Faunal Assemblage Paleoclimatology

FIGURE 6.19 Venn diagram illustrating conditions that are ideal (top) and nonideal (bottom) for calibrating a transfer function. In an ideal situation, the calibration data set C encompasses the range of all biological and environmental conditions that exist in the down-core data set D. In a non-ideal situation the calibration data set C does not reflect all the biological and environmental conditions that are represented within the down-core data set D and a no-analog condition results (shaded area). U is the universe of all biological and environmental conditions both today and in the past (Hutson, 1977).

data set itself. The "modern" faunal assemblages are generally derived from core-top samples that may represent a depositional period of several thousand years, due to bioturbation and disturbance during core recovery (Emiliani and Ericson, 1991). Indeed, the chronological heterogeneity of core-top samples was considered by Imbrie and Kipp (1971) to be the largest single source of error in their paleoenvironmental reconstructions ("most [core-top samples] represent... the last 2000 to 4000 years and. .. . some may contain materials deposited in the age range 4000-8000 years B.P."; Imbrie and Kipp, 1971). Furthermore, it is not unusual for modern océanographie parameters to be poorly known, commonly being based on interpolation between observations that are both short and geographically sparse (Levitus 1982). This is a particular problem in remote areas where sea-surface temperature and/or salinity gradients are strong, and may result in paleotemperature estimates for certain regions that are in error by several degrees. However, this is probably close to the magnitude of uncertainty associated with modern values, particularly in areas where significant changes of sea-surface temperature have occurred, even during the brief period of modern instrumental observations (Wahl and Bryson, 1975; Levitus, 1989). Under such circumstances the selection of a modern calibration value to equate with the core-top faunal assemblage is somewhat problematic, though by no means a problem unique to marine data (Bradley, 1991).

Of all the multivariate approaches to the quantification of former marine climates, the methodology of Imbrie and Kipp (1971) has been most widely applied.

In their original study, an attempt was made to reconstruct sea-surface temperature variations at a core site (V12-122) -150 km south of Haiti. To achieve this, the species composition of core-top samples from 61 sites in the Atlantic Ocean (and part of the Indian Ocean) were used as the basic "modern fauna" data set. As a first step, Imbrie and Kipp reduced the number of independent variables in this data set by the use of principal components analysis. Principal components analysis is an objective way of combining the original variables into linear combinations (eigenvectors) that effectively describe the principal patterns of variation in a few primary orthogonal components, leaving the less coherent aspects ("noise") for the last few components (Sachs et al., 1977). Thus, Imbrie and Kipp were able to condense much of the spatial variation of species abundance in 61 core-top samples from the Atlantic Ocean into five principal components or assemblages, which accounted for almost all of the variance in the original data set. By mapping the relative contribution of each component to the variance of each core-top sample, it was clear that four of these assemblages were related to temperature variations near the sea surface and could be simply described as subtropical, transitional, subpolar, and polar assemblages (Fig. 6.20). A fifth assemblage was more related to oceanic circulation around the subtropical high-pressure cells and was termed the gyre margin assemblage.

The next step was to utilize the relative weightings of each assemblage (factor scores) at each site to predict sea-surface temperatures. A stepwise multiple regression procedure was used, with temperature as the dependent variable and the factor scores as independent variables (predictors). In this way, an equation was derived that parsimoniously described sea-surface temperature in terms of the relative importance of the factor scores at each site. In the case of winter temperatures, for example, the following calibration equation was derived:

where A, B, C, and D refer to the four major assemblages (tropical, subtropical, subpolar, and polar) and K is a constant.18 This equation explained 91% of variance in the modern winter sea-surface temperature observations (Fig. 6.21).

At this stage, the modern faunal data set had been calibrated in terms of sea-surface temperatures. It was then necessary to transform the fossil (down-core) faunal variations from core V12-122 into relative weightings of the major faunal assemblages already defined. Finally, these values were entered into the calibration equation to produce paleotemperature estimates. These are shown in Fig. 6.22 together with 8180 measurements on the foraminifera Globigerinoides ruber from the same core (Imbrie et al., 1973). A sequence of cooler episodes can be seen separated by warmer periods when temperatures approached modern (core-top) values and the cooler periods generally coincide with high 8lsO values (and vice versa). However, the changes in water temperatures can only account for a small fraction (-20%) of the isotopic change and, in fact, provide support for the view that global ice-volume changes are manifested mainly in the 8lsO record

18 Gyre margin assemblage was not considered in this analysis.

90 90 90 90 90

90 90 90 90 90

dominated by a particular assemblage (polar, subpolar, transitional, subtropical, or gyre margin—off Equatorial West Africa) (Molfino et a/., 1982).
terpolaced values) vs those estimated from faunal assemblages in 61 core-top samples using factor analysis and transfer function methods (Imbrie and Kipp, 1971).

