Impact of sea ice variability on the oxygen isotope content of seawater under glacial and interglacial conditions



Records of temporal variability in the oxygen isotopic composition of biogenic carbonates (δ18Oc) from ocean sediment cores inform our understanding of past continental ice volume and ocean temperatures. Interpretation of δ18Ocvariability typically neglects changes due to factors other than ice volume and temperature. Here we investigate whether glacial-interglacial changes in sea ice, which fractionates seawater during its formation, could shift the isotopic value of seawater—in the deep ocean (affecting benthic foraminiferal δ18Oc and thereby potentially impacting oxygen isotope based sea level reconstructions) or in surface waters (affecting planktic foraminiferal δ18Oc). We simulate glacial and interglacial states with the isotope-enabled University of Victoria Earth System Climate Model and perform a global analysis. Distinct patterns of sea ice production emerge for the glacial versus interglacial states. We find no substantive shift in δ18Ow in model deep or bottom waters due to the simulated interglacial-glacial sea ice variability. Small isotopic shifts due to sea ice variability are concentrated in the model's surface waters of the Northern Hemisphere, specifically in the Labrador Sea and northeastern North Atlantic.

1 Introduction

Oxygen isotope content (δ18O defined as the ratio of math formulato math formula, relative to a standard ratio) measured in biogenic carbonates derived from ocean sediment cores constitutes a key paleoproxy, its record of temporal and spatial variation having provided a wealth of knowledge informing past ocean and climate conditions. Variations in mean seawater δ18O on time scales of 103 to 105 years result from changes in continental ice volume. However, ocean organisms secrete carbonate shells in temperature-dependent equilibrium with their local seawater environment. Assessing variations in marine carbonate δ18O involves the deconvolution of the processes recorded in the isotopic signal, thereby allowing the estimation of changes in past continental ice volume (and by extension, past global mean sea level) [Chappell and Shackleton, 1986; Shackleton, 1987; Labeyrie et al., 1987], sea water paleotemperature [Shackleton, 1974; Shackleton et al., 1977], and paleosalinity [Maslin et al., 1995; Rohling, 2000; LeGrande and Schmidt, 2011].

As described by Waelbroeck et al. [2002], the isotopic changes (thus, Δδ) recorded by benthic foraminifera (Δδ18Ob) reflect both changes in the isotopic composition of seawater (Δδ18Ow) and changes in temperature (Δδ18Otemp). The changes in the isotopic composition of seawater can be further decomposed into changes in the mean isotopic state of the ocean (Δδ18Oicevol) (a function of how much depleted ice is stored on the continents and the δ18O of that ice) and changes in the local seawater isotopic composition (Δδ18Olocal), such that

display math(1)

[Waelbroeck et al., 2002, their equation (1)].

Local seawater δ18O depends on the isotopic content of its water source and the sum of any upstream isotopic contributions. The balance of surface evaporation and precipitation of source water (when the water parcel was last in contact with the atmosphere), additions of river runoff, ice sheet melt, and sea ice brine and melt, as well as ocean circulation all may affect δ18Olocal[Rohling and Bigg, 1998]. Each of these processes contributing to δ18Olocal may change in time and space. By extension, assuming any factor to be constant can introduce error to paleoreconstructions. Neglecting a variable component of seawater δ18O when interpreting isotopic records from ocean sediments may, in effect, superimpose error upon the resulting paleoreconstruction. This may especially be important at high latitudes where changes in surface water δ18O are dominantly related to salinity changes as opposed to temperature changes. The alternative is to acknowledge process variability, where possible. Here we investigate sea ice variability on glacial-interglacial time scales and the extent to which the isotopic signature of sea ice in seawater may fluctuate.

Sea ice formation is accompanied by fractionation of stable water isotopes, such that newly formed sea ice is enriched relative to seawater and the expelled sea ice brine is depleted [O'Neil, 1968]. Thus, sea ice growth represents the only process by which changes in seawater salinity and isotope content are negatively correlated (see discussion in Hillaire-Marcel and de Vernal [2008]). Sea ice is dynamic and may form (and expel depleted brines) in one location and melt (depositing enriched meltwater into surface waters) in a distant location. Additionally, the density gradient between brines and sea ice melt can result in enriched meltwater accumulating within the surface layer, while depleted brines sink to the depth of the pycnocline, from where they are eventually mixed downward through the water column [Hillaire-Marcel and de Vernal, 2008]. The presence or absence of sea ice meltwater or brines may therefore produce a sizeable isotopic shift in surface waters [cf. Tan and Strain, 1980].

In regions of deep water formation influenced by sea ice brines (e.g., present-day Greenland-Iceland-Norwegian (GIN) Seas and Weddell Sea), variable sea ice brine production holds the potential to shift deep water δ18Ow. For example, large changes in sea ice production in the Arctic Ocean and North Atlantic region could shift the δ18Ow of North Atlantic Deep Water, and sea ice changes around Antarctica could influence the isotopic content of bottom waters throughout the global ocean. A deep water isotopic shift due to sea ice changes could propagate error to sea level estimates based on oxygen isotope records. Additionally, benthic foraminiferal δ18Obprecipitated in ambient waters influenced by variable sea ice brines may incorporate a sea ice isotopic component (and similarly for planktonic foraminiferal δ18O, a variable sea ice brine or meltwater component). The isotopic partitioning resulting from sea ice growth may also influence surface waters, as sea ice melt water and brines are distributed throughout the global ocean.

