Oxygen isotope data from planktonic and benthic foraminifera, on a high-resolution age model (44 14C dates spanning 17,400 years), document deglacial environmental change on the southeast Alaska margin (59°33.32′N, 144°9.21′W, 682 m water depth). Surface freshening (i.e., δ18O reduction of 0.8‰) began at 16,650 ± 170 cal years B.P. during an interval of ice proximal sedimentation, likely due to freshwater input from melting glaciers. A sharp transition to laminated hemipelagic sediments constrains retreat of regional outlet glaciers onto land circa 14,790 ± 380 cal years B.P. Abrupt warming and/or freshening of the surface ocean (i.e., additional δ18O reduction of 0.9‰) coincides with the Bølling Interstade of northern Europe and Greenland. Cooling and/or higher salinities returned during the Allerød interval, coincident with the Antarctic Cold Reversal, and continue until 11,740 ± 200 cal years B.P., when onset of warming coincides with the end of the Younger Dryas. An abrupt 1‰ reduction in benthic δ18O at 14,250 ± 290 cal years B.P. likely reflects a decrease in bottom water salinity driven by deep mixing of glacial meltwater, a regional megaflood event, or brine formation associated with sea ice. Two laminated opal-rich intervals record discrete episodes of high productivity during the last deglaciation. These events, precisely dated here at 14,790 ± 380 to 12,990 ± 190 cal years B.P. and 11,160 ± 130 to 10,750 ± 220 cal years B.P., likely correlate to similar features observed elsewhere on the margins of the North Pacific and are coeval with episodes of rapid sea level rise. Remobilization of iron from newly inundated continental shelves may have helped to fuel these episodes of elevated primary productivity and sedimentary anoxia.
 Here we investigate the mechanisms driving climate and environmental change in the high-latitude North Pacific by placing detailed isotopic and sedimentologic observations of upper ocean properties from the Gulf of Alaska (GoA) into a global context using a high-resolution radiocarbon-based age model. We focus on a site from the continental slope of SE Alaska, south of Kayak Island. Three cores, jumbo piston core EW0408-85JC, its trigger core EW0408-85TC (59°33.32′N, 144°9.21′W, 682 m water depth), and an adjacent multicore EW0408-84MC2 (59°33.30′N, 144°9.16′W, 682 m water depth) provide a complete, high-resolution record of the last deglacial transition through the Holocene in this region (Figure 1).
2. Study Area
2.1. Regional Oceanography
 Cyclonic motion of the subarctic gyre drives circulation in the Gulf of Alaska [Stabeno et al., 2004]. The southern boundary of this gyre, the West Wind Drift, diverges as it approaches the continental shelf of North America and its northward branch becomes the Alaska Current (AC; Figure 1). The AC dominates flow along the southwestern Alaska continental slope and is eventually diverted south around the Kenai Peninsula, beyond which it is termed the Alaskan Stream.
 Roughly parallel to the AC but largely confined to the continental shelf is the Alaska Coastal Current (ACC), a wind- and buoyancy-forced coastal jet with flow velocities occasionally in excess of 50 cm s−1 and mean annual transport of ∼106 m3 s−1 [Royer, 1982; Johnson et al., 1988; Stabeno et al., 1995]. Although primarily wind driven, the ACC is enhanced by a baroclinic response to a coastal freshwater discharge to the Gulf with an annual average >23,000 m3 s−1, equivalent to the flow of the Mississippi River [Royer, 1981, 1982]. This freshwater flux is delivered via a series of small, mountainous drainages, which experience high precipitation rates (2–6 m yr−1) due to adiabatic cooling of warm moist air associated with the cyclonic storm systems of the Aleutian Low [Weingartner et al., 2005]. Runoff peaks in the fall, and this yields an abrupt, shallow (<50 m) halocline of salinity contrast >3 over the shelf; the spring halocline is deeper (>100 m) and less abrupt (contrast of ∼1) [Stabeno et al., 2004].
 During fall, winter, and spring, strong cyclonic winds promote onshore surface Ekman transport and downwelling on the shelf, along with storm-induced vertical mixing [Childers et al., 2005]. During summer the onshore winds relax, allowing for brief periods of coastal upwelling in this dominantly downwelling system [Stabeno et al., 2004].
 Primary productivity in the Gulf of Alaska is limited by low light and deep surface mixed layers in the winter, but large algal blooms occur over the shelf with the return of the sun in the early spring. Productivity remains relatively high through early summer [Stabeno et al., 2004]. Nitrate, silicic acid, and phosphate nutrients come mostly from the subsurface ocean, via open-ocean upwelling, onshore Ekman transport, tidal pumping, and storm or eddy mixing [Childers et al., 2005]. Nutrients delivered by the fluvial system include iron and silicic acid [Stabeno et al., 2004]. Although reasonable estimates suggest that fluvial sources of reactive iron are sufficient to support shelf productivity, dissolved concentrations are limited (∼1–3 nM) by a low availability of organic ligands, required to maintain iron in solution [Lippiatt et al., 2010]. Relatively little data are available on the cycling of iron in this system, although it appears shelf processes and surface water discharge may play a role in regulating surface-ocean iron concentrations [Stabeno et al., 2004; Schroth et al., 2009; Wu et al., 2009].
 The continental shelf in SE Alaska is about 25 km wide near Kayak Island. The mean depth is 140 m above a shelf break at ∼220 m. Sediments exposed on the shelf include poorly sorted, glacially derived diamicton, glacial-marine sand and silt, and hemipelagic mud [Molnia and Carlson, 1978; Carlson, 1989].
 Core EW0408-85JC is located on the continental slope between two distinct sediment dispersal regions, the Copper River Shelf (59°30′N–60°15′N, 145°W–141°30′W) and the Bering-Malaspina Shelf (59°30′N–60°N, 147°W–145°W), and thus may be sensitive to fluctuations in terrestrial input from either source [Jaeger et al., 1998]. The Copper River Shelf includes the estuary and the delta of this large river, as well as regions north and west of Kayak Trough (∼145°W); for example, Hinchinbrook Sea Valley and Prince William Sound are both distal depocenters of Copper River sediment [Molnia, 1986; Milliman et al., 1996; Jaeger et al., 1998]. The Bering-Malaspina Shelf is mostly supplied with sediment from smaller rivers draining the Bering and Malaspina piedmont glaciers.
