A 3029-m-long deep ice core extending nearly to bedrock has been drilled at the very top of the Greenland ice sheet (Summit) by the Greenland Ice-core Project (GRIP), an international European joint effort organized by the European Science Foundation. The ice core reaches back to 250,000 yr B.P. according to dating based partly on stratigraphic methods and partly on ice-flow modeling. A continuous and detailed stable isotope (δ18O) profile along the entire core depicts dramatic temperature changes in Greenland through the last two glacial cycles, including abrupt climatic shifts during the Eem/Sangamon Interglaciation, which is elsewhere recorded as a warm and stable period. The stratigraphic continuity of the Eemian layers has therefore been scrutinized. New ice core studies, comprising cloudy band observations, deconvolution, and frequency analyses, lead to the conclusion that the climate instability suggested during the Eem Interglaciation in Greenland is likely to be real, though no conclusive evidence is available. Whereas latitudinal displacements of the North Atlantic Ocean current are considered the immediate cause of the glacial climate instability, longitudinal displacements may be the immediate cause of the Eemian instability. If so, the Eemian climate changes will be much subdued outside the Arctic region and will probably only be recognizable in sedimentary sequences of high sensitivity and temporal resolution.

The climatic evolution during the Eemian and the Holocene in western Europe is compared with the sea-surface conditions in the Norwegian Sea and with the oxygen-isotope-derived paleotemperature signal in the GRIP and Renland ice cores from Greenland. The records show a warm phase (ca. 3000 yr long) early in the Eemian (substage 5e). This suggests that the Greenland ice sheet, in general, recorded the climate in the region during this time. Rapid fluctuations during late stage 6 and late substage 5e in the GRIP ice core apparently are not recorded in the climatic proxies from western Europe and the Norwegian Sea. This may be due to low resolution in the terrestrial and marine records and/or long response time of the biotic changes. The early Holocene climatic optimum recorded in the terrestrial and marine records in the Norwegian Sea-NW European region is not found in the Summit (GRIP and GISP2) ice cores. However, this warm phase is recorded in the Renland ice core. Due to the proximity of Renland to the Norwegian Sea, this area is probably more influenced by changes in polar front positions which may partly explain this discrepancy. A reduction in the elevation at Summit during the Holocene may, however, be just as important. The high-amplitude shifts during substage 5e in the GRIP core could be due to Atlantic water oscillating closer to, and also reaching, the coast of East Greenland. During the Holocene, Atlantic water was generally located farther east in the Norwegian Sea than during the Eemian.

Two speleothems from a coastal lowland and an alpine cave site in northern Norway grew in isotopic equilibrium during Termination II and marine isotope stage 5 (150,000-80,000 yr B.P.), as dated by the U-series technique. The stable isotope record (δ18O, δ13C) displays a time resolution down to ∼20 yr. The δ18Oc signal in the two speleothems appears to be mainly dominated by the meteoric signal; i.e., there is a positive relationship between δ18Oc and temperature [∂(δ18Oc/∂T > 0]. The FM-2 couplet, a Younger Dryas type two-step structure in Termination II at 132,000 ± 5000 and 129,000 ± 5000 yr is prominent in the speleothem record. The record correlates well with details and gross features of the GRIP ice core and confirms the time scale and the unstable climate proposed for substage 5e. The isotopic response in Termination II is delayed relative to the Devil's Hole record and is in accord with the SPECMAP chronology, but cooling at the end of substage 5e seems to be synchronous between Devil's Hole and the Norwegian flowstones.

Alluvial, fluvial, and organic deposits of the last interglaciation are exposed along numerous river terraces in northeast Siberia. Although chronological control is often poor, the paleobotanical data suggest range extensions of up to 1000 km for the primary tree species. These data also indicate that boreal communities of the last interglaciation were similar to modern ones in composition, but their distributions were displaced significantly to the north-northwest. Inferences about climate of this period suggest that mean July temperatures were warmer by 4 to 8°C, and seasonal precipitation was slightly greater. Mean January temperatures may have been severely cooler than today (up to 12°C) along the Arctic coast, but similar or slightly warmer than present in other areas. The direction and magnitude of change in July temperatures agree with Atmospheric General Circulation Models, but the 126,000-year-B.P. model results also suggest trends opposite to the paleobotanical data, with simulated cooler winter temperatures and drier conditions than present during the climatic optimum.

