Coupled impacts of sea ice variability and North Pacific atmospheric circulation on Holocene hydroclimate in Arctic Alaska

Arctic Alaska lies at a climatological crossroads between the Arctic and North Pacific Oceans. The modern hydroclimate of the region is responding to rapidly diminishing sea ice, driven in part by changes in heat flux from the North Pacific. Paleoclimate reconstructions have improved our knowledge of Alaska ’ s hydroclimate, but no studies have examined Holocene sea ice, moisture, and ocean − atmosphere circulation in Arctic Alaska, limiting our understanding of the relationship between these phenomena in the past. Here we present a sedimentary diatom assemblage and diatom isotope dataset from Schrader Pond, located ∼ 80 km from the Arctic Ocean, which we interpret alongside synthesized re- gional records of Holocene hydroclimate and sea ice reduction scenarios modeled by the Hadley Centre Coupled Model Version 3 (HadCM3). The paleodata synthesis and model simulations suggest the Early and Middle Holocene in Arctic Alaska were characterized by less sea ice, a greater contribution of isotopically heavy Arctic-derived moisture, and wetter climate. In the Late Holocene, sea ice expanded and regional climate became drier. This climatic transition is coincident with a documented shift in North Pacific circula- tion involving the Aleutian Low at ∼ 4 ka, suggesting a Holocene teleconnection between the North Pacific and Arctic. The HadCM3 simulations reveal that reduced sea ice leads to a strengthened Aleutian Low shifted west, potentially increasing transport of warm North Pacific water to the Arctic through the Bering Strait. Our findings demonstrate the interconnectedness of the Arctic and North Pacific on multimillennial timescales, and are consistent with future projections of less sea ice and more precipitation in Arctic Alaska. both datasets coupled possible multimillennial patterns among these other datasets. Potential teleconnections during of and systems longer timescales, help predictions their contin- ued coevolution future. Here we present Holocene diatom assemblage and oxygen isotope ( δ 18 O diatom ) datasets from Arctic Alaska, which we interpret in terms of past hydroclimatic change. Our results show that Holocene variability in δ 18 O diatom at Schrader Pond (SP) in the northeastern Brooks Range was driven by changes in moisture source associated with fluctuating Arctic sea ice extent. We also present a data − model comparison, featuring a synthesis of Holocene hydroclimate and sea ice reconstructions from regional terrestrial and marine sites, together with coupled atmosphere − ocean model simulations, which supports our interpretation of δ 18 O diatom variability. Our data highlight a prominent shift in terrestrial hydroclimate and sea ice in the region, concomitant with a well-documented shift in North Pacific hydroclimate at ∼ 4 ka (24). The timing of these near-synchronous shifts suggests an Arctic − Pacific teleconnection has been present over the Middle to Late Holocene, emphasizing the important role of both sea ice and lower-latitude ocean − atmosphere dynamics in the past and future of the Arctic.

Arctic Alaska lies at a climatological crossroads between the Arctic and North Pacific Oceans. The modern hydroclimate of the region is responding to rapidly diminishing sea ice, driven in part by changes in heat flux from the North Pacific. Paleoclimate reconstructions have improved our knowledge of Alaska's hydroclimate, but no studies have examined Holocene sea ice, moisture, and ocean−atmosphere circulation in Arctic Alaska, limiting our understanding of the relationship between these phenomena in the past. Here we present a sedimentary diatom assemblage and diatom isotope dataset from Schrader Pond, located ∼80 km from the Arctic Ocean, which we interpret alongside synthesized regional records of Holocene hydroclimate and sea ice reduction scenarios modeled by the Hadley Centre Coupled Model Version 3 (HadCM3). The paleodata synthesis and model simulations suggest the Early and Middle Holocene in Arctic Alaska were characterized by less sea ice, a greater contribution of isotopically heavy Arcticderived moisture, and wetter climate. In the Late Holocene, sea ice expanded and regional climate became drier. This climatic transition is coincident with a documented shift in North Pacific circulation involving the Aleutian Low at ∼4 ka, suggesting a Holocene teleconnection between the North Pacific and Arctic. The HadCM3 simulations reveal that reduced sea ice leads to a strengthened Aleutian Low shifted west, potentially increasing transport of warm North Pacific water to the Arctic through the Bering Strait. Our findings demonstrate the interconnectedness of the Arctic and North Pacific on multimillennial timescales, and are consistent with future projections of less sea ice and more precipitation in Arctic Alaska.
