ENSO-like Forcing on Oceanic Primary Production During the Late Pleistocene

See allHide authors and affiliations

Science  28 Sep 2001:
Vol. 293, Issue 5539, pp. 2440-2444
DOI: 10.1126/science.293.5539.2440


Late Pleistocene changes in oceanic primary productivity along the equator in the Indian and Pacific oceans are revealed by quantitative changes in nanoplankton communities preserved in nine deep-sea cores. We show that variations in equatorial productivity are primarily caused by glacial-interglacial variability and by precession-controlled changes in the east-west thermocline slope of the Indo-Pacific. The precession-controlled variations in productivity are linked to processes similar to the Southern Oscillation phenomenon, and they precede changes in the oxygen isotopic ratio, which indicates that they are not the result of ice sheet fluctuations. The 30,000-year spectral peak in the tropical Indo-Pacific Ocean productivity records is also present in the Antarctica atmospheric CO2 record, suggesting an important role for equatorial biological productivity in modifying atmospheric CO2.

Interannual variability of the thermocline depth is a characteristic feature of the equatorial Pacific Ocean that has a strong influence on oceanic primary production (1). Most of the time, the tropical trade winds transport surface waters westward. This lifts the thermocline along the equator in the central and eastern Pacific but deepens the thermocline in the western Pacific, where warm surface waters pile up against the Australian and Asian continents. In the Indian Ocean, an analogous pattern exists but with a shallow thermocline in the west (the Somalian upwelling) and a deep thermocline in the east (2). When trade winds relax, the slope of this basin-wide thermocline changes intensely. The thermocline rises in the western Pacific and in the eastern Indian Ocean, whereas it deepens on the other sides of these two basins [i.e., causing a typical El Niño–Southern Oscillation (ENSO) event in the Pacific]. Although the two oceans appear to be climatically independent of each other (2, 3), a synchronism of equatorial wind intensity has been observed between the Pacific and Indian oceans on decadal scales (4, 5). A shoaling of thermocline produces an increase in primary production. Thus, the ENSO-related thermocline slope variability has been recognized as responsible for the “largest known natural perturbation of the global carbon cycle” on decadal scales (1). Here, we document long-term variations in the slope of the equatorial thermocline on the basis of patterns of changing primary productivity during the last two glacial cycles and its relation to global climate and orbital forcing.

We analyzed the coccolithophore community preserved in nine deep-sea cores taken from the tropical Indian and Pacific oceans (Fig. 1) (6). A precise chronology was obtained by tuning oxygen isotope records from foraminifera in most of the cores to the SPECMAP stacked record (7) so that our sampling achieves a temporal resolution of 0.7 to 3 thousand years (ky). Six records span at least the last 240 ky, two shorter cores (M38 and M40) the last 180 ky, and one (M74) the last 150 ky. Most of the coccolithophore species live in near-surface waters where light for photosynthesis is abundant. Phytoplankton also requires nutrients, and thus thrives where a shallow thermocline brings subsurface nutrients into the upper euphotic zone. In contrast, the coccolithophoreFlorisphaera profunda lives in the deep-photic zone, where nutrients are relatively abundant but light is rare (8). Where the thermocline is deep, total primary productivity is low, and the dominant coccolith species in fossil assemblages is F. profunda (9). As productivity increases, the relative abundance of F. profunda decreases. Thus, estimates of primary productivity (ePPs) are made from counts of the relative abundance of F. profunda (%Fp) (6,10). The data are given as productivity estimates because, within tropical systems, productivity is highly correlated with the depths of the nutricline and the thermocline (1, 11, 12).

Figure 1

(A) Map showing core locations and sea level anomalies relative to a 3-year mean recorded by Topex-Posidon during El Niño 1997 (cycle 193); scale is on the right. (B) Record of primary production from each core estimated on the basis of human (blue or red curves) and automated (neural network) counts (36)(dashed lines) of coccoliths. (C) Spectral analysis (multitaper method) of ePP records. Amplitude spectra are in orange; the significance values of the peaks (F test) are in blue. Shaded areas indicate the orbital periods of 100, 41, and 23 to 19 ky. Significant peaks with a period of about 30 ky are also indicated.

The ePP records (Fig. 1) show large amplitude variability with regional similarities and contrasts that are explored using empirical orthogonal functions (EOFs). EOFs reduce the set of observations to a small number of spatiotemporal patterns (an EOF series) that can adequately describe the data (13, 14). It also provides statistics (loading) on how each original record contributes to the EOF series. Two significant EOF series (EOF-1 and EOF-2) have been extracted for the last 180 ky from the eight cores (Figs. 2, 3; the maximum duration represented by cores M38 and M40 is 180 ky). They describe 60% of the total variance in all eight cores. EOF-2 reflects a straightforward mechanism and is discussed first.

