Climate and Vegetation History of the Midcontinent from 75 to 25 ka: A Speleothem Record from Crevice Cave, Missouri, USA

See allHide authors and affiliations

Science  04 Dec 1998:
Vol. 282, Issue 5395, pp. 1871-1874
DOI: 10.1126/science.282.5395.1871


Four Missouri stalagmites yield consistent overlapping records of oxygen and carbon isotopic changes and provide a climate and vegetation history with submillennial resolution from 75 to 25 thousand years ago (ka). The thorium-230–dated records reveal that between 75 and 55 ka, the midcontinental climate oscillated on millennial time scales between cold and warm, and vegetation alternated among forest, savanna, and prairie. Temperatures were highest and prairie vegetation peaked between 59 and 55 ka. Climate cooled and forest replaced grassland at 55 ka, when global ice sheets began to build during the early part of Marine Oxygen Isotope Stage 3.

Records of midcontinental climatic conditions throughout early parts of the last glacial period are scarce, and of these, dating has been problematic but is needed to allow meaningful comparison with the marine record of global climatic changes. Much of this period is beyond reliable radiocarbon control, which extends back to only 40 to 45 ka. We applied high-precision234U-230Th dating to the δ13C and δ18O profiles of four calcite stalagmites from Crevice Cave, southeastern Missouri, USA (37°45′N, 89°50′W) (Fig. 1). The234U-230Th chronometer extends the datable range to 500 ka (1) and has been successfully applied to corals (2) and inorganic calcite precipitates such as groundwater veins (3) and speleothems (4). Crevice Cave lies near the present-day ecotone between deciduous forest and tall-grass prairie and is sensitive to climatic change. The site is ideal for recording glacial-age climate, because Laurentide ice lobes advanced to within several hundred kilometers of the cave during the last glacial period.

Figure 1

Presettlement vegetation of the Midwest. Red line marks the maximum extent of Wisconsinan glaciation. Shown are the locations of Crevice Cave, pre-Wisconsinan kettle basins in Illinois, and the late-Wisconsinan (15 to 14 ka) Des Moines Lobe in Iowa.

Crevice Cave is Missouri's longest known cave and contains more than 45 km of mapped passages. The stalagmites we studied were found naturally broken within a stream-level passage approximately 25 m below the ground surface. The four stalagmites, labeled CC∖DBL-L, CC∖DBL-S, CC∖C, and CC∖E, were separated from one another by 100 to 200 m and were located approximately 1 km from the nearest known entrance (5). When split and polished, stalagmites appeared pristine and gave no indication of postformational alteration.

We milled powders for δ13C and δ18O analysis (6) with a 0.5-mm dental burr at a sampling interval of 1 to 3 mm along the central growth axis of each stalagmite. Sample sizes for U-Th analysis were 150 to 300 mg, so that the number of years averaged in a sample approached that of the U-Th analytical error (100 to 300 years) (7). We obtained 38 U-Th analyses for the four stalagmites, using thermal ionization mass spectrometry (Table 1) (8). For each stalagmite, all ages are in correct stratigraphic order, suggesting that these dense calcite stalagmites have remained closed to U and Th migration and that our age determinations are accurate. Ages for δ13C and δ18O values are assigned by linear interpolation between dated intervals (9).

Table 1

Thorium-230 ages of stalagmite subsamples. Distances are from the base of each stalagmite to the nearest half millimeter. Errors in age are 2σ and include analytical errors and uncertainties in initial 230Th/232Th, propagated through the age equation. For the two subsamples with errors larger than a thousand years, the error is dominated by uncertainties in initial230Th/232Th. All other subsamples have corrections for initial 230Th that are smaller than the analytical error. Half-lives are those used in (2). A full data table with isotope ratios and concentrations is available at theScience Web site (8).

View this table:

Values of δ13C for the Crevice Cave stalagmites vary from –9.7 to –1.8 per mil (Fig. 2). Despite differences in growth rate, color, banding features, and trace-element concentrations among the four stalagmites, trends in δ13C values are similar and imply that different hydrologic characteristics between specific drip pathways (10) cannot account for the long-term changes in speleothem δ13C values. Instead, these trends likely reflect a more pervasive influence, such as the changing δ13C composition of the overlying soil and vegetation.

Figure 2

Carbon isotope profiles for the Crevice Cave stalagmites and interpreted vegetation changes. The times of major vegetation transitions are highlighted by the dashed lines.

