Multiproxy data from Little Swift Lake, an alpine lake in southwestern Alaska, provide evidence for pronounced late glacial and Holocene environmental change. An alpine glacier upvalley of Little Swift Lake retreated following the Younger Dryas chronozone, as evidenced by sedimentological changes in the lake record. Glacier retreat was accompanied by local and regional vegetation changes, including the expansion of
Late glacial climate was remarkable for its rapid, high-magnitude fluctuations, and Holocene climate for its relative stability. Both extremes hold clues about how the climate system works: late glacial paleoclimate studies reveal how quickly climate can switch gears. Studies of Holocene paleoclimate demonstrate the range of natural climatic variability given near-modern configuration of ice sheets and oceanic and atmospheric circulation.
The late glacial and Holocene climate history of the North Pacific region is poorly known compared with the North Atlantic. Although changes in the North Atlantic region were probably greater in magnitude during this time, an increasing number of records indicate that the northeast Pacific region also experienced significant environmental changes. Such changes included climate reversals during the Younger Dryas chronozone (YD, ∼12,900 to 11,600 cal yr B.P. as defined by Alley ; cal yr B.P. indicates calendar years before 1950) (e.g., Engstrom et al., 1990; Mathewes et al., 1993; Peteet and Mann, 1994; Patterson et al., 1995; Hu et al., 1995, 1998, 2002; Brubaker et al., 2001; Mann et al., 2002; Briner et al., 2002), an early Holocene “thermal maximum” (e.g., Hu et al., 1996, 1998), and Neoglacial glacier advances (e.g., Calkin, 1988).
Here we reconstruct the major features of late glacial and Holocene paleoenvironmental change at an alpine site in southwestern Alaska, a region closely linked to the Bering Sea. We present lacustrine sedimentological evidence for glacier fluctuation, along with a continuous pollen record of vegetation changes, from Little Swift Lake in the Ahklun Mountains.
The Ahklun Mountains (Fig. 1A) are a large glaciated massif lying beyond the late Wisconsin extent of the Cordilleran Ice Sheet complex. Past research in the Ahklun Mountains has focused on the extensive Pleistocene stratigraphic and geomorphic records of this repeatedly glaciated region (e.g., Kaufman et al., 1996, 2001; Briner and Kaufman, 2000; Briner et al., 2001; Manley et al., 2001). Yet the Ahklun Mountains, containing many peaks over 1500 m a.s.l. and more than 100 extant glaciers, also preserve largely unstudied records of late glacial and Holocene glacier fluctuations.
The northwestern Ahklun Mountains include some of the highest cirques in the range, extensive well-preserved moraine sequences, and numerous lakes. Little Swift Lake (60°12′52″N; 159°45′57″W; 572 m a.s.l.) is a small (∼0.7 km2 surface area), 24-m deep drift-impounded basin capturing the drainage of a deeply carved ∼10-km-long glacial valley (Figs. 1, 2). The drainage basin has no glaciers today, but extant glaciers occupy cirques 15 km to the southeast.
Little Swift Lake is at the mouth of a valley tributary to the Crooked Creek valley (Figs. 1, 2). Bedrock within the lake's catchment includes granodiorite, rhyolite tuff, and mixed sedimentary and volcaniclastic rocks, with no carbonate or carbonaceous bedrock reported (Box et al., 1993). The lake is dammed by a lateral moraine that was deposited along the right margin of the trunk glacier that occupied the Crooked Creek valley during the late Wisconsin (Briner et al., 1999; Axford, 2000). The lake is downvalley from a sequence of alpine-glacier end moraines that were deposited by local valley glaciers. At least three late glacial and Holocene ice positions are represented (Axford, 2000; Figs. 1C, 2). Although glaciofluvial sediment produced by small glaciers in the high upper reaches of the valley may have been captured by several ponds upvalley of Little Swift Lake, the lake should contain sedimentary evidence of larger glaciers, such as the ca. 8.5-km-long glacier that deposited a prominent late glacial end moraine 1.3 km upvalley of the lake (Figs. 1C, 2).