The last interglacial high sea-level stand, the Pelukian transgression of isotope substage 5e, is recorded along the western and northern coasts of Alaska by discontinuous but clearly traceable marine terraces and coastal landforms up to about 10 m altitude. The stratigraphy indicates that sea level reached this altitude only once during the last interglacial cycle. From the type area at Nome, to St. Lawrence Island in the Bering Sea, to the eastern limit of the Beaufort Sea, Pelukian deposits contain extralimital faunas indicating that coastal waters were warmer than present. Amino acid ratios in molluscs from these deposits decrease to the north toward Barrow, consistent with the modern regional temperature gradient. Fossil assemblages at Nome and St. Lawrence Island suggest that the winter sea-ice limit was north of Bering Strait, at least 800 km north of its present position, and the Bering Sea was perennially ice-free. Microfauna in Pelukian sediments recovered from boreholes indicate that Atlantic water may have been present on the shallow Beaufort Shelf, suggesting that the Arctic Ocean was not stratified and the Arctic sea-ice cover was not perennial for some period. In coastal regions of western Alaska, spruce woodlands extended westward beyond their modern range and in northern Alaska, on the Arctic Coastal Plain, spruce groves may have entered the upper Colville River basin. The Flaxman Member of the Gubik Formation on the Alaskan Arctic Coastal Plain was deposited during marine isotope substage 5a and records the breakup of an intra-stage 5 ice sheet over northwestern Keewatin.

The last interglaciation (substage 5e) provides an opportunity to examine the effects of extreme orbital changes on regional climates. We have made two atmospheric general circulation model experiments: P+T+ approximated the northern hemisphere seasonality maximum near the beginning of 5e; P-T- approximated the minimum near the end of 5e. Simulated regional climate changes have been translated into biome changes using a physiologically based model of global vegetation types. Major climatic and vegetational changes were simulated for the northern hemisphere extratropics, due to radiational effects that were both amplified and modified by atmospheric circulation changes and sea-ice feedback. P+T+ showed mid-continental summers up to 8°C warmer than present. Mid-latitude winters were 2-4°C cooler than present but in the Arctic, summer warmth reduced sea-ice extent and thickness, producing winters 2-8°C warmer than present. The tundra and taiga biomes were displaced poleward, while warm-summer steppes expanded in the mid latitudes due to drought. P-T- showed summers up to 5°C cooler than present, especially in mid latitudes. Sea ice and snowpack were thicker and lasted longer; polar desert, tundra, and taiga biomes were displaced equatorward, while cool-summer steppes and semideserts expanded due to the cooling. A slight winter warming in mid latitudes, however, caused warm-temperate evergreen forests and scrub to expand poleward. Such qualitative contrasts in the direction of climate and vegetation change during 5e should be identifiable in the paleorecord.

The record of glacier fluctuations in western Scandinavia, as reconstructed from continental data, has been correlated with records of ice-rafted detritus (IRD) from well-dated sediment cores from the Norwegian Sea covering the past 150,000 yr B.P. The input of IRD into the ocean is used as a proxy for ice sheet advances onto the shelf and, thus, for the calibration of a glaciation curve. The marine results generally support land-based reconstructions of glacier fluctuations and improve the time-control on glacial advances. The Saalian ice sheet decayed very rapidly approximately 125,000 yr B.P. In the Early Weichselian, a minor but significant IRD maximum indicates the presence of icebergs in isotope substage 5b (especially between 95,000 and 83,000 yr B.P.). Reduced amounts of calcareous nannofossils indicate that surface waters were influenced by meltwater discharges during isotope substages 5d and 5b. An extensive build-up of inland ice began again during isotope stage 4, but maximum glaciation was reached only in early stage 3 (58,000-53,000 yr B.P.). Marine sediments have minimum carbonate content, indicating strong dilution by lithogenic ice-rafted material. Generally, the IRD accumulation rate was considerably higher in stages 4-2 than in stage 5. A marked peak in IRD accumulation rates from 47,000 to 43,000 yr B.P. correlates well with a second Middle Weichselian ice sheet advance dated by the Laschamp/Olby paleomagnetic event. Minimum ice extent during the Ålesund interstade (38,500-32,500 yr B.P.) and several glacial oscillations during the Late Weichselian are also seen in the IRD record. Of several late Weichselian glacial oscillations on the shelf, at least four correspond to the North Atlantic Heinrich events. Ice sheet behavior was either coupled or linked by external forcing during these events, whereas internal ice sheet mechanisms may account for the noncoherent fluctuations.

