Dating Ice Cores

One of the most important problems in any ice-core study is determining the age-depth relationship. Many different approaches have been used and it is now clear that very accurate timescales can generally be developed for at least the last 10,00012,000 yr if accumulation rates are high enough. Prior to that, there is increasing uncertainty about the age of ice, but new approaches are constantly improving age estimates, allowing comparisons with other proxy records to be made with more confidence (see Section 5.4.5). Furthermore, many of the methods that have been investigated in order to improve the dating of ice cores have themselves produced important paleoclimatic information. Some of the principal methods used and their paleoclimatic implications are now reviewed.

5.3.1 Radioisotopic Methods

Several different radioactive isotopes have been analyzed in ice cores in an attempt to provide quantitative chronological methods for dating ice. These include: 10Be, 14C, 36C1, 39Ar, 81Kr, and 210Pb (Stauffer, 1989). At present, however, apart from 210Pb and 14C analysis, radioisotopic dating of ice and firn is not a routine operation and other stratigraphie techniques are generally preferred.

The 210Pb (half-life: 22.3 yr) is washed out from the atmosphere as a decay product of 222Rn (see Fig. 3.21). It has been used successfully in studies of snow accumulation over the last 100-200 yr, providing an important perspective on the very short accumulation records otherwise available in remote parts of Antarctica and Greenland (Crozaz et al., 1964; Dibb and Clausen, 1997). The AMS 14C dates on C02 enclosed in air bubbles in ice can be obtained from ice samples as small as 10 kg (equivalent to a conventional ice core -1.5 m in length) though precision is improved with larger samples (Andrée et al., 1986). Unfortunately, the dates on C02 obtained may differ from the age of the enclosing ice by hundreds, or thousands, of years because of the time delay before gas bubbles become entirely sealed from the atmosphere (see Section 5.4.3). This problem limits the value of 14C dates on ice core samples.

5.3.2 Seasonal Variations

Certain components of ice cores show quite distinct seasonal variations, which enable annual layers to be detected. These can then be counted to provide an extremely accurate timescale for as far back in time as these layers can be detected. Where uncertainties exist in one seasonal chronology, a comparison with other parameters enables accurate cross-checking to be accomplished, thereby reinforcing confidence in the timescale produced (Hammer et al., 1978). For example, annual layer counts (back to 17,400 yr ago) have been carried out on the GISP2 ice core from Summit, Greenland, using a combination of, inter alia, visual stratigraphy, electrical conductivity measurements (ECM), laser light scattering (from dust) oxygen isotopes, and chemical variations in the ice (Meese et al., 1995, 1997). When compared to the independently derived chronology from the nearby GRIP ice core, the two records match to within 200 yr back to 15,000 calendar yr B.P. (Taylor et al., 1993a).13 However, at greater depths the counts diverge significantly as the difficulty of unequivocally identifying annual layers increases. In this section, the different types of information used in layer counting are discussed.

Visual stratigraphy: visual stratigraphy provides a "first cut" at identifying annual increments in an ice core. Cores are examined on a light table to identify changes in crystal structure and the presence of dust layers. In the GISP2 ice cores, a distinctive coarse-grained depth hoar layer, characteristic of each summer, can be seen (Alley et al., 1997a). In cores from the Quelccaya ice cap, Peru, a pronounced dust layer, which is diagnostic of conditions from May-August, permits the counting of annual layers (Thompson et al., 1985).

8180: Because of the greater cooling that occurs in winter months, much lower 8lsO concentrations are found in winter snow than in summer snow. This results in a very strong seasonal signal that can be used as a chronological tool, providing accumulation rates are reasonably high (>25 cm water equivalent per year), wind scouring of snow is not severe, and no melting and refreezing of snow and firn has occurred. In effect, the annual layer thickness can be identified by counting each couplet of high and low 8lsO values from the top of the core downward (Fig. 5.11). Unfortunately, at increasing depths in polar ice sheets the amplitude of the seasonal signal is reduced until it is eventually obliterated. In the upper layers, where density is <0.55 g cm-3, this results from isotopic exchange between water vapor and firn. In lower, denser layers, where air channels are closed off, obliteration results from diffusion of water molecules within the ice. This process is accelerated due to thinning by plastic deformation as the annual layers approach bedrock; thinning increases isotopic gradients in the ice, making molecular diffusion more effective in obliterating the seasonal variations (see Fig. 5.11).

