29
University of Groningen Laser spectrometry for stable isotope analysis of water van Trigt, R IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below. Document Version Publisher's PDF, also known as Version of record Publication date: 2002 Link to publication in University of Groningen/UMCG research database Citation for published version (APA): van Trigt, R. (2002). Laser spectrometry for stable isotope analysis of water: biomedical and paleoclimatological applications Groningen: s.n. Copyright Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons). Take-down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim. Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum. Download date: 07-06-2018

04. Chapter 4 · See Figure 4.1. In its simplest form, this model assumes that all water evaporates in tropical. Chapter 4 114 ... water vapour is deposited on smaller flakes and

  • Upload
    donhan

  • View
    215

  • Download
    2

Embed Size (px)

Citation preview

University of Groningen

Laser spectrometry for stable isotope analysis of watervan Trigt, R

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite fromit. Please check the document version below.

Document VersionPublisher's PDF, also known as Version of record

Publication date:2002

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):van Trigt, R. (2002). Laser spectrometry for stable isotope analysis of water: biomedical andpaleoclimatological applications Groningen: s.n.

CopyrightOther than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of theauthor(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons thenumber of authors shown on this cover page is limited to 10 maximum.

Download date: 07-06-2018

4Glaciological application

Chapter 4

112

Glaciological measurements

113

4. Glaciological applicationIsotope ratio measurements of water are widely being used in the study of the past climate. The

“proxy climate signal” that is contained in the isotope ratios is brought about by isotope fractionation

effects that occur in the meteoric water cycle. Since the magnitude of the effects is dependent on

climate indicators, especially on the local cloud temperature, isotope ratio measurements along depth

profiles of ice cores (natural precipitation archives) can be used to reconstruct Earth’s paleoclimate. By

now, several deep ice cores have been drilled both in Antarctica and on Greenland and from the

measurement results, much has been learned about the history of Earth’s climate. Here, we demonstrate

the first application of the new laser spectrometry (LS) method in ice core measurements and make a

first attempt to interpret the results. In this chapter an introduction to ice core research in general, the

measurements of ice core samples and the interpretations of the results will be presented. The latter

part is largely based on a paper submitted to “Annals of glaciology” (Van Trigt 2001b).

4.1 Introduction

4.1.1 Equilibrium and kinetic fractionation

As explained in Chapter 1, two kinds of isotope fractionation processes are distinguished:

Equilibrium and kinetic fractionation. Most often, in natural processes a combination of these two is

found, although some processes can be considered as being purely equilibrium (see also Chapter 1). The

fractionation ε for evaporation of water under equilibrium conditions is –9.71‰ and –78.4‰ at 20 ºC

for 18O and 2H, respectively (Majoube 1971). This implies that, after equilibrium is established, the

vapour is 9.71‰, respectively 78.4‰, depleted in the respective isotope abundances compared to the

water it is in equilibrium with. For kinetic fractionation it is much harder to measure accurate values,

since it is not easy to entirely separate the effect from its equilibrium counterpart. Moreover, it is often

difficult to accurately and quantitatively describe the physical processes leading to the kinetic fraction

under consideration. To give an indication of a kinetic process: diffusion of water vapour through dry air

has values for ε of about –27‰ for δ18O and –23‰ for δ2H (Merlivat 1978).

4.1.2 The Rayleigh process

The simplest model that can be used for the description of isotopic behaviour in the hydrological

cycle is the worldwide distribution of water vapour via the Rayleigh process, also known as Rayleigh

distillation. See Figure 4.1. In its simplest form, this model assumes that all water evaporates in tropical

Chapter 4

114

regions (Dansgaard 1964, Mook 2001). Average ocean water is defined as 0‰ (with respect to

VSMOW) and therefore in principle the composition of the vapour can be calculated, if a value for the

relative contributions of kinetic and equilibrium fractionation is assumed. This vapour will be isotopically

lighter than the ocean water. Subsequently, the water vapour is transported to higher latitudes. Due to

the prevailing lower temperatures, condensation will take place and rainfall will occur. During rainout,

fractionation will occur again (condensation is the opposite process from evaporation), thus further

depleting the remaining vapour in the heavier isotopomers. This process continues up to arrival at the

poles, where the last vapour freezes out as snow.

This very simple model already produces reasonable qualitative results in interpreting stable

isotope signals of precipitation and can be used to provide insight in the physical processes. Many

refinements to this very coarse model are possible and have indeed been made (e.g., Mook 2001).

Nowadays complicated atmospheric General Circulation Models (GCMs) are used to model the climate

system and to simulate isotope signals (Hofmann 2000). The transport, evaporation and condensation

phenomena in these GCMs are modelled in a much more reliable way; still they are in principle based on

Rayleigh processes.

Figure 4.1: Schematic representation of the Rayleigh process of the 18O depletion of water vapour when

flowing away from the Equator. For 17O and 2H similar plots can be drawn.

Glaciological measurements

115

4.1.3 Meteoric Water Line

For 2H and 18O (and 17O) the above described systematics of isotopic fractionation are very

similar. This implies a positive correlation between the 2H and 18O isotope concentrations, or abundance

ratios. Friedman (1953) was the first to report a relation between these isotopes for precipitation from

various parts of the world. Later it was quantified by Craig (1961a) as:

δ δ2 188 10H O= ⋅ + (4.1)

This relationship is known as the Meteoric Water Line (MWL). The MWL is a worldwide average

(therefore Global MWL or GMWL are also used). On a regional scale its slope and intercept may differ

from the standard values as found in Equation 4.1. Still the GMWL is useful as a starting point for further

interpretation of hydrological stable isotope data. Moreover it can help in understanding the different

processes that occur in the hydrological cycle.

The slope of 8 of the GMWL can be understood by first assuming equilibrium conditions in

evaporating and condensing water vapour. The ratio of the respective equilibrium fractionation factors of2H and 18O is slightly higher than 8; the slope decreases to the GMWL value of 8 because of a remaining

kinetic component in the evaporation process, which has nearly the same fractionation factor value for

both isotopes. It is believed and understandable that this kinetic influence appears most prominently

during the evaporation, where wind and humidity play important roles. In clouds, where condensation

takes place gradually, isotopic equilibrium is easily established. In the formation of snowflakes, however,

water vapour is deposited on smaller flakes and an additional kinetic effect is expected (Jouzel 1984,

Souchez 2000). For local MWLs slopes between 5 and 8 are being found.

