Sub-Antarctic Marion Island's glacial history has acted as a control on abiotic terrestrial processes and the colonization and distribution of biotic species found on the island today. Recent chronological studies have shown an early deglaciation of the island and identified new geomorphological features associated with past ice dynamics. These permit a reassessment of ice extent during and after the island's last local glacial maximum. In this paper, we provide a revised reconstruction of the island's palaeo-ice extent by using a geomorphology-based approach to delineate palaeo-ice margins and demarcate possible glacial basins. The model presented here provides the needed spatial context for future studies on the variations in the distribution of species (e.g. microorganisms and plant species) and abiotic processes and forms (e.g. soil development and periglacial landforms). In addition, it highlights areas that require improved geophysical assessment in order to produce a more complete island-scale reconstruction of former ice extents (e.g. the west coast).

Glacial lake sediments exposed at two sites in Skagit Valley, Washington, encase abundant macrofossils dating from 27.7 to 19.8 cal ka BP. At the last glacial maximum (LGM) most of the valley floor was part of a regionally extensive arid boreal (subalpine) forest that periodically included montane and temperate trees and open boreal species such as dwarf birch, northern spikemoss, and heath. We used the modern distribution and climate of 14 species in 12 macrofossil assemblages and a probability density function approach to reconstruct the LGM climate. Median annual precipitation (MAP) at glacial Lake Concrete (GLC) was ~50% lower than today. In comparison, MAP at glacial Lake Skymo (GLS) was only ~10% lower, which eliminated the steep climate gradient observed today. Median January air temperature at GLC was up to 10.8°C lower than today at 23.5 cal ka BP and 8.7°C lower at GLS at 25.1 cal ka BP. Median July air temperature declines were smaller at GLC (3.4°C–5.0°C) and GLS (4.2°C–6.3°C). Warmer winters (+2°C to +4°C) and increases in MAP (+200 mm) occurred at 27.7, 25.9, 24.4, and 21.2–20.7 cal ka BP. These changes accord with other regional proxies and Dansgaard–Oeschger interstades in the North Atlantic.

The late Quaternary development of the Lower Volga region of Russia is characterized by an alternating influence of marine and continental environments resulting from fluctuations in Caspian Sea level during the last glaciation. However, sediments deposited under continental conditions have received very little research attention compared to the under- and overlying marine deposits, such that even their origin is still in debate. Detailed magnetic mineralogical analyses presented here show clear similarities to loess. The results suggest that climate during the time of loess deposition, the Atelian regression (27–80 ka, MIS 4–3), was dry and cool, similar to the modern-day Northern Caspian lowland. The magnetic properties recorded in the loess-paleosol sequences of the Lower Volga also point to short episodes of potentially more humid and warmer climate during the late Atelian. The new findings in regard to the local Caspian climate and environmental evolution support decreased river discharge from the Russian Plain and Siberian Plain as the dominant factor causing the low Caspian sea level stand during the Atelian, although local-regional climate changes might have had an additional influence.

Isotopic and geochemical evidence of paleoclimates, especially for the last glaciation, has been obtained from deep confined groundwaters of southern India. The δ13C, δ18O, chloride, and deuterium analyses of groundwaters show distinct excursions inferred to be related to climatic variations. The arid climatic episode associated with the last glaciation (18,000 ± 2000 yr B.P.) is conspicuously identified by signatures of relatively enriched δ13C (−10 to −12‰ PDB) and δ18O (−5.3 to −4.8‰ SMOW) values, and high chloride concentration (80 to 160 mg/l). The transition from an arid to humid period ca. 12,000–8000 yr B.P. is shown by a decreasing trend in the δ13C (−9.5 to −17‰) and δ18O (−4.5 to −6.3‰) contents of groundwaters. The late Holocene (since 4000 yr B.P.), marked by a more humid but unstable climate, is identified by further depletion of δ13C (−13 to −20‰) and δ18O (−5.2 to −6.3‰). Similar variation between δ18O and chloride values in confined groundwaters further demonstrates two distinct climatic excursions (arid and humid) governed by the “amount effect.” This is the first time that isotopic and geochemical signatures related to changing paleoclimates have been identified in the confined groundwaters of the southern Indian landmass.