Determining the spatial structure of δ18Ow shifts due to interglacial-glacial sea ice variability is the goal of this study. Representative interglacial and glacial climates are chosen, as we assume the greatest sea ice variability under the current continental configuration occurs on millennial or multimillenial time scales and relates to glacial-interglacial shifts. Of course sea ice varies on many other time scales (e.g., daily, interannual, interdecadal), but fluctuations of relatively short duration would not be expected to influence the isotopic content of deep or bottom waters and therefore benthic δ18Oc. While no direct paleoproxy exists for the rate of sea ice production, the impact of sea ice variability may be explored in a model. Specifically, whether interglacial-glacial changes in sea ice can produce a shift in δ18Ow and whether the magnitude of this shift could require revisitation of reconstructions based on planktic or benthic δ18Oc can be tested in a model. We construct two sets of climate states—interglacial and full glacial—and investigate (i) changes in sea ice between these climates, (ii) the three-dimensional isotopic signature of sea ice in sea water in each climate, and (iii) the shift in oxygen isotope content between these distinct sea ice regimes in surface, intermediate, deep, and bottom waters.

1.1 Sea Ice Growth and Brine Formation

When new sea ice forms in open water, freezing produces ice crystals, which congelate into a layer of ice. The resulting sea ice consists of a complex network of (fresh) ice, pockets of brine, solid salts, and air. During freezing, brine is partly concentrated in bubbles in new sea ice and eventually expelled to the underlying water. The ice may continue to grow at the ice-water interface through accretion to the ice base. As the sea ice grows thicker, the complex network of brine inclusions in the ice may drain, reducing the salinity of first-year ice. Sea ice salinity varies within an ice floe (characteristic ice salinity profiles of first-year ice in the Weddell Sea exhibit values ranging from above 14 to below 4, as shown by Eicken [1992]) and depends on the balance of brine inclusion and desalination processes. Brine inclusion is a function of the growth rate and temperature of the ice [Eicken, 1992; Cox and Weeks, 1975, 1988]. Desalinization encompasses brine expulsion, gravity drainage, and flushing [Untersteiner, 1967; Lake and Lewis, 1970] and results in even fresher multiyear ice—Cox and Weeks [1974] observed a mean salinity of 2.0 and 3.0 for different types of Arctic multiyear ice.

1.2 Sea Ice Growth and Isotopic Fractionation

As seawater freezes, fractionation takes place between water molecules. With its larger mass, the molecule math formulaexhibits a lower vibrational frequency and zero-point energy (relative to math formula). The heavier molecule is therefore slightly preferred within the solid ice structure. In fresh water, this leads to a freezing induced fractionation of 3.0‰ at equilibrium, such that ice is enriched (by 3‰) [O'Neil, 1968]. During the formation of sea ice, ice is enriched by up to a maximum of 3‰ and the expelled sea ice brine is depleted by the equivalent amount.

In fact, the magnitude of fractionation may vary with the rate of ice growth, such that larger fractionation occurs with slower ice growth [Eicken, 1998]. As discussed by Ekwurzel et al. [2001], a range of fractionation factors for newly formed sea ice have been determined in the field based on the δ18O values in ice and the underlying seawater. For example, Melling and Moore [1995] found a mean fractionation of 2.5‰ in the Beaufort Sea, Macdonald et al. [1995] measured a fractionation of 2.6±0.1‰ in the Arctic, and Eicken [1998] observed a maximum 2.7‰ fractionation in the Weddell Sea. Pfirman et al. [2004] found a ∼2‰ fractionation at the base of Arctic multiyear ice, while Ekwurzel et al. [2001] determined that Arctic modern conditions could theoretically result in fractionation ranging from 1.5‰ to 2.7‰.

One important consequence of the isotopic fractionation occurring during sea ice growth is that oxygen isotopes may be utilized in combination with salinity to trace water mass sources, estimate the sea ice component of water masses, and elucidate freshwater cycling processes that produce an observed water mass [e.g., Redfield and Friedman, 1969; Östlund and Hut, 1984; Bauch et al., 1995; Frew et al., 2000; Bauch et al., 2005; Dodd et al., 2009; Yamamoto-Kawai et al., 2009; Cox et al., 2010].