In cores where seasonal isotopic differences are still preserved down to dense firn and ice layers, further smoothing due to molecular diffusion is so slow that the signal may then be preserved for thousands of years. This does not occur in most of

13 For the GISP2 ice core, multi-parameter dating cross-checked with various independent reference horizons, suggests that the age of the ice is known to within <1% for the last 2000 yt; increasing to 2% by -40,000 yr B.P., to 10% by 57,000 yr B.P. and up to 20% by~110,000 yr B.P. (Meese et al., 1997).

CAMP CENTURY

X

0-2

.35

5-7

.35

115

.36

500

.20

776

.16

1000

.074

1061

.054

1079

?

1198

3300 -30

3300 -30

FIGURE 5.1 I The 8I80 variations in snow and firn at different depths in ice core from Camp Century, Greenland.The S and W indicate interpretations of summer and winter layers, respectively. As the ice sinks towards the base of the ice sheet, the annual layer thickness (\) is reduced due to plastic deformation.Within a few years, short-term 8lsO variations are obliterated by mass exchange in the porous firn.With increasing age, the amplitude of the seasonal delta cycle is reduced to 2%o. As annual layers become thinner, the seasonal 5I80 gradients increase and molecular diffusion in the ice smoothes out the intra-annual variations. Eventually, seasonal differences are obliterated entirely (Johnsen et ai, 1972).

17000

FIGURE 5.1 I The 8I80 variations in snow and firn at different depths in ice core from Camp Century, Greenland.The S and W indicate interpretations of summer and winter layers, respectively. As the ice sinks towards the base of the ice sheet, the annual layer thickness (\) is reduced due to plastic deformation.Within a few years, short-term 8lsO variations are obliterated by mass exchange in the porous firn.With increasing age, the amplitude of the seasonal delta cycle is reduced to 2%o. As annual layers become thinner, the seasonal 5I80 gradients increase and molecular diffusion in the ice smoothes out the intra-annual variations. Eventually, seasonal differences are obliterated entirely (Johnsen et ai, 1972).

Antarctica, though, because of low accumulation rates (generally <25 cm water equivalent per year), which result in the seasonal signal being "lost" at relatively shallow depths. In many cases, removal of seasonal (or indeed annual) accumulation by wind scouring may occur, destroying any seasonal signal entirely. On temperate glaciers and ice caps, where snowmelt and percolation of meltwater takes place, it is also impossible to detect a reliable seasonal isotopic signal. In these conditions, seasonal differences in both 8D and S180 are rapidly smoothed out (within a few meters of the surface) due to isotopic exchange as the ice recrystallizes (Arna-son, 1969).

Microparticles and glaciochemistry: Detailed studies of microparticulate matter and ice chemistry (major ions and trace elements) in ice cores from Antarctica and Greenland reveal pronounced seasonal variations (Fig. 5.12). In Greenland, microparticles increase to a maximum in late winter-early spring, presumably as a result of a more vigorous atmospheric circulation at this time of year. Conversely, microparticle frequency minima are generally observed in autumn. There are similar seasonal variations in various cations and anions (e.g., sodium, calcium, nitrate, chloride) with spring concentrations of these ions commonly greater than at other times of the year (Hammer, 1989).

Compared to the diffusion rate of water molecules, which leads to obliteration of the seasonal 8lsO record at depth, diffusion of microparticles and metallic ions is essentially zero. Hence the counting of seasonal variations may allow dating of ice back to late Wisconsin time, or perhaps even earlier. This approach is particularly useful in areas where accumulation rates are so low that seasonal isotopic differences are rapidly lost at depth. For example, in parts of Antarctica, the concentration of sodium ions (Na+) varies markedly, due to pronounced seasonal changes in the influx of marine aerosols (Herron and Langway, 1979; Warburton and Young, 1981). At Vostok, in eastern Antarctica, Na+ concentrations reach a maximum in summer layers, due to sublimation of snow, leaving higher residual ionic concentrations (Wilson and Hendy, 1981). These variations are visible far below the level at which seasonal 8lsO variations become obliterated, and can even be detected at -950 m depth in the Vostok ice core.