Radiocarbon (14C) dating of total soil organic matter (SOM) often yields results inconsistent with the stratigraphic sequence. The onerous chemical extractions for SOM fractions do not always produce satisfactory 14C dates. In an effort to develop an alternative method, the pyrolysis-combustion technique was investigated to partition SOM into pyrolysis volatile (Py-V) and pyrolysis residue (Py-R) fractions. The Py-V fractions obtained from a thick glacigenic loess succession in Illinois yielded 14C dates much younger but more reasonable than the counterpart Py-R fractions for the soil residence time. Carbon isotopic composition (δ13C) was heavier in the Py-V fractions, suggesting a greater abundance of carbohydrate- and protein-related constituents, and δ13C was lighter in the Py-R fractions, suggesting more lignin- and lipid-related constituents. The combination of 14C dates and δ13C values indicates that the Py-V fractions are less biodegradation resistant and the Py-R fractions are more biodegradation resistant. The pyrolysis-combustion method provides a less cumbersome approach for 14C dating of SOM fractions. With further study, this method may become a useful tool for analyzing unlithified terrestrial sediments when macrofossils are absent.

In the midwestern United States, large rivers draining the Laurentide Ice Sheet (LIS) were the most important sources of Peoria Loess, deposited during the last glaciation. Loess deposition near those rivers may have responded primarily to ice-sheet dynamics rather than direct effects of climatic change. In contrast, it has been proposed that thick Peoria Loess on the central Great Plains was derived mainly from unglaciated landscapes northwest of the main loess deposits. In this study, transport directions inferred from more than 600 measurements of Peoria Loess thickness in Nebraska are used to test the hypothesis that much of the Peoria Loess on the Great Plains is nonglaciogenic. A strong northwest to southeast thickness trend indicates that most Peoria Loess in Nebraska was transported from one or more unglaciated northwestern source areas rather than from glacially influenced river floodplains. The Missouri River (draining the LIS), the Platte River (draining alpine glaciers), and the Elkhorn River (unglaciated basin) were secondary sources. Their contribution is not detectable beyond a distance of 40–60 km. Peoria Loess deposition on the central Great Plains was largely a direct response to climatic change in the unglaciated source region.

Pollen and charcoal records from two large, shallow lakes reveal that throughout most of the past 50,000 yr Noel Kempff Mercado National Park, in northeastern lowland Bolivia (southwestern Amazon Basin), was predominantly covered by savannas and seasonally dry semideciduous forests. Lowered atmospheric CO2 concentrations, in combination with a longer dry season, caused expansion of dry forests and savannas during the last glacial period, especially at the last glacial maximum. These ecosystems persisted until the mid-Holocene, although they underwent significant species reassortment. Forest communities containing a mixture of evergreen and semideciduous species began to expand between 6000 and 3000 14C yr B.P. Humid evergreen rain forests expanded to cover most of the area within the past 2000 14C yr B.P. coincident with a reduction in fire frequencies. Comparisons between modern pollen spectra and vegetation reveal that the Moraceae-dominated rain forest pollen spectra likely have a regional source area at least 2–3 km beyond the lake shore, whereas the grass- and sedge-dominated savanna pollen spectra likely have a predominantly local source area. The Holocene vegetation changes are consistent with independent paleoprecipitation records from the Bolivian Altiplano and paleovegetation records from other parts of southwestern Amazonia. The progressive expansion in rain forests through the Holocene can be largely attributed to enhanced convective activity over Amazonia, due to greater seasonality of insolation in the Southern Hemisphere tropics driven by the precession cycle according to the Milankovitch Astronomical Theory.

Pollen, foraminifer, dinocyst, and coarse lithic high-resolution analyses and δ18O measurements have been carried out for the last-glacial section of marine core MD95-2042 located near the southwestern margin of the Iberian Peninsula. The pollen data indicate a high frequency of vegetational changes on the adjacent continent during this period, suggesting a climatic variability very similar to that of the Dansgaard-Oeschger cycles recorded by the Greenland ice cores. The detailed direct correlation of the terrestrial and marine proxy data from core MD95-2042 indicates a three-phase pattern of Heinrich events in land and ocean environments. The first and last phases of the H5 and H4 events are characterized by a mild and humid climate in southwestern Europe, probably associated with the European origin of the ice-rafted detritus. The middle phase exhibits a cold and dry climate in Iberia linked with the maximum input of ice-rafted detritus. This phase seems to correspond with the Laurentide ice-sheet surges. Between the Heinrich events, several cold and dry periods on land are correlated with stades of the Dansgaard-Oeschger cycles. The impact of the Dansgaard-Oeschger stades in southwestern Europe seems to be preferentially connected to the cold winter air masses reaching this mid-latitude region.