1.3 Glacial-Interglacial Sea Ice Variability

Considering glacial and interglacial climate states, one would expect significant changes in patterns of sea ice extent, volume, and rates of growth and melt—both spatially (e.g., for the Arctic, North Atlantic, and Southern Oceans) and temporally (e.g., shifts in the seasonal cycle of sea ice processes at a given location). For example, the present-day Arctic exhibits a large annual cycle of ice growth and melt, whereas during the Last Glacial Maximum (LGM) the Arctic Ocean would have been permanently ice covered, with significantly thicker ice and lower in situ rates of ice growth and melt. Regions further south not subject to present-day sea ice would have seen seasonal ice cover. For example, evidence suggests that winter sea ice extended to ∼55°N in the central and eastern North Atlantic and to 40°N along the coast of North America [Kucera et al., 2005; de Vernal et al., 2005]. Glacial-interglacial differences in sea ice seasonality and spatial patterns have been investigated using microfossil-based transfer functions, including dinoflagellate cysts and diatoms, in the northern North Atlantic [de Vernal et al., 1994; de Vernal and Hillaire-Marcel, 2000; Rochon et al., 1998], the Southern Ocean [Crosta et al., 1998] and northern Pacific [Sancetta, 1983; de Vernal and Pedersen, 1997].

Sea ice variability may encompass changes in areal extents of summer and winter ice, ice thickness, rates of ice production, and locations of ice growth and melt. The potential for variable sea ice production to shift δ18Owbetween late Holocene, Heinrich event, and Last Glacial Maximum (LGM) conditions has been investigated in the Labrador Sea pycnocline by Hillaire-Marcel and de Vernal[2008], who tested for the presence of off-equilibrium isotopic excursions (i.e., offsets from carbonate precipitated at equilibrium in the ambient water) in the polar planktic foraminifer Neogloboquadrina pachyderma. In the modern Arctic, off-equilibrium values on the order of −1 to −3‰ are found (in the Northern Greenland Sea and Western Arctic, respectively) [Hillaire-Marcel and de Vernal, 2008]. Given that approximately −1‰ of the offset is due to a species-dependent vital effect [Bauch et al., 1997], the remainder may be related to the incorporation of depleted brines and thereby linked to sea ice production. As in modern N. pachyderma, off-equilibrium isotopic excursions were also associated with Heinrich events (with no associated change in surface salinity) but were not evident during the LGM. Hence, high sea ice production and brine addition to the pycnocline were likely characteristic of Labrador Sea conditions during both late Holocene and Heinrich conditions, while a different (low) sea ice production regime characterized the LGM [Hillaire-Marcel and de Vernal, 2008].

Sea ice brines increase water mass density and contribute to deepwater formation [Redfield and Friedman, 1969]. To what extent sea ice brines may have played a larger role as a mechanism of North Atlantic deep water production during stadials, and in particular, Heinrich events, during the last glacial (60 and 10kyr BP) [Dokken and Jansen, 1999; Vidal et al., 1998] is a matter of controversy. Dokken and Jansen [1999] proposed that freshwater additions to surface water during last glacial stadials caused overturning circulation to cease, and instead brine formation functioned as the main deepwater formation mechanism in the North Atlantic. This idea, which became known as the sea ice brine hypothesis, was invoked to explain the observed isotopic depletion in both planktonic and benthic foraminiferal δ18O values in the Nordic Seas. Processes with the potential to shift last glacial δ18O in polar North Atlantic benthic foraminifera were investigated by Bauch and Bauch [2001], who concluded that only by invoking high rates of sea ice production on a seasonally ice-free shelf in the Barents Sea could a −1‰ benthic shift result in the Nordic Seas, and other processes (i.e., not brine formation) remained more likely. While the balance of evidence calls into question the sea ice brine hypothesis for the Nordic Seas during Heinrich events in particular [Bauch and Bauch, 2001; Stanford et al., 2011], the possibility of sea ice changes shifting subsurface seawater δ18O may be applicable elsewhere, especially, as discussed by Stanford et al. [2011], in cold, high-salinity waters [Rasmussen and Thomsen, 2010]. The extent to which sea ice variability holds the potential to produce a deepwater isotopic shift has not been explored in a general circulation model.

Disentangling the individual roles of processes contributing to local seawater δ18O and assessing how each may vary through time (e.g., between different climate states) is a problem uniquely suited to oxygen isotope-enabled climate models. Using an isotope-enabled coupled climate model, we attempt to characterize the isotopic signature of sea ice in seawater under two climate end-members (interglacial and full glacial) and investigate the potential role of sea ice variability in shifting local seawater δ18O.

2 Model and Methodology

We employ the University of Victoria Earth System Climate Model, version 2.9, with oxygen isotopes represented (as math formula) in all model subcomponents containing water, including the ocean general circulation model, atmospheric model, land surface model, and dynamic-thermodynamic sea ice model. Horizontal resolution of each model subcomponent is 3.6° (zonal) by 1.8° (meridional). Non-isotope model physics are described by Weaver et al. [2001] and Meissner et al. [2003], while the implementation of oxygen isotopes is fully documented in Brennan et al. [2012]. Since sea ice variability between interglacial and glacial states is the topic of interest, we summarize the sea ice model in the following section.