Little Swift Lake is located within the transition between maritime and continental climate, in an area of discontinuous permafrost. The lake is surrounded by Betula shrub tundra, graminoid herb tundra, and steep, unvegetated rocky slopes. Dwarf birch (Betula nana) and diverse willows (Salix spp.), ferns (Polypodiaceae), sedges (Cyperaceae), grasses (Poaceae), club mosses (Lycopodium), and Sphagnum mosses dominate moist low-lying vegetation communities. Moraine ridge crests and highlands support sparse alpine tundra, with lichens, heaths (Ericaceae, e.g., labrador tea [Ledum decumbens], crowberry [Empetrum nigrum]), and herbs. Alder (Alnus) is not present within the lake's catchment, but Alnus thickets occupy creek beds ∼5 km away. Spruce (Picea) grows 20 km to the northwest (where small Picea glauca stands inhabit river margins), and in boreal forest lowlands 50 km to the east.
Core LS-A, a 575-cm-long sediment core, was recovered from Little Swift Lake using a modified Nesje corer (Nesje, 1992) in 14.7 m of water. Magnetic susceptibility (MS) was measured on unsplit core segments at 2-cm increments using a Sapphire Instruments loop detector. Loss-on-ignition (LOI) at 550°C was used to estimate the organic matter content of 1-cc samples taken at 5-cm increments (Dean, 1974; Bengtsson and Enell, 1986). Sediment grain-size distribution was analyzed for 1-cm-thick samples taken at 5-cm increments using a Coulter LS230 laser detection particle-size analyzer. Grain-size samples were pretreated with H2O2, NaOH, and sodium hexametaphosphate to remove organic material, remove biogenic silica, and disaggregate sediments, respectively. Laboratory tests revealed a problem measuring the abundance of sand-sized grains in samples, due to the evolution of sand-sized air bubbles in the cold water used to circulate sediment through the analyzer. The problem was later corrected by installing a holding tank to store water at room temperature. Because measurements of sand percentages for this study are unreliable, we report only the unaffected clay:silt ratio.
Pollen samples were processed at Northern Arizona University's Laboratory of Paleoecology according to conventional procedures (e.g., Faegri and Iversen, 1975). An extended acetolysis time of 4 min improved the efficacy of saffranin staining. Lycopodium tracer tablets were added to samples to allow for calculation of pollen concentrations. All samples were sieved through a 9-μm screen following acetolysis to remove clays. Processed samples were mounted in silicon oil. A sum of 200 to 350 terrestrial pollen grains was identified per sample. Pollen abundances were calculated as percentages of the terrestrial pollen sum, and spore abundances as percentages of pollen plus spores. Pollen zones were determined using a stratigraphically constrained incremental sum of squares cluster analysis (the CONISS program packaged with TILIA software; Grimm, 1987) based on pollen and spore taxa with abundances >2%.
Nine AMS 14C ages were obtained on mixed samples of hand-picked terrestrial and aquatic chitinous and floral macrofossil fragments >150 μm diameter. A tenth age—paired with a mixed sample—was obtained on terrestrial plant parts. δ13C was accounted for in age calculations. Radiocarbon ages were calibrated to calendar years using CALIB 4.3 (Stuiver and Reimer, 1993).