One difficulty in microparticle and trace element analysis is to ensure that the sample size selected is small enough to detect intra-annual changes. Near the surface, this is not a big problem, but in ice from very deep ice cores (where the actual thickness of an annual layer is not accurately known) intense lateral and vertical compressive strain may result in dust layers being merged together so that they cannot be adequately distinguished. This is particularly true if the strain rates of dirty ice and of clean ice are very different, as suggested by Koerner and Fisher (1979). Non-destructive laser light scattering can be used to produce a continuous record of dust variations in an ice core (Ram and Illing, 1995) but the problem of identifying each annual layer remains. This can lead to an underestimation of ice age at depth if the microparticle variations observed represent several years rather than seasonal variations. Fortunately, independent corroboration of age estimates can generally be achieved using multiple indicators, though difficulties increase greatly at depth.

321-

S042 (jieq/kg)

- Visible i

Stratigraphy

—M-

1

I I ' , i, '

" /

h*

(1207 f wi jvji

i i i i

K

i i i i

III!

I < I I I

i i i i

FIGURE 5.12 A section of an ice core from Site A, central Greenland, showing interannual variations in some parameters commonly used in dating. Seasonal peaks are seen in most records, though occasional uncertainties are apparent. Usually such uncertainties can be resolved by cross-referencing the records. In this way seasonal counting can be used to date the upper sections of ice cores (Steffensen, 1988).

Electrical conductivity measurements (ECM): ECM provides a continuous record of ice acidity by recording the ability of ice to conduct an electrical current. A current with a large potential difference is passed between two electrodes in contact with the surface of the ice core (1250 V was used on the GRIP and 2100 V on the GISP2 ice cores). When the ice contains strong acids from volcanic eruptions, ECM is high; layers containing alkaline continental dust, or ammonia (e.g., from biomass burning) have low ECM (Taylor et al., 1993a, b). Changes in deposition of CaC03 dust are associated with large changes in ECM, reflecting changes in the source region and/or transport and deposition processes. Pronounced changes in ECM characterize the transitions from cold, glacial periods to warmer interstadials in the GISP2 ice core (see Figs. 5.26a and b). Figure 5.13 shows in detail the ECM record at the transitions marking the beginning and end of the Younger Dryas period (-12,850 and 11,670 calendar yr before present in this record).

5.3.3 Reference Horizons

Where characteristic layers of known age can be detected, these provide valuable chronostratigraphic markers against which other dating methods can be checked. On the short timescale, radioactive fallout from atmospheric nuclear bomb tests in the 1950s and 1960s can be detected in firn by measuring the tritium content (or gross 3 activity). As the timing of the first occurrence of these layers is fairly well

Calendar Years B.P.

Calendar Years B.P.

Depth (m) Calendar Years B.P.
Cause Younger Dryas

FIGURE 5.13 Electrical conductivity measurements (ECM) in the GISP2 ice core from Summit, Greenland, at the transitions to (below) and from (above) the Younger Dryas cold episode. Annual layer thickness varies from ~6 cm yr'in warmer intervals (high ECM) to ~3 cm yr"1 in colder intervals (low ECM). Counting of annual layers is not based on ECM alone, but involves multiple parameters (Taylor et al., 1993b).

FIGURE 5.13 Electrical conductivity measurements (ECM) in the GISP2 ice core from Summit, Greenland, at the transitions to (below) and from (above) the Younger Dryas cold episode. Annual layer thickness varies from ~6 cm yr'in warmer intervals (high ECM) to ~3 cm yr"1 in colder intervals (low ECM). Counting of annual layers is not based on ECM alone, but involves multiple parameters (Taylor et al., 1993b).

known (spring 1953 in Greenland and February 1955 over much of Antarctica, reaching maximum levels in 1963) they can be used as marker horizons for snow accumulation studies, facilitating regional surveys of net balance over the last few decades (Crozaz et al., 1966; Picciotto et al., 1971; Koerner and Taniguchi, 1976; Koide and Goldberg, 1985).