EOF-2 accounts for 19% of the total variance and opposes essentially the western Pacific with the central Indian Ocean (6), and it clearly fluctuates with the 23-ky cycle of Earth's precession (Fig. 2E). However, the eastern Pacific ePP records show weak precession cycles in comparison with the 41- and 100-ky cycles, and consequently they do not contribute significantly to EOF-2. We repeated the EOF analyses on 23-ky bandpass-filtered ePP records. This filtering technique suppresses the influence of the large glacial-interglacial variability on the eastern Pacific productivity records. The first component of this new test (EOF-1f) approximates that of EOF-2, although it gives a much clearer picture of the opposition in dynamic between the western and eastern Pacific ePPs (Fig. 2) (6). This cyclic opposition can also be seen when comparing basin average regional records (Fig. 2C). Contemporary interannual variability in ENSO shows a similar spatial opposition in the altimetric anomaly (Fig. 1), thermocline depth, and hence productivity fluctuations in the equatorial Pacific (1) and equatorial Indian oceans (2). Thus, the opposition we found is likely to reflect similar variations in equatorial thermocline slope, although on a much longer time scale (15).

Figure 2

(A) The 500-year average NINO3 index from Zebiak-Cane model forced with Milankovitch solar variations (16) (solid red line). EOF-1 from the 20-ky bandpass-filtered ePP series (dashed blue line) with chronology based on U-Th date from Gallup et al. (19) (red dots) and Henderson and Slowey (20) (black squares). The data in (A) are shifted by 5 ky to match visual thermocline dynamics expressed in (B) and (C) (shaded areas correspond to high EOF-1f values). (B) EOF-2 from eight raw ePP records on reverted scale (red) and EOF-1 (EOF-1f) from the 20-ky bandpass-filtered ePP series (blue). (C) Productivity records added by pairs of cores: western Pacific (M38 added to M40 in dark blue), central Indian Ocean (M63 added to M48 in red), and eastern Pacific (20-ky filters of W84 added to 20-ky filter of core R13 in light blue). (D and E) Spectral analysis (multitaper method) of ENSO model (D) and EOF-2 (E). Amplitude spectra are in red; the significance values of the peaks (F test) are in light blue. Shaded areas indicate the orbital periods of 100, 41, and 23 to 19 ky.

Interestingly, in every core analyzed, ePP records lead by about 2 ky the δ18O records in the precession band (6). This lead indicates that insolation is the direct cause of ePP fluctuations, and that polar ice volume variations (δ18O) do not significantly affect the low-latitude climate dynamics in that time scale. An ENSO modeling experiment over the past glacial-interglacial cycles has shown that changes in low-latitude insolation induce variations in the frequency and intensity of ENSO events (16). A very good agreement is found between the predictive ENSO model and EOF-2-1f time series, although an almost constant shift of about 5 ky occurs between the model and the EOFs (Fig. 2). This offset may partly be explained by the use of the SPECMAP chronology for our age model (7), because new radioisotope dates point to significantly older ages for most of the isotopic record than previously estimated [compare (17–20) with (7)]. Calibrating the δ18O chronology based solely on the recent U/Th dates, the EOFs offer a better synchronism with the ENSO model (Fig. 2A). Although chronology is beyond the objective of this work, it appears that older ages in the SPECMAP chronology would reconcile model predictions and observed/measured data. Hence, the observed long-term variations of the equatorial thermocline slope would seem to be related to variations in the intensity of ENSO events driven by precession-related insolation changes.

EOF-1 accounts for 40% of the total variance. It characterizes productivity variations common to all sites (Fig. 3) (6). Because it contains significant periods close to the 100-ky (eccentricity) and 41-ky (obliquity) bands, and because EOF-1 broadly covaries with the SPECMAP stacked oxygen isotope record (Fig. 3), EOF-1 expresses the dominant influence of glacial variability on equatorial productivity. High EOF-1 values (i.e., high-productivity conditions) occur during maximum glacial conditions, low EOF-1 values (i.e., low-productivity conditions) during interglacial times. EOF-1 varies in phase with δ18O in the eccentricity band and lags it in the obliquity band, indicative of a significant response of low-latitude productivity dynamics to glacial-interglacial variability. These spectral characteristics are shown by each ePP series (6), except the central Indian Ocean one (10). The process that increases primary production during glacial in the equatorial Indo-Pacific remains uncertain. It cannot correspond to an increase in the strength of zonal winds, because that would produce an increase of the slope of thermocline in the Pacific, a fact that is not observed for EOF-1 (6). Such a general rise of nutrient in the equatorial upper photic zone could be related to a global increase of nutrient concentration in the intermediate waters that upwell into the photic zone. More likely it results from a weakened and/or shoaled stratification that permits an increased flux of new nutrients. A decrease of equatorial sea surface temperatures (14, 21, 22) relative to the intermediate water temperatures during glacial periods would weaken the stratification, and the thermocline would shoal (23). Dry weather conditions in the western Pacific (24) would further reduce stratification because a reduction of heavy rains would reduce fresh surface water inputs and its strong specific halocline. It has been shown that the global climate variability modulates the ENSO (23) by altering the background state and in particular the depth of the thermocline: Warmer climate increases the frequency and strength of the El Niño. The general fluctuations of the thermocline depth as recorded by EOF-1 are in complete agreement with those of the influence of the global climate on low-latitude oceanography.

