Technical Comments

The California Current, Devils Hole, and Pleistocene Climate

Science  05 Apr 2002:
Vol. 296, Issue 5565, pp. 7
DOI: 10.1126/science.296.5565.7a

Herbert et al. (1) are to be commended for their convincing and important correlation of mid- to late-Pleistocene California Current sea surface temperatures (SSTs), California coastal vegetation, and the paleotemperature record in vein calcite at Devils Hole, Nevada. Their intriguing notion linking California Current SSTs and the Laurentide ice sheet, however, is open to question, as is their assertion that the work resolves some pending inconsistencies with the orbital theory of Pleistocene climate.

Herbert et al. (1) noted a trend of SST warming in the California Current 10 to 15 thousand years (ky) before each of the past five deglaciations. They considered that early warming to be a regional signal caused by “collapse” of the California Current triggered by growth of the Laurentide ice sheet. Noting a strong linear correlation (r = 0.7 to 0.78) of the Devils Hole oxygen isotope (∂18O) time series with their SSTs, Herbert et al. (1) concluded that the early, pre-deglaciation warming exhibited by this continental record must also reflect a regional, not a global, signal. There are several reasons to question their regional-signal thesis. SST warming 5 to 15 ky before the last deglaciation occurred at a minimum of 13 locations in both hemispheres of the Pacific and Atlantic Oceans [figure 2 in (2)]. Levine et al. (3) documented similar early SST warming, before the penultimate deglaciation, at 15 locations in the Indian, Pacific, and Atlantic Oceans. Such widespread early SST warming, at latitudes ranging from 40°N to 44°S, might be attributed to regional or local causes that fortuitously occurred before the deglaciations; however, a global explanation for such behavior appears much more plausible and should be sought.

In addition, the atmospheric circulation model for the last glacial maximum (4) cited by Herbert et al. (1) to explain the collapse of the California Current is not supported by sedimentological studies; Muhs and Zarate have noted that “in North America, including the midcontinent, the northwestern United States, and Alaska, paleowinds derived from eolian sediments do not agree with climate model simulations” (5). And although they highlighted a strong linear correlation between their SSTs and the Devils Hole time series, Herbert et al. (1) failed to note the strong correlation of Devils Hole with another well-studied paleotemperature record, the Vostok, Antarctica, ice core, 114 degrees of latitude away. These linear correlations range from r = 0.61 to r = 0.92, depending on which of the several Vostok chronologies is utilized (6, 7). If Devils Hole is a regional signal responding, like California Current SSTs, to growth of the Laurentide ice, why is it strongly correlated with Vostok? All of these arguments, taken together, call into question the Laurentide ice sheet–based explanation offered by Herbert et al. for the collapse of the California Current and the ensuing early SST warming off the coasts of southwest Oregon and California.

Believing that the early warming at Devils Hole before deglaciation had a regional rather than a global cause, Herbert et al. (1) went on to assert that the Devils Hole record thus “does not pose a fundamental challenge to the orbital (‘Milankovitch’) theory of the Ice Ages.” In making that assertion, however, they failed to inform their readers that the Devils Hole study they cited (8) actually raised not one but five challenges to the orbital theory, three of which have credence solely by virtue of the robust and replicated (9, 10) thermal ionization mass spectrometry (TIMS) uranium-series dating of the Devils Hole ∂18O time series. The five challenges posed in that study (8) included (i) the early warming before deglaciations, already discussed; (ii) various studies of uranium series–dated corals that indicated that sea level was at or near modern levels as early as 132 to 135 thousand years ago (ka), a finding not easily reconcilable with insolation forcing of the penultimate deglaciation; (iii) the so-called stage 11 problem—that is, the occurrence of a prominent deglaciation and subsequent glaciation in the absence of significant variations in high-latitude summer insolation during the period from ∼450 to ∼350 ka; (iv) interglacial periods recorded at Devils Hole that are twice as long as prescribed by the orbitally tuned SPECMAP global ice volume record; and (v) the observation that glacial-interglacial cycles become aperiodic and of increasing duration as the record approaches the present day (8).


Response: Our group (1) and Winograd's Devils Hole team agree on many important aspects of late Pleistocene climate change along the western margin of North America. Our marine records reinforce the contention from Devils Hole that temperatures in the region rose toward interglacial levels during times that would be conventionally dated as glacial maxima and glacial terminations. Work recently presented by Winograd and his colleagues explicitly likens the timing of the Devils Hole record to our alkenone-based SST reconstruction from Santa Barbara Basin (2). Our interpretation of regional paleotemperature and paleoenvironmental data, however, differs from that of Winograd in two substantial ways.

1) We believe that the conjunction of alkenone SST records with benthic ∂18O data acquired on the same samples unequivocally demonstrates that a strong temporal offset exists between glacial maxima (defined by benthic ∂18O) and regional SST. That finding does not depend on the precise chronology assigned to samples. Papers published by the Devils Hole group asserted that the Devils Hole temperature record could be used to date the marine ∂18O sequence. That claim requires that regional surface temperatures be synchronous with ice volumes, and is falsified at a number of marine sites off the California margin discussed in our study (1). Regional temperatures departed from the benthic ∂18O curve in precisely the time intervals seized on by the Devils Hole group to question the orbitally tuned chronology of the marine isotope record. In his comment, Winograd continues to confuse surface temperatures with ice volume and therefore ignores a major point in our paper.

2) Winograd continues to assume that a single, globally applicable paleotemperature curve describes the evolution of late Pleistocene climate. Thus, he remarks on the similarity of the Devils Hole record to isotopic estimates of past temperatures recorded in the Vostok ice core from Antarctica. Recent efforts to synchronize the Vostok curve to the marine record using the ∂18O of air trapped in Vostok ice, however, do not support the chronology most favorable to the Devils Hole record (3, 4). Finding many regional paleotemperature records similar to Devils Hole does not prove Winograd's case; instead, the rule is disproved by the exceptions, which are numerous. For example, it is now well known that the Vostok temperature record is significantly offset in time from Greenland ice core records (5–7). Our own data showed how the marine SST response differed along along the California margin. Our Baja location showed next to no offset of SST from the benthic ∂18O record, whereas sites in the southern California region had the largest anomalies. Other examples of regional heterogeneity of SST relative to ∂18O come from the work of Kirst et al. (8) along the Angola margin. There, alkenone-based SSTs moved perfectly in phase with ∂18O at some sites, while other locations showed early warmings similar to those detected in our work. The classic spectral study of SSTs in the North Atlantic by Ruddiman and McIntyre (9) found systematic differences in the relative power of the Milankovitch frequencies as a function of latitude. In an analogous manner, our grid of cores allowed us to argue that the “anomalous” SSTs during glacial maxima have a strong regional fingerprint—that is, we are dealing with signal, not noise.

Advances in understanding the full suite of climatic events that accompanied glacial-interglacial cycles will require loosening the hold of the “template” concept. Some records, such as benthic ∂18O, should from first principles provide records that are very nearly globally sychronous. But we should move away from equating global change with synchronous change as we study many other climatic fields. That most late Pleistocene records resemble each other to a large degree should not blind us to meaningful regional deviations that may point toward the still unknown dynamics that move climate so dramatically between glacial and interglacial states. Our study's conclusion that the glacial collapse of the California Current system came from perturbations to the wind fields over the North Pacific caused by the Laurentide ice sheet seems valid with our present data set, but could be revised when a better geographic distribution of past SSTs becomes available. We stand by our basic conclusion that regional climate processes are important and are expressed in the preserved paleoclimate record, and that the Devils Hole record is best viewed as just such a regional signal.


Navigate This Article