On a much longer timescale, other reference horizons have resulted from major explosive volcanic eruptions. Violent eruptions may inject large quantities of dust and gases (most importantly hydrogen sulfide and sulfur dioxide) into the stratosphere where they are rapidly dispersed around the hemisphere. The gases are oxidized photochemically and dissolve in water droplets to form sulfuric acid, which is eventually washed out in precipitation. Hence, after major explosive volcanic eruptions, the acidity of snowfall increases to levels significantly above background values (Hammer, 1977). By identifying highly acidic layers resulting from eruptions of known age, an excellent means of checking seasonally based chronologies is available (Fig. 5.14). For example, variations in electrical conductivity (a measure of acidity) along a 404 m core from Crête, Central Greenland, reveal a record that closely matches eruptions of known age (Hammer et al., 1978, 1980). The core was originally dated by a combination of methods, primarily seasonal counting (Hammer et al., 1978). This enabled the acidity record to be checked against historical evidence of major eruptions during the last 1000 yr (Lamb, 1970) confirming that the timescale developed was extremely accurate. Once major acidity peaks have been identified they can be used as critical reference levels over the entire ice sheet. For example, the highest acidity levels in the last 1000 yr in Greenland ice resulted from the eruption of Laki, Iceland, in 1783. At Crête, the only acidity peak of greater magnitude in the last 2000 yr resulted from another Icelandic eruption (Eldgja) at A.D. 934 ± 2, providing two very distinct reference layers (Hammer, 1980). Similarly, a major eruption of Huaynaputina (Peru) in February A.D. 1600 provides a diagnostic reference horizon in conductivity records from the Quelccaya and Huascarân (Peru) ice cores, as well as in Antarctica (Delmas et al., 1992; Cole-Dai et al., 1995) (see Fig. 5.38). Further discussion of the volcanic record in ice cores can be found in Section 5.4.4.

Volcanic dust (tephra) from large eruptions may also provide chronostrati-graphic horizons if the chemical "fingerprint" of the layer can be correlated between the different sites. In the GISP2 (Greenland) ice core, for example, volcanic particles from an Icelandic eruption 52,680 ± 5000 yr ago can be matched with the Z2 tephra found in many marine sediment records from the North Atlantic (Ruddi-man and Glover, 1972; Kvamme et al., 1989; Zielinski et al., 1997). Similarly, volcanic particles in the Younger Dryas section of the Dye-3 and GISP2 ice cores have the same geochemical signature as the Vedda ash (14C-dated at 10,320 yr B.P.), which is widely distributed in northwest Europe and the North Atlantic (Mangerud et al., 1984; Johnsen and Dansgaard, 1992; Birks et al., 1996; Zielinski et al., 1997).

Another important reference horizon is provided by "spikes" in the 10Be record found in some polar ice cores. The 10Be is a cosmogenic isotope, produced in the upper atmosphere, which eventually settles, or is washed out, to the earth's surface (see Section 5.4.1). Two large increases in 10Be, far above background levels, are

FIGURE 5.14 Mean acidity of annual layers from A.D. 553 to A.D. 1972 in the ice core from Crête, central Greenland. Acidities above the background (1.2 ixequiv. H+ per kg of ice) are due to fallout of acids, mainly H2SOv from volcanic eruptions north of 20°S.The ice core is dated with an uncertainty of ± I yr in the past 900 yr, increasing to ± 3 yr at A.D. 553, which makes possible the identification of several large eruptions known from historical sources (e.g., Laki, Iceland, l783;Tambora, Indonesia, 1815; Hekla, Iceland, 1104).Also seen is the signal from the Icelandic volcano Eldgja, which was known to have erupted shortly after A.D. 930. Note the low level of volcanic activity recorded from A.D. 1100 to A.D. 1250 and from A.D. 1920 to A.D. 1960. Considerably higher levels of volcanic activity occurred from A.D. 550 to A.D. 850 and from A.D. 1250 to A.D. 1750 (Hammer etal., 1980).