 In the late Pleistocene, the upper reaches of the Copper River basin were occupied by >9000 km2 ice-dammed Glacial Lake Atna [Nichols and Yehle, 1969; Williams and Galloway, 1986]. During deglaciation, failure of the ice dam is thought to have produced freshwater megafloods in the Gulf of Alaska between ∼26,000 and ∼15,500 years B.P., and perhaps additional younger events that are not yet well dated [Wiedmer et al., 2010]. The lake ultimately drained out the Copper River by 10,270–11,090 cal years B.P. [Rubin and Alexander, 1960].
 The northwestern lobe of the Cordilleran Ice Sheet was present on the shelf during the Last Glacial Maximum (LGM), although the full extent of ice cover is poorly known [Molnia, 1986; Manley and Kaufman, 2002]. As recently as the early 1900s, the tidewater Guyot Glacier in Icy Bay released ice-rafted debris directly into the sea [Molnia, 1986]. Terrestrial records suggest that regional LGM expression lasted from 23,000 to 14,700 cal years B.P., with evidence for millennial-scale cooling and transient glacial readvances during deglaciation [Engstrom et al., 1990; Mann and Peteet, 1994; Briner et al., 2002; Hu et al., 2006]. Retreat of the Bering Glacier off the continental shelf following the LGM is not well dated, but peat was accumulating in parts of the Bering foreland as early as early as 16,000 cal years B.P. [Peteet, 2007]. Timing of advances and retreats of land terminating valley and cirque glaciers in the early Holocene are also poorly constrained [Barclay et al., 2009]. Advances are estimated between 8600 and 6700 cal years B.P. in southwestern Alaska [Ten Brink, 1983; Yehle et al., 1983], and between 12,300 and 7570 cal years B.P. in the Aleutian Islands [Thorson and Hamilton, 1986]. Regional late-Holocene glacial readvances are dated at 3300–2900 cal years B.P. and 2200–2000 cal years B.P. [Barclay et al., 2009].
3.1. Sediment Depths
 The 1124 cm jumbo piston core EW0408-85JC (59°33.32′N, 144°9.21′W, 682 m water depth) was spliced to its 210 cm trigger core and to adjacent 56 cm multicore EW0408-84MC2 (59°33.30′N, 144°9.16′W, 682 m water depth) using visual correlation of point-source magnetic susceptibility and gamma attenuation density data measured at sea. The sediment-water interface was recovered in the multicore. The trigger core overpenetrated the seafloor and did not recover an interval equivalent to 0–13 cm in the multicore. The jumbo piston failed to recover the uppermost 150 cm of sediment. Based on these observations, we define the centimeters-below-seafloor (cmbsf) scale, which adds 150 cm to the nominal depths in EW0408-85JC and 13 cm to EW0408-85TC, but retains raw measured depths in EW0408-84MC2.
 Benthic and planktonic foraminifera were picked from the >150 μm sediment fraction. The planktonic species Neogloboquadrina pachyderma (sinistral) and Globigerina bulloides were analyzed for radiocarbon using accelerator mass spectrometry (AMS). Although these two species were combined for most depths to increase sample size and minimize analytical error, the two species were dated separately at 759 cmbsf. The resulting age of N. pachyderma was 65 ± 50 years older than that of G. bulloides, close to the combined measurement uncertainty for the individual species. Benthic foraminiferal 14C analyses were run as mixed species, although care was taken to avoid agglutinated and deep infaunal species such as Globobulimina affinis.
 Radiocarbon analyses were performed at the Keck AMS facility, University of California at Irvine. A total of 44 measurements were performed for 40 samples: 3 benthic/planktonic pairs from core EW0408-85TC, 34 planktonic samples from core EW0408-85JC, and three benthic samples from core EW0408-84MC2. The raw radiocarbon dates were converted to a calendar age scale with using the Marine09 curve in CALIB 6.0 [Stuiver and Reimer, 1993; Reimer et al., 2009]. To account for regional surface water reservoir ages [McNeely et al., 2006], a constant reservoir age anomaly (ΔR) of 470 ± 80 years was applied to all planktonic foraminiferal dates. A value for ΔR of 990 ± 100 years was applied to the benthic foraminiferal dates, based on water column data at equivalent water density horizons in the Gulf of Alaska [Sabine et al., 2005]. These reservoir age corrections are consistent with the 14C age differences observed in the paired benthic/planktonic dates within the trigger core. The calendar calibrated dates for the overlapping portions of the multi, trigger, and jumbo piston core form a smooth progression without age reversals on the composite cmbsf scale derived in section 3.1.
3.3. CT Scans
 Computerized tomographic (CT) density measurements were performed at the Oregon State University College of Veterinary Medicine using a Toshiba Aquilion 64 Slice. Scans were collected at 120 kVp and 200 mAs. For visualization purposes, the resulting images were processed with a “sharp” algorithm to generate sagittal and coronal images every 4mm across the core. Downcore and across-core pixel resolution within each slice is 500 μm. The cores were scanned in ∼60 cm segments and then joined into a composite image using Adobe Photoshop software (Figure 2). The pixel intensities of the resulting compilation were calibrated as a high-resolution proxy for sediment density by applying a second-order polynomial regression between the pixel values and the calibrated shipboard gamma ray attenuation density measurements (r = 0.91, n = 1122, p < 0.0001; Figure 2).
3.4. Stable Isotope Measurements
 Raw sediment subsamples of 15 cm3 (1 cm thick, quarter core) or 30 cm3 (2 cm thick, quarter core, for samples also subjected to 14C analyses) were collected for stable isotopic measurements at 5 cm intervals (in bioturbated units) and 1 cm intervals (in laminated units) and wet-sieved at 125 μm. Planktonic foraminifera (N. pachyderma, sinistral coiling, and G. bulloides) and benthic foraminifera (Uvigerina peregrina, Cibicidoides wuellerstorfi, and Nonionella sp.) were picked from the >150 μm size fraction. Care was taken during hand picking to select specimens that were as clean as possible, with no visual evidence for diagenetic overgrowths, which can sometimes be recognized by yellowish discoloration. Foraminiferal samples were ultrasonically cleaned prior to stable isotope analyses, but otherwise untreated.
