Owens Lake is located in the Great Basin of the western United States between the central Sierra Nevada and the Inyo-White mountains. Maximum precipitation along the Sierra Nevada is associated with the annual north–south progression of the polar jet stream, with cool-season orographic precipitation from North Pacific sources supplying >99% of the runoff reaching the Owens basin13,14.

Core OL84B was obtained from the Owens basin in 1984 using a modified Livingstone piston corer (Fig. 1)15. A series of accelerator mass spectrometry (AMS) 14 C dates on bulk organic carbon were used to set the chronology of the core section used in this study ( Fig.2). The 14 C-based records of Owens Lake were converted to calendar years using 230 Th–234 U and 14 C ages of corals16,17,18,19. The 14 C data indicate sediment hiatuses at 2.25 and 9.20m. The upper surface of the 9.20-m hiatus contains features indicating desiccation and subaerial exposure, including a lag deposit (1–3mm thick) of frosted quartz grains. The 2.25-m hiatus is marked by a coarse sand. Sediments examined in this study are thus bounded by two desiccation events that occurred 18.3–15.8 and 6.7–4.5kyr ago (Fig. 3)20. The younger hiatus, previously noted in a core taken from higher elevation in the Owens basin21 seems to have occurred at about the same time as hiatuses noted in cores from the Mono Lake basin22. The older hiatus appears to coincide with major declines in the levels of Mono Lake and Lake Lahontan23,24.

Figure 1
figure 1

Map of North America showing locations of sediment core OL84B and ice core GISP2.

Figure 2: Radiocarbon-age model for core OL84B between depths of 225 and 9.20m. Filled circles indicate 14C analyses on bulk organic carbon used in construction of the age model.
figure 2

Vertical dashed lines indicate location of hiatuses in sedimentation. All age data (kyr) presented in the text of this Letter are in calendar years.

Figure 3: δ18O, total inorganic carbon (TIC) and pollen records for core OL84B.
figure 3

D1 to D4 are relatively dry intervals; W1 to W4 are relatively wet intervals. The timing of the Oldest Dryas (OD1)/Bølling (B): Older Dryas (OD2)/Allerød (A): Younger Dryas (YD)/Holocene boundaries and the Inter-Allerød Cold Period (IACP) are based on ice-layer counts from GISP234. Cold intervals in the ice-core record and dry periods in the Owens Lake record are indicated by black intervals. Note that the 14C age model for core OL84B is accurate to only a few hundred years. This implies that the δ18O record from OL84B could lead or lag the δ18O record from GISP2; that is, oscillations present in both cores cannot be demonstrated to be synchronous. In the pollen record shown, A indicates Artemisia ; C, Chenoam (Chenopodiaceae + Amaranthus) and J, Juniperus .

To determine the degree of variability in the hydrological balance of Owens Lake between 15.8 and 6.7kyr, we analysed a continuous set of 120 samples for δ18O and total inorganic carbon (TIC) ( Fig. 3). Each sample integrated 60yr of time and was repeatedly washed with distilled–deionized water to remove soluble salts. The isotopic analyses were done on the TIC fraction.

In lakes that oscillate between closed and open hydrological states, δ18O and TIC values tend to decrease with decreasing residence time of water in the lake basin20. The δ18O value of a lake is a function of the overflow:inflow ratio. When this ratio approaches unity, the δ18O value of a lake approaches the δ18O value of inflow; when this ratio approaches zero, the δ18O value of lake water becomes highly enriched owing to evaporative concentration of the heavy isotope of oxygen (18 O) in lake water. For Owens Lake, calcites precipitated during high overflow:inflow conditions would have δ18O values approaching 15‰ (relative to the VSMOW standard); in contrast, calcites precipitated during hydrological closure would have δ18O values approaching 30‰ (ref. 20 ).

During overflow, losses of Ca2+ (and CO32−) increase with increase in the overflow:inflow ratio; therefore, if the flux of detrital silicates remains roughly constant, the TIC fraction becomes a proxy for change in hydrological balance. During hydrological closure, all Ca2+ that reaches a lake eventually precipitates as CaCO3. The TIC fraction in a closed lake tends to decrease with increasing lake size because the flux of suspended sediment increases exponentially with increasing discharge to the lake, diluting the carbonate precipitate25. Glaciation of a lake basin's catchment area also can markedly increase the flux of detrital silicates to a lake, completely masking the TIC signal20.

The δ18O data obtained in this study indicate four extremely dry (closed-basin) intervals (D1 to D4) that occurred between 15.8 and 6.7kyr (Fig. 3). The older three dry intervals are centred at 15.1, 13.2 and 12.2kyr ago; the youngest dry interval (D4) marked the beginning of the Holocene (11.3kyr) and extends to Desiccation II. The presence of prismatic cracking in the reddish D3 interval suggests the existence of a soil formed during subaerial exposure of lake sediments. Relatively wet intervals (W1 to W4) precede each of the dry events. Interval W1 was not recorded in the δ18O data set because the overflow:inflow ratio was too high to permit saturation with CaCO3. Interval W2 contains two secondary peaks (W2a andW2b).

