Abrupt Tropical Vegetation Response to Rapid Climate Changes

See allHide authors and affiliations

Science  25 Jun 2004:
Vol. 304, Issue 5679, pp. 1955-1959
DOI: 10.1126/science.1092995


Identifying leads and lags between high- and low-latitude abrupt climate shifts is needed to understand where and how such events were triggered. Vascular plant biomarkers preserved in Cariaco basin sediments reveal rapid vegetation changes in northern South America during the last deglaciation, 15,000 to 10,000 years ago. Comparing the biomarker records to climate proxies from the same sediment core provides a precise measure of the relative timing of changes in different regions. Abrupt deglacial climate shifts in tropical and high-latitude North Atlantic regions were synchronous, whereas changes in tropical vegetation consistently lagged climate shifts by several decades.

In order to evaluate the relative roles of high and low latitudes in initiating and propagating abrupt global climate changes, we need precise information regarding the relative timing of abrupt changes in different regions. Dating uncertainties, however, are typically too large to constrain the timing of the briefest decadal events in records from different sites (14). Another approach is to identify high- and low-latitude climate proxies in the same high-resolution record, and determine the relative timing of changes stratigraphically (36). For instance, increased methane concentrations attributed to the expansion of tropical wetlands (7) have been measured in air trapped in Greenland ice and used to infer shifts to warmer and/or wetter tropical climate during the abrupt Glacial/Bølling and Younger Dryas/Preboreal transitions (3, 4). Temperature changes over Greenland were also reconstructed from the same samples using nitrogen and argon isotopes, allowing the precise identification of relative timing for rapid changes between the tropics and high latitudes. The tropics were found to lag Greenland by 20 to 30 and 0 to 30 years for the Bølling and Preboreal warmings, respectively (3, 4), favoring a North Atlantic trigger at least for the Bølling event (4). In similar studies, radiocarbon measured in planktonic foraminifera from tropical Cariaco basin sediments was shown to have increased steeply during the onset of Younger Dryas cooling (5, 6). Atmospheric concentration of cosmogenic beryllium-10 does not show a similar increase during the Younger Dryas onset (8), suggesting that the 14C increase was not caused by changes in production rate, but that it instead reflects an abrupt decrease in high-latitude North Atlantic Deep Water (NADW) formation and export (5, 6, 9). In addition, relative reflectance (gray scale) and laminae thickness from the same sediments reveal rapid shifts in Cariaco upwelling and trade-wind intensity caused by shifts in the mean position of the Intertropical Convergence Zone (2, 5, 6). Direct comparison of Cariaco radiocarbon and gray-scale data showed that high- and low-latitude climate shifts during the onset of Younger Dryas cooling were synchronous within 10 years (6), and thus allows either a North Altantic or a tropical trigger for this rapid climate cooling. If abrupt deglacial warming and cooling events were manifestations of the same millennial-scale shifts in global climate, this subtle discrepancy in timing between Cariaco and Greenland data must be resolved before we can understand the mechanisms responsible for abrupt climate change.

The delayed increase in atmospheric methane following abrupt warmings may have been caused by the release of gas hydrates rather than expansion of tropical wetland vegetation (10). However, studies have shown fluctuations in deglacial tropical moisture balance similar in timing to the Bølling/Allerød and Younger Dryas oscillations in the North Atlantic region (1117), supporting tropical wetlands as the source for the atmospheric methane signal. Specifically, detailed pollen records from Central and northern South America (11, 12), including the Cariaco basin watershed (13, 14), show that vegetation shifted between predominantly dry grasslands during the Glacial and Younger Dryas periods, and wet montane forest during the Bølling/Allerød and Preboreal periods. The lag of the methane increase behind Greenland warming may have occurred because the time scale necessary for tropical wetland expansion and development of anoxia following a climate shift may be longer than previously thought (4, 18). However, the response time for changes in vegetation following abrupt shifts in climate is currently unknown, and the precise temporal relation between tropical vegetation shifts and climate changes in tropical and high-latitude North Atlantic regions remains a major question. We address this question by using multiple-proxy records of terrestrial vegetation change from vascular plant biomarkers preserved in Cariaco basin sediments, providing direct comparisons to climate proxies from the same core in order to evaluate precisely the timing of tropical vegetation change relative to local and regional climatic forcing (18).