The 10 calibrated ages from core LS-A range from 1480 ± 65 cal yr B.P. on a sample from 20 to 25 cm depth to 13,110 ± 85 cal yr B.P. from the interval 545 to 550 cm depth (Table 1; Fig. 3). As described herein, correlations between the Little Swift Lake tephra stratigraphy and the regional tephra stratigraphy suggest that the radiocarbon ages are erroneously old, even though we reject the ages obtained from 508-513 cm and 545-550 cm depth, which contradict the basal age and apparently greatly overestimate the age of the core. A thick pumiceous tephra at 208 to 232 cm depth is most likely the Aniakchak tephra, which is prominent in lake records throughout the Ahklun Mountains (Kaufman et al., 2003) and has been described elsewhere in western Alaska (Riehle, 1985; Beget et al., 1992). Its age is well constrained at ∼3600 ± 100 cal yr B.P., based upon over a dozen 14C ages (Beget et al., 1992; Waythomas and Neal, 1998). A minimum limiting age of 4250 ± 100 cal yr B.P. from 202 to 207 cm depth in LS-A, immediately above the tephra, is apparently ∼650 yr too old. A more mafic tephra at 318 cm depth in LS-A dates to ∼6680 cal yr B.P., based upon a second-order polynomial interpolation between 8 of the 10 radiocarbon ages (Fig. 3). This tephra correlates with a ∼6100 cal yr B.P. tephra found in nearby Arolik and Waskey lakes (Levy et al., 2003; Kaufman et al., 2003). The uppermost age of 1480 cal yr B.P. from 20 to 25 cm depth is probably also too old. Although our coring procedure did not capture the sediment-water interface, based on previous experiences with the same equipment, and assuming sedimentation rates throughout the core are a fair approximation of sedimentation rates over the last millennium, it is very unlikely that the core top is really 1000 yr old. We conclude that 14C ages from the upper half of LS-A are ∼600 yr too old. The polynomial trend through our original radiocarbon ages would estimate the age of the base of LS-A at 13,200 cal yr B.P. (Fig. 3), but a ∼12,200 cal yr B.P. tephra found in cores from nearby Arolik and Nimgun lakes (Kaufman et al., 2003) is absent in LS-A; thus ages from the base of LS-A may be ∼1000 yr too old.
In order to assign reasonable ages to events recorded in the core stratigraphy, we use an age model that corrects for offset in the radiocarbon chronology. We assume an offset of 600 yr for ages less than 10,000 cal yr B.P., and an offset of 1000 yr for older ages. 10,000 cal yr B.P. is an arbitrary cut-off, but is consistent with evidence that ages near the base of the core are 1000 yr too old. The ages immediately above and below this cut-off are ∼2000 yr apart, so there is no sudden jump in the chronology.
A second-order polynomial trend line (y = 0.012x2 + 13.574x + 518.45) through the eight adjusted ages has an R2 value of 0.995, and agrees well with the tephrochronology (Fig. 3). Pollen correlation provides another test: Striking regional palynological changes occurred at Ongivinuk (Hu et al., 1995), Idavain (Brubaker et al., 2001), and Nimgun (Hu et al., 2002) lakes (Fig. 1) at the end of the YD chronozone; the same changes (including a sharp decline in Cyperaceae and increase in Poaceae and Betula) date to 11,430 cal yr B.P. in Little Swift Lake using our adjusted age model. Based on comparisons with the regional tephra and pollen stratigraphies, we estimate the uncertainty of our model to be ± 200 yr.
The shell of a living snail collected from the lake in 1998 has a radiocarbon activity of 100.1% Modern, less than atmospheric values of the 1990s. This indicates that there is a source of 14C-depleted dissolved inorganic carbon (DIC) to the lake. Possible sources of 14C-depleted DIC include groundwater, dissolution of rocks within the lake's watershed (although no carbonate or carbonaceous bedrock has been reported from Little Swift Lake's drainage), and decomposition of organic soil horizons within the lake's watershed or old organic matter within the lake (e.g., Abbott and Stafford, 1996).
The statistically indistinguishable paired ages obtained on terrestrial plant material and mixed aquatic and terrestrial material at 20 to 25 cm depth (Table 1) indicate that there must also be an influx of 14C-depleted terrestrial macrofossils to the lake. Reworked organic carbon from old soils can contribute depleted carbon to arctic lakes (Abbott and Stafford, 1996), but old soils should have been removed from this watershed during the extensive late Wisconsin glaciation, and we found no buried soils in deep river bluff exposures. Thus the source of old terrestrial macrofossils to the lake is unknown. We note that we would have been unaware of problems with this chronology, had it not been for comparison with tephras of known age. Similar unexpected chronological problems at other arctic lakes may go unrecognized, emphasizing the importance of reliable chronostratigraphic markers.