The distribution of oxygen isotopes in the model seawater for the preindustrial has been compared against the Seawater O18 Dataset [Schmidt et al., 1999] in Brennan et al. [2012]. The model captures the observed global and regional patterns in δ18Ow (see their Figure 7), although high-latitude surface waters are not as depleted in the model as the data set. Small model-data δ18Owdiscrepancies are unlikely to affect the conclusions of the current study, which is focused on the sea ice contribution of δ18Ow in isolation, as determined by differencing model simulations.

2.1 Sea Ice Model

The standard thermodynamic-dynamic sea ice model is based on work by Maykut and Untersteiner [1971] and the zero-layer ice model by Semtner [1976] and employs the lateral growth and melt parameterization of Hibler [1979]. Model sea ice manifests elastic viscous plastic dynamics, the elastic viscous ice rheology based on work by Hibler [1979] and the plastic component contributed by Hunke and Dukowicz [1997]. The two-category sea ice model operates upon the domain of ocean grid cells at the subgrid level, such that each grid cell is characterized by an open water areal fraction (ao) and an ice-covered areal fraction (ai), with the two categories summing to one (ao+ai=1). Sea ice is assumed to form as a horizontally uniform slab, and snow may accumulate on top of the ice as a single horizontally uniform layer. The layers of sea ice and overlying snow may each vary in height, although the snow layer is limited to a thickness of 10m. The top surface of the sea ice or snow layer may sublimate to the atmosphere. At the ice-ocean interface, ice may grow via accretion or decrease via melt (ablation) (depending upon the balance of ocean heat and ice diffusive fluxes). Brine pockets within sea ice are not explicitly represented, and model sea ice is considered as fresh for purposes of freshwater exchange with the ocean model. Brine rejection is parameterized in the model, such that when sea ice forms, expelled brine is added to the underlying water. A determination of water column stability is performed, and if instability exists then convective vertical mixing ensues [Duffy et al., 1999; Weaver et al., 2001]. The thermodynamic and dynamic equations used in the standard sea ice model are summarized in Weaver et al. [2001].

When sea ice forms in the model, the ice is enriched by 3.0‰ relative to its seawater source. This fractionation factor is identical to that employed in other isotope-enabled coupled models (for example, the Goddard Institute for Space Studies (GISS) [Schmidt, 1999] and Global ENvironmental and Ecological Simulation of Interactive Systems (GENESIS)-MOM [Mathieu et al., 2002] models), although it is slightly larger than the values observed in field studies discussed above (see section 1.2). Snowfall that accumulates on top of sea ice in the snow layer retains its isotopic content separately. Isotopes in sea ice and the overlying snow layer may be transferred to the atmosphere via sublimation and to the surface ocean via melting (no fractionation occurs during either sublimation or melting). The implementation of oxygen isotopes is fully described in Brennan et al. [2012].

2.2 Boundary Conditions for Interglacial and Glacial Climates

To simulate the interglacial and glacial climate states, the model is integrated for 5kyr under preindustrial (year 1800) and LGM (21kyr BP) boundary conditions. These include equivalent CO2 at 284 and 190 ppm, orbital parameters corresponding to year 1800 and 21kyr BP, respectively, and present-day monthly mean wind fields superimposed with wind stress anomalies calculated dynamically as a function of surface pressure anomalies [Weaver et al., 2001] for both climate states. Both interglacial and glacial simulations employ ocean bathymetry and sea level corresponding to modern conditions with a closed Bering Strait (the Bering Strait being too shallow and narrow to meet the minimum number of model grid cells required to simulate flow). The glacial-interglacial sea level change (∼−130 m, less than 3.5% of the average depth of the global ocean) is neglected. The choice of modern ocean bathymetry and sea level in the glacial simulation is based on the experimental design in order to difference the three-dimensional oceanic fields of δ18Ow between interglacial and glacial states.

For the glacial climate, LGM land ice is simulated after the ICE-4G reconstruction of Peltier [1994]. The ICE-4G implementation includes permanent ice shelves extending from the coast over multiple ocean points in the Arctic and Antarctic, preventing ocean-atmosphere exchange and sea ice at those grid cells. ICE-4G includes a reconstruction every 1kyr, from 21kyr BP to present, and is a practical choice for time evolution of ice sheet height and model integrations. In contrast, the more recent ICE-5G reconstruction consists of one static reconstruction at 21kyr BP. While equilibrium simulations are the current focus, the ICE-4G 21kyr BP reconstruction is preferred such that time-evolving extensions to the study can be accommodated at a future point. Differences between ICE-4G and ICE-5G—changes in ice sheet coverage in northern Siberia and along the southeastern edge of the Scandinavian Ice Sheet, and a redistribution of ice thickness (to west central Canada, with slight reductions over central Greenland and the above mentioned areas)—are, in all likelihood, not large enough to have a significant impact on LGM ocean circulation or patterns of sea ice production. For the purposes of this study, ICE-4G provides a sufficient ice sheet distribution to simulate the glacial state.