Certain classes of organic compounds that are sequestered in sediments can be traced directly to specific sources, such as aquatic versus terrestrial vegetation, or woody versus herbaceous land plants (19). In particular, long-chain (C25-C33) n-alkanes exhibiting an odd-carbon-number predominance, and C24-C32 n-alkanoic acids and n-alcohols exhibiting an evencarbon-number predominance [e.g., high Carbon Preference Index (CPI)] (20), are synthesized nearly exclusively by vascular plants as components of epicuticular leaf waxes, whereas shorter chain homologs may have a large marine component (19, 21, 22). The preservation of these lipids in sediments is generally good (19), and their analysis in sediment cores can be used to study vegetation change or patterns of atmospheric circulation (2328). Although these compounds have been shown to be transported long distances by eolian processes (24, 28), studies of lake and marine sediments show the predominant source of plant waxes to be local vegetation in the immediate watershed (23, 2527). Thus, a near-shore marine basin such as Cariaco is expected to have the greatest concentration of plant waxes derived from local sources and not from long-distance transport.

Stable carbon isotopes in vascular plant biomarkers can be used to distinguish between vegetation using different metabolic pathways, such as C3 versus C4 carbon fixation (29). The C4 pathway is used by plants in arid environments to limit water loss and results in less isotopic fractionation (heavier δ13C values) than the C3 pathway (29). Although leaf waxes are typically depleted in 13C relative to bulk tissue, this isotopic distinction is preserved. Those leaf waxes from C4 plants typically have δ13C values of about –23‰, whereas C3 leaf wax δ13C values are about –34‰ (30, 31). Such a large isotopic signal can readily be discerned in biomarkers from dry grassland versus more humid forest vegetation sources, even in marine sediments (26, 27). In addition to the information available from carbon isotopes, distributional variations in plant biomarker homologs also record differences between “arid” and “humid” vegetation sources. For example, atmospheric dust samples collected in transects along the West African coast reveal changes in the chain-length distributions of n-alkanes responding to aridity in the source regions (26, 27). Similarly, n-alkanes from southern Asian paleosols and Bengal Fan sediments show chain-length variability that correlates strongly with n-alkane δ13C, reflecting the Miocene expansion of C4 grasses (31). Multiple independent proxies for terrestrial vegetation can thus be used to provide a check against factors other than source vegetation that may bias either record.

Cariaco sediment samples were obtained from high–deposition rate piston core PL07-58PC (2, 6), taken on the saddle between eastern and western subbasins where interferences from turbidites and sediment slumping are minimal (32). The sediments are annually laminated, and a floating varve chronology between 15,000 and 10,000 calendar years before present (15 to 10 ka B.P.) was anchored to an absolute age scale by cross-correlating variations in foraminiferal 14C with 14C in the newly revised Preboreal Pine tree-ring chronology (33, 34). Organic carbon concentrations throughout the last deglaciation are ∼2 to 5%, indicating good preservation of biomarker compounds. We investigated n-alkanoic acids because of their high concentrations and because n-alkanes showed evidence of contamination (e.g., low CPI) from migration of fossil hydrocarbons in the region. Concentrations of individual long-chain (≥C24) n-alkanoic acids ranged from 0.5 to 5 μg per gram of dry weight. Molecular isotopic measurements were made on C14-C32 homologs (32).

The δ13C values for C30 and C32 n-alkanoic acids were enriched by 4 to 5‰ during the Glacial period relative to the Preboreal and Holocene, reflecting a greater proportion of waxes from C4 grasslands during cold, dry climates (Fig. 1A). This trend is absent from shorter-chain C16 n-alkanoic acids, likely to derive predominantly from marine sources (35) (fig. S1). The δ13C data show rapid shifts coincident with the Glacial-Bølling and Younger Dryas–Preboreal transitions (Fig. 1A), indicating that C3 vegetation biomass favoring humid conditions expanded abruptly at those times. Also, δ13C variability coincides with some centennial-scale variability, including early Bølling and Preboreal cool/dry oscillations at ∼14.4 ka B.P. and ∼11.3 ka B.P., respectively (Fig. 1A).

Fig. 1.

Analysis of n-alkanoic fatty acid (FA) leaf waxes from Cariaco basin sediments. (A) δ13C values of long-chain (C32) n-alkanoic acids (red circles) are enriched during Glacial (GL) and Younger Dryas (YD) climate periods, consistent with the increased presence of arid C4 grasslands. Depleted values consistent with more humid C3 forest vegetation occur during wetter Bølling/Allerød (B/A) and Preboreal (PB)/Holocene periods. Similar results were observed for C30 chain-length FAs (fig. S1). (B) An average chain length (ACL) index for C24-C28 n-alkanoic fatty acids (red circles) (36) shows the same deglacial oscillations as δ13C, but includes a rapid transition to drier C4 vegetation at the beginning of the Younger Dryas. Both biomarker indices are plotted with sediment reflectance (gray scale), a proxy for upwelling and trade-wind intensity (2) (blue lines). Abrupt transitions between climatic periods are indicated by gray bars. Error bars are 1σ.