FIGURE 6.22 Winter sea-surface paleotemperature estimates (right) and 8lsO values (left) based on Caribbean core VI2-122. The sea-surface temperature estimates are derived from transfer functions, in the manner shown schematically in Fig. 6.18 (Imbrie et at., 1973).

(see Section 6.3.1). From this reconstruction of Caribbean Sea paleotempera-tures, it appears that sea-surface temperatures in this region have been predominantly cooler than today over the duration of the core record (-560,000 yr) with winter temperatures as much as -7.5 °C lower than today around 430,000 yr B.P. (Imbrie et al., 1973).

Since Imbrie and Kipp's pioneering work, there have been several attempts to refine and improve on their methodology (Ruddiman and Esmay, 1987; Dowsett and Poore, 1990). An alternative strategy was proposed by Prell (1985), who used a modern analog technique (MAT) to find the modern (core-top) assemblages that most closely resemble each fossil assemblage. Prell uses a statistical measure (a similarity coefficient) to quantify how closely each modern assemblage is to the fossil one. The modern sea-surface temperatures of the "top 10" modern assemblages are then used in a weighted average, to estimate the paleo-SST. This approach is developed further for the Atlantic Ocean, by Pflaumann et al, (1996), who demonstrate a very high degree of skill (r2 = 0.99) in estimating modern SSTs for both summer and winter seasons over the entire range of Atlantic temperatures, from -1.4 °C (in winter at high latitudes) to +28.6 °C (in summer in equatorial regions). This augurs well for reliable paleo-SST estimates when this approach is applied to down-core foram assemblage records, providing differential dissolution of paleo-assemblages has not biased their representativeness.

One of the most rewarding and interesting applications of Imbrie and Kipp's methodology has been use of the technique to provide a synoptic view of paleo-oceanographic conditions in the past. By applying transfer functions to samples from a particular time horizon in many different cores, it is possible to reconstruct and map marine climates as they were at that time. This was one of the major objectives of the CLIMAP project, which focused attention on marine conditions at 18,000 yr B.P. (CLIMAP Project Members 1976, 1981, 1984). The date of 18,000 (radiocarbon) yr B.P.19 was selected as the time of the last maximum continental glaciation (LGM) defined by maximum 8lsO values during isotope stage 2 (Shack-leton and Opdyke, 1973). Using transfer functions derived for each of the major world oceans, February and August sea-surface temperatures have been reconstructed for this period (Table 6.3). Most studies relied mainly on foraminiferal assemblage data, but in areas where siliceous fossils predominate (e.g., in the South Atlantic and Antarctic Oceans) the technique has also been applied to radiolarian assemblages (Lozano and Hays, 1976; Morley and Hays, 1979) and to diatoms (Pichon et al., 1992; Ko? Karpuz and Schrader, 1990). In the Pacific Ocean, where preservational characteristics of carbonate and siliceous fossils vary significantly from one area to another, it has been found advantageous to develop transfer functions based on four major microfossil groups (coccoliths, foraminifera, Radiolaria, and diatoms) to achieve optimum paleotemperature reconstructions (Geitzenauer et al, 1976; Luz, 1977; Moore, 1978; Sancetta, 1979; Moore et al, 1980). Although the paleotemperature maps for 18,000 yr B.P. are of interest alone, it is perhaps of most interest to use the reconstructions to produce maps of differences in tempera-

19 This 14C date is approximately equivalent to -21,000 calendar yr B.P. (see Section 3.2.1.5).

H TABLE 6.3 Sea-surface Paleotemperature Reconstructions for 18,000 Yr B.R

Area

Principal faunal groups used

Major reference

North Atlantic

Foraminifera

Kipp (1976), Mclntyre et al. (1976)

South Atlantic

Radiolaria

Morley and Hays (1979)