Arctic sea ice | Alaska hydroclimate | Aleutian Low | oxygen isotopes | Holocene R apidly rising Arctic air and sea surface temperatures have resulted in the reduced annual duration and extent of Arctic sea ice (1), which in turn drives the ice−albedo feedback leading to amplified warming in the Arctic (2). These reductions in sea ice are projected to continue in future decades (3) and have important implications for Arctic terrestrial hydroclimate, as sea ice extent and duration impact the seasonality, type, and amount of precipitation in this region (4). Recent studies have also suggested teleconnections between the extent and duration of Arctic sea ice and midlatitudinal storm tracks (5,6), as well as synoptic-scale processes involving the Aleutian Low atmospheric pressure cell (AL) (7,8) and ocean−atmosphere circulation in the Bering Strait (9)(10)(11), which might link North Pacific hydroclimate directly to changes in Arctic sea ice. While recent observations show the influence of North Pacific climate on Arctic sea ice, little is known about their long-term dynamics or their coupled influence on hydroclimate in the western Arctic.
Our understanding of past hydroclimate in Arctic Alaska is based in part on stable isotope reconstructions that reflect changes in the oxygen (δ 18 O) and hydrogen (δD) isotope composition of water. δ 18 O has proven particularly useful for studying both current (12,13) and past (14)(15)(16)(17)(18)(19)(20) hydroclimate in the region, because it is sensitive to climate and environmental variables. As a result, δ 18 O has been used as a paleoclimate proxy for precipitation source (16), effective moisture (14), and temperature (20) in Arctic Alaska. Interpretations of these paleoclimate datasets have considered the impact of Holocene changes in AL variability (15,16,18), but they have not been used to examine the influence of Holocene Arctic sea ice variability on western Arctic climate, despite well-established sea ice conditions for this time period (e.g., ref. 21). The influence of sea ice extent on δ 18 O in various climate archives has been demonstrated in Arctic Alaska during the Pleistocene−Holocene transition (19), as well as in Greenland during the Holocene (22) and the Last Interglacial period (LIG) (23), suggesting that sites adjacent to seasonally ice-free Arctic waters can be sensitive recorders of sea ice conditions.

Significance
Our new record of hydroclimatic change over the past ∼10,000 years in Arctic Alaska reveals that periods of reduced sea ice result in isotopically heavier precipitation derived from proximal, Arctic moisture sources. This systematic relationship is supported by isotope-enabled model simulations and a compilation of regional paleoclimate records. Periods of sea ice reduction also correspond to regionally wetter conditions. A transition from generally wetter conditions (∼10,000 y to 4,000 y ago) to relatively drier conditions (∼4,000 y ago to present) appears related to a recognized shift in North Pacific circulation involving the Aleutian Low atmospheric pressure cell, where less sea ice is associated with strengthened low pressure shifted to the west.
In light of increasing evidence from both data and models for a modern connection between North Pacific circulation and Arctic sea ice (5)(6)(7)(8), as well as the demonstrated influence of North Pacific (15,16,18) and Arctic (13,19) ocean−atmosphere systems on past and present terrestrial hydroclimate conditions, it appears that northern Alaska lies at a climatological crossroads within the western Arctic. This means that paleoclimate records from Arctic Alaska are especially well situated for studying the effects of both changing Arctic sea ice and North Pacific circulation. However, existing paleoclimate datasets from this region have not been interpreted in the context of such a coupled system, and little has been done to synthesize possible multimillennial patterns among these and other datasets. Potential teleconnections during the Holocene must be explored, because this paleoclimate context is important for understanding the coevolution of Arctic and Pacific hydroclimate systems on longer timescales, which could help clarify predictions of their continued coevolution in the future.
