m
- 20
.
- 25
......
\ ..
a
....
3650
rank by precipitation amounl
....
.,......
....
-" " "
20
,.../
-':'"
15
102
to
E
Fig. 2. Average trajectory paths and density if trajectory originsfor the 6 days prior to precipitation at ( a) Casey and ( b)
Vostok locations. Dark regions show higher density in arbitrary units if trajectory origins per unit area. An asterisk
marks model locations.
0
5
c:
...'"
Q)
0
..,;::l
...'"
Q)
-5 0.
E
~
- JOQ.)
gical mean interpolated to the appropriate time of the year.
Thus, for each of the 350 Antarctic points, a 10 year time
series of both the precipitation and surface temperature
was available.
Ordering the precipitation totals at each point by
amount from lowest to highest shows a distribution of an exponential nature, common in precipitation data (e.g. Noone
and Stern, 1995). This highlights the fact that only a small
percentage of precipitation events have high precipitation
totals. The shaded region of Figure 3 shows the average of
these distributions. ,,ye plot the surface-temperature anomaly corresponding Lo th e precipitation amounts and then
average across the 350 points.
Figure 3a shows a clear monotonic trend in the associated surface-temperature anomalies. We see that the wet400
"...
~
-15:::l
m
-20
-25
b
910
rank by precipilalion amount
Fig. 3. M ean precipitation amount ( shaded region ) and
corresponding mean surface-temperature anomaly (curve).
The x axis is ranked by precipitation amount from lowest to
highestJor ( a) all data (3650 days ) and (b ) seasonal comparison (910 days). Surface-temperature anomaly is marked
on the right axis in o ~ and precipitation is on the lift axis in
mm a'.
Noone and Simmonds: Precipitation analysis and interpretation qf ice-core data
ra diatio n by the cha nged clo ud conditions. This question
can be a nswered simply by p erforming experiments in a
GeM in which we prescribe cloud cover of give n amounts.
Specificall y, we set the cloud concentration at all three cloud
levels to have, first, 100 % a nd, second, 0 % a nd perform a
3 year integration. T he close simil a rit y in th e temperature
trends res ulting from these two runs (not shown) suggests
tha t indeed th e u nderl ying mecha nism is wa rm advection
rather than radiation associated with greater cloud
a mounts. This conclusion is supported by the earlier result
that days with p recipi tation have a predomin ant north erl y
air stream advecting moistu re to the precipitati on site. Simila rl y, the reduction in the temperature trend for inl and sites
results from both weaker a dvection and a drier ai r mass.
The trend seen in Figure 3 has a n obvious co nsequence
for recreating pa laeoclimates from ice-core records. Specifically, as it is the d ays with high-precipitation events which
a re recorded in the ice cores, a nd these days are shown to
have sig nificantl y higher th an ave rage temperatures, the
core isotope data will refl ect these warmer temperatures.
DISCUSSION
The naive interp retation of the above result suggests that
a ny reconstruction of palaeoclimates from core isotope d ata
will overestimate the temperatures if one does not include
this warm bias. H owever, the relationship between the isotope concentratio n a nd temperature is determin ed by emp irical means fo r a give n site or a rea. Th erefore, the bias on
a ny give n d ay has a lready been included in the regression.
We note tha t such a method foc uses on the annu al m ean
temperatu re, and that d aily informati on associated with individua l precipita tion events is lost.
The appropri ate interpretati on is that th e co re data represent conditions tha t a re in some way abnorm a l. This is
true in the simple case of surface temperature a nd in the
more complex case of the synoptic situation as seen above
with the example of MSLP a noma li es. Th at is, the ice-core
sig nal will refl ect a circulation regime that is significantly
different from the mean conditions.
An importa nt consider atio n that foll ows from thi s is the
tempora l va riation of the bi as. We saw in Fig ure 3b that th e
temperature a nomali es on precipitation d ays in the autumn
a nd winter differ significantly from th ose in the warmer
summer and spring months. "Ve attribute this difference to
more intense baroclinic acti vity leading to stronge r norther ly advec tion of wa rm, moist air during autumn and
winter. As very littl e is known about th e ci rcul ation conditions of various pa laeoclim ates, it is difficult to say whether
such a seasonal variation would prevail. Simil a rl y, it is likely
tha t the stre ng th of circulati on was different from th at of tod ay. Thi s is importa nt insomuch as we cannot be sure that
the co nditions under which acc umul a ti on can occ ur have
the same bi as over time.