Speleothem δ13C values are linked to vegetation because cave drip waters first pass through the overlying soil, and soil organic matter is derived from vegetation. Large variations in soil organic matter δ13C values result from differences between the C3 and C4 photosynthetic pathways; C3 plants have δ13C values from –32 to –22 per mil, averaging ∼–27 per mil, whereas C4 plants have values from –16 to –9 per mil, averaging ∼–12 per mil (11). C4 plants are typically warm-season grasses found in tropical and temperate grasslands, whereas C3 plants are mostly trees, cool-season grasses, and forbs. Although δ13C values are modified as infiltrating water dissolves carbonate bedrock enroute to underlying caves (12), the relative contribution of C3 and C4 plants is maintained. Preserved δ13C signatures may thus serve as indicators of the prevalence of forests versus grasslands through time.

This relationship between vegetation type and speleothem δ13C values has been confirmed and calibrated in a number of studies on well-dated Holocene speleothem, pollen, plant-macrofossil, and sedimentary organic matter sequences from northeast Iowa (13). At Cold Water Cave, stalagmite δ13C values varied from –9.1 to –4.2 per mil and tracked (for several thousand years) the δ13C values of nearby sedimentary organic matter as well as transitions from deciduous forest to prairie to oak savanna, as revealed in nearby alluvial deposits at Roberts Creek. These high-resolution studies also revealed that rapid, decade-scale transitions in vegetation initiate gradual, millennial-scale turnover of the soil organic matter, at least for the deep soils of Iowa.

Sediment records from small basin lakes in south-central Illinois (14) are less than 200 km from Crevice Cave and provide the best opportunity for a midcontinental, pollen-based vegetation history during the early Wisconsinan glacial period. These basins formed as kettles during late-Illinoian time (∼140 ka) and escaped obliteration because of less extensive Wisconsinan ice. The pollen records (14, 15) show marked variations from conifer to temperate forest, with prominent intervals of prairie or prairie/forest border vegetation (16). Thus, these records document conspicuous vegetation and climate changes in the midcontinent during the Wisconsinan period, but the timing is poorly known because of the inability to date these types of sediments precisely and accurately.

Our data show several oscillations in δ13C values that we interpret as transitions between forest and grassland environments (Fig. 2). The issue of millennial-scale soil organic matter storage is important in these interpretations, because the storage phenomenon dampens the magnitude of initial speleothem δ13C response to the actual vegetation change. Thus, the δ13C trends are gradual, and the timing of the actual vegetation change is represented not necessarily by some instantaneous value of δ13C but instead by a robust inflection in the δ13C profile.

The most notable feature of the Crevice Cave vegetation history is that grassland-type environments dominated between 71 and 55 ka, during Marine Oxygen Isotope Stage 4 and the early part of Stage 3, and that forest persisted from 55 to 25 ka (17). Between 59 and 55 ka, high δ13C values suggest that the cave region was a grassland of high C4 plant biomass.

Whereas speleothem δ13C is sensitive to vegetation conditions, δ18O is linked to meteoric water. For the modern midlatitudes, the δ18O value of mean annual precipitation (MAP) in continental interiors is observed to be largely a function of mean annual atmospheric temperature (MAT) (18). The modern empirical relation between MAP δ18O values and MAT is about 0.6 per mil/°C. Other factors, however, may obscure this relation; precipitation derived from different air masses often has distinct δ18O signatures, and seasonal variations in the δ18O value of precipitation can be pronounced even when derived from a single air mass. Thus, any strict paleotemperature reconstruction must assume that neither the moisture source nor the seasonality of precipitation has varied significantly through time. Caves are well suited for such reconstructions, because ambient temperatures in poorly ventilated areas of deep caves (>11 m) are stable year-round and reflect the mean surface atmospheric temperature over several years (19).

Crevice Cave δ18O values (Fig. 3) range from ∼–3.6 to –5.0 per mil. Although many aspects of atmospheric circulation during the last glacial period remain unknown, we consider the simplest scenario in which this variation primarily reflects a change in MAT (20). In this scenario, the MAP δ18O–MAT relation of 0.6 per mil/°C is counteracted by the –0.26 per mil/°C fractionation between calcite and water (21) during crystallization, resulting in a net speleothem δ18O–MAT relationship of ∼0.35 per mil/°C. Thus, the Crevice Cave δ18O range of ∼1.4 per mil represents a MAT range of approximately 4°C. The warmest temperatures were around 57 ka, and the coldest were at ∼46 and 41 ka, culminating a cooling trend initiated at 55 ka.