Carbon isotopes of soil organic matter (SOM) were used to evaluate and establish the chronology of the vegetation dynamics of an ecosystem presently composed of savannas surrounded by forests. The study was carried out on a 200-km transect along highway BR 319, on the border of Amazonas and Rondônia states, in southern Amazon, Brazil. Large ranges in δ13C values were observed in SOM collected from profiles in the savanna (−27 to −14‰) and forest regions (−26 to −19‰), reflecting changing distribution of 13C-depleted C3 forest and 13C-enriched C4 savanna vegetation in response to climate change. These results indicate that from about 17,000 to 9000 14C yr B.P., the study area was covered by forest vegetation. Between approximately 9000 and 3000 14C yr B.P., savanna vegetation expanded at the expense of the forest. Although the expansion of savanna did not occur with the same intensity along the study transect, this process was very clearly registered by 13C-enrichment in the SOM. Since 3000 14C yr B.P., the carbon isotope data suggest that forested regions have expanded. This study adds to the mounting evidence that extensive forested areas existed in the Amazon during the last glaciation and that savanna vegetation expanded in response to warm and dry conditions during the early to middle Holocene.

Moraine chronology is combined with digital topography to model deglacial rates of paleoglacier volumes in both the Huancané Valley on the west side of the Quelccaya Ice Cap and the Upismayo Valley on the northwest side of the Cordillera Vilcanota. The fastest rates of deglaciation (39×10−5 to 114×10−5 km3 yr−1 and 112×10−5 to 247×10−5 km3 yr−1 for each valley, respectively) were calculated for the most recent paleoglaciers, corresponding to the last few centuries. These results are consistent with observations in the Venezuelan Andes showing high rates of deglaciation since the Little Ice Age. These rates also fall within the range of 20th century rates of deglaciation measured on the Quelccaya Ice Cap (29×10−5 to 220×10−5 km3 yr−1, Brecher and Thompson, 1993; Thompson, 2000). These results imply that rates of deglaciation may fluctuate significantly over time and that high rates of deglaciation may not be exclusive to the late 20th century. Equilibrium line altitude (ELA) depressions for the ice volumes of the last glaciation modeled here were computed as 230 m for the Quelccaya Ice Cap and 170 m for the Cordillera Vilcanota. Maximum ELA depressions are lower than previously published: <500 m for the Cordillera Vilcanota and <400 m for the Quelccaya Ice Cap. These lower values could imply a topographic control over paleoglacier extent.

The late Quaternary paleoclimate of eastern Beringia has primarily been studied by drawing qualitative inferences from vegetation shifts. To quantitatively reconstruct summer temperatures, we analyzed lake sediments for fossil chironomids, and additionally we analyzed the sediments for fossil pollen and organic carbon content. A comparison with the δ18O record from Greenland indicates that the general climatic development of the region throughout the last glaciation–Holocene transition differed from that of the North Atlantic region. Between ∼ 17 and 15 ka, mean July air temperature was on average 5°C colder than modern, albeit a period of near-modern temperature at ∼ 16.5 ka. Total pollen accumulation rates ranged between ∼ 180 and 1200 grains cm− 2 yr− 1. At ∼ 15 ka, approximately coeval with the Bølling interstadial, temperatures again reached modern values. At ∼ 14 ka, nearly 1000 yr after warming began, Betula pollen percentages increased substantially and mark the transition to shrub-dominated pollen contributors. Chironomid-based inferences suggest no evidence of the Younger Dryas stade and only subtle evidence of an early Holocene thermal maximum, as temperatures from ∼ 15 ka to the late Holocene were relatively stable. The most recognizable climatic oscillation of the Holocene occurred from ∼ 4.5 to 2 ka.