Here we present Holocene diatom assemblage and oxygen isotope (δ 18 O diatom ) datasets from Arctic Alaska, which we interpret in terms of past hydroclimatic change. Our results show that Holocene variability in δ 18 O diatom at Schrader Pond (SP) in the northeastern Brooks Range was driven by changes in moisture source associated with fluctuating Arctic sea ice extent. We also present a data−model comparison, featuring a synthesis of Holocene hydroclimate and sea ice reconstructions from regional terrestrial and marine sites, together with coupled atmosphere−ocean model simulations, which supports our interpretation of δ 18 O diatom variability. Our data highlight a prominent shift in terrestrial hydroclimate and sea ice in the region, concomitant with a well-documented shift in North Pacific hydroclimate at ∼4 ka (24). The timing of these near-synchronous shifts suggests an Arctic−Pacific teleconnection has been present over the Middle to Late Holocene, emphasizing the important role of both sea ice and lower-latitude ocean−atmosphere dynamics in the past and future of the Arctic.

Climatic Setting and Controls on δ 18 O diatom
The data presented in this study are from SP (69.36°N, 145.08°W; 869 m above sea level), a deep (18.5 m) kettle hole in Pleistocene moraine sediments (SI Appendix, Fig. S1). SP is located in the Arctic National Wildlife Refuge within the catchment of Lakes Peters and Schrader (Neruokpuk Lakes) (Fig. 1A), which has been the focus of a recent 4-y environmental monitoring effort (25). Climate in Arctic Alaska, including the SP region, is highly seasonal, with the majority of annual precipitation falling in the summer months (SI Appendix, Fig. S2 and Table S1). There is also a strong seasonal pattern in the oxygen isotope composition of precipitation (δ 18 O precip ), with higher δ 18 O precip values during the summer and lower values in the winter (SI Appendix, Fig. S2 and Table S1). Storms arriving at SP are derived from either proximal Arctic moisture sources, such as the Beaufort and Chukchi Seas (hereafter classified as "Arctic"), or from the more distal Bering Sea or Gulf of Alaska (hereafter classified as "Pacific") ( Fig. 1B).
To characterize the isotope composition of meteoric water arriving at SP, precipitation samples (n = 28) were collected throughout the Lake Peters/Schrader catchment and analyzed for δ 18 O and δD (26). Precipitation source waters in the Arctic oceans are cold and have lower δ 18 O (-2.9‰ at 71.5°N, 150.5°W in the Beaufort Sea) compared to seawater in the North Pacific (-0.8‰ at 55.5°N, 150.5°W in the Gulf of Alaska) (27) (Fig. 1C). However, back-trajectory modeling (Fig. 1C) shows storms that travel from an Arctic source generally have higher δ 18 O (μ = -17.4‰ ± 4.2, n = 15) compared to storms arriving from the Pacific (μ = -21.4‰ ± 4.9, n = 13) (SI Appendix, Fig. S3B; T = 2.3, P = <0.03), a phenomenon that has also been documented in Utqia _ gvik, ∼500 km northwest of SP on the Arctic Alaskan coast (28). Arctic-derived moisture arriving at SP likely has higher δ 18 O due to the shorter distillation path of storms traveling from Arctic sources, which results in less rainout of 18 O, and ultimately higher δ 18 O precip ( Fig. 1D and SI Appendix, Fig. S3B). This effect has been previously proposed to explain changes in δ 18 O of wood cellulose from Arctic Alaska during the Younger Dryas-Early Holocene transition (19), and in glacier ice from Greenland during the LIG (23). This distillation path effect, along with changes in condensation temperature, is likely also responsible for the pronounced seasonal change in δ 18 O precip : The retreat of sea ice in the warmer months exposes proximal Arctic Ocean surface water that fuels summer moisture, with warmer air temperatures also promoting higher δ 18 O precip values (29). δ 18 O of SP water (δ 18 O pond ) plots along the local meteoric water line (slope = 6.74; SI Appendix, Fig. S3D), indicating that δ 18 O pond reflects changes in δ 18 O precip and is largely unaffected by evaporation (SI Appendix). SP also has an outflow and sits within unconsolidated moraine sediments that likely allow for high groundwater throughflow, suggesting the residence time of water in the pond is relatively short, and supporting the notion that δ 18 O pond changes primarily in accordance with shifts in δ 18 O precip (30). Given the controls on δ 18 O precip outlined above, a primary influence on δ 18 O pond is the source of moisture (Arctic versus Pacific) arriving at SP.