 Stable isotopic measurements were performed at the Oregon State University College of Oceanic and Atmospheric Sciences (OSU/COAS) Stable Isotope Mass Spectrometer Facility using a Kiel III carbonate preparation device connected to a Thermo-Finnigan MAT-252 mass spectrometer. The data were corrected to the accepted PDB scale using an internal lab calcite standard (Wiley) and the international calcite standard NIST-8544 (also known as NBS-19). External precision (±1 standard deviation) of the Wiley standard of similar weight to the foraminiferal samples, run on the same days, was ±0.02‰ for δ13C and ±0.04‰ for δ18O (n = 230). Average NBS-19 values for these runs (δ13C = +1.93 ± 0.02‰, δ18O = −2.20 + 0.06‰, n = 51) were comparable to the accepted VPDB values for NBS-19 (δ13C = +1.95‰ δ13C, δ18O = −2.20‰) [National Institute of Standards and Technology, 1992]. To allow direct comparison to U. peregrina, empirical species corrections of +0.64‰ and +0.1‰ were applied to the δ18O values of C. wuellerstorfi and Nonionella sp., respectively. Similarly, an empirical species correction of −0.2‰ was applied to the δ13C values of C. wuellerstorfi.
4.1. Chronology and Accumulation Rates
 Based on the calendar-corrected radiocarbon chronology, the composite sediment sequence at site EW0408-85JC continuously spans the last >17,400 years B.P. in the Gulf of Alaska and documents the Pleistocene/Holocene transition (Figure 3). The timings of the features described in the following sections are derived from this age model, with uncertainties propagated from the nearest bracketing radiocarbon dates.
 Within the limitations of the dated increments, sediment accumulation rates vary from a low of 9 ± 2 cm/kyr between 13,770 and 12,220 cal years B.P., reflecting hemipelagic sedimentation during a time equivalent to the late Allerød and early Younger Dryas chronozones of Northern Europe, to values of >500 cm/kyr prior to 17,050 cal years B.P., during an interval of ice-proximal marine sedimentation composed dominantly of glacial silt with dropstones. Average postglacial sedimentation rates are 56 cm/kyr, and thus potential bioturbation-induced dating biases [e.g., Broecker et al., 1984] are expected to be <150 years (roughly one 10 cm mixed-layer depth) in most of the record. Within the interval of high sediment accumulation rates during deposition of the glacial-marine diamicts, two 14C age reversals of <300 years occur with overlapping error bars; all other ages increase with depth (Table 1). We use direct linear interpolation between radiocarbon dates to generate the age model, except in the high-sedimentation rate interval, where we use a linear best fit through the dates that increased with depth, a compromise that falls within the 1σ uncertainties of all but one of the calibrated dates in this sedimentary unit (Figure 3). Sediment accumulation rates decrease by almost an order of magnitude (from >500 cm/kyr to ∼90 cm/kyr) by 875 cmbsf (16,890 ± 90 cal years B.P.) within the interval of ice-proximal sedimentation, approximately 45 cm (2100 ± 390 years) prior to the onset of hemipelagic sedimentation at the site.
Table 1. Radiocarbon Dates From Multi, Trigger, and Jumbo Coresa
Multi, trigger, and jumbo cores are abbreviated MC, TC, and JC, respectively.
Calculated from mean planktonic 14C ages for JC and TC samples and benthic 14C ages for MC samples, using CALIB v.6.0 [Stuiver and Reimer, 1993] with the Marine09 calibration curve. For planktonic samples (JC and TC) ΔR = 470 ± 80 years, and for benthic samples (MC) ΔR = 900 ± 100.
 At depths >831 cmbsf, sediments are composed of terrigenous silt and clay with pebbles, which we infer to be glacial dropstones and other ice-rafted debris (Figures 2 and 3). Marine fossils, including benthic and planktonic foraminifera, are present in this interval and demonstrate the marine character of sedimentation. The radiocarbon chronology implies very rapid sediment accumulation in the oldest recovered interval, leading us to infer that this lithology represents ice-proximal glacial-marine sedimentation.
 The glacial-marine unit is overlain at 831 cmbsf (interpolated to 14,790 ± 380 cal years B.P.) by a relatively low-density laminated unit 34 cm thick. The contact between these units is sharp, but with no visible evidence for erosion and no apparent discontinuity in the radiocarbon chronology. The preserved laminations (approximately 140 laminae of millimeter to submillimeter thickness based on CT scans and visual descriptions) span 1800 ± 420 years in the calendar-corrected radiocarbon age model, and appear to reflect decadal-scale variability. Average sediment accumulation rates in this interval are ∼25 ± 12 cm/kyr. The lithology of the laminated interval is diatom-rich mud; darker laminae are predominantly terrigenous silt, whereas lighter, lower density laminae have higher concentrations of siliceous and calcareous microfossils. Opal (biogenic silica) was measured using a spectrophotometric wet-alkali extraction method [Mortlock and Froelich, 1989], and is more thoroughly discussed by J. A. Addison et al. (Productivity and sedimentary δ15N variability for the last 17,000 years in the Gulf of Alaska, submitted to Paleoceanography, 2011). Opal content of the sediment increases during the laminated interval from ∼3 wt % to 10 wt %, and it is likely that fluctuations in the ratio of biogenic to lithogenic sediment drive bulk density variability throughout EW0408-85JC (Figures 4a and 4b).
 At 797 cmbsf (interpolated age 12,990 ± 190 cal years B.P.) the upper contact of the laminated unit is lightly bioturbated, grading into a 37 cm (∼1830 year) interval of higher density. This interval contains mottled terrigenous silt and clay, with opal and calcium carbonate. A secondary, sublaminated unit (with light burrow mottling) of low density and with opal content of 10 wt % occurs between 745 and 760 cmbsf (11,160 ± 130 to 10,750 ± 220 cal years B.P.). At depths <745 cm, sediment composition is bioturbated hemipelagic mud for the remainder of the Holocene; fine-scale variations in the lithology of this sediment is described by Addison et al. (submitted manuscript, 2011).
4.3. Stable Isotopes
 Benthic foraminiferal (U. peregrina) δ18O ranges from maximum values of 4.9‰ (17,050 ± 350 cal years B.P.) to minimum values of 3.1‰, a range of 1.8‰ (Figure 4d). Mean glacial and late Holocene values (4.66 ± 0.05‰ and 3.27 ± 0.02‰, respectively, n = 7, standard error of the mean) are essentially identical to values reported from the same laboratory at 980 m depth from the Oregon margin [Mix et al., 1999] suggesting that this site captures the full glacial-interglacial transition. The decrease from Pleistocene values of benthic foraminiferal δ18O begins at 16,650 ± 170 cal years B.P., with a transient −1‰ excursion to minimum values centered at 14,250 ± 240 years B.P. (Figure 4d). The δ18O variations recorded by U. peregrina are reproduced by analyses of both C. wuellerstorfi and Nonionella sp. in 18 samples between 720 and 820 cmbsf.