The intercept of 10 in the GMWL is another consequence of the kinetic contribution in the

evaporation of (ocean) water (Kendall 1998). In local MWLs, the variations found in this intercept are

larger than in the slope.

4.1.4 Climate signal

From the model it follows that the degree of depletion compared to ocean water is dependent

on the temperature difference between the source region (in this coarse model the tropics) and the

precipitation region. Since summer-winter temperature differences tend to be larger at longer distances

from the equator, the yearly cycle shows a larger amplitude in higher altitude regions. As an example,

this seasonality is observed in the 18O and 2H isotope abundance ratios from precipitation in The

Netherlands. Figure 4.2 shows the monthly mean δ18O values of all rain and snow in three stations in

Chapter 4

116

The Netherlands between 1981 and 1995. The difference in summer and winter values is at most 3‰

for 18O, the yearly average is about –7‰. The same seasonality can be seen in precipitation in polar

regions, as an example data from Nord (Greenland) are shown (Figure 4.3). Here the summer-winter

variance is much larger, up to 15‰. The yearly average value is around –25‰. For 2H comparable

figures can be plotted. Nowadays the International Atomic Energy Agency (IAEA) and the World

Meteorological Organization (WMO) collaborate in collecting data on isotope ratios of precipitation in the

Global Network for Isotopes in Precipitation (GNIP). This database holds data from more than 500

stations world-wide, analysed by over 200 laboratories, going back to as far as 1961 (Araguas-Araguas

2000, IAEA 2001, http://isohis.iaea.org).

Figure 4.2: The average seasonal cycle of δ18O in precipitation in Groningen (53.14º N), Beek (50.54º N)

and Wieringerwerf (52.52º N) as analysed at the CIO. The plotted points are averages of the δ18O values

determined for the precipitation for that particular month over a range of 15 years (1981 – 1995). The

error bars indicate the deviations in the mean and are therewith a measure for the interannual

variability. The points have been fitted with a two-harmonics curve (CIO Scientific Report 1995-1997,

original data also available at the GNIP database).

Glaciological measurements

117

Figure 4.3: Average (1961 – 1972) 18O depletion in monthly precipitation in Nord (81.60º N), data taken

from the GNIP database, data analysed by University of Copenhagen, Copenhagen, Denmark.

The most obvious difference between Figure 4.2 and 4.3 is that average values are lower in

polar regions than in moderate climates (such as Groningen) due to continuing Rayleigh distillation

(Dansgaard 1964). This trend is reflecting lower average local temperatures and is referred to as the

latitude effect. Other effects that can be deduced from observations are the altitude effect (more

negative values at increasing surface elevation), the continent effect (more negative values and larger

seasonal signals further from the coast) and the precipitation effect (more negative values in periods

with more precipitation). All of these effects can be directly understood in a qualitative sense from the

Rayleigh distillation model. Like the latitude effect, the altitude effect is related to average local

atmospheric temperatures. The continent effect is caused by the gradual depletion of the atmospheric

water vapour during its journey over land. The precipitation effect is again, loosely, coupled to the local

temperature. The existence of these different influences on the isotope signal imply that the isotope

signal is certainly not a “perfect” climate measure, but rather a powerful “proxy” to climate (Lajtha

1994).

Chapter 4

118

4.1.5 Paleotemperatures (climate)

The same isotope information as in present day precipitation is in principle conserved in the

kilometers thick ice layers in the Arctic and Antarctic regions. After all, these layers can be regarded as

natural archives for (hundreds of) thousands of years of precipitation. The old ice can therefore provide

a proxy for past climate and climate changes. For recent times (the upper ice layers) we find the same

seasonality as in our local measurements on precipitation. See for example Figure 4.4. For deeper layers

with older ice, the resolution is not sufficient (due to compression of layers and to diffusion) to reveal

seasonality. Still, the average over one or more years provides us with valuable information about the

past climate, with still a better time resolution than can be obtained with other types of archives (e.g.,

pollen or ocean sediments). The typical resolution that can be obtained in e. 10,000 year old samples is

in the order of a few years, or better.

Figure 4.4: Part of the δ18O depth profile along the GISP ice core. The y-scale is from –25‰ to –35‰

with respect to VSMOW. Dark coloured peaks indicate summer periods. The age is centered around

1325 years AD. The seasonal cycles can be clearly observed [J. Glac. Vol. 20, No. 82, p 12, 1978].

With the above in mind, many studies have been done on ice cores (e.g., Dansgaard 1989,

Grootes 1993) drilled on selected locations in Greenland and Antarctica. For these locations it is

important that the layers are stacked in a well-organised way. The high pressure caused by the younger

snow makes the oldest (deepest) layers to be pressed to the sides of an ice area. The local ice dynamics

should thus be well-understood in order to be able to determine the age of the layer. After drilling a

deep core, the ice is stored for later measurements, among which the isotope ratio measurement in the

laboratory.

An interesting ice coring effort was made by a number of countries at Vostok station in

Antarctica in an attempt to construct a climate record up to 420.000 years ago (Petit 1999). This long

period allows scientists to study the past four glacial-interglacial cycles. Another example is a joint

European effort, the European Project for Ice Coring in Antarctica (EPICA). Again, the aim is to

reconstruct past changes in climate and atmospheric composition with high resolution. A linkage and

comparison with the Greenland Ice Core Project (GRIP) will be made to determine whether the changes

Glaciological measurements

119

observed in Greenland were global events or more regional ones (see also Mazaud 2000). Alternatively,

ice cores are being taken on smaller ice caps (Canadian Arctic, Spitsbergen) and in alpine regions

(Himalaya, Andes, Alpes) where the ice history is not going back so far. As an example, the exploration

of permanent glaciers for past hydrological and environmental parameters in the Alps can be mentioned

(Stichler, 2000).

