Info

Temperature (°C)

FIGURE 5.6 Annual 8I80 in precipitation in relation to mean annual temperature at the same site, based on data from the International Atomic Energy Authority (Jouzel et al., 1994).

relationship for locations in the extra-tropics (regions with mean annual temperatures <15 °C) is

but this varies geographically (Jouzel et al., 1987b; 1994). In the colder regions of Antarctica, the slope of this relationship is greater (-0.8, on average). At low latitudes, 8180 is not related to temperature, but is more a function of precipitation amount (Dansgaard, 1964; Rozanski et al., 1993). Significant departures from the regression may be expected in certain circumstances (Hage et al., 1975): (a) if precipitation occurs in an area where very stable (i.e., inversion) conditions are common, surface temperatures will be lower than expected from the regression (and conversely, 8lsO values will appear anomalously high); and (b) if local precipitation is derived from water that was re-evaporated from a source with an already low 8lsO content (e.g. freshwater lake or snow cover) the mean annual 8lsO value may fall below the regression line, perhaps by as much as 10%o for complete re-evaporation of precipitated water) (Koerner and Russell, 1979).

5.2.4 Calibrating Sl80 for Paleotemperature Reconstruction

A number of studies in Antarctica have shown that there is a strong relationship between 8180 of snowfall and temperature on a daily (storm event) basis. However, strong surface-based temperature inversions essentially decouple the surface from the atmospheric circulation above the ice cap; temperatures are often much lower at the surface (<Pg) than at the top of the inversion (4>.) and on average, = 0.67<Pg - 1.2 (Jouzel and Merlivat, 1984). Hence, there is a much stronger relationship between 0 (or cloud temperatures above the inversion) and 8lsO, than between surface temperature and the isotopic content of snowfall. This was first demonstrated by Picciotto et al. (1960), who found a relationship between the mean temperature within precipitating clouds at King Baudouin Station, Antarctica (Fig. 5.7) and the isotopic composition of snowfall at the surface (8lsO = 0.9T + 6.4, where Tis in the range +5 °C to -30 °C). Subsequently, Aldaz and Deutsch (1967) conducted a similar study at the South Pole, in which isotopes in snow samples collected during the course of a year were compared with temperatures at the surface and up to 500 mb. They found a relationship between 8lsO values and condensation level temperatures (t) such that 8lsO = 1.4i + 4.0 (where t is in the range -25 to -50 °C). These results were confirmed by Jouzel et al. (1983), who found that correlations between mean annual 8D and surface temperature at the South Pole were not as good as with temperatures just above the inversion layer.

These process-based studies are important in testing the statistically based relationships evident in long-term measurements from around the globe (see Fig. 5.6) with data from the polar ice sheets. However, with few exceptions (Steffensen, 1985) studies of 8180 in polar regions rely on the spatial relationship, which has been observed between SlsO in surface snowfall and mean annual temperature derived from 10-m depths (where the annual temperature cycle has been damped to near zero). Such data are derived from a geographically extensive network of ice

TEMPERATURE (°C)

FIGURE 5.7 Isotopic composition (8I80) in snowfall compared to the corresponding temperature in the precipitation cloud (Picciotto et al., I960).

cap sites. In looking at mean annual or mean monthly temperatures and corresponding 8lsO values, the difficult problems associated with the development of precipitation and the processes occurring in clouds on a storm-to-storm basis are avoided. In effect, it is assumed that mean condensation temperature and mean annual temperature vary in parallel. Why this should be so is hard to understand; most snowfall on polar ice sheets results from a small number of synoptic events occurring on only a fraction of days per year (generally <50%) so mean annual temperature, which is greatly influenced by strong surface inversions in dry winter months, should have little in common with 8lsO values in the ice cores (Peel et al., 1988). Nevertheless, empirical observations do show that mean annual 8lsO values and mean annual temperature are strongly correlated in the spatial domain, as shown in Fig. 5.6. Whether this relationship can be applied to the temporal domain, to convert variations in 8lsO or 8D over time to changes in mean annual temperature, is an important issue (and one that also applies to many other paleo-