An index of average chain length (ACL) (36) for C24-C28 n-alkanoic acid homologs also shows abrupt changes during millennial-scale deglacial oscillations (Fig. 1B). Thus, tropical vegetation biomass in the Cariaco region shifted rapidly between arid grassland and wet forest during deglaciation, in phase with gray scale and the characteristic trend of climate oscillations in the high-latitude North Atlantic. ACL records an abrupt shift to drier C4 vegetation during the onset of the Younger Dryas, whereas δ13C changes more gradually (37).

We have chosen to construct high-resolution vegetation records by using a molecular (isotopic) approach because biomarkers provide information independent from, and complementary to, pollen records. For example, the molecular and isotopic composition of leaf waxes are likely indicators of vegetative biomass, rather than reproduction (pollen). In some circumstances, plants can be induced to produce pollen under conditions of stress rather than optimum growth, leading to potential ambiguity in interpreting abrupt pollen shifts (38). Similarly, some vegetation living near the limit of its range may be able to survive for years without reproducing, providing a continuous biomarker signal but no pollen. To the extent that plant waxes record biomass of exposed leaf surface area, there is the potential for resolving extremely rapid responses to climate change.

Biomarkers from terrestrial vascular plants must be transported from the continental to the marine environment before deposition in sediments. For plant waxes, there is abundant evidence that this residence time is short, and that shifts recorded in biomarkers from annually laminated Cariaco basin sediments will not substantially lag the corresponding shifts in terrestrial vegetation. Plant wax compounds are transported primarily as aerosols via eolian pathways and are not likely to experience extended periods suspended in the atmosphere (24, 2628). This has been demonstrated with aerosols collected over 3 years in an air-monitoring system in Bermuda that showed a strong seasonal signal in the molecular and isotopic composition of North American land-plant–derived n-alkanoic acids, n-alcohols, and n-alkanes (28). The observed seasonality indicates that leaf waxes are introduced into the atmosphere mainly by wind ablation off living vegetation rather than by remobilization of detrital waxes during soil deflation (28) and implies a transport time on the order of days to weeks, much shorter than the time span encompassed by individual Cariaco sediment samples (∼10 to 15 years).

Previous work has shown that trade-wind strength in the Cariaco region intensified during the Younger Dryas and Glacial periods (2, 39) and that transport of dust from arid African source regions to Cariaco sediments increased during Glacial intervals (40). We therefore investigated the possibility that the Cariaco C4 vegetation signal during Glacial and Younger Dryas climates was not caused solely by changing tropical vegetation, but also resulted from increased long-distance transport and mixing of leaf waxes between arid northwest Africa and humid northern South America. A plot of δ13C versus ACL for Cariaco basin leaf waxes reveals a significant linear correlation (r = 0.64, P < 0.001), with “interstadial” and “stadial” end members clearly delineated (Fig. 2). If the Cariaco “stadial” vegetation signals were an artifact of mixing “interstadial” leaf waxes with those transported from arid African regions, the potential African contaminant would lie to the right of the Cariaco Glacial end member along the mixing line. However, Holocene leaf waxes from dust and marine sediments off the coast of northwest Africa (26) lie away from the mixing line normal to the Cariaco distribution (Fig. 2). The Cariaco stadial vegetation signals therefore cannot be explained as an artifact of mixing interstadial leaf waxes between tropical sources with different aridities and C4 plant concentrations, but must rather reflect true changes in northern South American vegetation during cold, dry Glacial and Younger Dryas climates (41).

Fig. 2.

Comparison of δ13C for C32 and ACL for C24-C28 n-alkanoic fatty acids (FAs) from Cariaco basin core PL07-58PC. δ13C and ACL show a significant positive correlation (r = 0.64, P < 0.001), indicated by the dashed line. Asimilar correlation and regression line are seen between ACL and C30 fatty acid δ13C (r = 0.51, P < 0.001) (fig. S2). Data points from warm/wet B/A and PB/Holocene “interstadial,” and cool/dry GL and YD “stadial” periods, identified on the basis of gray scale for the same core, are plotted as red and blue circles, respectively. The same indices measured on n-alkanoic acid leaf waxes from Holocene African C4 vegetation are shown for C32 FAs (green triangles) and C30 FAs (inverted green triangles) (26). The majority of African data points lie normal to the South American data distribution and regression line and are inconsistent with potential contamination of Cariaco “stadial” C4 signals with long-distance transport and mixing of “interstadial” leaf waxes during cold, windy climates (41).