Eolian sand is extensive over the Great Plains of North America, but is at present mostly stabilized by vegetation. Accounts published by early explorers, however, indicate that at least parts of dune fields in Nebraska, Colorado, Kansas, New Mexico, and Texas were active in the 19th century. Based on an index of dune mobility and a regional tree-ring record, the probable causes for these periods of greater eolian activity are droughts, accompanied by higher temperatures, which greatly lowered the precipitation-to-evapotranspiration ratio and diminished the cover of stabilizing vegetation. In addition, observations by several explorers, and previous historical studies, indicate that rivers upwind of Great Plains dune fields had shallow, braided, sandy channels, as well as intermittent flow in the 19th century. Wide, braided, sandy rivers that were frequently dry would have increased sand supplies to active dune fields. We conclude that dune fields in the Great Plains are extremely sensitive to climate change and that the potential for reactivation of stabilized dunes in the future is high, with or without greenhouse warming.

Paleotemperatures for the last glacial maximum (LGM) have been derived from noble gases dissolved in 14C-dated groundwater of the Ojo Alamo and the Nacimiento formations in the San Juan Basin, northwestern New Mexico. The difference in mean annual (ground) temperature between the Holocene and the LGM was determined to be 5.5 ± 0.7°C. A practically identical result, 5.2 ± 0.7°C, has been obtained previously from the Carrizo aquifer in southern Texas. This suggests that the southwestern United States was uniformly cooler during the LGM and that the mean annual temperature gradient along a transect from the Gulf of Mexico to northwestern New Mexico has been unchanged since the LGM. The noble gas paleotemperatures are supported by paleoecological evidence in the region. The Holocene/LGM temperature difference of 5.4°C indicates that a simple lapse rate calculation may be applied to convert the 1000-m glacial depression of snowlines in the Colorado Front Range into a temperature decrease. A continental temperature change of 5.4 ± 0.7°C is inconsistent with a temperature change of about 2°C determined for the surface waters of the Gulf of Mexico.

Compositions of δD, δ18O, 14C, and 3H are distinct in unconfined and confined groundwaters beneath the nonglaciated U.S. High Plains and reflect differences in average paleoclimatic conditions between the Holocene and the Middle to Late Wisconsinan. δD and δ18O are more depleted in confined groundwaters than in unconfined groundwaters beneath the southern and central Plains but are more enriched in confined than in unconfined groundwaters beneath the northern Plains. The range in stable isotopic composition of meteoric water north-to-south across the High Plains was apparently smaller during the Middle to Late Wisconsinan than it has been during the Holocene. The greater δD and δ18O depletion of confined groundwater beneath the southern and central Plains agrees with isotopic and noble gas data on other paleowaters that suggest Middle to Late Wisconsinan recharge temperatures were cooler than average Holocene temperatures by 5 to 8°C. The greater δD and δ18O enrichment of confined groundwater beneath the northern High Plains adds to previously reported evidence for stable isotope enrichment in glacial age precipitation across the northern Plains to the eastern United States. Additional research is needed to evaluate how patterns and mechanisms of moisture transport and other climatic variables might explain isotopic composition of paleorecharge during the last glacial period.

Radiocarbon dating of three organic matter fractions (total, humic acid, and residue) isolated from late Quaternary buried soils of the central Great Plains reveals that there often are considerable differences among, but no consistent order to, the ages of fractions. For late Holocene soils, the residue fraction or the total fraction generally produces the oldest age; for late Pleistocene soils, however, no fraction was consistently the oldest. The absence of a consistent sequence of fraction ages is attributed to postburial contamination of soils. When bulk samples from the same soil were split and sent to two laboratories, different radiocarbon ages were usually obtained. The variability in radiocarbon ages of soil organic matter confirms that caution should be taken when using radiocarbon ages obtained from different laboratories to make regional stratigraphic correlations.