 Planktonic foraminiferal δ18O values range from 3.7‰ (17,330 ± 270 cal years B.P.) to 1.3‰ (3600 ± 190 cal years B.P.; Figure 4c). This glacial-to-interglacial range of 2.4 ‰ is substantially larger than that measured in the benthic foraminifera. The decrease from Pleistocene values begins at 16,650 ± 170 cal years B.P., and as in the benthic record, includes an approximately −1‰ δ18O excursion between 14,710 ± 280 and 13,770 ± 120 cal years B.P. Both the onset of and recovery of the event are abrupt, and the peak of expression is centered at 14,250 ± 240 cal years B.P. The planktonic event includes two episodes of δ18O depletion, reproduced in both planktonic species and centered at 822.5 cmbsf (14,340 ± 390 cal years B.P.) and 814.5 cmbsf (14,150 ± 290 cal years B.P.) respectively. The benthic δ18O excursion occurs between these planktonic events (Figure 5). This fine structure is likely not an artifact of bioturbation, as the event occurs within the laminated interval.
 The two planktonic species, N. pachyderma and G. bulloides, reproduce δ18O values within 0.3‰ for most of the record, with two exceptions (Figure 4c). The first is during the interval of high δ18O between 12,940 ± 190 and 10,850 ± 220 cal years B.P. Following the shift to low δ18O values of ∼1.5‰ in the early Holocene, δ18O increases in both species to 2.0‰ by 7770 ± 120 cal years B.P., followed by a decrease to the most depleted values in the record (1.3‰) at 3600 ± 190 cal years B.P. Planktonic foraminiferal δ18O values then increase, reaching 1.8‰ by 2520 ± 180 cal years B.P., after which the values from N. pachyderma and G. bulloides again diverge (Figures 4c and 5a).
 Detailed interpretation of the benthic δ13C values is complicated by the shallow infaunal behavior of U. peregrina, the species for which a complete record exists (Figure 4g). Other stable isotope records from the North Pacific are typically dominated by this species, and their δ13C values have been evaluated cautiously as benthic water mass tracers, with possible productivity overprints [e.g., Lund and Mix, 1998; Mix et al., 1999; Okazaki et al., 2010]. Although C. wuellerstorfi is relatively rare, where present its δ13C values approximately reproduce those of U. peregrina, with an average offset of 0.2‰. Benthic δ13C in core EW0408-85JC is relatively high (−0.7‰) in the glacial interval (∼17,400 cal years B.P.), falling to a low of ∼−1.2‰ between 16,720 ± 170 and 15,340 ± 380 cal years B.P. Values then climb to a high of −1.05‰ centered at 14,710 ± 280 cal years B.P., before falling to the most depleted interval of the deglacial (−1.4‰) between 13,770 ± 120 and 11,550 ± 200 cal years B.P. By 10,310 ± 180 cal years B.P., benthic δ13C has rebounded to −0.7‰, and then gradually declines until 7420 ± 120 cal years B.P. (−1.3‰). From this point, benthic δ13C slowly increases throughout the remainder of the Holocene, reaching −0.8‰ by 1090 ± 160 cal years B.P.
 The δ13C values of N. pachyderma are consistently elevated (average offset of 0.51 ± 0.15‰) relative to those of G. bulloides (Figure 4f). The lowest δ13C values of the record (−0.8‰ for G. bulloides, −0.17‰ for N. pachyderma) occur at 16,640 ± 150 cal years B.P. Values increase to maxima at 14,710 ± 280 years B.P. (0.06‰ for G. bulloides, 0.3‰ for N. pachyderma), before decreasing again. Both species return to high δ13C values between 11,850 ± 240 and 10,880 ± 180 cal years B.P. before falling to a depleted plateau (−0.55‰ for G. bulloides, 0.09‰ for N. pachyderma) between 10,380 ± 150 and 9120 ± 240 cal years B.P. Recovery from these δ13C minima occurs by 6630 cal years B.P. (0.34‰ for G. bulloides, 0.78‰ for N. pachyderma). The highest δ13C values of the record are observed in the late Holocene; 0.57‰ for G. bulloides, and 0.96‰ for N. pachyderma.
5.1. Regional Deglaciation of the Northwest Cordilleran Ice Sheet
 Decreasing sedimentation rate within the interval of glacial-marine sedimentation likely documents early stages of glacial stagnation or retreat near 16,900 cal years B.P. Starting at 16,650 ± 170 cal years B.P., planktonic foraminiferal δ18O decreases rapidly (and benthic foraminiferal δ18O decreases more slowly), consistent with warming temperatures and the input of meltwater from retreating outlet glaciers. The abrupt contact between ice-proximal sediments and the overlying diatom-rich laminated interval at 14,790 ± 380 cal years B.P. records the retreat of Pleistocene tidewater glaciers onto land or behind fjord sills as global sea level rose.
 The abundance of opal, mostly from diatoms, and relatively high biogenic flux in the laminated intervals implicates high surface productivity as a cause of the laminations, which are preserved by benthic anoxia (also reflected by high molybdenum concentrations [Barron et al., 2009; Addison et al., submitted manuscript, 2011]. Benthic foraminiferal δ13C changed very little at the onset of laminations, suggesting no major changes in benthic water masses (Figure 5), implying that sedimentary anoxia was due primarily to changing productivity rather than changes in subsurface ocean circulation.
 By subtracting the benthic δ18O record from the N. pachyderma planktonic δ18O, we remove the potential isotopic signature of global ice volume that is common to both records and are left with the regional signature of isotopic gradients (δ18OP-B; Figure 4e) associated with local temperature and salinity stratification between surface and local bottom water [Lopes and Mix, 2009]. Anomalously low δ18OP-B values of −0.8‰ from 14,710 ± 280 to 12,950 ± 190 cal years B.P., indicate strong upper-ocean stratification, likely associated with high freshwater input during deglaciation. Superimposed upon this structure, the abrupt planktonic δ18O depletions centered at 14,340 ± 390 cal years B.P. and 14,150 ± 290 cal years B.P. (Figure 5) may reflect two discrete episodes of very high freshwater input to the Gulf of Alaska.