Norwegian and Greenland Seas

Foraminifera

T. Kellogg (1975, 1980)

Diatoms

Ko£ Karpuz and Schräder (1992)

Caribbean and equatorial Atlantic

Foraminifera

Prell etal. (1976)

Western equatorial Atlantic

Foraminifera

Be et al. (1976)

Eastern equatorial Atlantic

Foraminifera

Gardner and Hays (1976)

Indian Ocean

Foraminifera

Hutson (1978), Prell and Hutson (1979), Prell etal. (1980)

Antarctic Ocean

Radiolaria

Lozano and Hays (1976), Hays (1978)

Diatoms

Pichon etal. (1992)

Pacific Ocean

South

Foraminifera

Luz (1977)

North and South

Coccoliths

Geitzenaueretal. (1976)

North

Diatoms

Sancetta (1979, 1983)

North and South

Radiolaria

Moore (1978)

North and South

All four groups (synthesis)

Moore et al. (1980)

World Ocean (summary)

Foraminifera Radiolaria, Coccoliths

CLIMAP Project Members (1976)

ture between modern conditions and those at 18,000 yr B.P. Such maps are shown in Figs. 6.23-6.29 and are discussed briefly in the following sections. But first, a few caveats are neccessary.

One aspect of the CLIMAP SST reconstructions that has been controversial from the outset is the apparent lack of a significant temperature change in low latitude ocean surface temperatures at the last glacial maximum (LGM) (Prell, 1985). This conclusion does not fit well with other evidence from the tropics, such as the much lower snowlines recorded in mountainous areas (Selzer, 1990). If SSTs remained more or less constant, yet temperatures at an altitude of 4-6 km fell, an increase in lapse rate in the lower atmosphere is implied, but this is very difficult to envision (Webster and Streten, 1978). Furthermore, general circulation models can not reproduce an appropriate drop in snowline without sea-surface temperatures 5-6 °C lower than modern values (Rind and Peteet, 1985).

Recently, several independent lines of evidence have converged to challenge the veracity of low latitude CLIMAP SST estimates. Analysis of strontium/calcium ratios (see Section 6.8.5) in corals dating 10,200 B.P. from 16° S in the central Pacific indicate that temperatures were ~5 °C cooler at that time (Beck et al., 1992). A similar result was obtained with corals of LGM age from Barbados (Guilderson et al., 1994). Temperature changes of this magnitude are supported by studies of noble gases in groundwater from tropical Brazil. The concentration of noble gases (Ne, Ar, Kr, and Xe) in groundwater is largely a function of the temperature at the water table where dissolution occurs. By comparing noble gas concentrations in radiocarbon-dated groundwaters of Holocene and of glacial age, a temperature difference of 5.4 ± 0.6 °C was estimated. Additional evidence for lower tropical temperatures at the LGM comes from S180 measurements in an ice core from the high mountains of Peru (Huascaran: 6050 m elevation), which shows that snowfall was ~8%o lower in 8lsO in glacial times compared to the Holocene (Thompson et al., 1995b). This indicates much lower temperatures in the lower troposphere at that time. Finally, Miller et al., (1997) estimate that temperatures in central Australia were at least 9 °C colder in glacial times compared to the Holocene, based on the extent of racemization of amino acids in 14C-dated emu eggshells. As race-mization is a function of age and temperature, by controlling for age, paleotem-peratures can be calculated (see Section 4.2.1.4). These very diverse lines of evidence all point to significantly cooler temperatures in tropical and subtropical latitudes at the LGM.

Other lines of evidence are more in line with the original CLIMAP estimates (Table 6.4). For example, several recent studies use the temperature dependence of long chain alkenones (synthesized by planktic primnesiophyte algae — see Section 6.5) as a means of reconstructing SSTs (Brassell et al., 1986). These studies generally find that LGM SSTs were only 1-2 °C lower in the western and eastern tropical Pacific (Ohkouchi et al., 1994; Prahl et al., 1988) and central equatorial Atlantic (Sikes and Keigwin, 1994) but up to -2.5 °C lower in the central equatorial Indian Ocean (Rostek et al., 1993; Bard et al., 1997) and 2-3 °C lower in the eastern tropical Atlantic (with colder episodes as much as 4-5 °C lower for periods of a few hundred years at a time, corresponding to enhanced flow of the cool Canaries current during Heinrich events) (Zhao et al., 1995). In addition, a reassessment of the LGM SSTs using a large database of modern core-top samples to identify optimum down-core analog assemblages (the "modern analog technique") led Prell (1985) to conclude that the original CLIMAP estimates were not biased by methodology and required no drastic revision. This was also the view of Thunell et al., (1994) who used a high-quality core-top data set from the western Pacific; they found that LGM SSTs were generally within 1 °C of modern values between 20° N and 20° S, indicating that the western Pacific Warm Pool has existed since at least the LGM, and probably throughout the last glacial-interglacial cycle.