Most of core LS-A (0 to 547 cm depth) is massive to laminated medium-brown to gray-brown gyttja, with only subtle changes in appearance (Fig. 4). The gyttja is banded throughout, with finer (millimeter-scale) laminations at the base of this interval. Magnetic susceptibility (MS) within the gyttja is low (1.9 × 10−5 to 3.6 × 10−5 cgs), with little variation except for a subtle upward-increasing trend. The gyttja exhibits only minor changes in organic content (7 to 11% LOI), with a zone of highest LOI between 380 and 455 cm depth. Clay:silt decreases (i.e., the sediment becomes coarser grained) from the bottom of the gyttja (maximum of 0.76 clay:silt at 545 cm depth) to the top (minimum of 0.26 clay:silt at 50 cm depth).
A 24-cm-thick, light brownish-gray tephra at 208 to 232 cm depth is composed of sand- and silt-sized pumice grains, bubble-wall glass shards, and fine mafic mineral grains. There is also a dark brown fine sandy tephra, ca. 0.2 cm thick, at 318 cm depth, and a light gray fine sandy tephra of similar thickness at 398 cm depth. Each tephra layer is coincident with a prominent peak in MS and a zone of low LOI.
The core terminates in medium gray silty clay, extending from an abrupt transition at 547 cm depth (∼11,400 cal yr B.P.) to the base of the core at 575 cm (∼12,200 cal yr B.P.). (Note that we did not meet resistance in coring, so the base of LS-A does not represent the beginning of the sedimentary record in Little Swift Lake.) The basal gray mud has a lower clay:silt ratio (0.50 to 0.57) than the brown gyttja immediately above (0.76 at 545 cm depth) (Fig. 4), and low LOI (3 to 4%). The MS of the basal unit is an order of magnitude higher than that of the gyttja, reflecting its low organic content, coarse grain size, and probably distinct mineralogy. None of the sediment in core LS-A fizzes in contact with dilute HCl.
Modern pollen deposition was estimated by analyzing pollen from the top 0.5 cm of a short gravity core that captured the undisturbed sediment-water interface near core LS-A. The modern pollen rain is dominated by Alnus (44%), Cyperaceae (20%), and Betula (19%), with low percentages (<5%) of Poaceae, Picea, Artemisia, Salix, Ericales, and numerous herb taxa. Monoletes, Lycopodium, and Sphagnum are the most abundant spores.
Pollen assemblages from Zone LS1 (575 to 547 cm depth; ∼12,200 to 11,400 cal yr B.P.) include abundant Cyperaceae (>45%), and the core's highest percentages of Salix, Artemisia, Tubuliflorae, and Saxifragaceae (Fig. 5), all of which decrease toward the top of this zone. Betula pollen increases slightly from its lowest value in the core (11%). Spores (including Lycopodium, Sphagnum, monoletes, and Equisetum) and Pediastrum cell nets are rare or absent (Fig. 5).
Zone LS2 (547 to 485 cm depth; ∼11,400 to 9800 cal yr B.P.) is defined by a Poaceae peak reaching a maximum abundance of 34%. Cyperaceae decrease to about half their earlier abundance. Ericales and Brassicaceae are relatively abundant throughout much of LS2. Alnus pollen first appears, but remains rare. Monolete spores and Pediastrum rapidly increase, and Pediastrum, Sphagnum, and Equisetum reach their maximum abundances in this zone.
Within Zone LS3 (473 to 390 cm depth; ∼9800 to 7600 cal yr B.P.), Poaceae pollen declines to its earlier low percentage (∼10%). Betula, Sanguisorba, and monolete spores reach peak abundance, with Betula reaching maximum abundance (44%) at 428 cm depth (∼8500 cal yr B.P.). Artemisia, Pediastrum, and Equisetum decline. Picea pollen first appears at 399 cm depth (∼7800 cal yr B.P.).