A huge advantage of the use of these isotope records is the high resolution of the archives and

the high degree of certainty that the record is sequential and complete. Other natural archives often

suffer to a greater extent from diffusive effects. A complication of ice core archives is the uncertainty on

how to use the isotopic composition of ice sheets as a paleothermometer or, in other words, how to

relate the measured isotope ratios with past temperature. These questions have already been subject to

discussion for a long time (e.g., Mix 2001). In our present climate it is straightforward to calibrate the

isotope thermometer in many different regions on a local scale (depending on local altitude, latitude,

continental and precipitation effects), by measuring both atmospheric temperature and isotope ratios of

precipitation over a certain period (e.g., data from the GNIP database). The isotope ratios turn out to be

linearly dependent on atmospheric temperature and for the other climate parameters, relationships can

be found as well (Dansgaard 1964). This is also referred to as the spatial isotope/surface temperature

relationship. Dansgaard has already found a good correlation, valid for coastal and polar locations with

an average change in δ18O of 0.7‰ per degree Celsius (Dansgaard 1961) and this value was later

confirmed for Greenland (Johnsen 1989). For Antarctica, values of 9‰ per degree Celsius for 2H are

estimated (Salamatin 1998). If the assumption is made that these relations did not change in time, it is

possible to translate paleo isotope abundances into temperatures via the so-called transfer functions.

Indeed, for the past few hundred years in Greenland a significant correlation of stable isotope

concentrations and both local as well as more regional meteorological and climatic parameters exists

(White 1997): It can be concluded that under the present climate the assumptions on stability of the

influence of the parameters hold. Over longer time scales, however, many complicating factors exist that

have not yet been fully understood. These remain uncertainties in the input of the General Circulation

Models (GCMs) which are used to model the paleotemperatures from measurement results. Amongst the

uncertainties are changes in (1) surface altitude, (2) seasonal distribution of precipitation, and (3) the

evaporative origin of the moisture in time (Jouzel 1997, Werner 2000). It might well be too blunt an

approximation to just using one fixed number for relating isotope ratios with temperature. Indeed,

strong evidence exists that it is not correct to use the spatial relations over the entire time scale spanned

by the ice core. Direct temperature measurements in the ice core boreholes suggest that local surface

paleotemperatures were much lower than predicted by the results from ice core measurements. From

these borehole temperatures it was concluded that at the time the last glacial period reached its lowest

Chapter 4

120

temperatures (the so-called last glacial maximum, LGM) the average temperature in Greenland appears

to have been 22ºC colder than today (Johnsen 1995). This is almost double the difference derived on

the basis of older/initial analyses of ice core data. Current insights in the isotopic make-up of the ice

sheet using glacial circulation models are siding with the borehole derived temperature. A remaining

question and point of discussion is what both methods do really measure; ice cores are primarily

sensitive to temperatures in the atmosphere and clouds at the top of the inversion layer at the moment

the precipitation fell, while boreholes reflect more directly the average local surface temperatures. It is

believed that at the LGM the precipitation was more concentrated in the summer and that the

temperature inversion was stronger than at present day. Thus, the values do not necessarily contradict

each other.

Another problem in reconstructing climate history based on ice core measurements is to

determine the exact age of the deep ice. In modern interpretations a number of parameters is

simultaneously used for dating. In the upper layers one can measure seasonality in isotope ratio signals

and thus count layers, comparable to counting tree rings. In deeper, more compressed layers, however,

due to diffusion the signal has almost disappeared (and the yearly slices of ice become too thin due to

compression). The classical approach to dating is then calculating the age/depth relationship using ice

flow and ice-accumulation models (Lorius 1985). Although ice flow models have substantially been

refined in the course of years, several alternative techniques have also been presented. Among those

are radiocarbon dating of old atmospheric CO2 (Van der Wal 1994) and measuring the CH4 concentration

(Blunier 1998), both trapped in bubbles in the ice. Another method is the counting of layers using a

systematic combination of parameters, such as visual stratigraphy, electrical conductivity, laser-light

scattering from dust, volcanic signals (also dated by e.g., deep-sea isotope records), and major ion

chemistry signals. For example, a core with a length of over 3000 m has been dated in this manner up to

160.000 years BP (Meese 1997). Uncertainties are typically a few percent, but up to 20% for the deepest

layers. And yet another means is to fit the major features of the stable isotope signal to Milankovich

oscillations of the earth’s orbit which have a known frequency (Salamatin 1998).

The interpretation of ice core information is a continuing debate, but as our understanding

increases, more and more of the information about the past climate will be disclosed.

4.1.6 Deuterium excess

The so-called “deuterium excess” d was defined by Dansgaard (1964) as:

d H O= − ⋅δ δ2 188 (4.2)

Glaciological measurements

121

and it can be considered to be a measure of the difference in behaviour between 18O and 2H, or the

contribution of kinetic isotope fractionation effects to the formation of the precipitation. For the GMWL

(per definition) a value of 10‰ is found. Local MWLs often have slopes that differ in time (over the

year, but also over ages or millennia). The value of deuterium excess for local measurements is per

definition (4.2) calculated with a fixed regression of δ18O versus δ2H with a slope of 8, and thus, when

the true slope (ratio) for some reason changes in time, the calculated deuterium excess changes. As an

example, Figure 4.5 shows the trend for the deuterium excess for Groningen precipitation between 1964

and 1996.

Figure 4.5: Deuterium excess for Groningen precipitation. Its trend is compared with the NAO index for

the period 1964 – 1996. Although some long term correlation seems likely, evidence for interannual

variability correlation is lacking.

A comparison with the North Atlantic Oscillation (NAO, a quasi-periodic change in sea surface

temperature and atmospheric moisture in the North Atlantic) is made in this plot as well. It is likely that

some correlation exists between the two since most of the Groningen precipitation originates from the

Northern Atlantic ocean, but it is only observable in the long-term trend and not in the interannual

Chapter 4

122

variability. The average seasonal cycle in Groningen of the deuterium excess (detrended) is shown in

Figure 4.6. Here, a clear and significant pattern exists. Its interpretation, however, is not

straightforward.