To evaluate the relative timing between Cariaco basin records of climate and vegetation change, we used a stepped cross-correlation approach (6, 32, 42). We focused initially on the onset of the Younger Dryas event, because it can be identified clearly in Cariaco proxies documenting shifts in North Atlantic (Δ14C) and tropical (gray scale) climate, as well as tropical vegetation (ACL) (43) (Fig. 3). Applying this technique to Δ14C and gray-scale records confirms previous results that high- and low-latitude climate shifts were synchronous (lag of 0 ± 15 years) at that time. Analysis of ACL versus Δ14C shows that tropical vegetation change lagged North Atlantic climate by 50 ± 25 years (44) (Fig. 3). Unfortunately, Δ14C data across the Younger Dryas/Preboreal and Glacial/Bølling transitions do not show a distinct North Atlantic Deep Water (NADW) signal (6), and tropical versus high-latitude comparisons are not possible for these events. However, cross-correlating ACL with gray scale at both of these rapid shifts to warmer and/or wetter conditions reveals that Cariaco vegetation lagged local tropical climate change by at least 25 ± 15 years (44) (Fig. 4), and confirms a vegetation lag of 50 ± 25 years at the Younger Dryas onset (Fig. 3). Comparison of leaf wax δ13C to ACL at the abrupt Bølling and Preboreal transitions shows synchronous changes, a lag of 0 ± 15 years; comparison of leaf wax δ13C to gray scale shows the same vegetation-climate lag as ACL, 25 ± 15 years (44) (Fig. 4). The time scale of vegetation response to rapid climate change thus appears to depend on the rate of change in climate (45). This may indicate the ability of some plants to adapt for a longer period, before being replaced by other species, when faced with gradual rather than rapid climate changes.

Fig. 3.

Details of abrupt change at the beginning of the Younger Dryas interval in multiple proxies from Cariaco basin sediment core PL07-58PC. (A) Planktonic foraminiferal Δ14C, indicating a rapid decline in high-latitude NADW formation (6, 46). (B) Percent reflectance (gray scale), recording a sharp increase in Cariaco upwelling and tropical trade-wind intensity (2). (C) Average chain length for C24-C28 n-alkanoic acid leaf waxes, showing a shift in local Venezuelan vegetation from wetter C3 forest to arid C4 grassland. The timing of the rapid shift in each record is indicated by a gray line. Lagged cross-correlation analysis shows that high- and low-latitude climate shifts were synchronous within ± 15 years, but vegetation change in northern South America lagged climate by 50 ± 25 years (44). Error bars are 1σ.

Fig. 4.

Details of abrupt changes at the GL/BO (right) and YD/PB (left) transitions recorded in multiple proxies in Cariaco basin core PL07-58PC. (A and D) Percent reflectance (gray scale), recording abrupt declines in upwelling and trade-wind strength. (B and E) δ13C for C32 n-alkanoic acid leaf waxes. (C and F) ACL index for C24-C28 n-alkanoic acids. Both biomarker indices record abrupt shifts from arid C4 grassland to wetter C3 forest biomass at the GL and YD terminations. Lagged cross-correlation analysis shows that vegetation lagged local tropical climate changes by 25 ± 15 years for both events (44). Vertical gray bars indicate timing of rapid climate shifts based on gray scale. Error bars are 1σ.

Although the Cariaco basin varve chronology is anchored to newly revised treering ages (33) by matching 14C variations between ∼10 and ∼12.5 ka B.P. (34), independent (i.e., not 14C) climate proxies can be compared within the interval of overlap. The end of the Younger Dryas event in tree rings is placed at 11,590 ± 0 years B.P. and in Cariaco at 11,580 ± 16 years B.P. (33, 34). Taken together, these data and our new results suggest that climate shifts in both high-latitude and tropical North Atlantic regions were synchronous, and that tropical vegetation change lagged abrupt climatic shifts by several decades. This has profound implications for the interpretation of very rapid climate changes from vegetation records and may help reconcile the lag of atmospheric methane behind Greenland temperature change with Cariaco evidence for synchronous climate shifts at high and low latitudes.

Supporting Online Material

Materials and Methods

Figs. S1 and S2

Data available at the World Data Center-A for Paleoclimatology (

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article