For both the interglacial and glacial simulations, we use the present-day ocean bathymetry (ICE-4G ocean bathymetry is not used), the same river routing scheme (all runoff from a given river basin is distributed along the coastal points associated with that basin, as shown by Figure 2 in Weaver et al. [2001]). As such, Arctic surface freshwater balance could respond in the model to changes in the amount and timing of glacial runoff delivered via river routing but not a change in the river routing itself or changes in bathymetry or Bering Strait. Arctic surface freshwater balance could also respond to changes in the sea ice regime (i.e., the permanent Arctic sea ice cover in the glacial simulation).

2.3 Modeling Approach

A set of four simulations allows us to isolate the sea ice component of seawater δ18O for the interglacial and glacial climates and calculate the difference in this field between these climate states. From the interglacial and glacial equilibrium states, we perform a pair of simulations that are identical except that during sea ice formation isotopic fractionation either remains on or is turned off. If isotopic fractionation does not occur during sea ice formation, there is no isotopic signature of sea ice in the global ocean. Both interglacial and glacial simulation pairs (with fractionation on and off during ice formation, called “frac-on” and “frac-off” hereafter) are integrated for 3kyr. Table 1 summarizes the experimental design and model simulations.

Table 1. Model Simulations and Experimental Design
BoundaryFractionation DuringDifference of SimulationsDifference of Interglacial
ConditionsSea Ice FormationWith Fraction On and Offand Glacial Seawater δ18Of(seaice)
Interglacial (Preindustrial, year 1800)OnSeawater math formulaIsotopic shift
 Off in seawater
   due to sea ice
Glacial (LGM, 21kyr BP)OnSeawatermath formulavariability

By differencing the pair of simulated seawater oxygen isotope fields (“frac-on” minus “frac-off”), we determine the seawater δ18O field that is due to sea ice for the given climate. For example, glacial seawater δ18O due to sea ice, math formula, is calculated as the difference of the pair of glacial seawater oxygen isotopic fields, math formula. Then, by taking the difference of the seawater oxygen isotopic field due solely to sea ice for the interglacial and glacial climate states (math formula), we estimate the local shift in seawater δ18O attributable to sea ice variability. We perform the analysis for the sea surface (the top ocean model level with a depth of 50m), averaged over the upper water column (the top four ocean model levels, 0–380m), averaged over intermediate waters (ocean model levels 5–12, or 380–2580m), averaged for deep water (ocean model levels 13–19, corresponding to the depth range 2580–6080m), and for the deepest level in the ocean model (kmax), representing bottom water.

The modeling approach, as developed, aims to determine whether changes in sea ice between distinct climate regimes (interglacial and glacial) may shift δ18Ow in bottom water, deep water, intermediate water, surface water, or the sea surface. It bears emphasizing that while LGM surface forcing conditions are used to simulate the glacial state, the modeling objective is the representation of a characteristic glacial climate and not the LGM in particular.

3 Results

3.1 Model Interglacial and Glacial Sea Ice

Monthly mean September and March sea surface temperature and the corresponding maximum and minimum sea ice and snow areal fractions (superimposed by a 1/4-degree coastline by Pawlowicz [2011]) are plotted for the model's Northern and Southern Hemispheres in Figure 1. The modeled interglacial monthly mean sea ice is similar to the corresponding monthly climatological sea ice distributions (1979–1987) constructed from satellite passive-microwave observations (as in Figures1.16–1.19 in Wadhams [2000] after Gloersen et al. [1992], not shown). The sea ice model performance is evaluated in Weaver et al. [2001], where the present-day seasonal climatology for sea ice and snow cover (considered together and separately) is demonstrated to be ranked highly against that of other coupled atmosphere-ocean GCMs (their Table 3). Discrepancies include the formation site of North Atlantic Deep Water which is shifted slightly southward in the model. This in turn allows sea ice cover to persist too far south, since less heat is available to reach the Greenland-Iceland-Norwegian seas (as shown in Figures 18a and 19 in Weaver et al. [2001]).

Figure 1.

Model September and March (a) monthly mean sea surface temperatures (°C) and (b) areal fraction of Northern and Southern Hemisphere sea ice and snow in the preindustrial (left) and LGM (right) simulations. Model areal fraction of snow and sea ice in Figure 1b is superimposed with a 1/4-degree coastline by Pawlowicz (2011).

In the model, the maximum sea ice thickness (>5 m) occurs in the Canadian Archipelago region of the Arctic, and the mean ice thickness in the Arctic Ocean is approximately 1m. In contrast, modeled Antarctic average sea ice thickness is approximately 40cm. A maximum thickness of Southern Hemisphere ice (≤2.1 m) is found near 75°W. The simulated seasonal cycle of Northern Hemisphere sea ice area produces a maximum in early March and a minimum in early September. The Southern Hemisphere cycle reaches its peak in mid-September and its minimum in mid-February. The modeled seasonal cycle is very similar in timing to that observed in the climatology [Gloersen et al., 1992; Wadhams, 2000].