Although not amplitude-locked, the TIC data parallel oscillations in δ18O throughout the entire time period (Fig. 3). This parallelism indicates that detrital silicates (glacial rock flour) did not obscure the TIC signal, supporting other studies which indicate that the Sierras were essentially deglaciated by 15–14kyr ago26,27. The amplitude of TIC variability increases between 8.8 and 6.7kyr ago: this suggests that maxima in TIC were caused by carbonate precipitation in extremely shallow lakes. Under these conditions, carbonate precipitation masked the input of detrital silicates whose flux decreased with decreasing inflow of the Owens River. The time of transition to shallow oscillating conditions (8.8kyr) was coeval with the rapid disappearance of Early Holocene woodlands from the Chihuahuan, Sonoran and Mojave deserts in the southwestern United States 9kyr ago28. This change to a drier type of vegetation also indicates an increase in aridity of the southwestern United States after 9kyr.

The oscillatory behaviour of TIC was terminated by Desiccation II, which lasted from 6.7 to 4.5kyr ago (Fig. 3). Transition to wetter conditions after 4.5kyr at Owens Lake is consistent with pollen data from meadows in the Sierra Nevada that indicate a transition at 4.7kyr to peat and other plant types that require abundant soil moisture29.

Pollen was extracted from 17 sediment samples from OL84B to determine if rapid changes in the hydrological balance of Owens Lake also were reflected in the vegetation community that surrounded the lake. When the climate was wet and Owens Lake occupied most of its basin, juniper (Juniperus) or sagebrush (Artemisia) should have composed much of the nearby vegetation. When the climate was dry and Owens Lake was small, the Chenoam (Chenopodiaceae and Amaranthus) group—which includes drought-tolerant salt-resistant desert taxa—should have populated the saline playa abandoned by the retreating lake.

The pollen data indicate that Juniperus was abundant between 15.6 and 14.6kyr ago; thus, it was very wet during the period that followed Desiccation I. After 15.0kyr, Juniperus declined rapidly and then recovered somewhat during subsequent wet phases W3 and W4 ( Fig. 3). Peaks in the Artemisia/Chenoam (A/C) ratio, which indicate relative wet conditions, occurred during minima in the δ18O and TIC records (Fig. 3). Thus all three proxies of climate variability yield a consistent picture of oscillations in the hydrological balance of Owens Lake.

Comparison of the δ18O record of change in the hydrological balance of Owens Lake with the GISP2 record of the Oldest Dryas/Bølling/Older Dryas/Allerød/Younger Dryas and Holocene cold/warm oscillations indicates a remarkable degree of similarity in the number, duration, and timing of climate regimes in both records between 17.0 and 11.2kyr ago (Fig. 3). There are five North Atlantic cold regimes centred at 16.4, 14.8, 14.0, 13.2 and 12.3kyr ago and five dry regimes in western North America centred at 16.9, 15.1, 14.2, 13.2 and 12.2kyr ago. Although the errors associated with our age model (±500yr; Fig. 2) make it impossible to demonstrate absolute synchroneity between the two records, it is clear that both records attest to a similar number of large and abrupt climate oscillations during the last glacial termination. We argue that, in general, Atlantic cold events (for example, the Younger Dryas occurred during dry intervals in western North America (for example, D3); also, warm events in the Atlantic region (for example the Bølling) occurred during wet intervals in western North America (for example, W2a). The last wet phase (W4) occurred during the last Greenland warm peak.

With the advent of the Holocene, linkage of the climate regimes of the North Atlantic and western North America weakened and perhaps disappeared. Before the Holocene, relatively dry conditions occurred in western North America when the North Atlantic region was relatively cold. During the Early and Middle Holocene this relationship was reversed; that is, western North America was relatively dry and the North Atlantic region was relatively warm.

The duration, timing and similar number of climate oscillations in western North America and the North Atlantic region, indicated by this and other studies20, suggests a climate-change link during the last glacial termination throughout at least part of the Northern Hemisphere. Errors inherent in our age model do not allow us to completely rule out an oceanic linkage; however, recent climate simulations more strongly support the concept of atmospheric forcing11,12. In agreement with these studies, we suggest that oscillations in wetness and temperature in western North America were linked to oscillations in the strength and pattern of the North Atlantic thermohaline circulation through its effect on sea surface temperature and atmospheric water content. Rapid climate oscillations in the North Atlantic regions have been attributed to sudden changes in the rate and location of thermohaline overturn30,31,32,33. We propose that cooling of the North Atlantic, resulting from a decrease in thermohaline circulation, caused a downstream cooling of the North Pacific11, which in turn decreased the temperature and moisture content of air passing over the middle latitudes of western North America12.