climate proxies). Cuffey et al. (1994, 1995) used borehole temperatures at GISP2 to examine this question. The down-hole temperature profile represents the thermal history of the site that is, in a sense, buried with the accumulating snow. The down-core 8lsO record represents another measure of that thermal history. By developing a model that optimizes the fit between these two records, Cuffey et al. were able to resolve the long-term relationship (over the past 600 yr) between 8180 and temperature (8180 = 0.53T — 18.2%o). However, longer-term changes (over the last glacial-interglacial cycle) using a deeper borehole record produced a solution of 8180 = 0.33T - 24.8%o. These differences must be considered in terms of the temporal scale, or frequency domain of interest. The large range in estimates of the slope (a) of the 8lsO-temperature relationship may be related to the various timescales being considered. On the timescale of individual storm events a = >1; for monthly to annual values, a = -0.65, and even lower values of a seem to be appropriate when considering changes over longer timescales. Thus, in comparing changes averaged over decades to millennia, 8lsO = 0.5T and over even longer periods of time (several to tens of millennia) 8lsO = 0.35T (Boyle, 1997). On this basis, Cuffey et al. (1995) interpreted the overall glacial to interglacial temperature change at GISP2 as +14-16 °C, considerably greater than previous estimates (which had relied on 8lsO = 0.65T). For short-term (abrupt) changes in 8lsO, higher values of a are probably more appropriate; thus Dansgaard et al. (1989) used a = 0.65 when interpreting abrupt changes in 8lsO, giving mean annual temperature changes of -7 °C within 50 yr. Such an interpretation implicitly assumes that the factors that influence modern patterns operated essentially unchanged (or at least with the same overall gradients) as they do today; changes in moisture source, or in the seasonal distribution of precipitation are not considered to be important. However, a change in the timing of precipitation could be particularly important, as mean temperatures change in the Spring and Fall by several degrees Celsius per week; a shift in precipitation events by only a few weeks could result in large changes in 8lsO without any real change in mean annual temperature (Steig et al., 1994). Similarly, changing source regions for precipitation could also account for the abrupt isotopic shifts (Charles et al., 1994; Kapsner et al., 1995). Indeed, the rapid changes seen in Greenland ice cores during the last glaciation are associated with changes in other parameters (e.g., dust and Ca++ levels), which suggests that the precipitation source regions did vary.

Many other factors affecting 8lsO at a location today have not been constant over time. During glacial periods, ice thicknesses gradually increased on many ice sheets, resulting in lower 8lsO values at the surface because of the increase in elevation. However, this may not have been the case everywhere; there is evidence that both the GISP2 and Vostok sites were lower in the LGM (due to lower accumulation rates during colder times) so this could have led to less adiabatic cooling and higher 8lsO levels (Lorius et al., 1984; Cuffey and Clow, 1997). More extensive sea ice during glacial periods would effectively have increased distance to moisture sources, leading to lower 8lsO values in isolated continental interiors (Kato, 1978; Bromwich and Weaver, 1983). Furthermore, during glacial periods the isotopic composition of ocean water itself changed (a 8lsO value of ~l.l%o higher than today) due to the storage of water depleted in lsO in large ice sheets (Labeyrie et al., 1987; Shackleton, 1987). Finally, as temperatures fell to very low levels in some regions, 8lsO would decrease more rapidly for a given drop in temperature, because of the curvilinear nature of the 8180-temperature relationship (Fig. 5.8). Any interpretation of isotopic values in ice cores must consider all these factors, which undoubtedly have had an effect on the isotopic composition of high-latitude precipitation over time.

One approach to understanding how these various factors have interacted to influence the isotopic content of precipitation in the past is to use general circulation models (GCMs) with isotopic tracers in the hydrological cycle (Joussaume et al., 1984). Results from simulations with modern boundary conditions compare very well with observations (Jouzel et al., 1987b, 1991). Running the models with glacial age boundary conditions suggests that there was little change in 8lsO equa-torward of 40° N or 50° S at the last glacial maximum (LGM) but that there were large decreases in 8lsO at higher latitudes (Fig. 5.9) (Joussaume and Jouzel, 1993; Jouzel et al., 1994; Charles et al., 1994).

FIGURE 5.8 Relationship between oxygen isotope ratio (8I80) and temperature of condensation level in samples of Antarctic precipitation. Curves A and B are based on empirical observations at King Baudoin base and Amundsen-Scott Station. Curves I and 2 are theoretical using different assumptions about the fractionation of lsO at very low temperatures (Aldaz and Deutsch, 1967).