The oxygen isotope composition of diatoms (δ 18 O diatom ) is controlled by the δ 18 O of water in which they form their silicabased frustules, a process thought to occur in isotopic equilibrium (31). The temperature-dependent fractionation between diatom silica and lake water is relatively small (−0.2‰/°C) (32), meaning it is often subsumed by larger fluctuations in δ 18 O pond . Large fluctuations in δ 18 O pond can occur as a result of changes in the balance of precipitation inputs and evaporative losses (P/E), or because of changes in δ 18 O precip (30). At SP, our data suggest that the main control on δ 18 O pond is δ 18 O precip , which in turn is influenced by the proportion of proximal, Arctic-sourced moisture, where less sea ice results in higher δ 18 O precip . Therefore, we interpret variability in δ 18 O diatom to reflect changing moisture source at SP, similar to other paleoclimate datasets from archives proximal to seasonally open Arctic waters (19,23).
To assess the relationship between sea ice and δ 18 O precip at SP established through modern isotope observations, we use a series of previously published experiments performed with the isotopeenabled version of the Hadley Centre Coupled Model Version 3 (HadCM3), a coupled atmosphere−ocean model (23). These experiments simulate the effect of different sea ice reduction scenarios on δ 18 O precip in the Arctic during the LIG, 125,000 y before present. Because they simulate changes in sea ice during the LIG, the reduction scenario experiments have different orbital and greenhouse gas boundary conditions compared to the preindustrial period (PI) (23) (SI Appendix, Table S2). While isotope-enabled simulations of the relatively warm Middle Holocene are available for comparison with the PI (e.g., ref. 33), they do not isolate the influence of Arctic sea ice reduction on the climate system; the HadCM3 LIG simulations are the only existing model experiments that do this. Despite the difference in time period and the exaggerated sea ice reduction compared to inferred Early and Middle Holocene conditions, these LIG simulations allow us to test and explore the influence of sea ice on δ 18 O precip and other climate variables in our observational datasets, and evaluate the sensitivity of Arctic hydroclimate. Furthermore, we are able to examine the difference between the PI and LIG control simulations, allowing us to isolate the influence of LIG orbital and greenhouse gas forcings.
To compare our observational isotope data with the simulations, we use two modeled scenarios: low and high reduction in sea ice extent (SI Appendix, Table S2). We also compare the LIG and PI control simulations, as sea ice is minimally reduced during the LIG compared to the PI (SI Appendix, Table S2). In these simulations, sea ice concentration is reduced throughout the Arctic Ocean, including in the Beaufort Sea (23). The model output shows that reduced sea ice results in higher δ 18 O precip at SP and across Arctic Alaska compared to the PI control simulation ( Fig. 2 A-C and SI Appendix, Fig. S4 and Table S3). The model experiments mimic the trend observed in our measured δ 18 O precip data, and support the conclusions drawn from our modern and paleo oxygen isotope datasets. This relationship between reduced sea ice and δ 18 O precip is also evident when comparing Middle Holocene (6 ka) and PI simulations from the isotope-enabled version of the Max Planck Institute Earth System Model (MPI-ESM) (33), although the positive δ 18 O precip anomalies are less pronounced than those in the LIG sea ice reduction experiments (SI Appendix, Table S3).