Figure 1 also maps the glacial monthly mean September and March sea ice and snow areal fraction for each hemisphere in the model. Under glacial conditions, perennial ice cover is simulated for the entire Arctic Ocean, in the northern Labrador Sea, and in the northeastern North Atlantic (north of ∼70°N), while winter sea ice extends significantly further south. The modeled LGM winter sea ice edge appears consistent with available recent reconstructions [Kucera et al., 2005; de Vernal et al., 2006]. Northern Hemisphere glacial sea ice volume at its peak is three times that of the interglacial (about 4.5×104 km3versus 1.5×104 km3, respectively).

In the Southern Hemisphere, glacial sea ice covers at least 20% more area at its maximum than under interglacial conditions (approximately 2.5×107 km2versus 2.1×107 km2). Compared to a range of LGM winter and summer ice edge reconstructions for the Southern Hemisphere summarized in Gersonde et al. [2005], the modeled winter ice distribution is visually similar to the reconstructions except for the region south of Australia (90°E to 150°E) where the model simulates less ice, while the modeled summer ice edge is more similar to the EPILOG reconstruction (albeit with less ice simulated in the region of 30°W to 10°E) [Gersonde et al., 2005] than that of CLIMAP [1981].

3.2 δ18Osw Due to Sea Ice for Interglacial and Glacial Climates

The interglacial and glacial seawater δ18O fields due to sea ice (math formulaand math formula) are shown in Figure 2. A positive (i.e., isotopically enriched) sea ice contribution to seawater δ18O corresponds to regions dominated by a net addition of sea ice meltwater, while a negative (i.e., isotopically depleted) sea ice contribution is associated with a net addition of sea ice brine. For example, in the interglacial sea surface, a net brine addition is present in the Labrador Sea and adjacent to Arctic and Antarctic coasts, while a net meltwater addition is apparent in the Arctic Chukchi Sea (Figure 2a). A similar pattern of sea ice isotopic contribution is found when the signal is averaged over the upper water column (Figure 2b). In the model circum-Antarctic, significant brine production results in the largest negative isotopic contribution from sea ice to deep water (Figure 2d, −0.0135‰, found at 65.7°S, 48.6°W). Interglacial bottom waters receive only very diluted isotopic contributions from sea ice, except in the Labrador Sea and Arctic coastal regions, which are dominated by net brine addition (Figure 2e).

Figure 2.

Annual mean seawater δ18O due to sea ice in equilibrium interglacial (left) and glacial (right) climates in the (a, f) model sea surface (top 50m), (b, g) surface waters (averaged over the top 380m), (c, h) intermediate water (380–2580m averaged), (d, i) deep water (>2580 m averaged), and (e, j) bottom water (deepest ocean model level, kmax). Red (blue) regions in the figure correspond to a positive (negative) isotopic component and therefore a net addition of sea ice meltwater (brine).

The glacial sea surface exhibits a markedly different pattern of sea ice meltwater and brine additions in the Northern Hemisphere from the interglacial. The Labrador Sea net brine addition is larger, while the net meltwater addition is focused between southern Greenland and northern Europe, entirely displaced from the Arctic basin (Figure 2f). This pattern extends throughout the surface water column (averaged over the top 380m) (Figure 2g). Glacial deep water contains its largest isotopic contribution from sea ice brine production (−0.0167‰ at 62.1°S, 45°W) in the circum-Antarctic, which appears to be slightly enhanced relative to the interglacial state (Figure 2i). In glacial bottom waters, the most significant depleted (net brine) contribution is found in the Labrador Sea, while the largest enriched (net melt) contribution is in the North Sea (Figure 2j).

Table 2 lists the maximum positive and negative sea ice isotopic contribution to seawater δ18O for the simulated interglacial and glacial sea surface (top 50m), surface waters (0 to 380m, averaged), intermediate water (380 to 2580m, averaged), deep water (>2580 m, averaged), and bottom water (corresponding to the deepest ocean model level). The magnitude of isotopic contributions by sea ice processes to the sea surface and surface waters is small (<0.14‰), while that to deep water is smaller again by an order of magnitude (<0.02‰).

Table 2. Upper and Lower Bounds of the Sea Ice Isotopic Contribution to δ18Ow (in Units of Permil): Interglacial, Glacial, and Interglacial-Glacial Shift (IG-G)
IG-G shift     
max site65.7°N, 171°W65.7°N, 171°W74.7°N, 63°W35.1°S, 41.4°W65.7°N, 171°W
min site62.1°N, 1.8°W76.5°N, 91.8°E74.7°S, 189°W71.1°S, 19.8°W76.5°N, 91.8°E

3.3 Interglacial-Glacial Shift in δ18Osw Due to Sea Ice

The shift in local seawater δ18O due to interglacial-glacial sea ice variability is shown in Figure 3, where positive (negative) values correspond to a more enriched (depleted) interglacial sea ice isotopic contribution to seawater. The interglacial-glacial shifts due to sea ice variability are very small in intermediate water (380–2580m, averaged) and deep water (>2580 m, averaged): ≤0.05‰ and 0.01‰, respectively. Similarly, isotopic shifts in bottom water are insignificant throughout the global ocean, except for the shallow Arctic shelf regions, the Bering Strait, and in Baffin Bay west of Greenland. Only in high-latitude surface waters of the Northern Hemisphere, namely the Labrador Sea and the northeastern North Atlantic, are small shifts (<0.13‰) found.