At the base of Zone LS4 (390 to 0 cm depth; ∼7600 cal yr B.P. to present), Alnus pollen rapidly increases from 2 to 25%. Above the increase in Alnus, Zone LS4 is relatively stable, with abundant Alnus, Betula, and Cyperaceae pollen, and lower but significant percentages of Salix, Ericales, Artemisia, and Poaceae. Ericales are relatively abundant and increase slightly throughout LS4. Picea pollen reaches a maximum abundance of 5% at 18 cm depth (∼800 cal yr B.P.). Monolete spore and Pediastrum percentages decline throughout LS4, whereas Lycopodium increases slightly.
LATE GLACIAL ENVIRONMENTAL CHANGE
The record from core LS-A begins during a period of climatic amelioration, characterized by increasing Betula shrub tundra and decreasing Salix, Artemisia, and Tubuliflorae. These vegetation changes correlate with similar changes that occurred during a period of warming and increased moisture at the end of a YD-age cold reversal at nearby Grandfather and Nimgun lakes (Hu et al., 1995, 2002), and hint that Little Swift Lake may have experienced a similar reversal to colder conditions and herb-dominated tundra.
There is evidence that glacier retreat accompanied vegetation change at the end of the YD chronozone at Little Swift Lake. Considering the proximity of the lake to upvalley late glacial moraines (Briner et al., 1999; Axford, 2000), we interpret the basal, inorganic gray silt in LS-A as ice-proximal sediment transported to the lake by glacial meltwater streams. An upvalley glacier retreated (probably to a position behind an upvalley pond) ∼11,400 cal yr B.P., as recorded by the transition to organic clay-rich mud.
Alternatively, these sedimentological changes may record an abrupt increase in lake and watershed productivity without the influence of an upvalley glacier. However, diverse additional data support our original interpretation: The distinct bulk major-element geochemistry of the core's basal unit, exceptionally rich in Mg, Na, K, Al, and Ti, suggests an abundance of clastic minerals and a unique provenance. Geochemical analyses of likely sediment sources within the lake's watershed indicate that a granodiorite, which outcrops only in upvalley cirques, is the most likely source (Carey et al., 2000). The low LOI and lack of Pediastrum cell nets within the basal unit suggest low aquatic production, which could result from a variety of conditions, but is consistent with high turbidity and low water temperatures due to meltwater influx. The near-disappearance of the diatoms Hannaea arcus and Nitzschia spp. from the lake at ∼11,400 cal yr B.P. further supports decreased meltwater flux to the lake (I. Gregory-Eaves, pers. comm., 2002); Hannaea arcus is found in cold, flowing water and Nitzschia species thrive in turbid, unproductive, river-influenced arctic lakes (e.g., Ludlam et al., 1996; Hay et al., 1997).
There is widespread evidence from southwestern Alaska indicating a reversal to cold, dry conditions during the YD, followed by warming (Peteet and Mann, 1994; Hu et al., 1995, 2002; Brubaker et al., 2001; Hu and Shemesh, 2003). Briner et al. (2002) reported evidence for a probable YD-age glacier readvance at Waskey Lake, 50 km south of Little Swift Lake. It is possible that there was a similar readvance above Little Swift Lake, but we cannot evaluate this possibility with available data. It does appear that Little Swift Lake experienced warming at the end of the YD comparable to that at lower-elevation sites in southwestern Alaska.