 Assuming meltwater δ18O of ∼−30 to −40‰ (SMOW), consistent with the isotopic values recorded in the Mount Logan ice core during the last termination [Fisher et al., 2008], and no temperature change, a −0.8‰ δ18O excursion would be equivalent to lowering the surface ocean salinity by ∼0.7. Today, a halocline of >2 exists between the sea surface and 680 m depth at the site (Figure 1) driven by high freshwater input along the margin of 23,000 m3 s−1 (annual average) [Royer, 1982], so as a first approximation, the surface-ocean δ18O excursion during deglaciation could have been induced by modest additional freshwater inputs of half to one third the modern freshwater flux (i.e., ∼8000–12,000 m3 s−1). The laminated interval is not uniquely associated with the interval of inferred high freshwater input, as the laminations continue for ∼800 years after the end of the planktonic δ18O excursion (Figure 5).
 An abrupt low δ18O excursion in the benthic foraminifera at 14,250 ± 290 cal years B.P. occurs within the laminated interval, between the two low δ18O events in the planktonic foraminifera (Figure 5b). This event is unlikely to be explained solely as a subsurface temperature excursion, as it would require an implausible 5°C increase in bottom water temperature at a paleodepth of 580 m (accounting for ∼100 m lowered sea level). Following the assumptions above, the benthic δ18O excursion implies a salinity reduction of ∼1 at the seafloor. Three possible mechanisms for this observation include: deepening of the modern halocline associated with the Alaska Coastal Current, hyperpycnal flow associated with an abrupt meltwater event, and/or local formation of isotopically depleted brines associated with sea ice expansion.
 Deepening of the modern halocline at this site (Figure 1), which presently reaches a spring maximum of between 100 and 200 m depth, to more than 580 m paleodepth, would require a very large increase in freshwater inputs. Although the amount is difficult to quantify, deepening of the regional halocline by a factor of 4 implies at least a proportional increase in freshwater flux of that amount, and perhaps much more (i.e., a factor of 16 to account for a factor of 4 depth increase coupled to a factor of 4 offshore extension). Given modern freshwater inputs of 23,000 m3 s−1 [Royer, 1982], driving the benthic foraminiferal δ18O excursion via halocline depression would imply a seasonal increase of this flux to at least 92,000 m3 s−1, and perhaps as much as 370,000 m3 s−1. To produce an event of the duration observed in the laminated sediments, ice sheet melting would have to intermittently reach this rate for a period of >300 years. Rates of melting on the low end of this range have been observed off South Greenland [Rignot and Kanagaratnam, 2006], which may be a fair analog for the retreating NW Cordilleran at the end of the Pleistocene. Runoff of this magnitude (∼105 m3 s−1) sustained for six months per year over the period of laminations, would imply a loss of 6%–30% of the Cordilleran Ice Sheet volume [Peltier, 1994].
 Benthic foraminiferal δ18O is lowest during a period of relatively high planktonic δ18O between the transient lows centered at 14,340 ± 390 cal years B.P. and 14,150 ± 290 cal years B.P. (Figure 5). Differences in the fine structure of the benthic and planktonic δ18O anomalies suggest that local meltwater and thickening of the halocline did not cause the benthic anomaly. Other potential mechanisms include nonlocal hyperpycnal flows, or brine formation. To reach depths of >580 m, a low-salinity hyperpycnal flow must be charged with sediment [Imran and Syvitski, 2000; Aharon, 2006]. Although the radiocarbon dates indicate a transient increase in sedimentation rate at the time of the benthic δ18O anomaly, there is no sedimentological evidence for an increase in the fraction of lithogenic sedimentation associated with the event. A plausible source for freshwater inputs capable of generating hyperpycnal flows is the Copper River, a known conduit for megafloods associated with draining of glacial Lake Atna as recently as ∼10,000 years B.P. [Ferrians, 1989; Shimer, 2009; Wiedmer et al., 2010]. If this is the source of the low benthic δ18O anomaly at Site EW0408-85JC, the silt-rich portion of the flow must have been focused in the submarine canyons near the source, allowing some of the entrained low-salinity, low δ18O waters to diffuse across the slope to the location of the core without an accompanying sediment load.
 Formation of brine water on the shelf by salt rejection from seasonal sea ice has been invoked to explain negative benthic δ18O excursions of similar magnitude observed in the Nordic Seas [Bauch and Bauch, 2001] and in the Northwest Pacific [Gebhardt et al., 2008], although this inference is controversial [Rasmussen and Thomsen, 2009]. As sea level at the time of the benthic excursion was ∼100 m lower, it follows that ocean salinities were ∼0.8 higher than modern values. The planktonic isotopic excursion in this interval suggests that regional freshwater input decreased surface salinities by ∼0.7 at this time, so surface water salinities during the excursion were roughly similar to today. In order to sink to the depths of the core site, brine-enhanced water would have had to reach salinities of ∼35, implying ∼8% surface water conversion to sea ice. An abrupt increase in winter sea ice formation in the Gulf of Alaska may have been facilitated by the decrease in surface salinity associated with regional meltwater input. Sea ice-related diatoms peak in EW0408-85JC (30%–40% of the total diatom assemblage) during the Bølling period (Figure 5) [Barron et al., 2009]. Caissie et al.  also note high abundance of sea ice diatoms in the SE Bering Sea at about this time. Cooler surface temperatures associated with sea ice formation may also help to explain the concurrent minor enrichment in the planktonic δ18O at the height of the benthic excursion. Thus, either the hyperpycnal flow or the brine rejection mechanisms remain as plausible explanations of the low benthic foraminiferal δ18O anomaly near 14,250 ± 290 cal years B.P.
 By 12,990 ± 190 cal years B.P., laminations were no longer preserved and accumulation of authigenic manganese increased (Addison et al., submitted manuscript, 2011), implying that oxygen was again present at the seafloor. The increase in bulk density of the sediment at this time, accompanied by an increase in sedimentation rate (Figure 2) and a visually apparent increase in the >150 μm lithologic size fraction, imply increased ice-rafted debris at the site. These multiple lines of evidence suggest regional glacial readvance. The δ18O values in both planktonic species reach their greatest post-Pleistocene enrichment in this interval, suggesting that the annual mean temperature of the surface ocean was colder and/or more saline than at any time following the initial retreat of the Northwest Cordilleran ice sheet off the continental shelf. These findings are congruent with other terrestrial/lacustrine evidence for cooling and alpine glacial readvance in Alaska during the Younger Dryas interval [Engstrom et al., 1990; Hajdas et al., 1998; Briner et al., 2002; Hu et al., 2006].