Figure 4.6: Average seasonal cycle in the deuterium excess, the measured data were detrended and

averaged over the years. The points have been fitted to a two-harmonics curve. The error bars are a

measure for the interannual variability (CIO Scientific Report 1995-1997)

A changing regime of evaporation in the source area (caused by changing humidity, wind, or

waves or by a seasonal variation of the source region) will alter deuterium excess values, because the

relative kinetic contribution to the evaporation process will change. A change in the form of precipitation

(e.g., snow instead of rain) or other processes in the clouds can influence the kinetic contribution and

therewith the deuterium excess signal in a similar way (Ciais 1994). Calculations in which it was tried to

derive individual contributions of possible factors have been made (Jouzel 1982). Another example is the

Law Dome shallow ice core in Antarctica. Here, seasonal δ18O and δ2H cycles were found to reflect the

local temperature, but the deuterium excess signal is shifted four months backwards in phase (Delmotte,

2000). From this, the different sources of the precipitation in the different seasons were identified. A

Glaciological measurements

123

comparable lag is seen in high-altitude regions of the Greenland ice sheet and also in this case it was

possible to draw conclusions concerning the origin of the water vapour (Dansgaard 1989).

For polar regions it is now widely accepted that deuterium excess is above all affected by (1) the

temperature of the moisture source and (2) the absolute humidity in the source region of the

precipitation (Fisher 1991). Complex GCMs can nowadays predict these factors quite well for the present

day situation. However, relatively simple Rayleigh-type models can do this too, under most

circumstances (Armengaud 1998). Still, we have to keep in mind that these models all start with the

well-known present-day circumstances. The same models do not (yet) succeed in a reliable

reconstruction of the past climate using reverse-modelling of the paleo 2H and 18O isotope ratio signals,

let alone the deuterium excess. Furthermore, both the simple and the complicated models can only

simulate large scale effects, while measurements are always done on a local scale (Jouzel 1996).

Nowadays, in many studies deuterium excess values have been determined, providing

information additional to that of 18O or 2H values alone. The extra information that becomes gradually

available in this way has not been fully exploited yet.

4.1.7 Traditional ice core isotope measurements

Stable isotope ratio measurements are usually performed on dedicated isotope ratio mass

spectrometers (IRMS). For measuring the stable isotope abundance ratios of 18O/16O and 2H/1H in water,

extensive sample pre-treatments are necessary. Traditionally, off-line methods are used. In the case of

deuterium measurements, water is reduced to H2 gas over hot uranium (Bigeleisen 1952) or zinc

(Friedman 1953, Coleman 1982). In the case of 18O ratio measurements, the isotope signal in water is

often transferred to CO2 of known isotopic composition by equilibration, often referred to as the

Epstein/Mayeda technique (Epstein 1953). These techniques and some alternatives are described in

more detail in the introduction of this thesis (Chapter 1).

It requires an enormous effort to analyse an isotope depth profile over the entire length of a

typical ice core, since the traditional techniques are laborious and ice coring delivers many thousands of

samples. Therefore a number of techniques has been developed in order to automate the traditional off-

line techniques. For example, for δ18O, on-line automatic equilibration systems have been built (Johnsen

1997), and also for δ2H measurements the traditional method has been automated (Vaughn 1998). Both

methods are based on traditional techniques, but are optimised and automated to handle a larger

number of samples.

More recently, new on-line continuous-flow (CF) techniques have been developed that use

different approaches. For deuterium measurements, equilibration of hydrogen gas with water using a

Chapter 4

124

catalyst and alternative on–line reduction methods coupled to continuous flow IRMS (CF–IRMS) are used

(Meijer 1999 and references therein, Brand 1996). As a catalyst for the H2–H2O equilibrium reaction,

platinum is used (Horita 1988, Coplen 1991). Reducing materials reported in the literature include

chromium (Gehre 1996) uranium (Vaughn 1998, Hopple 1998) and zinc (Socki 1999). On–line pyrolysis

of many different sample types, water included, coupled with CF–IRMS is another promising

development (Begeley 1997). For 18O, however, the problems are smaller, and consequently less efforts

have been taken to improve the existing automated systems, based on traditional Epstein/Mayeda

processes. Still, some alternatives were published: again on–line pyrolysis (CO is formed) coupled with

CF–IRMS (Kornexl 1999, Wang 2000), or on–line isotopic exchange with CO2 bubbles in a long capillary

at elevated temperatures (Leuenberger 2001).

From all these new techniques the best results report precisions of about 0.05‰ for δ18O and

about 0.6‰ for δ2H and these are comparable to the best precisions attainable with traditional

methods. However, international interlaboratory comparisons in which selected laboratories perform

measurements in ring tests, show larger spreads than the mentioned values. When calculating

deuterium excess the situation gets even worse, because there is no correlation between the deviations

of the isotopes. In other words, a laboratory that gets somewhat lower than average results for one

isotope might give slightly higher values for the other. Therefore Meijer (1999) reports the spread of

deuterium excess in an interlaboratory comparison to be almost ±4‰ (2σ). Note that this is an

important observation for comparison of deuterium excess results of different laboratories, but not

necessarily for the observation of trends.

4.2 Groningen ice core measurements

The text in this paragraph is based on a paper published in “Annals of Glaciology” (Van Trigt 2001b).

4.2.1 Abstract

We report on the first application of a new technique in ice core research, based on direct

absorption infrared laser spectrometry (LS), for measuring 2H, 17O, and 18O isotope ratios. The data is

used to calculate the deuterium excess d (defined as δ2H - 8·δ18O) for a section of the Dye-3 deep ice

core around the Bølling transition (14,500 BP). The precision of LS is slightly better than that of most

traditional methods for deuterium, but not for the oxygen isotopes. The ability to measure δ17O is new

and is used here to improve the precision of the δ18O determination. Still, the final precision for δ18O

remains inferior to traditional isotope ratio mass spectrometer (IRMS). However, its accuracy may be

Glaciological measurements

125

better, as the LS measurements are not affected by sample contamination by, e.g., the drilling fluid.

Therefore, deuterium excess was calculated from a combination of the LS and IRMS isotope

determinations.

4.2.2 Introduction

Isotope ratio measurements of δ18O and δ2H of water have been and are being widely used in

the study of the past climate. The “proxy climate signal” that is contained in the isotope ratios is brought

about by isotope fractionation effects that occur in the meteoric water cycle. Since the magnitude of

these effects is dependent on climate indicators (especially the local cloud temperature at the time of

precipitation) isotope ratio measurements on ice cores can be used as a temperature proxy (Dansgaard

1964). By now, several deep ice cores have been drilled, both in Antarctica and on Greenland, and

analysed for a variety of parameters, such as electrical conductivity, dust, chemical constituency and

isotope concentrations. From these measurements, much has been learned about the paleoclimate.