FIGURE 5.14 Mean acidity of annual layers from A.D. 553 to A.D. 1972 in the ice core from Crête, central Greenland. Acidities above the background (1.2 ixequiv. H+ per kg of ice) are due to fallout of acids, mainly H2SOv from volcanic eruptions north of 20°S.The ice core is dated with an uncertainty of ± I yr in the past 900 yr, increasing to ± 3 yr at A.D. 553, which makes possible the identification of several large eruptions known from historical sources (e.g., Laki, Iceland, l783;Tambora, Indonesia, 1815; Hekla, Iceland, 1104).Also seen is the signal from the Icelandic volcano Eldgja, which was known to have erupted shortly after A.D. 930. Note the low level of volcanic activity recorded from A.D. 1100 to A.D. 1250 and from A.D. 1920 to A.D. 1960. Considerably higher levels of volcanic activity occurred from A.D. 550 to A.D. 850 and from A.D. 1250 to A.D. 1750 (Hammer etal., 1980).

seen in the Vostok ice core around 35,000 and 60,000 yr B.P. (Raisbeck et al., 1987). The reason for these peaks is not clear; they may have resulted from a change in primary cosmic ray flux, or a reduction in solar or geomagnetic modulation of cosmic rays penetrating the atmosphere (Baumgartner et al., 1998), or even from a super nova. Whatever the cause, these anomalies can be seen in many ice cores and can be used as chronostratigraphic markers. For example, the 35 ka B.P. 10Be spike is found in ice cores from Vostok, Dome C, and Byrd (Antarctica), enabling these records to be properly aligned (Figure 5.19). There is also a 10Be peak in the Camp Century ice core from Greenland and a spike of 36C1 (also a cosmogenic isotope) in the Guliya (western China) ice core at about the 35 ka level, confirming the age models applied to these records and allowing them to be aligned with those from Antarctica (Reeh, 1991; Beer et al., 1992; Thompson et al., 1997). However, the 60 ka 10Be anomaly is less pronounced and has not been as useful a marker. Interestingly, a second 10Be spike seen in the Camp Century ice core, if ascribed to the 60 ka event, would force a major revision in the chronology of this record, and of the Dye-3 record (in southern Greenland) with which it was correlated, making both series much shorter than envisioned by Dansgaard et al. (1982). Reeh (1991) argues that this is in fact the case, because the ice sheet was considerably smaller in the last interglacial, so that higher 8lsO values prior to 70 ka B.P. (in his revised chronology) were due to a smaller, lower ice sheet. This argument is supported by the studies of Koerner (1989) and Letreguilly et al. (1991). The counterargument is that the "60 ka" B.P. spike in Camp Century is not reliable (based on only a single high value) and that the chronology of Dansgaard et al. (1982) in fact fits much better with other proxy records and with reasonable flow model assumptions, which place the "60 ka" horizon closer to 95 ka B.P. (Beer et al., 1992; Johnsen and Dansgaard, 1992; Reeh, 1991). This controversy nicely illustrates the difficulties of dating ice at depth, particularly in Greenland where there may have been dramatic changes in the ice sheet configuration over the last 150,000 yr. Clearly, having unequivocal stratigraphic markers would be extremely helpful in resolving such controversies. Other approaches, using gases in ice cores, are discussed further in Section 5.4.5.

As noted earlier, the best approach to identifying annual layers is a composite one, using 8lsO profiles, microparticles, variations in conductivity, and reference horizons. In this way, questionable sections of one record may be resolved by reference to the others. This multiparameter approach was adopted by Meese et al. (1995, 1997) in dating the upper section of the GISP2 ice core. Having established the annual chronology, it was then possible to calculate accumulation rate changes over time, given certain assumptions about vertical strain since deposition and density variations down core (Meese et al., 1994). Figure 5.15 shows the accumulation record from Summit, Greenland, derived in this way for the last 11,500 yr. Accumulation was considerably lower in the last glacial period, increasing by >30% from -12 ka to 9 ka B.P. Thereafter, the record (subjected here to 100 yr smoothing) reveals only minor changes (± 5% on this timescale). Interestingly, a drop in accumulation at -8200 B.P. corresponds to very low 8180 values in many ice cores, as well as a sharp reduction in CH4 (see Fig. 5.33), indicating that a significant and abrupt, large-scale climatic change occurred at this time (Alley et al., 1997c).

0 0

Post a comment