Figure 3.

Shift in annual mean seawater δ18O (Δδ18O) due to interglacial-glacial variability in sea ice in the (a) model sea surface (top 50m), (b) surface waters (averaged over the top 380m), (c) intermediate water (380–2580m averaged), (d) deep water (>2580 m averaged), (e) and bottom water (deepest ocean model level, kmax). Red (blue) regions in the figure correspond to a positive (negative) isotopic shift, and therefore an interglacial state with increased sea ice meltwater (brine) addition and/or decreased brine (meltwater) addition.

Table 2 summarizes the largest positive and negative isotope contributions from sea ice for each analyzed depth interval (sea surface, upper water column, intermediate water, deep water, and bottom water), along with the maximum positive and negative isotope shift between interglacial and glacial climate states (i.e., the difference of interglacial minus glacial δ18Of(seaice)). Variation of the isotopic shift with depth is shown in Figure 4 for six Northern Hemisphere high-latitude locations where sea surface anomalies are particularly pronounced (see inset map for locations).

Figure 4.

Interglacial-glacial shift (Δ) in the annual mean sea ice component of δ18Ow (permil) at locations indicated on the inset map: 62.1°N, 1.8°W (black diamond), 76.5°N, 91.8°E (black circle), 74.7°N, 167.4°W (red square), 69.3°N, 59.4°W (red star), 62.1°N, 59.4°W (black star), and 65.7°N, 171°W (black square).

4 Discussion

That sea ice varied throughout geologic time between glacial and interglacial climates is unquestionable. Changes in sites of sea ice growth and brine rejection, sites of sea ice melt, and rates of sea ice production at a given location can shift the isotopic signature of sea ice processes in seawater. In the model, we indeed observe differences in interglacial versus glacial sea ice extent, volume, and seasonal cycle of growth and melt (not shown). In order to properly address the issue of isotopic shifts in deep or bottom waters (where benthic foraminifera reside) due to changes in sea ice, it is necessary for the model to represent two distinct types of processes—surface processes and the transport of those surface waters to the deep. It is appropriate to ask whether a coarse resolution model can reasonably advect surface signals to depth. That the model can simulate important global water masses (NADW, AIW, AABW), with deep water forming in the expected regions (North Atlantic and the Weddell Sea) and an overturning circulation that falls within the range of other models [Cao et al., 2009; Eby et al., 2009; Rahmstorf et al., 2005], suggests that the transport of surface signals to depth is robust.

Model results depict negligible shifts in deep water δ18O, indicating essentially no impact from glacial-interglacial sea ice variability on benthic δ18Oc and, therefore, little possibility of sea ice variability contributing error to ice volume (and sea level) reconstructions. Bauch and Bauch [2001] concluded that the sea ice brine hypothesis for shifting benthic carbonate δ18O in northern polar regions is possible but highly unlikely. The results of this study are consistent with that finding: here, the modeled isotopic signature of sea ice is insignificant in intermediate and deep water. A geographically limited impact in the model sea surface and surface waters is found in the northern North Atlantic, where small shifts in δ18Ow may influence planktic δ18Oc. Outside of the northern North Atlantic, little impact of glacial-interglacial sea ice variability is detected in surface waters.

To further gauge the potential effect of the above results on paleoreconstructions, we assess the surface seawater isotopic content in the model at the grid cells located nearest each North Atlantic sediment core site (between 50°N–85°N and 65°W–20°E, N=64), primarily those employed by Glacial Atlantic Ocean Mapping (GLAMAP 2000) [Pflaumann et al., 2003]. The sediment core sites, listed in the supporting information, are mapped in Figure 5. In the case of multiple cores corresponding to the same model grid cell, only one core is selected (to avoid sampling the same model location more than once).

Figure 5.

Locations of North Atlantic sediment cores. Indicated cores are listed in the supporting information.

An additional glacial equilibrium simulation is performed without any seawater enrichment. At the model locations corresponding to the North Atlantic sediment core sites, glacial surface water is enriched by 1.11±0.01‰ (mean ±1σ) due to the accumulation of depleted ice on continents. This is demonstrated in Figure 6 by the offset between the glacial simulation with enriched seawater (black circles) and the glacial simulation with unenriched seawater (gray circles).

Figure 6.

Seawater δ18O sampled from the model grid cell nearest each North Atlantic sediment core (listed in the supporting information) in the interglacial and glacial simulations. The two glacial simulations are with and without ocean isotopic enrichment due to continental ice volume.

In the southernmost region of the domain, interglacial seawater δ18O is consistently more enriched than the (unenriched) glacial: across sites, the offset is positive and of a similar magnitude. In contrast, the northernmost region displays variability in both sign (i.e., enrichment or depletion) and magnitude of the interglacial-glacial offset. The greatest variability, however, is found in the central region. This pattern is indicative of a dominant role for cryosphere-ocean interactions in shifting seawater δ18O north of ∼58°N, while a different set of processes, such as atmosphere-ocean interaction (for example, via changes in evaporation and precipitation) acts to shift seawater δ18O south of ∼58°N.