However, in practically every single case only δ18O or δ2H has been measured; rarely both isotopes have

been measured simultaneously and then only in a small section of the core, basically due to the cost and

time–consuming nature of these measurements.

From 1979 to 1981 the deep ice core at a location named Dye–3 (South Greenland) was drilled

by a team of Danish, Swiss and American scientists. It was part of the well–known Greenland Ice Sheet

Program (GISP). The total length of the core amounted to 2037 m until bedrock (Dansgaard 1982).

From, among others, δ18O measurements on this core, the paleoclimate has been reconstructed (see

Figure 4.7). A most interesting event was found at a depth of 1786 m (Figure 4.8), where an abrupt shift

in δ18O (and nearly all other parameters studied) was located. Since then, this Younger Dryas/PreBoreal

(YD/PB) transition has been examined in great detail (Dansgaard 1989). The δ18O level in the core

between 1784 m and 1788 m shifted upwards by 5‰ within a 50 year period. Based on present day

spatial δ18O - temperature relations, Dansgaard and co–workers supposed that this indicates a 7°C

temperature rise. Later it was argued that this value should be as high as 15°C, based on bore hole

temperature calibration of the δ18O values in Central Greenland (Johnsen 1995, Cuffey 1995). The age of

the ice at 1786 m below surface was dated at 10,720 ± 150 year BP by counting annual layers in δ18O

and electrical conductivity of the core. A more precise date of 11,500 ± 70 years BP for this transition

has been obtained from the GRIP core by counting annual layers in several high resolution chemical and

isotope profiles (Johnsen 1992). This event defines the end of the last glacial period (Weichselian

glaciation) and was preceded by a complex structure of rapid climatic shifts. The YD/PB transition is the

last transition in a climate oscillation, named the Bølling/Allerød–Younger Dryas (B/A–YD) oscillation. The

Chapter 4

126

observed shift in δ18O at the onset of the Bølling period has equal magnitudes as the YD/PB transition,

thus indicating similar enormous climate changes on a short time scale.

Figure 4.7: Example of the δ18O analysis of the Camp Century (Greenland) ice core. A 120,000 period is

covered on the y-axis. Different periods are marked in the Figure (reproduced from Dansgaard, 1973).

Glaciological measurements

127

Figure 4.8: a: Radiocarbon dated δ18O profile along a 4 m long sediment core from the Gerzensee

(Switzerland); b: δ18O profile along 150 m of the deep Dye–3 ice core. The 1700 – 1850 m depth interval

spans the entire pleistocene to holocene transition, including the Bølling/Allerød–Younger Dryas

oscillation; c: Concentration of continental dust; d: Detailed δ 18O record through the

Younger–Dryas–Pre–Boreal transition, a strong shift in 50 years is observed; e: Deuterium excess of the

same period, the transition occurs in 20 years; f: Dust concentration, shows the same shift as deuterium

excess. Figure is reproduced from Dansgaard (1989).

Deuterium excess d, defined by Dansgaard (1964) as:

d H O = − ⋅δ δ2 188

can be considered a measure of the difference in behaviour between 18O and 2H, or the contribution of

non–equilibrium isotope fractionation effects to the entire hydrological cycle. A changing regime of

evaporation in the source area will alter deuterium excess values because the relative non–equilibrium

contribution to the evaporation process will change. For polar regions it is now widely accepted that

deuterium excess is above all affected by (1) the temperature of the moisture source and (2) the

absolute humidity in that region (Johnsen 1989, Fisher 1991, Armengaud 1998). For example, for Law

Chapter 4

128

Dome, Antarctica, seasonal δ18O and δ2H cycles were found that both reflect the local temperature, while

the deuterium excess signal is four months backwards shifted in phase (Delmotte 2000). From this the

most likely sources of the precipitation were identified, as well as their seasonal dependence. A

comparable phase–lag is seen in high–altitude regions of the Greenland ice sheet and here too

information concerning the origin of the water vapour is obtained (Johnsen 1989).

For the YD/PB transition in the Dye–3 ice core, the δ2H profile has been measured as well

(Dansgaard 1989). Having both isotope profiles for this section, the deuterium excess could be

calculated (see Figure 4.8). It showed a shift of about –5‰, starting at the same time as the δ18O–shift,

but reaching a new stable value about twice as fast as δ18O. The time–scales of the d and δ18O changes

were initially calibrated at 20 and 50 years, using dating work done by Hammer and co–workers (1986),

who claim a 2 cm annual layer thickness in the YD and a 3 cm thickness shortly after the YD/PB

transition. More recent insight is based on a comparison with the well–dated GRIP ice core, yielding

mean annual layer thicknesses of 1.7 cm for the early PB, 0.7 cm in the YD, 0.9 cm in the Allerød,

0.95 cm in the Bølling and 0.45 cm in the pre–Bølling period. We estimate the accuracy of these figures

to be close to 10%. They are in fair agreement with annual high resolution PIXE data from sections of

the Dye–3 core (Hansson 1993). This makes it necessary to revise the time scale of the YD/PB climate

shift upwards to 50 and 100 years, for the deuterium excess and δ18O transitions respectively. These are

still very fast climate changes. A possible explanation is that the sea–ice cover retreated rapidly due to

the return of the North Atlantic current, thus creating a vast area of initially cold surface water as an

additional source of moisture (Dansgaard 1989). The immediate cause is believed to be the return of the

North Atlantic Current to higher latitudes and an associated northward shift of the polar front (Bond

1995, Broecker 1995, Ruddiman 1981).

From all isotope (and other) evidence it can be concluded that the climate in the last glacial

period has shown abrupt and radical changes in ocean circulation, polar front position, storminess,

humidity, atmospheric temperature and evaporation conditions. The δ18O data from Dye–3 have been

confirmed and validated by measurements on other cores, such as GRIP (Dansgaard 1993) and GISP2

(Grootes 1993), but so far this is not true for the deuterium data in the last glacial period.