FIGURE 5.8 Relationship between oxygen isotope ratio (8I80) and temperature of condensation level in samples of Antarctic precipitation. Curves A and B are based on empirical observations at King Baudoin base and Amundsen-Scott Station. Curves I and 2 are theoretical using different assumptions about the fractionation of lsO at very low temperatures (Aldaz and Deutsch, 1967).

FIGURE 5.9 Differences between modern observed 8lsO in annual precipitation (using the same data as in Figs. 5.5 and 5.6) and simulated values for the last glacial maximum (LGM). Largest differences are associated with the polar ice sheets; in the intertropical zone, LGM values are actually higher than modern values, possibly reflecting a direct effect of higher 8lsO values in the ocean at that time. However, the tropical values are inconsistent with 8lsO in LGM groundwater in North Africa. Further data are needed to resolve this discrepancy (Jouzel et al„ 1994).

5.2.5 Deuterium Excess

On a global scale, fractionation of oxygen and hydrogen during evaporation and precipitation processes approximates a well-defined relationship, whereby SD = 88lsO +10. This defines what is termed the meteoric water line (Craig, 1961c), which in effect characterizes the "normal" equilibrium conditions that exist between 8lsO and 8D. The offset value (10 in this case) is termed the deuterium excess (d, where d = 8D - 88lsO). However, deuterium excess varies under non-equilibrium conditions, providing information not available from 8lsO and 8D alone.

Values of d vary because of kinetic effects occurring at both the evaporation and condensation stages of the water cycle. During evaporation, higher rates of diffusion to the water surface of molecules containing light isotopes leads to an increase in the light isotope content of water vapor, relative to the water source. This effect is in addition to the fractionation effect caused by the lower vapor pressure of water containing heavier isotopes (discussed in the preceding section). However, because of the different masses of HDO and H2180, the kinetic effect is slightly greater for H2180 than for HDO, resulting in changes of d when conditions deviate from equilibrium. Thus, when there is strong mixing of the surface waters (by higher wind speeds) or when relative humidities increase (reducing evaporation rates) or when water temperatures decrease (also reducing evaporation rates) the kinetic effect is reduced and values of d in precipitation will be lower. Thus, Jouzel et al. (1982) interpreted low values of deuterium excess (d = 4%o) in the pre-Holocene

90 N

60 N

30 N

30 S

90S 180

FIGURE 5.9 Differences between modern observed 8lsO in annual precipitation (using the same data as in Figs. 5.5 and 5.6) and simulated values for the last glacial maximum (LGM). Largest differences are associated with the polar ice sheets; in the intertropical zone, LGM values are actually higher than modern values, possibly reflecting a direct effect of higher 8lsO values in the ocean at that time. However, the tropical values are inconsistent with 8lsO in LGM groundwater in North Africa. Further data are needed to resolve this discrepancy (Jouzel et al„ 1994).

120W

120E

90 N

60 N

30 N

30 S

90S 180

120W

120E

section of an ice core from Dome C, Antarctica, as indicative of higher relative humidities and/or higher wind speeds in the water vapor source regions than during the Holocene (when d = 8%o). Increased levels of Na+ (from sea salt) and aeolian dust confirmed that wind speeds and humidities were probably higher in glacial time than in the Holocene.

A model of the isotopic fractionation of precipitation, based on the Rayleigh distillation process described earlier, works well at temperatures above —10 °C, but under the colder conditions common in Greenland and Antarctica, the simple model cannot explain the observed 8D - 8lsO relationship, or the isotope-temperature gradients. At such low temperatures, clouds become supersaturated with respect to ice and the transition from water vapor in clouds directly to ice crystals is the principal mode of precipitation formation. However, the lower molecular diffusivity of water molecules containing heavier isotopes (HDO, H2lsO) leads to the preferential condensation of isotopically light water molecules on the ice crystals. This is analogous to the kinetic fractionation effect that occurs at the ocean surface during the process of evaporation. By taking this additional kinetic effect into account, a much better fit between modeled and observed deuterium excess values, and a realistic isotopic-temperature relationship is obtained for snow forming at very low temperatures (Jouzel and Merlivat, 1984). Unfortunately, snowfall deposited on polar ice sheets has generally undergone a complex history; the transition temperature at which clouds become primarily made up of ice crystals rather than water vapor is not well known, yet this significantly influences the value of d (Fisher 1991, 1992).