Such a view is strongly contested by Emiliani and Ericson (1991), who examine several lines of evidence which call into question the CLIMAP SST estimates at low latitudes. For example, they note that certain species of foram, with well-defined temperature tolerances, can be used as indicators of threshold temperatures in the past. Thus, the absence of Pulleniantina obliquiloculata and Spbaeroidinella dehis-cens from equatorial Atlantic and Caribbean sediments of LGM age suggests that winter temperatures fell below 18.5 °C, a drop of at least 7-8 °C from modern val-

H TABLE 6.4 Paleotemperature Estimates for the Last Glacial Maximum

Location

Lat

Long

AT (°C)

Reference

Based on planktonic forams"

Western Pacific (many sites)

22° N-20° S

120° E-165" E

<2

Thunnell et al. (1994)

Based on alkenones h

E. Equatorial Atlantic

23° W

1.8

Sikes and Keigwin (1994)

N.E. Equatorial Pacific

1° N

139° W

1.3

Prahl et al. (1989)

W. Equatorial Pacific

3.5° N

142° E

<1.5

Ohkouchi etal. (1994)

Off N.W. Africa

19° N

20° W

3-4

Zhao etal. (1995)

Off N.W. Africa

19° N

20° W

~3

Chapman etal. (1996)

Off N.W. Africa

21° N

18.5° W

3-4

Eglington etal. (1992)

Indian Ocean (a transect)

20° N-20° S

-36° E

0.5-2.5

Bard et al. (1997)

N.E. Atlantic

48.3° N

25° W

3-4

Madureira etal. (1997)

N.E. Atlantic

56° N

12.5° W

5

Sikes and Keigwin (1996)

Central N. Atlantic

43.5° N

30.4° W

4-5

Villaneuva etal. (1998)

Based on Sr/Ca or U/Ca

Huon Peninsula, New Guinea

>3

Aharon and Chappell (198

Huon Peninsula, New Guinea

>5-6

Min etal. (1995)

Vanuatu, southwestern Pacific

>4-5

Min et al. (1995)

Vanuatu, southwestern Pacific

15.5° S

167° E

>6.5

Beck etal. (1997)

Barbados

-5

Guilderson etal. (1994)

Barbados

4-5

Min etal. (1995)

Based on noble gases

Brazil

5.4

Stute etal. (1995)

Based on amino acid

racemization

Central Australia

-9

Miller etal. (1997)

" Using Modern Analog Technique for foram assemblages.

h Up-to-date estimates can be found at http://NRG.NCLAC.UK:8080/CLIMATE/Art.htm

ues (see Fig. 6.6). They argue that the primary reason for the "incorrect" CLIMAP SST estimates is the probability (noted earlier) that many of the "modern" core-top samples were in fact not representative of truly modern sediments, but were contaminated by early Holocene faunas, which represent quite different oceanographic conditions in many tropical regions. If this is so, they argue, then not surprisingly any further analysis of this data set is only likely to confirm the original (erroneous) conclusions. However, this cannot be true of the high resolution data set used by Thunell et al., (1994) in which core-tops were demonstrably modern.

Finally, a thorough review of planktic isotope data from the tropical oceans led Broecker (1986) to conclude that the LGM 8lsO values are consistent with the relatively small changes in SST revealed by the CLIMAP studies. This conclusion has been reinforced by Stott and Tang (1996), who examined 8lsO in individual planktic forams from the Holocene and the LGM in the tropical Atlantic. The majority of their samples showed that the tropical Atlantic SSTs were ~2 °C cooler at the LGM than in the Holocene. However, the presence of individuals with significantly higher 8lsO values (i.e. indicative of colder conditions) may reflect short-term episodes like those found by Zhao et al., (1995) in their alkenone studies.