The dramatic expansion of Poaceae at this site and others in southwestern Alaska after the YD (Hu et al., 1995, 2002; Brubaker et al., 2001) is difficult to interpret because Poaceae are ecologically diverse. We note that post-YD Poaceae expansion was most dramatic at Little Swift, Ongivinuk, Nimgun, and Idavain lakes, which are distinguished from other sites in that they receive most of their moisture from Bristol Bay rather than the Gulf of Alaska. One explanation is that Poaceae expansion resulted from aridity in the Bristol Bay region during post-YD warming. The postglacial transgression of the shallow Bering Shelf was not yet complete at the end of the YD (Fairbanks, 1989; Elias et al., 1996), and all of these sites were significantly more continental than at present. The Bristol Bay shoreline probably transgressed to a near-modern location between 10,000 and 9000 cal yr B.P. (Manley, 2002), coincident with the decline of Poaceae at these sites.
POSTGLACIAL TREE AND SHRUB EXPANSION
Pollen records from southwestern Alaska record the gradual postglacial expansion of Betula, Alnus, and Picea to their modern ranges (e.g., Ager, 1983; Hu et al., 1995, 1996, 2002; Brubaker et al., 2001). Betula was already abundant at the beginning of the Little Swift Lake record ∼12,200 cal yr B.P., consistent with the establishment of Betula shrub tundra by ∼13,600 cal yr B.P. at Nimgun Lake (Hu et al., 2002) and elsewhere in the Ahklun Mountains (Brubaker et al., 2001; Hu et al., 2002). Alnus rapidly increased from less than 5% of the pollen rain at Little Swift Lake to more than 25% ∼7600 cal yr B.P., roughly coincident with the rapid regional expansion of Alnus at many sites throughout southwestern (Hu et al., 1996; Brubaker et al., 2001), central (Bigelow and Edwards, 2001), and northern Alaska (Anderson and Brubaker, 1994). Middle and late Holocene Alnus percentages were 10 to 25% lower than the modern value of 44%.
Picea expanded southwestward across Alaska for several thousand years after deglaciation, apparently reaching southwestern Alaska after other parts of eastern Beringia. The Ahklun Mountains region may have been one of the last areas to be colonized. Picea did not become important at Grandfather Lake until after ∼4500 cal yr B.P. (nearly 5000 yr after Picea colonized the area around Farewell Lake ∼400 km away in the northwestern Alaska Range) and reached modern abundance at Grandfather Lake ∼2000 cal yr B.P. (Hu et al., 1995; Brubaker et al., 2001). Windblown Picea pollen percentages at Little Swift Lake reached modern values ∼3500 cal yr B.P. Picea groves were probably never much closer to the lake than at present, given that the highest Holocene percentages of Picea pollen are less than 1% higher than modern values.
HOLOCENE ENVIRONMENTAL CHANGE
After ∼9800 cal yr B.P., Little Swift Lake experienced an early-Holocene period of warm, moist conditions. Abundant Betula shrubs, Sanguisorba, and Polypodiaceae thrived on land, and high terrestrial and aquatic production and/or stable soils resulted in the deposition of organic-rich lake sediments. Betula pollen percentages reached a Holocene maximum of 44% at ∼8500 cal yr B.P. Early Holocene warmth at Little Swift Lake coincides with peak warmth ∼9500 to 8900 cal yr B.P. at Farewell Lake ∼400 km to the northeast, as inferred from ostracode trace-element geochemistry (Hu et al., 1998). Farewell Lake experienced a coincident peak in Betula abundance (Hu et al., 1996).
Warmth presumably resulted from the tail end of a peak in summer insolation (Berger and Loutre, 1991) at high northern latitudes combined with a weakening of the Laurentide glacial anticyclone (e.g., Heusser et al., 1985; Bartlein et al., 1991, 1998). As Bering Sea sea-surface temperatures warmed, sea-ice extent was reduced (Sancetta et al., 1984), decreasing regional albedo and increasing moisture availability. The primary moisture source was also more proximal to the Ahklun Mountains and Little Swift Lake because the Bering Sea had transgressed much of its shelf by this time.