In the last years we have developed a new technique for measuring isotope ratios in our

Groningen laboratory (Kerstel 1999). The method is conceptually different from the existing methods

that are all based on IRMS. Instead, our apparatus uses an infrared laser to measure the direct

absorption spectrum of gaseous water in order to obtain its isotope ratios (δ2H and δ18O, as well as

δ17O). We have already shown its application in the biomedical field (Van Trigt 2001a, 2001c). This

technique, apart from being elegant, is potentially very fast and can easily be automated. Advantages of

Glaciological measurements

129

the new method over the traditional ones include the absence of sample preparation. In fact, even

volatile contaminants do in practically all cases not interfere with the measurement, due to the very high

selectivity obtained by high–resolution infrared spectroscopy. We directly obtain isotope ratios for

deuterium and both oxygen isotopes. The δ18O measurement is not (yet) as accurate as with

conventional techniques, but further progress is foreseen. However, for δ2H we already achieve a higher

precision than with traditional methods, while the measurement of δ17O is new. Although it is known that

for all natural, meteoric water samples a fixed relationship between 17O and 18O holds and thus, in

principle, no new information can be derived from the 17O signal (Meijer 1998), the 17O measurement

can be used together with this fixed relationship as a check on the δ18O data, and possibly to improve its

precision.

Here we demonstrate the application of the newly developed method to the measurement of the 18O/16O

and 2H/1H isotope abundances in water. As a real–world test on glaciological samples we have

performed a detailed investigation of the deuterium excess in the Bølling transition in the Dye–3 deep ice

core.

4.2.3 Methods

4.2.3.1 Measurements

The Laser Spectrometer (LS) technique is based on direct absorption spectrometry, using a small

section (~1.3 cm-1) in the 2.7 µm region of the infrared absorption spectrum of water. This section

contains rotational-vibrational transitions for all four isotopomers of interest (i.e., 1H16O1H, 1H17O1H,1H18O1H, and 2H16O1H). For water samples with natural isotope abundances the absorption strengths of

these transitions are of the same order of magnitude and, although the spectral features are close to

each other, they are well resolved. We can use the low-pressure, gas phase, infrared absorption

spectrum for isotope ratio determinations since the intensities of the transitions are a direct measure of

the abundances of the corresponding isotopomers.

To record an absorption spectrum we scan a tunable, single mode laser (a Color Centre Laser or

FCL, Burleigh) from 3664.05 cm-1 to 3662.70 cm-1 in about 5000 steps. For each step of the laser we

record the laser power before and after the passage through the gas cells using phase sensitive

detection. The spectra of the water samples in the four multiple-pass gas cells are thus recorded

simultaneously. A 10 µl liquid water sample is injected into the cells, assuring a final (partial) pressure of

the water vapor of about 13 mbar, well below the saturation vapor pressure. One of the four gas cells

always contains a working standard, while the others contain either reference water or an unknown

Chapter 4

130

sample. For each sample injection eight successive scans were recorded. A full measurement cycle,

including introduction of the sample, takes about 40 minutes. Since we have four gas cells, we measure

three samples (or standards) in one run together with the working standard. Where duplicate

measurements did not agree to within 3 times the mean standard deviation, an extra measurement was

made. This was needed for typically 10% of the samples. In the Bølling transition section of the core,

measurements were performed in fourfold. The error due to memory effects amounts typically to less

than 5% of the difference in δ-value between previous and current sample. In this study this error is

generally smaller than the analytical error. Special care had to be taken only after measuring VSMOW or

SLAP, because their isotope ratios differ significantly from that of the samples and the international

reference standard, GISP. In these cases the new samples were injected and removed once, before the

actual measurement commenced.

4.2.3.2 Standards

As in traditional IRMS, LS needs a working standard to compare the samples with, in order to

obtain reliable isotope ratio determinations. We chose a working standard as close as possible to the

expected sample values, namely a mixture of old “leftover” batches of Greenland Ice Sheet Precipitation

(GISP). Initially, the isotope ratio of this mixture was not known exactly, since fractionation might have

occurred during storage of the different bottles over the years. Still, we later found that its value was

close (just 2.5‰ higher for δ2H) to the values of fresh GISP.

As reference materials for the calibration of the system, we used fresh VSMOW, SLAP and GISP.

The use of primary calibration standards is defendable in this stage of the work, largely thanks to the

very small amounts of water that are used.

The ratio of measured standards to samples for this project was about 1:3, Table 4.1 shows the

numbers in more detail.

Table 4.1: Total number of single measurements made on the different waters for the entire Dye-3

measurement project. In all cases Old GISP mixture was used as the working standard. For all samples,

isotope ratios for all three isotopomers were acquired.

Old GISP mix GISP SLAP VSMOW Samples

Total # of

measurements178 57 53 69 807

Glaciological measurements

131

4.2.3.3. Samples

We measured 279 water samples of the Dye–3 ice core. Their age varies from 9200 year BP to

14,700 year BP, thus including both the YD/PB transition and the Bølling transition. As stated in the

introduction, these samples have been previously measured for δ18O over the entire core, but not for

δ2H. The YD/PB transition has been extensively studied for deuterium excess, as well as for other climate

indicating parameters (Dansgaard 1989). In those experiments all water was used up and we could

therefore not include this particular section in our current programme. The depth resolution of the

sampled ice–core section between the depths of 1730 m and 1812 m is 55 cm, 27.5 cm, 11 cm or 5 cm,

depending on the desired resolution for the specific period. In the Bølling transition the resolution is

5 cm, corresponding to roughly 7 years per sample.

4.2.3.4 Calibration

We apply a calibration procedure to scale our raw measurement results to the internationally

accepted values of the calibration materials VSMOW and SLAP, complying with the procedure

recommended by the IAEA (Gonfiantini, 1984). It should be noted that in IRMS several types of

corrections are necessary as well, but these are not as well understood and usually much bigger in

magnitude than those in LS.

The different gas cells exhibit different zero-point offsets. These turn out to be primarily

associated with the optical alignment of the instrument. Because the alignment is very stable we can

easily and reliably correct for these offsets. The values are around zero with a magnitude of the order of

one per mil.