P. Valdes and others (p ersonal communicati on, 1997)
indicate from simul ations with a GeM that there is a reduction in streng th of polar ba roclinicity during th e L as t
G lacia l M aximum . Foll owing th e above a rgument, we
wo uld ex pect a reduced slope in the temperat ure bias. This
wo uld resul t in a n underestim ate of pa laeotemperatures
from the regression relation deri ved from experimental d ata
coll ected under tod ay's conditi ons. Thi s scena ri o is incomplete, in th at altering the circul ation strength will a lso intro-
duce changes in oth er complex atm ospheric processes like
evaporati on, transport a nd precipitation.
It m ay be possible to qu antify, to some extent, the differences be tween var ious globa l circul ati on regimes by considering the vari ati on in acc umulati on rates in ice cores. One
could focus on times when there is a shift from glacial to interglacial periods. If there is a pa rticula r peri od of cha nge that is
tempora lly consistent over a la rge geographical domain, it is
possible there was a sig nificant change in the circulation
regime a nd that empirical te mperature reconstruction would
be less reliable.
10 obtain detailed inform ation abo ut the vari a tions in
isoto pic content from a n ice co re on short time-scales wo uld
be very difficult, a nd indeed m ay be impossible. Th e u e of
fll0delling techniques, instead of observationa l data, m ay
yield a valu abl e approach. Running a n ensembl e of si mu lations th at re present a number of pa laeoclimate systems
would give a n indicati on of the range of circulation regimes
that m ay ex ist. Incorporating a hydrological cycle that
allows for the stable-isotopic forms of water wo uld then give
a n indication of the spati a l a nd te mporal vari ations in th e
isotopic concentrations, along with estim ates for accumul ati on rates. This wo uld p rovide a useful tool in linking the
measured isotope d ata in ice co res to the synoptic conditi ons
via the precipit ation signal. This wo uld a llow one to gain
more informati on than is give n simply by the m ean conditions.
CONCLU DING R EMARKS
It is of g reat importa nce to obtain a complete understa nding
of the precipi tatio n p rocesses leading to acc umulation
before analyzing ice-core data. The exa mpl e shown here
highli ghts how, without such knowledge, th e regress ion
techni q ues com mo nl y used to reconstruct pa laeo temperatures a re q uesti onable, in that the ass umed rela tionship
may be different under a pa laeoclima tic regime.
Through precipi tati on, it is possibl e to associate the
meas urements in th e 18 0 record with the prevailing synoptic condi tions of the day. As there is onl y one d ay in ten th at
exhibits a high-precipitati on event leading to sig nificant accumul ation, th ere will be a bi as in the ice-core compositi on
towa rd conditions that a re in so me way abnorm al. In th e
results given here, these conditions ar e warmer. Th e
increased synoptic activity shown in the MSLP a noma li es
is a lso not representati ve of th e "average" condi tions a round
a ny single Anta rctic site.
An extension of th e a rg ument of bi as in ice-co re records
presented ass umes a simila r bias wo uld have been evident in
a da tase t represe nting the circul ati on patterns of pa laeoclimate systems. Specificall y, one can infer from the ice-core
measurements only the condi tions of the abnorma l synoptic
situ ati on tha t lead to precipitation. Nonetheless, it m ay be
possible to assess th e expected vari ability of vari ous palaeoclim ates by co nsidering the extreme conditi ons expected
during precipitati on. G e M experiments th at recreate th ese
extreme conditi ons of precipi tation will also co ntain informati on about d a il y circulation that does not result in precipitati on. Th e difference between th ese two conditions will
give a n indicati on of the atm ospheric vari abili ty on a daily
time-scale prese nt under a given pa laeoclimatic circulation
regime. By consideration of a range of pa laeoclimates a
comp rehensive understa nding of the processes underlying
401
Noone and Simmonds: Precipitation analysis and interpretation of ice-core data
the isotopic deposition cou ld b e obtained. Palaeoclim ate reconstructions could then include estimates of the temporal
differences in the variability of th e global atmospheric circul ation, as well as the mean conditions that are currently
reported in the literature.