No doubt, this controversial subject will continue to be hotly debated as new evidence is brought to bear on the problem. But it is worth noting that the matter is of more than academic interest because the role of tropical SSTs is very important in understanding how global climates may change in the future. If large decreases in tropical SSTs did occur during glacial times, it implies that climate sensitivity to changes in greenhouse gases (and associated feedbacks) is at the high end of most estimates (-4.5 °C for a doubling of C02), otherwise the LGM forcing would not have produced such large changes (Crowley, 1994). Hence, resolving these differences in LGM SSTs is critical for a clearer understanding of the evolution of future climate. Meanwhile, when examining differences between modern SSTs and the CLIMAP reconstructions in Figs. 6.23-6.29 one should bear in mind that an ocean of controversy continues to surround LGM SST estimates at low latitudes.

One final point before the CLIMAP reconstructions are introduced; it is important to note that the maps are derived by interpolation between discrete sample points, which are often widely separated. This may lead to erroneous SST estimates being extrapolated over huge areas of the surrounding ocean (especially in the Pacific). This problem has been examined by Broccoli and Marciniak (1996), who find that the correspondence between GCM simulations of LGM paleotemperatures is significantly better when individual data points are compared, compared to using the interpolated paleo-SST maps. This should be borne in mind when examining Figs. 6.23-6.29; for example, in Fig. 6.23 the strongest gradients in temperature anomalies in the North Atlantic (located south of Newfoundland) are not well constrained by data (shown by the black dots) and result from the difference between an interpolated paleo-SST map and modern instrumentally measured conditions. This is not to say that they are wrong, but only that due caution is needed in interpretation of the maps.

6.4.1 North Atlantic Ocean

Situated between major ice sheets of the Northern Hemisphere (at 18,000 yr B.P.) the North Atlantic experienced the most significant changes in temperature of all oceanic areas (Figs. 6.23 and 6.24). At 18,000 yr B.P. August sea-surface temperatures were more than 10 °C cooler in a broad zone from 40-45° N in the west to 45-50° N in the east. This reflects a marked southward movement of polar and subpolar water at the time. In February, temperatures were less depressed off the North American coast (AT = 3-5 °C) but in the eastern North Atlantic temperatures were 6-12 °C cooler than today in a triangular area stretching from Scandinavia to Portugal. In both seasons, strong upwelling off the coast of northwestern Africa (presumably related to stronger Trade winds, see Section 6.7) resulted in signifi-

FIGURE 6.23 Difference between modern and 18,000 yr B.R August sea-surface temperatures. Contour interval is 2 °C. Cores used in paleotemperature study shown as dots. Paleotemperatures were derived from faunal assemblage transfer functions. Figure derived by subtracting map of modern temperatures from map of paleotemperatures (both containing interpolated data). Hence the estimates of paleotemperature increase since 18,000 yr B.P. contain large values in some areas, even though core data may not have been available from those areas (e.g., the western North Atlantic between 42° and 50° N). See text for discussion of uncertainties in tropical SSTs (Mclntyre et al„ 1976).

FIGURE 6.23 Difference between modern and 18,000 yr B.R August sea-surface temperatures. Contour interval is 2 °C. Cores used in paleotemperature study shown as dots. Paleotemperatures were derived from faunal assemblage transfer functions. Figure derived by subtracting map of modern temperatures from map of paleotemperatures (both containing interpolated data). Hence the estimates of paleotemperature increase since 18,000 yr B.P. contain large values in some areas, even though core data may not have been available from those areas (e.g., the western North Atlantic between 42° and 50° N). See text for discussion of uncertainties in tropical SSTs (Mclntyre et al„ 1976).

cantly cooler temperatures in that area (AT = 5-8 °C). Relatively minor temperature differences were apparent over most of the subtropical North Atlantic to the west, though February sea-surface temperatures in the Caribbean were 2-4 °C cooler than today at 18,000 yr B.P.