The raw measurement results are also corrected for small differences in gas cell pressure

(amount of water) between the reference and sample cells. This linear correction is very well understood

and can be calculated from simulated absorption spectra (Kerstel 1999). Moreover, the magnitude of the

corrections is small (typically below 0.1‰) for all isotopes.

The gas cell and isotope dependent scale expansion factors lie in a range from 0.98 to 1.02,

which is much smaller than what is usually seen in IRMS. The calibration procedure is described in more

detail elsewhere (Kerstel 1999). Note that the above scale corrections constitute a VSMOW/SLAP scale

normalization as prescribed by the IAEA (Coplen 1988). After removal of obvious outliers, the final

results are averaged for each sample.

Chapter 4

132

4.2.4 Results and Discussion

4.2.4.1 Measurement precision

An indication of the precision of the LS measurements is the single measurement standard

deviation (SD) of repeated measurements on the same sample. As each sample was measured only two

to four times, one sample will not provide reliable statistical information. Therefore we take the mean of

all calculated SD’s as a measure. We then find the single measurement precision to be ~0.6‰ for δ2H,

~0.5‰ for δ18O and ~0.3‰ for δ17O. The statistical spread of the standard deviations (histogram) is

fairly well represented by a Gaussian curve. These results are comparable to those obtained in analyses

based on repeated measurements of the same water sample (in particular VSMOW), which were carried

out in the framework of previous studies (Kerstel, 1999; Van Trigt 2001a). The better performance of

the LS system in the case of δ17O is attributed to the higher signal-to-noise obtained on the H17OH

spectral feature, compared to the H18OH line.

The relationship between δ18O and δ17O for meteoric waters established by Meijer and Li (1998),

enabled us to calculate values for δ18O from the measured δ17O. In the case of a linear fit forced through

zero for the inferred δ18O against the measured δ18O, we find a slope of 1.0023(20). We conclude that to

good approximation these inferred (indirect) and measured (direct) δ18O values may be treated as

duplicate determinations. The δ17O measurements thus serve as a check on the δ18O measurements and

may even be used to improve the precision of the latter by doubling the number of independent δ18O

determinations. We averaged the δ18O measurement and the calculated δ18O value (inferred from the

δ17O measurement), using the squared errors as weighing factors, resulting in a precision of the

combined determination of ~0.4‰. The combined result (i.e., the weighted mean) does not differ

significantly from the direct δ18O measurement.

4.2.4.2 2H and 18O isotope records

The depth profiles of the δ2H and δ18O records determined by means of LS are shown in Figure 4.9. They

show qualitatively the same behavior and the major transitions are clearly visible in both. As mentioned

before, samples from the interval between 1784 and 1788 m (YD/PB transition) were no longer

available.

Glaciological measurements

133

Figure 4.9: δ2H (GrLS2) and δ18O (GrLS18) depth profiles as measured with the Groningen LS apparatus.

As explained in the text, the water samples around 1785 m (the YD/PB transition) were used-up in the

original measurements (Reyk18, given here for comparison purposes) by Dansgaard and co-workers,

thus leaving a gap in the Groningen records. The 1989 deuterium measurements of Dansgaard and co-

workers (Saclay2) fit well in the gap in the GrLS2 record.

The LS δ18O record (GrLS18) can be compared directly to the old Reykjavik IRMS data (Reyk18),

also shown in Figure 4.9. The median δ-values for the two curves amount to –30.83‰ (279 samples)

and –30.89‰ (281 samples) for the GrLS18 and Reyk18 records, a strong indication that the data have

been properly calibrated. Closer inspection of the two isotope records reveals just one small section of

the core at the end of the Bølling transition in which the two records deviate. Figure 4.10 shows this

Bølling transition region in detail.

-40

-35

-30

-25

-20

-15

-400

-360

-320

-280

-240

-200

1740 1760 1780 1800

GrLS18Reyk18

GrLS2Saclay2

δδδδ18O

(‰

) δδ δδ2H

(‰)

Depth (m)

Chapter 4

134

Figure 4.10: Depth profiles for δ18O around the Bølling transition (about 14,500 years BP). Groningen LS

and IRMS data are shown, as well as the old Reykjavik IRMS data. The deviation of the old and new

measurements for the samples between 1809 m and 1810 m is clearly visible.

Between 1809 m and 1810 m (24 data points) the GrLS18 and Reyk18 records have median

delta-values that differ by 1.1‰. Such a big difference cannot be explained by fractionation (e.g.,

accompanying evaporation) during storage, transportation, or sample preparation. Contamination, most

likely by fragmentation of the drilling fluid in the ion-source of the mass spectrometer, would have

resulted in a higher IRMS value, not a lower one, with respect to laser spectrometry. Moreover, the

Reykjavik IRMS system is equipped with special cold-traps to prevent such contamination from having an

effect on the measurements. Also, it is highly unlikely to have affected a small section of the core only.

In fact, ice-core analyses of our Copenhagen laboratory (that, as the Groningen laboratory, does not

take such elaborate measures against drilling fluid contamination) in general show a systematic and

more or less constant offset, up to about 0.5‰. In conclusion, we have no explanation for the observed

-34

-33

-32

-31

-30

-29

-28

-27

1806 1807 1808 1809 1810 1811 1812

GrLS18GrMS18Reyk18

δδδδ18O

(‰

)

Depth (m)

Glaciological measurements

135

local discrepancy between the GrLS18 and the Reyk18 records. For this reason, we re-measured 74

samples belonging to the Bølling transition (1806 to 1813 m) by means of mass spectrometry in

Groningen. This partial record we will refer to as GrMS18. Its median value is 0.43‰ higher than the

GrLS18 record in the same depth range, tentatively attributed to drilling fluid contamination of the ice-

core, as also observed in the past in our Copenhagen laboratory. The GrMS18 record shows a

substantially smaller scatter, reflecting the higher precision of IRMS with respect to δ18O measurements

by means of laser spectrometry. After shifting the GrMS18 record downwards by 0.43‰, a nearly

perfect agreement with the GrLS18 record is obtained, as can be seen in Figure 4.10.