Twenty packrat (Neotoma) middens recovered from three sites (1265-1535 m) in the Alabama Hills, Inyo County, California, provide a ca. 31,450-yr record of vegetation change. Located ca. 7 km east of the Sierra Nevada, the middens document that Utah juniper (Juniperus osteosperma), Joshua tree (Yucca brevifolia), and bitterbush (Purshia tridentata) occupied the site between 31,450 and 19,070 yr B.P. Joshua tree and bitterbush departed by ca. 17,760 yr B.P., with cliffrose (Purshia mexicana) and joint-fir (Ephedra viridis) appearing. By 13,350 yr B.P., blackbush (Coleogyne ramosissima) and cholla (Opuntia echinocarpa) entered the record. Between 9540 and 7990 yr B.P., Utah juniper and other species now extralocal to the sites departed and modern components such as wolfberry (Lycium andersonii) and rubber rabbitbrush (Chrysothamnus teretifolius) appeared. The middle Holocene records little variation in plant macrofossil composition; however, pollen analysis reflects an increase in aquatic pollen types which might suggest more-open conditions. The transition to the modern vegetation associations at the sites occurred after ca. 2800 yr B.P. The record from the Alabama Hills correlates well with that of other regional vegetation data but documents conditions of increasing aridity earlier than many other packrat midden sites. A shift in understory vegetation between 19,070 and 17,760 yr B.P. may reflect a transition from glacial maximum to post-maximum conditions in the eastern Sierra Nevada.

An extensive late Quaternary fauna, including many extinct giant lemurs, has been collected recently in a 110+-km system of caves in the Ankarana Massif of northern Madagascar. AMS 14C dates for the acid-insoluble (collagen/gelatin) fraction of bones of the giant lemur Megaladapis (26,150 ± 400 and 12,760 ± 70 yr B.P.) confirm its presence in the area during the late Pleistocene and provide the first Pleistocene 14 C ages from bones of the extinct megafauna of the island. The first date from bones of the recently described extinct Babakotia radofilai (4400 ± 60 yr B.P.) shows that it was present in northern Madagascar in mid-Holocene times. A comparatively recent age of 1020 ± 50 yr B.P. for the extinct Archaeolemur indicates survival of this genus for at least a millennium after the first direct evidence for humans in Madagascar. This suggests that the island's "extinction window" may have represented a longer time span than would have been expected under the Blitzkrieg model of late Quaternary extinctions. A mid-Holocene age (4560 ± 70 yr B.P.) for a bone sample of the small extant lemur Hapalemur simus indicates that the disappearance of this now-restricted species from the Ankarana occurred after this date. New data from the Ankarana and other sites on the island add to the consensus that major biotic changes occurred on Madagascar in the late Holocene.

Initial excavations (1935-1941) in San Josecito Cave, Nuevo León, México revealed a rich late Quaternary fossil fauna. The San Josecito Cave fauna was viewed as a single time-related entity despite the lack of 14C dating. Renewed excavations (1990) recovered a series of 14 C samples related to the complex stratigraphy. Six 14C dates show that the recently excavated deposits range in age between 45,000 and 27,000 yr B.P. These ages suggest that the upper deposits, removed in the earlier excavations, date from 27,000 to 11,000 yr B.P.

Plant and vertebrate macrofossils in an early Holocene fossil packrat (Neotoma sp.) midden with a radiocarbon age of 8640 ± 100 14C yr B.P. are reported from the Chocolate Mountains, near the Salton Sea, Riverside County, California. An inventory of the midden has permitted a comparison of the modern flora and fauna of the site with that extant during the early Holocene. Whereas the biota had assumed most aspects of its modern Sonoran desert aspect by this date, statistically significant evidence of differences is attributed to an increased flow of surface water in Salt Creek, a high-standing, low-salinity Lake LeConte, and the late arrival of some characteristic Sonoran desert plants. These observations are consistent with models of significant fall-winter precipitation in the Sonoran Desert, although we cannot exclude alternative explanations.