Influence of Arctic Sea Ice on Regional Holocene Hydroclimate The sedimentary δ 18 O diatom dataset from SP covers the last ∼9.2 ka, and has a range of 8.5‰ (+19.3 to +27.8‰, n = 42; Fig. 3A). δ 18 O diatom is generally higher in the Early and Middle Holocene (9.2 ka to 4 ka; μ = +25.4‰, n = 20) compared to the Late Holocene (4 ka to present; μ = +23.4‰, n = 22). Given our observational and model-based evidence for the influence of changes in sea ice, moisture source, and δ 18 O precip on δ 18 O diatom , this suggests that Arctic sea ice was reduced during the Early and Middle Holocene, compared to the Late Holocene when it became more persistent. This first-order interpretation of the expansion of Arctic sea ice through the Holocene is consistent with previously published syntheses of sea ice reconstructions from data and models (34), and with proxy datasets from this region based on the biomarker index PIP 25 , which uses the sea ice proxy IP 25 and phytoplankton biomarkers to reconstruct past sea ice conditions (35). PIP 25 records indicate sea ice persisted for more months of the year during the Late Holocene in the Chukchi and eastern Siberian Seas (21) (Fig. 3 B and C), relative to the Early and Middle Holocene. The SP δ 18 O diatom dataset is also weakly negatively correlated (r = -0.32; P = 0.04) with the most proximal of these records, marine core ARA2B-1 from the Chukchi Sea (21) (Fig. 3 A and B and SI Appendix, Fig. S5), potentially further supporting the Holocene relation between sea ice extent, δ 18 O precip , and δ 18 O diatom at SP. Sea ice fluctuations may have also influenced the water isotopes at other lakes in the Brooks Range. For example, nearby Wahoo Lake/shelf (20) (Fig. 3D), Meli Lake (14) (Fig. 3E), and Tangled Up Lake (14) (Fig. 3F) are hydrologically open lakes that show a shift in inferred δ 18 O lake from higher values during the Early and Middle Holocene to lower values during the Late Holocene. These studies do not highlight the possible influence of δ 18 O precip or sea ice extent, focusing instead on the contributions of temperature and P/E. However, the influence of δ 18 O precip cannot be ruled out, due to the hydrologic setting of these lakes (30), and in light of the nearby proxy datasets that have been interpreted to reflect changes in δ 18 O precip (16,18,19).
The diatom flora from SP indicate that proximal sea ice conditions also influenced Holocene lake ice dynamics. Reduced ice cover in Arctic lakes results in enhanced light penetration, and encourages deeper mixing and nutrient cycling, all of which   creates a more favorable habitat for planktonic diatoms that dwell in the water column, and often results in an increase of the proportion of planktonic (compared to benthic) diatoms (36). Therefore, at SP, an increase in the proportion of planktonic diatoms is likely indicative of reduced lake ice (SI Appendix). The SP diatom assemblage (SI Appendix, Fig. S6) is dominated by planktonic species, although the proportion of planktonic diatoms decreases in the Late Holocene, most notably in the youngest sediments (Fig. 3G). This suggests that the ice-free season was longer in the Early and Middle Holocene, compared to the Late Holocene when the ice-free season shortened. The same pattern arises from an analysis of two of the most abundant planktonic diatom taxa in the dataset (SI Appendix, Fig. S6), which have different habitat preferences: Lindavia intermedia, which prefers a deeper mixing depth (37), and diatoms in the Lindavia rossii−comensis−tripartita complex (38) (hereafter "L. rossii complex"), which prefer shallower mixing depths (39). The abundance ratio of the L. rossii complex to L. intermedia therefore may reflect changes in mixing depth throughout the Holocene, where a lower (higher) index indicates conditions when the lake is less (more) persistently ice covered. A Late Holocene decrease in both this ratio (Fig. 3H) and the proportion of planktonic diatoms is broadly consistent with that of inferred proximal sea ice extent, suggesting that reduced (increased) sea ice is accompanied by a longer (shorter) ice-free season at SP, further demonstrating the connection between marine and terrestrial climate conditions. These changes in lake ice cover, sea ice, and inferred δ 18 O precip from SP and proximal sites are also evident in the first principal components axis (PC1) of these datasets (Fig. 3I), which explains 87% of the variance among these datasets.  