Figure 3

(A) Purple line, first component of the EOF; red line, SPECMAP stack; blue line, SPECMAP stack transformed in the following manner: the stack was filtered at 100, 41, and 20 ky. The 20-ky filter was inverted for interglacial periods (unshaded areas) and kept as initial in glacials and during terminations (shaded area). The three filtered series were then added together. At the bottom positive part of 30-ky Gaussian filters (37) (centered on 0.034 ky−1 with a band width of 0.004 with cutting at period of 26.6 and 34.3 ky) of EOF (purple), transformed SPECMAP (blue), 23-ky sine-wave frequency modulated by 100- and 40-ky cycle from SPECMAP (green) (38), and –CO2 from Vostok (orange) (30). (Bto D) Spectral analysis (37) (multitaper method) of SPECMAP (B), EOF-1 (C), and transformed SPECMAP (D). Amplitude spectra are in red; the significance values of the peaks (F test) are in light blue. Shaded areas indicate the orbital periods of 100, 41, and 23 to 19 ky. The extra 30-ky period is also indicated.

Finally, the spectra of EOF-1, and to a lesser degree of EOF-2, reveal significant peaks at 30 and 19 ky (Figs. 2 and 3). These spectral peaks are interpreted as reflecting the phase modulation of precession in ePP: The time series of EOF-1 seems to be in phase with those of SPECMAP during maximum glacial and deglaciation (stripped in gray in Fig. 3A) conditions, but in anti-phase during interglacial periods. To illustrate this, we inverted the precession signal of the SPECMAP record for each interglacial period. The spectral signature of this new series (transformed SPECMAP, Fig. 3A) shows a distinct 30-ky period and a shift of the precession period from 23 ky to 19 ky (Fig. 3D). These spectral signatures are those of EOF-1. Therefore, the 30-ky component that occurs in EOF-1 as well as in most of the ePP series (Fig. 1) corresponds to a cyclic phase shift of the precession parameter. Two forcings are acting in phase opposition at a 20-ky time scale and in amplitude opposition at a 100-ky time scale. The mechanism involved is uncertain, but could be linked to the boreal summer monsoon (BSM) dynamics. Present climate studies show that the strengths of BSM and ENSO are often correlated with stronger BSMs during La Niña events [for example, (25)]. But on Milankovitch time scales, this is not the case: BSMs intensify during interglacial periods (26), whereas we have shown that ePPs (La Niña equivalent) increase during glacial periods. It is therefore plausible that during interglacials, intense BSMs somehow altered the path of the thermocline depth dynamics described above and were responsible for that 30-ky period.

A 30-ky period has been found in series from the tropical Pacific [see discussion in (14)] and tropical Indian (27–29) oceans, but so far not from the tropical Atlantic. We speculate that the 30-ky cycle is restricted to the tropical Indo-Pacific Ocean (14), because the ENSO is characteristic of that area. However, the CO2 record from the Vostok ice core also reveals a 30-ky cycle (30). Cross-spectral analyses indicate that the productivity series are highly coherent with the CO2 record at the 30- and 23-ky periods, and that low CO2 values are associated with high productivity (Fig. 3) (6). It is therefore possible that, with primary production acting as a significant sink in the carbon cycle, the 30-ky record in global CO2 is the signature of ENSO-like control of biological production in the equatorial Indo-Pacific. This is consistent with a significant role of the low-latitude biological pump in controlling atmospheric CO2concentrations (31).

We have identified two independent forcings responsible for 60% of the long-term equatorial Indo-Pacific productivity dynamics. The first forcing concerns the response of the depth of the equatorial thermocline to global climatic variations. The second forcing is related to changes in the equatorial east-west thermocline tilt and is linked to the 23-ky period of Earth's precession. This precession-related variability could reflect the influence of low-latitude insolation on ENSO, as a predictive ENSO model stipulates (16). A similar dual “precession-glacial” forcing has been described recently on New Guinea corals (32). The 23-ky signal precedes ice volume variations by about 2 ky. Thus, long-term ENSO dynamics provide a possible causality for the growing body of evidence that low-latitude climates are early responders to orbital forcing (9, 10, 22, 33–35). Although minor, a 30-ky period is also evidenced. It is characteristic of the equatorial Indo-Pacific coherent with a similar period found in Vostok CO2 record. That coherency attests to the importance of biological carbon fixation in the equatorial Indo-Pacific in controlling variations of atmospheric CO2. Therefore, because of its early response and its possible effect on the carbon cycle, the 23-ky ENSO-like cycle is likely to have played a significant role in global climate dynamics.


View Abstract

Navigate This Article