REFERENCES
Argete, A. a nd I. Simmonds. 1996. Compa rison of tempora l cloud vari ability simulated by a GCM with observations from t he Nimbus-7 satellite.
Atm6sJera, 9 (1), 1-21.
Aristarain, A.]. , ]. J ouzel a nd M. Pou rchet. 1986. Past Anta rctic Penin sul a
clim ate (1850- 1980) deduced from a n ice core isotope record. CLimatic
Change, 8 (1), 69- 89.
Dansgaard, W. 1964. Stabl e isotopes in precipitation. Tellus, 16 (4), 436--468.
Deardorff, ]. W. 1977. A parameteri zation of gro und-surface moist ure content for use in atmospheric prediction models. J Appl. Meteorol. , 16(11 ),
1182- 1185.
Fels, S. B. a nd M.n Schwa rzkopf. 1975. T he simplifi ed excha nge approximation: a new method for radi ati ve transfer calculations. ] Atmos. Sci. ,
32 (7), 1475- 1488.
J ohnsen, S.]., 0. Da hl:Jensen, W. Da nsgaard a nd N. S. Gundestrup. 1995.
G reenl and pa leotemperat ures derived from GRIP borehole temperatu re a nd ice core isotop e profi les. TeUus, 47B (5), 624--629.
J ones, p. 0. , R . M arsh, T. M. L. Wigley and 0. A. Peel. 1993. D ecada l timescale links between Anta rctic Peni nsul a ice-co re oxygen-18, deuterium
and temperature. HoLocene, 3 (1), 14- 26.
J ouzel,]. and 14 others. 1996. Climatic interpretation of the recently extended
Vosto k ice core records. CLimate Dyn., 12 (8), 513-521.
K ato, K. 1978. Factors controlling oxygen isotopic compos ition of fa ll en
402
snow in An tarctica. Na ture, 272 (5648),46-48.
M a nabe, S., ]. Sm agorinsky and R . F. Strickler. 1965. Simul ated clim atology of a genera l circ ul ation model with a hyd rologic cycle. Mon. Weather
Rev. , 93 (12), 769-798.
Noo ne, n and H. Stern. 1995. Verification of ra infall forecasts from the Austra li an Bureau of Meteo rology's Globa l Assimi lation a nd Prognosis
(GASP) system. Aust. Meteorol. Mag., 44 (4), 275-286.
Reynolds, R .W. 1988. A real-time global sea surface temperature a nalysis.
] CLimate, 1 (I), 75-86.
Schwa rzkopf, M . n and S. B. Fels. 1991. The simplified excha nge method
revisited: an accurate, rapid method for computati on of infra red cooling rates and flu xes. J CeoMys. Res. , 96 (D 5), 9075-9096.
Simm ond s, 1. 1985. Analysis of the "spinup" of a genera l circul ation model.
J Geophys. Res., 90 (D 3), 5637- 5660.
Simm ond s, 1. 1990. Improvements in general circulation model performance in simul ating Antarctic climate. Antarct. Sci., 2 (4),287- 300. (C orrection: Antarct. Sci. , 3 (2), 1991, 230.)
Simmond s, I. a nd W. F. Budd. 1990. A simpl e pa rameterization of ice leads
in a general circulatio n model , and the sensiti v ity o f clim ate to change
in Antarctic ice concentration. Ann. Claciol., 14, 266-269.
Simmonds, !. a nd R. Law. 1995. Associa tions between Antarctic katabatic
fl ow and the upper level winter vortex. Int. ] CLimatal., 15 (4), 403- 421.
Sim mond s, 1. , G. Trigg and R. Law. 1988. Th e climatoLogy if the Melbourne Uni versiry generaL circuLation modeL. M elbourne, University of M elbourne. Depa rtment of M eteorology. (Publ ication 31.)
Smith, S. M., H ."V. Mena rd and G. F. Sha rman. 1966. vVorLd-wideoceandepths
and continentaL elevations averaged Jor areas approximating one degree squares 'If
latitude and Longitude. La J olla, CA, Scripps Institute of O ceanography.
(SIO Reference Report 65 -8.)
Taylor, R. B. , D.]. Barnes a nd A. M. Lough. 1995. On the inclusion of trace
materials into massive coral skeletons. I. Materials occ urring in the environment in short pulses. J Exp. Mm: BioI. Ecol., 185, 255-278.