Figure 3

Oxygen isotope profiles for the Crevice Cave stalagmites versus the normalized deep sea curve and chronology of Martinson et al. (29). Timing of key climatic events is highlighted by dashed lines.

Both the δ18O and δ13C records show a prominent event at 55 ka, indicating major climatic and environmental change. Age estimates for the basal Roxana Silt, a widespread loess blown from outwash plains of the first advance of Laurentide ice into the Upper Mississippi valley during the Wisconsinan glacial period (22–24), range from 45 to 55 ka (25,26). Pedogenic features of the Roxana Silt imply that climatic conditions became cooler and less conducive to soil formation as deposition of the loess progressed (24). Spruce charcoal that occurs throughout much of the Roxana Silt in Wisconsin (24), along with pollen evidence from northern Illinois (27), suggest that the Upper Mississippi valley was dominated by a conifer forest during the period of Roxana Silt deposition. Farther south at Crevice Cave, the δ13C values are also consistent with forest during this period.

The extreme conditions suggested by the Crevice Cave δ18O record around 46 and 41 ka may be represented by a recently described till that lies below the well-studied Des Moines Lobe till, of 15 to 14 ka, in Iowa (28). Wood from this till yielded a single14C age of 41.8 ± 1.6 ka, hinting that a Laurentide ice lobe already occupied central Iowa before ∼40 ka. Warmer climatic conditions eventually followed, as suggested by widespread evidence for episodes of rejuvenated soil-forming conditions throughout Illinois between ∼37 and 25 ka (23). The Crevice Cave δ18O record shows a jump at 37 ka to less-negative values and conditions that were approximately 2°C warmer.

Before 55 ka, climatic conditions appear to have oscillated more frequently. Several short-lived drops in δ18O values culminate at ∼74, ∼71, and ∼64 ka. These excursions represent cooling of about 2° to 3°C during Marine Oxygen Isotope Stage 4, a period of global ice buildup and presumably colder global temperatures (29). Instead of unidirectional cooling, the Crevice Cave δ18O record suggests that the midcontinental climate was characterized by cool/warm oscillations every 3000 to 7000 years. This frequency resembles that of the glacial-period climatic oscillations observed in North Atlantic marine sediments and ice cores, most notably the Heinrich events and Dansgaard-Oeschger cycles (30).

The chronology of the Crevice Cave record raises several questions about midcontinental and global climatic changes. From ∼59 to 55 ka, high δ13C values suggest that the site was a grassland of high C4 plant biomass, and the δ18O values suggest that temperatures were warmest for the period from 75 to 25 ka. Although well-dated records for comparison are sparse, a British flowstone known to have grown during times of relative warmth during the last glacial cycle grew at 57.9 ± 1.5 ka (31), suggesting that the warm interval from 59 to 55 ka that we have identified in Missouri may have affected a broad area of the globe. Midcontinental warming from 59 to 55 ka also seems compatible with the marine record, because the transition from Marine Oxygen Isotope Stage 4 to Stage 3 at ∼59 ka is marked by decreasing global ice volume (Fig. 3). Milankovitch forcing appears consistent with the warming, because Northern Hemisphere summer insolation peaked around 58 ka (32).

The rapid climatic cooling at 55 ka in central North America is more difficult to explain. Laurentide ice advance into the Upper Midwest at 55 ka, as suggested by some age estimates for the basal Roxana Silt (25), actually seems an unlikely event considering that the marine δ18O record suggests that global ice volume decreased from 61 to 55 ka. However, this scenario is complicated by uncertainties about the relationship between Laurentide and global ice volume, and Laurentide ice volume versus the frontal position of Laurentide ice. Taking the marine δ18O chronology at face value, we note that the climatic change at 55 ka at Crevice Cave coincides with the onset of global ice buildup during Stage 3. Thus, global and midcontinental climate may have both cooled synchronously at 55 ka. Given the relatively warm conditions from 59 to 55 ka suggested by both the Crevice Cave and the marine records, we speculate that the spread of glaciation into the Upper Mississippi valley may have lagged behind the climatic cooling at 55 ka by several thousand years. We therefore suspect that the basal Roxana Silt age estimates of 55 ka are somewhat too old and that the estimates of 50 ka may be closer to the true age.

  • * To whom correspondence should be addressed. E-mail: dora0011{at}


Stay Connected to Science

Navigate This Article