Models of isotopic fractionation taking into account kinetic effects during evaporation and snow formation can be used to constrain the uncertainties inherent in these processes, providing good simulations of observed variations (Jouzel and Merlivat, 1984; Petit et al., 1991; Fisher, 1992). Petit et al. (1991), for example, show that the observed relationship between d and 5D in East Antarctica could only be accounted for if initial sea surface temperature at the air mass source region was between 15 and 22 °C, corresponding to latitudes 30-40° S (Fig. 5.10). Changes in humidity in the source region are of secondary importance, but are of more significance at higher values of 8D (warmer, lower elevation sites in Antarctica). Of course, it is unrealistic to expect that the moisture source for Antarctic snow is confined to only one latitudinal band, and this factor has also been modeled. Simulations in which moisture was added to the air mass as it passed from 30-40° S to Antarctica give results that are consistent with observations, suggesting that up to 30% of moisture originates from areas with SSTs <10 °C (Petit et al., 1991). This model thus provides strong support that mid-latitudes are the primary moisture source for at least the interior of Antarctica, with smaller contributions from regions at higher latitudes.

On the other hand, experiments with the NASA/GISS general circulation model incorporating deuterium (which can be traced back to each ocean basin) suggest that the average source region SST for Antarctic mid-winter precipitation is considerably lower, in the range 9-14 °C (Koster et al., 1992). Further experiments on total annual precipitation may help to resolve these differences. Whatever the solution w in

Effect of latitude of moisture source on deuterium excess of snow on ice sheet

Average SD

-250

FIGURE 5.10 Model-derived estimates of how the latitude of the oceanic moisture source region influences the average 8D and d values in Antarctic precipitation (solid lines) compared to the observed 8D-d relationship in Antarctica (symbols — based on surface snow samples across a wide geographical region). Each model line uses a different supersaturation function (relating the supersaturation of water vapor [with respect to ice] to the snow surface temperature) optimized to produce the best fit with observed data.The latitudes 30-40° S correspond to SSTs of-15-21° S (Petit et al„ 1991).

is, small seasonal variations in d (± 5%o ) in snowfall at the South Pole seem to indicate that the process of delivering snow to this location involves fairly consistent pathways from the moisture source, year-round. Similar conclusions were reached by Johnsen et al. (1989) for the higher elevations of Greenland, though Fisher (1992) found that the isotopic content of snow at the summit region of the ice sheet (Crête) was not compatible with a single moisture source, and that a mixture of moisture from the east (2/3) and west (1/3) seemed probable. This conclusion is similar to that reached by Charles et al. (1994) using a GCM with isotopic tracers in the model's hydrological cycle. When glacial age boundary conditions were imposed on the model, source regions changed; the southern part of the ice sheet was dominated by North Atlantic moisture sources, and the northern part by North Pacific moisture. With the incorporation of isotopic tracers into other general circulation models, it should be possible to gain further insights into fractionation processes occurring today, providing greater confidence in the interpretation of past changes observed in ice core records (Jouzel et al., 1993a).

Variations in deuterium excess in relation to 8lsO and 8D also shed light on the abrupt changes observed in Greenland ice cores during the LGM. Unlike the situation in the Antarctic Dome C record, during the coldest episodes of the LGM, d was no lower than in the Holocene (d = 8%o), suggesting that conditions in the moisture source were no different than today, with most of the moisture originating in the subtropical Atlantic. However, when there were abrupt shifts to warmer conditions (higher values of 8lsO) d became lower (by 4-5%o). Temperatures in the moisture source region must therefore have been lower in the milder episodes (and/or humidities and/or wind speeds were higher). Johnsen et al. (1989) explain this counterintuitive conclusion by suggesting that the milder periods of the LGM were associated with an abrupt shift in oceanic conditions in which the sea ice boundary rapidly retreated northward, revealing colder waters that acted as a local moisture source near to the ice sheet. Subsequently, as water temperatures increased, deuterium excess values became higher.

Variations in d will be better understood when a transect or network of ice cores across the major ice sheets becomes available, because unique explanations are generally not possible with only one record. Nevertheless, it is clear that a consideration of deuterium excess together with 8D and 8lsO will provide new insights into paleoclimatic conditions which cannot be obtained from dD or 8lsO alone.

0 0

Post a comment