Neogloboquadrina pachyderma δ18O values are relatively constant, 2.65 ± 0.05‰, between 12,940 ± 200 and 11,740 ± 200 cal years B.P. In contrast, G. bulloides δ18O values decrease by ∼0.3‰ through the same interval. At high latitudes today, N. pachyderma blooms at the height of summer and thus cannot shift to a warmer growth season, making them a faithful recorder of the summer conditions [Fraile et al., 2009]. In contrast, G. bulloides is sensitive to food availability [Ortiz et al., 1995; Tedesco et al., 2007; Fraile et al., 2009], and thus may shift its seasonal preferences to track phytoplankton blooms. The delay in δ18O increase in G. bulloides, and earlier return to lower δ18O values relative to N. pachyderma, may reflect a shift in the seasonality of the G. bulloides bloom. Thus we hypothesize the δ18O signature of G. bulloides in this interval reflects a combination of environmental and physiological shifts, and will favor N. pachyderma as a proxy of regional summer surface ocean salinity/temperature. Based on this species, the period from 12,940 ± 200 and 11,740 ± 200 cal years B.P. was a time of fairly stable cold and/or saline surface conditions. This observation is further supported by an increase in cold-water microfossil groups in EW0408-85JC [Barron et al., 2009].
 A return to sublaminated conditions occurs from 11,160 ± 130 to 10,750 ± 220 cal years B.P. (Figure 5a). This sublaminated interval has high opal content similar to the older fully laminated interval, although sediment accumulation rates during the younger event are lower, within the range of Holocene variability. High concentrations of biogenic opal, along with diatom species assemblages and the synchronous high concentrations of molybdenum and uranium [Barron et al., 2009; Addison et al., submitted manuscript, 2011] show that high surface-ocean productivity drove benthic anoxia during this event, similar to the earlier event. The δ18O and δ13C data again indicate no major changes in bottom water masses. The onset of sedimentary anoxia occurred a few hundred years prior to significant surface water warming or freshening (decrease in planktonic δ18O relative to benthic foraminiferal δ18O) suggesting primary production was not uniquely linked to local freshwater runoff.
 The Holocene interval of core EW0408-85JC records more modest environmental changes; planktonic δ18O suggests that the surface ocean gradually cooled or increased in salinity until 7770 ± 120 cal years B.P., followed by warming and/or freshening through the late Holocene. The late Holocene (<4000 cal years B.P.) displays greater millennial-scale variability than the mid-Holocene. During this time δ18OP-B increased for ∼600 years centered at 2500 ± 180 cal years B.P., accompanied by a minor but concomitant increase in benthic δ18O, possibly associated with regional terrestrial glacial advances dated between 3300 and 2900 cal years B.P. and 2200–2000 cal years B.P. [Barclay et al., 2009].
5.2. Links to Regional and Global Climate
 Warming and freshening of surface waters in the subarctic Northeastern Pacific at 14,710 ± 280 cal years B.P. coincides within analytical uncertainty with the current layer-count chronology for the onset of Bølling warmth in the Greenland NGRIP ice core records (14,740 ± 60 cal years B.P.) [Rasmussen et al., 2006], and is consistent with a rapid atmospheric teleconnection between the North Pacific and Atlantic [Broecker, 1994; Mikolajewicz et al., 1997; Hostetler et al., 1999]. The increase of δ18OP-B at 13,770 ± 120 cal years B.P. likely reflects the end of anomalous freshwater inputs, and perhaps cooling and/or drying conditions in the Northeast Pacific. This occurs within the Allerød interstadial event, preceding the onset of the Younger Dryas (Y-D) event of the North Atlantic by almost a millennium (12,890 ± 140 cal years B.P.) [Rasmussen et al., 2006].
 The early termination of the Bølling-Allerød warm interval observed in EW0408-85JC appears in a number of high-latitude North Pacific records. In Figure 6 we compare the planktic oxygen isotopic records from this study to Gulf of Alaska cores MD02-2489 (54°23.4′N, 148°55.26′W, 3640 m) [Gebhardt et al., 2008] and PAR87A-10 (54°21.8′N, 148°28.0′W, 3664 m) [Zahn et al., 1991], and Northwest Pacific core Vinogradov GGC-37 (50°25.2′N, 167°43.2′E, 3300 m) (Keigwin  on the chronology of Galbraith et al. ) through Termination 1. The low resolution of the isotopic data and age model of PAR87A-10 [Zahn et al., 1991] makes direct comparison with EW0408-85JC difficult, but represents a pioneering effort for the region. Similarly, although planktonic oxygen isotope records have been generated for Gulf of Alaska ODP Site 887 [McDonald et al., 1999; Galbraith et al., 2007], these provide little additional insight due to their low resolution through Termination 1.
 To facilitate comparison with EW0408-85JC, radiocarbon dates from PAR87A-10 and MD02-2489 have been recalibrated with CALIB 6.0 using a constant ΔR of 550 ± 250, following the logic applied to nearby ODP Site 887 by Galbraith et al. . In the case of MD02-2489, four age reversals (Figure 6) were excluded from the resultant age model, derived by linear interpolation between calibrated dates. Within dating uncertainties, all three high-resolution, high-latitude North Pacific records (GGC-37, MD02-2489, EW0408-85JC) show warming/freshening of the surface ocean coincident with the onset of the Bølling warm interval in the North Atlantic (Figure 6), providing support for a rapid atmospheric teleconnection between the two basins [Broecker, 1994; Mikolajewicz et al., 1997; Okumura et al., 2009]. However, these same records on their independent radiocarbon-based chronologies lack evidence for a warm/fresh surface ocean during the Allerød (Figure 6).
 An apparent North Pacific lead, relative to the North Atlantic, in the post-Bølling return to conditions typical of the Younger Dryas has been found elsewhere in the North Pacific [e.g., Mix et al., 1999; Ortiz et al., 2004; Cook et al., 2005; Dean et al., 2006], although these records are less well dated than core EW0408-85JC. If the lead of North Pacific cooling is an artifact of dating, it would imply a significantly higher reservoir age at that time, on the order of ∼1600 years for the surface ocean. This is implausible during a time of strong freshwater input and upper ocean stratification; if anything, the high input of freshwater with a reservoir age near zero would support a lower oceanic reservoir age at that time, implying that the apparent lead of North Pacific cooling prior to the onset of Younger Dryas conditions in the North Atlantic is a real phenomenon.