Several terrestrial and marine Holocene hydroclimate proxy datasets from Arctic Alaska reveal that changes in sea ice dynamics were also accompanied by a relatively wetter Early and Middle Holocene compared to a drier Late Holocene (Fig. 3 J-Q). At neighboring Lakes Peters and Schrader, higher sedimentary organic content is interpreted to reflect warmer conditions, which would reduce the extent of local mountain glaciers and decrease associated minerogenic input, and/or wetter conditions, which would increase allochthonous organic input (40). In several sediment sequences from Lakes Peters and Schrader, organic content is generally higher in the Early and Middle Holocene compared to the Late Holocene, most prominently in the oldest sediment sequence from Lake Peters (Fig. 3J) (40). This transition is also evident in Holocene δ 18 O datasets from lakes where P/E is a major control on δ 18 O lake . In these instances, higher δ 18 O represents increased evaporation and generally drier conditions. These records include δ 18 O chironomid from Qalluuraq Lake (17) (Fig. 3K) and δ 18 O Chara from Takahula Lake (15) (Fig. 3L), both of which indicate a wetter Early and Middle Holocene, and a drier Late Holocene.
The evolution of Hf and Nd isotope compositions (eHf and eNd) from marine sediments in the Beaufort (Fig. 3 M and N) and Chukchi Seas (Fig. 3 O and P) indicates that much of northern Alaska and northern Yukon were wetter during the Early and Middle Holocene, as evidenced by inferred higher Mackenzie River discharge, compared to the Late Holocene (41). Warmer, wetter summers during the Early Holocene have also been inferred at several sites in Arctic Alaska (42)(43)(44)(45), as has relatively warm Early Holocene July temperature (46), while increases in the extent of mountain glaciers during the Late Holocene point to cooler summer temperatures (47,48). These datasets suggest the increased precipitation of the Early and Middle Holocene might be associated with summer season moisture. This transition from a wetter Early and Middle Holocene to a drier Late Holocene is also evident in the first PC representing these datasets (Fig. 3Q), which explains 91% of the variance among the records. The relation between reduced sea ice extent and increased precipitation during the Early and Middle Holocene is consistent with HadCM3 output, which also shows wetter conditions with reduced sea ice (Fig. 2 D-F and SI Appendix, Table S3).

Implications for Arctic and North Pacific Ocean−Atmosphere Circulation
Our results from SP and the regional compilation of Arctic Alaska datasets reveal a coherent shift, from reduced sea ice and more precipitation in the Early and Middle Holocene to more sea ice and less precipitation in the Late Holocene (Fig. 3). This transition from wetter conditions (less sea ice) to drier conditions (more sea ice) in Arctic Alaska coincides with a previously documented shift at ∼4 ka in terrestrial hydroclimate throughout the northeastern Pacific continental margin (24, 49-54) (Fig. 3R). This shift has been associated with a strengthened AL in the Late Holocene (24,(49)(50)(51)(52)(53)(54) (Fig. 3R), which would have increased moisture flow from the North Pacific to southern Alaska and southwestern Yukon (55), and even as far north as Arctic Alaska (18). Evidence for increased Mackenzie River discharge from Beaufort and Chukchi Sea marine sediment sequences has previously been related to this shift in the AL (41).