Considered together, the two maps suggest that the North Atlantic is made up of two zones, an area of dynamic change from -40 to 50° N and a relatively stable zone to the south. Such a condition has apparently been characteristic of a much longer period than just the last 18,000 yr. Mclntyre et al., (1975) have reconstructed major water-mass boundaries along a meridional north-south transect through the Atlantic Ocean (at 20° W) for the past 130,000 yr (i.e., an interglacial-glacial cycle) and their study clearly indicates that the North Atlantic has been the most variable zone of

FIGURE 6.24 Difference between modern and 18,000 yr B.P. February sea-surface temperatures, derived as explained in Fig. 6.23. Contour interval is 2 °C (Mclntyre et al., 1976).

both hemispheres (Fig. 6.25). At 50° N, faunal assemblages have varied from being predominantly subtropical at 125,000 yr B.P. (the last interglacial) to polar from -35,000 to 15,000 yr B.P. Furthermore, changes of a similar nature have been observed in cores reaching back over 600,000 yr, indicating that seven complete glacial-interglacial cycles have occurred during this period (Ruddiman and Mclntyre, 1976).

6.4.2 Pacific Ocean

August sea-surface temperature differences were maximized in subarctic and equatorial regions (Fig. 6.26). In the area around Japan, temperatures were as much as 8 °C cooler at 18,000 yr B.P. due to a southward displacement of the warm Kuroshio current at that time and its replacement by subarctic water (Oyashio current). Less pronounced temperature depressions occurred in the Gulf of Alaska and southward along the California coast (AT = 2—4 °C), a fact which stands in marked contrast to conditions in the eastern Atlantic Basin. In equatorial regions, temperatures were cooler by 2-4 °C in a broad band, perhaps related to greater advection of cool waters into North and South Equatorial currents due to intensified tradewinds at the

FIGURE 6.25 Variations of Atlantic surface water masses along the 20° W meridian, from 60° S (left) to 60° N (right) over the last 130,000 yr (vertical axis). Principal ocean currents along the 20° W meridian are shown schematically at top. Water mass variations shown cover the last interglacial-glacial cycle. Note the large latitudinal variations in water masses in the North Atlantic (Mclntyre et al., 1975).

FIGURE 6.25 Variations of Atlantic surface water masses along the 20° W meridian, from 60° S (left) to 60° N (right) over the last 130,000 yr (vertical axis). Principal ocean currents along the 20° W meridian are shown schematically at top. Water mass variations shown cover the last interglacial-glacial cycle. Note the large latitudinal variations in water masses in the North Atlantic (Mclntyre et al., 1975).

120* 150" 1S0" 150" 120" 90s

120* 150" 1S0" 150" 120" 90s

Campo Oxidaci

FIGURE 6.26 Differences between modern and 18,000 yr B.P.August sea-surface temperatures In the Pacific Ocean. Dots show locations of cores used in paleotemperature estimates. Dark shading indicates 18,000 yr B.P. temperatures >4 °C cooler; intermediate shading 2—4 °C cooler; light shading 0-2 °C cooler than modern values. Areas where it was warmer at 18,000 yr B.P. than under modern conditions are unshaded. See text for discussion of uncertainties in tropical SSTs (Moore et at., 1980).

FIGURE 6.26 Differences between modern and 18,000 yr B.P.August sea-surface temperatures In the Pacific Ocean. Dots show locations of cores used in paleotemperature estimates. Dark shading indicates 18,000 yr B.P. temperatures >4 °C cooler; intermediate shading 2—4 °C cooler; light shading 0-2 °C cooler than modern values. Areas where it was warmer at 18,000 yr B.P. than under modern conditions are unshaded. See text for discussion of uncertainties in tropical SSTs (Moore et at., 1980).

time. In the South Pacific, cooler waters adjacent to the coast of South America and west of New Zealand are noteworthy (AT = 2-4 °C). It is also of interest to note that, in addition to areas of major cooling, large parts of the Pacific Basin appear to have been warmer at 18,000 yr B.P. than at present. In particular, core regions of subtropical high-pressure centers are reconstructed as having been 1-2 °C warmer than modern values. Temperatures along the eastern coast of Australia were also warmer (by up to 4 °C) perhaps due to enhanced equatorial flow from the stronger Equatorial currents. However, this view is contested by Anderson et al., (1989), who show that the CLIMAP paleo-SSTs in this region are too high. One potentially important problem in deriving paleo-SSTs from the tropical Pacific is the effect of differential dissolution, which tends to remove "warm" forams from the core-top assemblages. During the LGM, if dissolution in the Pacific was less, there may have been more "warm" forams preserved, leading to an erroneous view of overall temperature changes (Broccoli and Marciniak, 1996).