 The post-Bølling increase in δ18O in the North Pacific may reflect the blended influences of Northern Hemisphere (i.e., the Allerød and Younger Dryas) and Southern Hemisphere (i.e., the Antarctic Cold Reversal, ACR) [Blunier et al., 1997; Blunier and Brook, 2001] climates in the region. The influence of the ACR was previously inferred in subsurface North Pacific paleotemperature records [e.g., Mix et al., 1999]. The Southern Ocean is the formation site of North Pacific Deep Water, and recent evidence suggests that changes in Antarctic source water properties may be propagated rapidly to the high-latitude North Pacific via internal waves [Fukasawa et al., 2004; Masuda et al., 2010]. In this model, changes in the location and rate of deep water formation in the Southern Ocean reorganize isopycnal surfaces throughout the interior of the Pacific Basin, impacting the high-latitude North Pacific within a few decades [Masuda et al., 2010]. This suggests the potential for a geologically instantaneous teleconnection between the Antarctic and subsurface North Pacific, perhaps superimposed upon a Northern Hemisphere atmospheric teleconnection [Broecker, 1994; Mikolajewicz et al., 1997; Okumura et al., 2009]. When compared to the δ18O records from the EDML ice core in Antarctica [Ruth et al., 2007] and the NGRIP ice core in Greenland [Andersen et al., 2006; Rasmussen et al., 2006; Svensson et al., 2006], we find the planktonic oxygen isotope pattern of EW0408-85JC bears some similarity to both of these records (Figure 6), lending support to the idea that North Pacific climate records reflect both North Atlantic and Southern Ocean forcing [Mix et al., 1999].
5.3. Mechanisms for the Deglacial Productivity Events
 Hypotheses to explain the apparently synchronous onset of short-lived high-productivity events during the Bølling-Allerød interstadial event and in the early Holocene around the margin of the North Pacific include (1) basin-wide enhanced upwelling [e.g., Dean, 2007; Barron et al., 2009], (2) nutricline adjustment related to variability in the strength of the Atlantic Meridional Overturning Circulation (AMOC) [e.g., Schmittner, 2005; Galbraith et al., 2007; Schmittner and Galbraith, 2008], (3) aeolian fertilization associated with increased dust or volcanic ash delivery [e.g., Martin, 1990], (4) freshwater fertilization associated with deglacial meltwater flux to the North Pacific, and/or (5) fertilization associated with enhanced benthic iron flux from shelf sediments driven by deglacial sea level rise [e.g., Mix et al., 1999].
 In Hypothesis 1, the increase in oceanic fertility is attributed to bioavailable macronutrients and micronutrients associated with enhanced upwelling. Dean  hypothesized that the B-A warming in Greenland was transmitted from the Atlantic to the Pacific Ocean via more energetic Hadley and Walker circulations, which strengthen subtropical high-pressure cells, potentially increasing upwelling at the edges of the subtropical high. This wind-driven upwelling hypothesis, however, is difficult to apply to the SE Alaska margin, an area of coastal downwelling, and is not supported by the planktonic δ13C records. Upwelling zones typically have low δ13C in planktonic foraminifera such as G. bulloides [Sautter and Thunell, 1991; Ganssen and Kroon, 2000]. Although this in part reflects the presence of enhanced contribution of waters with low-δ13C dissolved inorganic carbon during upwelling, Ortiz et al.  attributed some of the δ13C data in foraminifera to metabolic disequilibrium effects related to temperature (warmth yielding lower δ13C relative to equilibrium) or food supply (high biomass yielding lower δ13C relative to equilibrium). For both reasons, the upwelling hypothesis for high productivity on the margins predicts a decrease in planktonic foraminiferal δ13C during the events. This interpretation is consistent with other deglacial planktonic δ13C foraminiferal data sets from the North Pacific [e.g., Hendy et al., 2004]. In core EW0408-85JC, both measured species of planktonic foraminifera, N. pachyderma and G. bulloides, record a 0.4‰ increase of δ13C during the deglacial productivity events, which appears to preclude the upwelling hypothesis as a general driver.
 In Hypothesis 2, the collapse of the Atlantic meridional overturning circulation at the end of the last ice age (during Heinrich Event 1, prior to Bølling-Allerød warmth) and again during the Younger Dryas cold event [McManus et al., 2004] traps nutrients in the North Atlantic, and depletes them in the rest of the upper ocean, which in turn decreases ocean productivity in the North Pacific [Schmittner et al., 2007]. In this biogeochemical model, phosphate and nitrate accumulate in the interior ocean and productivity rebounds about 600 years after the initial reduction as the nutrient inventory adjusts to the change in circulation. Accompanying the accumulation of nutrients in the interior Pacific predicted by this model, benthic δ13C should be depleted leading up to the rebound in productivity. In contrast, the benthic δ13C record of EW0408-85JC shows enrichment in δ13C associated with the onset of both the Bølling-Allerød and early Holocene productive intervals, seeming to contradict the global nutricline adjustment hypothesis as a cause of these events.
 In Hypothesis 3, the driver of the deglacial productivity events would be an enhanced supply of micronutrients such as iron to the marine environment via aeolian dust or volcanic eruptions. Dust is thought to be an important source of dissolved iron to surface waters in most areas of the open ocean [Martin, 1990; Duce and Tindale, 1991]. However, dust concentrations in marine sediment and ice core records suggest that dust fluxes were higher during glacial and stadial intervals, and lower in warmer interstadial or interglacial intervals; atmospheric supply during the Last Glacial Maximum may have been 5–50 times greater than modern day [De Angelis et al., 1987; Winckler et al., 2008; Martínez-Garcia et al., 2009]. This offset in timing between observations of high atmospheric dust content and the deglacial productivity highs of the North Pacific makes increased aeolian iron supply an unlikely driver of these particular events. Alternatively, volcanic ash may be a source of iron. This would require a period of enhanced volcanic activity sufficiently prolonged to drive anomalously high productivity for up to two millennia. We found no evidence for the presence of volcanic ash in the laminated silica-rich sediments of the high production intervals, so local ash fertilization cannot be invoked. Huybers and Langmuir  infer a 200%–600% increase in volcanism in areas experiencing deglaciation following the LGM, possibly related to decompression of the mantle associated with ablating ice sheets. However, their inferred increase in global volcanism does not begin until 12,000 cal years B.P., and peaks near 7000 cal years B.P. before subsiding to near-glacial levels in the modern day. This significantly postdates the larger B-A productivity event and has no convincing relationship to the shorter early Holocene productivity event. The enhanced volcanic fertilization scenario is unlikely to explain the abrupt onset of high productivity and anoxia associated with the deglacial laminated intervals.