The impact of the AL on terrestrial hydroclimate is most apparent during the winter (December−February [DJF]) months, when the pressure cell strengthens and influences the trajectory of North Pacific storms (55). In Arctic Alaska, the majority of winter precipitation originates from Pacific sources, as the Arctic ocean waters are frozen. Therefore, lower δ 18 O precip values in the Late Holocene would be associated with a strengthened AL and an increased contribution of winter moisture from the North Pacific. At some sites, the opposite relationship has been documented, where a stronger AL delivers moisture with higher δ 18 O to Arctic Alaska (18). This relationship may be present on a storm-by-storm basis, where AL variability alters δ 18 O precip of storms traveling from the North Pacific. In contrast, a multimillennial-scale shift from the dominance of Arctic versus North Pacific moisture would not capture this storm-specific variability, and would instead be characterized by the effect of Rayleigh distillation along the air mass path, described in this and other studies (19,23). An increased contribution of winter precipitation, occurring at lower air temperatures, would also lower δ 18 O precip values (29), compounding the distillation effect.
The contemporaneous occurrence of a strengthened AL and an increase in Arctic sea ice at ∼4 ka implies the presence of a teleconnection between the North Pacific and western Arctic during the Holocene. Such Pacific−Arctic teleconnections, which have the potential to drive climatic changes in both Arctic Alaska and lower latitudes, have been suggested in numerous studies focused on the 20th to 21st centuries and future climate (6,8,(56)(57)(58)(59)(60)(61). Several of these studies associate reduced Arctic sea ice with an intensified and northward-shifted AL (56-59), while others associate it with a weakened and/or westwardshifted AL (6,8,60,61). When the AL strengthens, the lowpressure center is often located near the Gulf of Alaska, whereas the pressure center of a weaker AL is generally located farther west, over the Aleutian Island chain (55). While often discussed together, the position and intensity of the AL are not unilaterally related; the pressure center associated with the AL may intensify or diminish in either its easterly or westerly position (62).
In HadCM3 simulations, the progressive depletion of Arctic sea ice causes the AL to intensify and expand, with the center of low pressure shifting westward (Fig. 4 A-C). These modeled changes in sea level pressure (SLP) are accompanied by increases in precipitation of higher magnitude during the summer (June−August [JJA]; Fig. 4 G-I) than the winter (DJF; Fig. 4 D-F) months. This suggests the additional moisture accompanying reduced sea ice cannot be attributed solely to an increase in storm activity associated with an intensified AL, which dominantly influences winter season precipitation. The additional moisture more likely results from a teleconnection between the North Pacific low-pressure system and Arctic sea ice, which increases Arctic moisture throughout the year (and most prominently in the summer). One possible mechanism for this feedback could be the Bering Strait inflow (BSI), another important feature of regional ocean−atmosphere circulation that has previously been related to the behavior of the AL (11). Specifically, a westward-shifted AL can cause a divergence in Ekman transport over the Aleutian Basin, resulting in increased sea surface heights on the eastern Bering shelf and a strengthened BSI (11). When strengthened, the BSI increases heat flux into the Arctic from the North Pacific, reducing sea ice (63,64). Therefore, the modeled westward shift of the AL could promote a reduction in Arctic sea ice as a result of the transport of relatively warm North Pacific waters into the Chukchi and Beaufort Seas via the BSI. Even though this mechanism is unlikely to fully explain the relationship between sea ice and the AL in the model simulations, as the BSI is not fully resolved in HadCM3 (65), it might explain the connection between less sea ice and a westward and/or weaker AL following the formation of the Bering Strait in its nearly modern form during the Early and Middle Holocene (9,66). The simulated response of the AL to sea ice reduction might also result from localized feedbacks, where sea ice loss results in vertical diffusion of upward turbulent heat fluxes, decreasing local SLP (57), consequently strengthening the AL.