The transition (increasing temperature with time; note that the time scale is from right to left)

occurs in about 1.5 m of ice. If we take the average annual layer thickness during the transition to be

equal to 0.7 cm, this corresponds to about two hundred years or twice as long as the YD/PB transition.

During the Bølling transition δ18O increases from about –33.5‰ to –28.5‰ and the δ2H signal increases

from about –260‰ to –220‰. This is similar to the YD/PB transition and suggests a similar

temperature rise of about 15°C. We also notice in the δ18O record 2‰-strong cold event at 1811.0 m

depth in the middle of the Bølling transition lasting only 25 years. This climatic event is not as clearly

depicted in other Greenland isotope records.

4.2.4.3 Deuterium Excess

For the samples analysed in this study, no deuterium depth profile was acquired previously, so there is

no data to compare to. Only for the YD/PB transition (the section between 1784.20 m and 1788.05 m)

δ2H has been measured (Dansgaard 1989). But, as explained before, this section is missing in our data-

set. However, as Figure 4.9 shows, the old δ2H record (Saclay2), measured previously in the stable-

isotope laboratory at Saclay, fits well in the “gap” in the present record (GrLS2), again demonstrating the

quality of the calibration of the data, as well as the integrity of the 2-decade old samples (δ2H in

particular is very sensitive to fractionation processes).

From other deuterium excess measurements on large numbers of ice core samples, as well as from

several laboratory ring-tests, we conclude that a typical deuterium excess precision, using conventional

IRMS techniques, amounts to about 1.8‰ (based on a precision of ~1‰ for δ2H and ~0.1‰ for δ18O).

Although some glaciological isotope laboratories claim a precision well below the figures used here,

these claims may prove to be exaggerated if it comes to the accuracy of the measurements, particularly

of the deuterium excess. Inter-laboratory comparisons carried out by the International Atomic Energy

Agency have demonstrated the difficulty of maintaining such high levels of accuracy across a number of

specialized isotope laboratories (Lippman 1999). This is especially reason for concern when two isotope

Chapter 4

136

measurements (δ2H and δ18O) are used to calculate the deuterium excess. We would therefore strongly

argue in favor of participation of the ice-core isotope community in similar ring-tests. The LS

measurements alone (precision of ~0.6‰ and ~0.4‰ for δ2H and δ18O, respectively) would yield a

precision for deuterium excess of about 3.8‰, which is of the same size as the expected natural

(climate) variations. We therefore calculate the deuterium excess during the Bølling transition using the

GrLS2 deuterium record together with the GrMS18 oxygen-18 record. The latter has been shifted

downwards by 0.43‰ in its entirety, in order to best overlap with the more accurate, well-calibrated,

GrLS18 record. As mentioned before, this procedure is justified by earlier observations of systematically

higher IRMS δ18O-values when no proper precautions are taken to prevent residual drilling fluid from

interfering with the measurements.

Figure 4.11 presents the deuterium excess depth profile in the range of 1806 m to 1813 m, around the

Bølling transition. The RMS deviation of the data points with respect to the smoothed curve amounts to

1.4‰. This equals the estimated uncertainty in the deuterium excess determination, based on the

measurement uncertainties in δ2H and δ18O (0.6‰ and 0.1‰, respectively). The curve indicates that

deuterium excess decreased by about 6‰ within a time span of about 70 years at the onset of the

warming. The residual structure on the curve, in particular the two small dips at 1811.5 m and

1809.5 m, fall within the measurement uncertainty and we hesitate to associate these with minor climate

events. The over all pattern is then rather similar to what has been observed previously for the YD/PB

transition (Dansgaard 1989) and we may indeed assume that the same mechanism that caused the

YD/PB transition was also operative during the Bølling transition. It would be interesting to compare the

Bølling isotope records to the other climate indicators, in particular dust. If indeed the rapid shift in

deuterium excess at the onset of the Bølling transition signals a northward shift of the polar front, in

response to a return of the North Atlantic current to higher latitudes, one would expect to see a

decrease in dust in parallel with the deuterium excess shift, indicative of a more humid, milder, and less

stormy climate.

The general picture emerging from these isotope and other studies is that the climate in the last

glacial period has shown abrupt and radical changes in ocean circulation, polar front position,

storminess, humidity, atmospheric temperature and evaporation conditions.

Glaciological measurements

137

Figure 4.11: Deuterium excess, d = δ2H - 8·δ18O, for the Bølling transition. The solid curve is obtained by

smoothing of the GrLS2/GrMS18 data and serves mainly to guide the eye. The RMS deviation of the data

with respect to the smooth curve is 1.4‰. The shift in deuterium excess at the Bølling transition is

about 6‰ as was found for the YD/PB transition 26 m higher up in the core (Dansgaard 1989).

4.2.5 Conclusions

Laser Spectrometry is a new and elegant way of measuring stable isotopes in ice core samples.

Its sample throughput is already quite high (50 sample/day) and can easily be increased further. The

single measurement precision obtained for δ2H measurements (~0.6‰) is very competitive with

traditional IRMS methods. For δ18O the precision (~0.5‰) is still almost one order of magnitude worse,

while the measurement of δ17O (~0.3‰) is new. The δ18O precision can be improved to ~0.4‰ when

δ18O and δ17O measurements are combined. When IRMS δ18O measurements and LS δ2H measurements

0

2

4

6

8

10

12

1806 1807 1808 1809 1810 1811 1812

GrLS2/GrMS18

d (

‰)

Depth (m)

Chapter 4

138

are combined, a precision of ~1.4‰ for deuterium excess measurements can be achieved, comparable

to IRMS-only. Where IRMS δ18O measurements can be severely affected by drilling fluid contamination, if

no proper precautions are taken, LS is virtually immune to such effects, thanks to its extremely high

molecular and isotopomer selectivity. If this contamination is present, the LS δ18O results are more

accurate (but not more precise) than those obtained by IRMS.

The YD/PB transition (11,500 BP) as measured by Dansgaard and co-workers is not the only sharp

transition at the end of the last glaciation. Some 3000 years earlier, the Bølling transition showed an

about equal temperature rise, in approximately two hundred years time. Deuterium excess shifted

similarly in about 70 years. Together, these observations indicate that the underlying mechanisms may

have been very similar during the onset of the Bølling interstadial and the YD/PB climate transition