Similar patterns of difference are observed in many areas in the February sea-surface temperature reconstructions (Fig. 6.27) with maximum temperature changes

FIGURE 6.27 Difference between modern and 18,000 yr B.R February sea-surface temperatures in the Pacific Ocean. Dots show location of cores used in paleotemperature estimates. Shading as in Fig. 6.26 (Moore eta/, 1980).

in the area east and north-east of Japan (AT = 6-8 °C). Cooler temperatures at this season could have resulted in sea-ice formation over an extensive area (Sancetta, 1983). In the Southern Hemisphere, cooling was relatively minor at 18,000 yr B.P., except in the Peruvian current off the western coast of South America (AT = 2-4 °C) and in the extreme south due to an expanded subpolar water mass. Again, a noticeable feature is the extensive area of warmer sea-surface temperatures at 18,000 yr B.P. centered over the subtropical high-pressure cells. It is particularly interesting to note the large extent of this positive anomaly in the Southern Hemisphere, associated with the poleward movement of the subtropical high-pressure center at this time of year, A corresponding southward shift in the Northern Hemisphere positive anomaly field is also apparent in the February maps compared to those for August. Such a pattern suggests a more intense Hadley cell circulation at 18,000 years B.P., with well-developed subtropical high-pressure centers. In these areas, adiabatic warming and clear skies would favor warmer sea-surface temperatures and, on the subtropical margins, trade winds and gyre margin ocean currents would be strengthened. All these factors fit together quite coherently in relation to the reconstructed paleotemperatures, which demonstrates that the overall reconstructions are at least internally consistent.

6.4.3 Indian Ocean

August sea-surface temperature anomalies reveal relatively minor differences between 18,000 yr B.P. and today (Fig. 6.28). Apart from areas associated with the eastern and western boundary currents, off the western coast of Australia and off southeastern Africa (the Agulhas Current) most areas at 18,000 yr B.P. were within 1 or 2 °C of modern values. It is interesting that temperatures in the Arabian Sea were ~1 °C warmer at 18,000 yr B.P., suggesting a weaker Southwest Monsoon flow at that time, resulting in less upwelling of cool water (Prell et al., 1980). However, this is not supported by alkenone evidence from off the Arabian Peninsula, which indicates glacial-interglacial temperature differences of >3 °C in spite of reduced upwelling at the LGM (which would tend to lessen the difference) (Emeis et al., 1995).

February maps reveal larger temperature differences, particularly in the area centered on 40° S, where northward movement of the Antarctic Convergence zone and associated subpolar water caused temperatures to be lower by 4-6 °C at 18,000 yr B.P. (Fig. 6.29). Compared to the other ocean basins, however, temperature changes in the Indian Ocean were relatively small (overall cooling of only -1.8 °C) and large areas off the coast of eastern Africa and in the Arabian Sea may have been slightly warmer at 18,000 yr B.P. than they have been in recent years.

AUGUST AT

LGM-MODERN

FIGURE 6.28 Difference between modern and 18,000 yr B.P. August sea-surface temperatures. LGM = last glacial maximum. Contour interval is I °C. Widely spaced diagonal lines indicate areas warmer at 18,000 yr BP than today. Closely spaced diagonal lines indicate areas at least 4 °C cooler than today at 18,000 yr B.P. See text for discussion of uncertainties in tropical SSTs (Prell et at., 1980).

AUGUST AT

LGM-MODERN

FIGURE 6.28 Difference between modern and 18,000 yr B.P. August sea-surface temperatures. LGM = last glacial maximum. Contour interval is I °C. Widely spaced diagonal lines indicate areas warmer at 18,000 yr BP than today. Closely spaced diagonal lines indicate areas at least 4 °C cooler than today at 18,000 yr B.P. See text for discussion of uncertainties in tropical SSTs (Prell et at., 1980).

FEBRUARY AT

LGM-MODERN

FIGURE 6.29 Difference between modern and 18,000 yr B.R (LGM) February sea-surface temperatures. Contour interval is I °C. Shading as in Fig. 6.2 (Prell et <j/., 1980).

FEBRUARY AT

LGM-MODERN

FIGURE 6.29 Difference between modern and 18,000 yr B.R (LGM) February sea-surface temperatures. Contour interval is I °C. Shading as in Fig. 6.2 (Prell et <j/., 1980).

0 0

Post a comment