The modeled deepening AL with decreased sea ice is consistent with projections of AL behavior under increased global warming, although these projections generally show that the low-pressure center also shifts to the north (56)(57)(58)(59). While the center of intense low pressure does not appear to migrate northward in the simulations used in this study (Fig. 4 A-C), negative SLP anomalies over Alaska, the Bering Sea, and the Arctic Ocean do indicate generally lower SLP farther north than is characteristic of present conditions. This decrease in high-latitude SLP is also consistent with a weakening of the Northern Hemisphere temperature gradient, as has been previously proposed for the Early Holocene and a warmer world (5). This connection between North Pacific SLP and the hemispheric temperature gradient might explain the increase in moisture at these higher latitudes (56,59), driven by the northward spread of low pressure, as midlatitudinal moisture decreases (5). It is therefore possible the connection between the AL and Arctic sea ice is related not only to future increases in Arctic moisture but also to future decreases in midlatitudinal moisture, as both these features of Northern Hemisphere hydroclimate have been demonstrated during periods of Holocene sea ice decline.
Geochronology. The core chronology for SCP16-2A was established using nine accelerator mass spectrometry (AMS) 14 C dates on terrestrial and aquatic plant and insect macrofossils (SI Appendix, Fig. S8 and Table S4). Sediment samples 1 cm thick were sieved at 150 μm, and macrofossils were picked, dried in an oven at 50°C, and identified under a Zeiss light microscope. Macrofossils were prepared and converted to graphite at NAU, and 14 C content was measured at the Keck-Carbon Cycle AMS facility at University of California, Irvine (UC Irvine). Uncalibrated 14 C dates were incorporated into an age model for the sediment sequence using the R package Bacon (v2.2) (68). Diatom assemblage. Sixteen 1-cm-thick samples from SCP16-2A were selected at 3-to 6-cm increments for diatom species analysis. These samples were treated with 30% H 2 O 2 and 70% HNO 3 to remove organic matter before creating slides for counting using Naphrax© mounting medium. A Zeiss light microscope was used to count 300 valves per slide along transects at 1,000×, and taxonomic classifications were made (SI Appendix, Fig. S6). Diatom oxygen isotopes. Samples for diatom oxygen isotope analysis (1 cm thick) were taken every 1 cm to 3 cm of core SCP16-2A (n = 42). Sampling avoided inorganic layers containing too few diatoms to purify. Purification consisted of a series of chemical digestions, sieving, and heavy liquid separations (69). All samples were visually inspected for contamination under a Zeiss light microscope, and 24 samples were analyzed with a Zeiss Supra 40VP variable pressure field emission scanning electron microscope. Both visual inspection and energy-dispersive X-ray spectroscopy demonstrate that contamination of purified biogenic silica by silt and clay is insignificant (SI Appendix, Fig. S9). δ 18 O diatom values were measured using the stepwise fluorination method (70)  Model Simulations. The isotope-enabled HadCM3 (71) was used to investigate the influence of sea ice extent on precipitation isotopes and hydroclimate. These LIG simulations apply different heat fluxes (from 0 W/m 2 to 300 W/m 2 ) to the bottom of Northern Hemisphere sea ice to explore the influence of water temperature on sea ice extent and its effects on large-scale oceanatmosphere circulation (23). The model simulations and heat flux experiments are described in detail by the study for which they were originally produced (23), and are summarized in SI Appendix, Table S2.
Analysis of Paleorecords. Chronological uncertainties for the diatom assemblage and δ 18 O diatom time series from SP, as well as all paleorecords in Fig. 3, were generated in GeoChronR (72) using the ensemble output from Bacon age models (which were generated for all previously published proxy datasets). The principal component analysis (PCA) of paleorecords shown in Fig. 3 I, Q, and R was completed by compiling the data for each record into 500-y bins for 10 ka to the present, where missing values were replaced by the mean of the dataset. The PCA was then performed on the correlation matrix of untransformed data for all records in each category [moisture source versus wetter/drier (73)]. For Fig. 3R, the PCA was performed on z-score values for the five included records (49)(50)(51)(52)(53).
Data Availability. The following datasets from a ∼10,000 y old sediment core will be deposited in Neotoma Paleoecology Database (https://www.neotomadb. org/): δ 18 O diatom , diatom assemblage, biogenic silica abundance, loss on ignition, and magnetic susceptibility data. These data are also available in the Dataset S1 for this manuscript.