 In Hypothesis 4, the deglacial retreat of the Cordilleran Ice Sheet would increase the supply of meltwater-delivered dissolved nutrients and glacial flour. Fluvial input today delivers both iron and silicic acid to the Gulf of Alaska [Stabeno et al., 2004] and iron in Alaskan glacial rock flour is ten times more soluble than iron in Chinese Loess or Saharan Dust [Schroth et al., 2009]. Megafloods associated with the draining of glacial lake Atna reached the Gulf of Alaska via the Copper River as recently as ∼10,000 years B.P. [Ferrians, 1989; Shimer, 2009], and could have transported glacial rock flour from the retreating Cordilleran Ice Sheet. However, regional increases in glacial flour delivery to the Gulf of Alaska cannot explain the apparently coincident productivity maxima around the margins of the North Pacific, and far from glaciated coastlines.
 In Hypothesis 5, the deglacial sea level rise could increase the supply of iron via remobilization of iron from sediments deposited on previously subaerial shelves. Dissolved iron is transported today from shelves in shallow subsurface waters [Johnson et al., 1999; Lam and Bishop, 2008]. These subsurface particulate iron maxima contain a relatively high proportion of reduced iron (Fe2+ comprises up to 25% of the total Fe) [Lam and Bishop, 2008]. Primary mineral phases dominated by Fe2+ are soluble in oxidizing water conditions [Moffett, 2001], increasing their bioavailability in the upper ocean. The large phytoplankton blooms artificially induced during the SERIES iron addition experiments in the Gulf of Alaska were seeded with ferrous sulphate (Fe2+SO4) [Boyd et al., 2004].
Mix et al.  first hypothesized remobilization of shelf sources of iron during sea level rise as an explanation for deglacial productivity events observed off of Oregon, citing resuspension of particulate iron as a potential fertilization mechanism. More recently, Severmann et al. [2006, 2008, 2010] described dissolved iron flux from suboxic continental shelf sediments. Combining these ideas, we suggest that abrupt sea level rise during deglaciation inundated previously exposed coastal plains on the continental shelf, which were depocenters of glacial, fluvial, and aeolian sediment. Suspension of these sediments would provide iron (and other) nutrients, which would in turn increase regional biological productivity near the ocean margins and drive anoxia in the shelf sediments. Iron release would then be further enhanced within these suboxic sediments [Severmann et al., 2008, 2010], fueling production in a positive feedback mechanism. Remobilization of sedimentary iron associated with sea level rise may be a particularly potent mechanism in the North Pacific, a region with volcanic margins, vast continental shelves and two large, shallow marginal seas (the Bering and Okhotsk).
 In support of this hypothesis, the onset of both North Pacific productivity highs are coincident with periods of rapid sea level rise. For example, Meltwater Pulse 1A, which was initiated at 14,690 ± 85 cal years B.P. [Weaver et al., 2003; Kienast et al., 2003; Deschamps et al., 2009] was coincident, within analytical error, with the onset of the older Gulf of Alaska laminated interval (14,790 ± 380 cal years B.P.). A less abrupt sea level rise known as Meltwater Pulse 1B at ∼11,300 cal years B.P. [Fairbanks, 1989; Bard et al., 1990, 1993], but reduced in rate by Bard et al. , coincides (within dating uncertainties) with the onset of the younger laminated interval in the Gulf of Alaska (11,140 ± 85 cal years B.P.).
 The detailed radiocarbon chronology of high-accumulation rate jumbo piston core EW0408-85JC and its associated trigger core and multicore provide an opportunity to place high-latitude North Pacific climate changes into a global chronological framework. Retreat of glaciers in the region of the northeastern Gulf of Alaska began by 16,650 ± 170 cal years B.P., based on apparent salinity reductions recorded in planktonic foraminiferal δ18O and a decrease in the rate of glacial-marine sediment accumulation. The transition from the ice-proximal to laminated hemipelagic sediments at 14,790 ± 380 years B.P. marks the retreat of glaciers either behind sills or onto land.
 An interval of low δ18O in planktonic foraminifera reflects regional freshwater input from retreating glaciers between 16,650 ± 210 cal years B.P. and 13,770 ± 120 cal years B.P. ± 200 cal years B.P. A more abrupt low δ18O event in the benthic foraminifera at 14,250 ± 290 cal years B.P. likely reflects an injection of low-salinity water to 580 m paleodepth driven by rapid glacial melt, perhaps due to low surface salinities that enhanced winter sea ice cover and attendant brine formation on the shelf. Alternatively, the benthic foraminiferal oxygen isotopic excursion could reflect hyperpycnal flows or a transient deepening of the halocline.
 Radiocarbon dates constrain the timing of deglacial warming and freshening of the Gulf of Alaska as coeval with the onset of Bølling interstadial warmth of the North Atlantic and Greenland. North Pacific cooling and/or an increase in surface salinities during Allerød interstadial time may reflect the influence of the Antarctic Cold Reversal, likely transmitted via the subsurface ocean.
 Productivity maxima drove sedimentary anoxia and laminated sediments occur between 14,790 ± 380 to 12,990 ± 190 cal years B.P. and between 11,160 ± 130 to 10,750 ± 220 cal years B.P. These events are similar to and likely correlative with, less precisely dated events observed around the rim of the North Pacific. The high-resolution chronology links these events to episodes of global sea level rise. We evaluate several hypotheses to explain these events, and conclude that remobilization of iron and other limiting nutrients from continental shelves and inundated estuaries during sea level rise may help to explain synchronous increases in productivity and anoxia that preserved laminated sediments in many places around the margins of the North Pacific.
 The authors gratefully acknowledge the crew and science party of cruise EW0408, Nick Pisias for valuable commentary and insight, John Southon and the staff of the UCI Keck AMS laboratory, Jennifer McKay, Andy Ross, and the OSU Stable Isotope Facility. We thank Maziet Cheseby, Bobbi Conard, Mysti Weber, and the OSU Marine Geology Repository. We thank Eric Galbraith and Ingrid Hendy for their assistance in assembling the planktic stable isotope data from extant North Pacific records. We also thank the two anonymous reviewers, whose thoughtful comments substantially improved the manuscript. This work was funded by National Science Foundation Paleoclimate grant ATM-0602395, National Science Foundation Geophysics grant EAR-0711584, and National Science Foundation Ocean Drilling grants OCE-0242084 and OCE-0351096.