The contribution of sea ice processes to the isotopic shift in surface seawater is plotted for each core site in Figure 7 (green bars), along with the total shift in seawater δ18O (gray bars). The calculation of the former is described in section 2.3. The latter is calculated as the difference between surface seawater δ18O from the interglacial and unenriched glacial simulations. Hence, the effects of seawater enrichment due to continental ice are not included, while all other interglacial-glacial shifts in the hydrologic cycle and ocean circulation are represented in the plotted shift in seawater δ18O.

Figure 7.

Interglacial-glacial shift in seawater δ18O (gray bars) and the sea ice component of the shift (green bars), sampled from the model grid cell nearest each North Atlantic sediment core (listed in the supporting information and mapped in Figure 5). The total interglacial-glacial shift in seawater δ18O is the difference between the interglacial and unenriched ocean glacial simulations; hence, the effects from ocean enrichment due to glacial continental ice volume are not included.

Indeed, sea ice processes have a negligible isotopic contribution to surface seawater Δδ18O at sites south of ∼58°N. Interestingly, sea ice has almost no impact on Δδ18O north of 70°N, with the exception of HU76-029-033 in Baffin Bay. This suggests that Δδ18O in the northern region of the domain results from changes in atmospheric forcing (e.g., evaporation, precipitation), discharge of freshwater from continents, and/or ocean circulation. In the central region of the domain (58°N–70°N) at sites where the sea ice contribution and the total observed seawater isotopic shift have the same sign, sea ice processes contribute between ∼7.7%and ∼26.5% of the observed seawater δ18O shift. At those sites where the sea ice component of the shift is in the opposite direction of the observed shift in seawater δ18O (e.g., 23056-2 or DSDP 336), the sea ice component may be sizeable, but competing effects cause local seawater δ18O to change in the opposite direction. These effects amount to an unknown combination of changes in evaporation, precipitation, freshwater discharge from land, or ocean circulation (potentially each with an isotopic shift of unknown magnitude). Similarly, at sites where the sea ice component has the same sign but is larger than the observed seawater Δδ18O (e.g., V28-56 or DSDP 337), again changes in other hydrological fluxes or circulation work to effectively cancel a portion of the contribution of sea ice processes to the isotopic shift at the site.

5 Conclusions

The current study presents a global model analysis of the contribution of sea ice to seawater isotope chemistry, and the potential for variable sea ice to shift local seawater δ18O. Interglacial and glacial sea ice regimes produce distinct spatial distributions of oxygen isotopes in sea water; however, the modeled contribution of sea ice to seawater δ18O for interglacial and glacial states is small (<0.14‰). Beyond the small shift (<0.13‰) found in model surface waters in a limited region (the Labrador Sea and northeastern North Atlantic), there is no evidence that interglacial-glacial sea ice variability may shift δ18Ow by more than a negligible amount. Between 58°N and 70°N, sea ice contributes only ∼7−26% of the total change in sea surface δ18Ow (after accounting for continental ice volume) at the GLAMAP North Atlantic sediment core sites [Pflaumann et al., 2003] sampled in the model. The implications include little role for sea ice induced changes in δ18Oc, both in surface waters and in deep waters (such as was proposed in the sea ice brine hypothesis).

We acknowledge the following potential caveats for the present analysis. First, the particular results depend on the distribution of sea ice, sea ice growth and melt, and the subsequent ocean transport of ice brine and meltwater in glacial and interglacial climates simulated by one model. Second, in addition to the effect of sea ice variability investigated here, the variability of other processes that can affect δ18Olocal, including additions of river runoff [e.g., Schlosser et al., 2002] and ice sheet meltwater, may obviously also contribute uncertainty. Third, here we consider variability in seawater δ18O, a signal that is recorded in biogenic carbonate concurrently with superimposed temperature effects and vital effects of the calcifying organisms. This latter biological aspect is not considered but could potentially introduce additional, nontrivial error. In fact, the dominant species of high-latitude foraminifera, N. pachyderma, thrives in very high-salinity waters and, as such, may be more likely to record seawater δ18O with a significant component of sea ice brine [Ravelo and Hillaire-Marcel, 2007]. Finally, we note that the presented results are applicable for the current continental configuration and for climate states within the range of atmospheric CO2 levels considered in this study. As discussed by Raymo [1994], Northern Hemisphere glaciation began approximately 2.5 Ma ago, initially with seasonal Arctic sea ice and later (∼1.6 Ma) with perennial sea ice cover, while perennial sea ice existed in the Weddell Sea after 2.46 Ma. Together these constraints may limit the time scale of relevance of the present results to possibly the last 2.4 Myr (Pliocene to modern).


We are grateful for support from the Canadian Foundation for Climate and Atmospheric Sciences, the Natural Sciences and Engineering Research Council of Canada CREATE program, and the Australian Research Council Future Fellowship program.