The end of the thermal maximum at Little Swift Lake was punctuated by a brief low in Betula, accompanied by relative increases in Artemisia, Salix, Sanguisorba, and Polypodiaceae, and decreases in Ericales and Pediastrum. These changes, which occurred ∼8000 cal yr B.P. and just before the local expansion of Alnus, may have resulted from cooling during a period of relatively moist climate. Previous researchers have suggested that the rapid regional expansion of Alnus resulted from increased effective moisture (possibly more winter snowfall) associated with climatic cooling (Anderson and Brubaker, 1994; Hu et al., 1995); and our pollen data, which suggest a cool, moist climate immediately preceding the expansion of Alnus, support this inference. Effective moisture continued to increase at other sites until ∼7000 cal yr B.P. (Hu et al., 1998; Edwards et al., 2001). It is possible that a small but distinct upvalley readvance moraine with a mean cosmogenic surface exposure age of ∼5500 cal yr B.P. (Briner et al., 1999; Axford, 2000) records a glacier advance that was driven by mid-Holocene cooling and increased effective moisture.
Widespread Neoglacial glacier advances were initiated after ∼4000 cal yr B.P. at sites throughout Alaska (e.g., Calkin, 1988; Calkin et al., 1998). In the Ahklun Mountains, glaciers in the Waskey Lake valley were reactivated 3100 cal yr B.P. (Levy et al., 2003). Although proxy data from Little Swift Lake do not necessarily suggest Neoglacial advances, the subtle upcore decrease in organic content and coarsening of sediment grain size may record late Holocene climatic deterioration. Glaciofluvial sediment from limited Holocene glacier advances probably would have been trapped by the string of ponds upstream of Little Swift Lake, so we cannot rule out the possibility that glaciers expanded at Little Swift Lake during this time. Alaskan pollen records generally offer little evidence for late Holocene climate changes, partly because it is difficult to resolve subtle changes in tundra composition. CONISS analysis identifies a change in the Little Swift Lake pollen assemblage ∼4800 cal yr B.P., with lower percentages of Betula and monolete spores, and more abundant Cyperaceae, possibly hinting at cooler temperatures in the late Holocene.
Our results indicate that both glaciers and alpine vegetation in southwestern Alaska responded to climatic changes at the end of the YD chronozone, a conclusion consistent with previous work. Paleoclimate records reveal a complex pattern of climate changes across Alaska during late glacial time. Southeastern Alaska apparently became cooler and wetter during the YD (e.g., Engstrom et al., 1990; Mathewes et al., 1993; Patterson et al., 1995). In contrast, southwestern Alaska experienced a reversal to cold, dry conditions (e.g., Peteet and Mann, 1994; Hu et al., 1995, 2002). The spatial heterogeneity of responses to probable North Pacific cooling suggests that local- and regional-scale factors, such as proximity to the Bering Sea, must have modulated responses to larger-scale, late glacial climate forcing. Evidence for temporal and spatial variability of moisture availability throughout the late glacial and Holocene is a reminder that predictions of future conditions in the Arctic should take into account the effects of local- and regional-scale influences on moisture balance.
We thank Feng Sheng Hu (University of Illinois) and R. Scott Anderson, Susan Smith, and Renata Brunner-Jass (Northern Arizona University) for advice regarding pollen analysis; Jason Briner (University of Colorado) for comments on this manuscript and assistance in the field; Kathleen Carey, Anna Jones, and Amy Moscrip for help in the field and laboratory; Al Werner (Mt. Holyoke College) for lending coring equipment; Irene Gregory-Eaves (Queens University) for analyzing diatoms; Jocelyn Turnbull (NSRL) for 14C sample preparation; and anonymous reviewers for helpful feedback on this manuscript. Togiak and Yukon Delta National Wildlife Refuges provided logistical support and access to field sites. This research was supported by NSF grants OPP-9529940 and EAR-9808593 (to Kaufman), a Geological Society of America (GSA) student research grant, a GSA Howard Award, a Mentor Grant from ARCO, and Utah State University's Eccles Fellowship (to Axford). This paper is NSF PALE/PARCS contribution No. 216.