Twentieth century sea‐ice trends in the Ross Sea from a high‐resolution, coastal ice‐core record

Abstract

We present the first proxy record of sea‐ice area (SIA) in the Ross Sea, Antarctica, from a 130 year coastal ice‐core record. High‐resolution deuterium excess data show prevailing stable SIA from the 1880s until the 1950s, a 2–5% reduction from the mid‐1950s to the early‐1990s, and a 5% increase after 1993. Additional support for this reconstruction is derived from ice‐core methanesulphonic acid concentrations and whaling records. While SIA has continued to decline around much of the West Antarctic coastline since the 1950s, concurrent with increasing air and ocean temperatures, the underlying trend is masked in the Ross Sea by a switch to positive SIA anomalies since the early‐1990s. This increase is associated with a strengthening of southerly winds and the enhanced northward advection of sea ice.

1 Introduction

Southern Ocean sea ice exerts a profound impact on physical and biological processes in Antarctic waters and plays a critical role in modulating global climate via its influence on ocean‐atmosphere heat exchange, global albedo, and ocean circulation [Yuan and Martinson, 2000]. Little is known about Antarctic sea‐ice variability before 1979, when near‐daily multichannel passive microwave imagery became available. Thus, an extended proxy record can provide important insights into the response of sea ice to warming of Southern Ocean waters [Jacobs et al., 2002] and can address uncertainties related to the response of the Southern Ocean and Antarctic margin to climate change, and particularly the role of sea ice as a feedback in this system [Mayewski et al., 2009]. As sea‐ice growth in the Ross Sea is one of the three main sources of Antarctic Bottom Water, a major component of the global ocean circulation system [Johnson, 2008], an understanding of the physical mechanisms underlying its variability is particularly important.

We investigate variability in sea‐ice area (SIA) in the Ross Sea between 1883 and 2012 using an ice‐core geochemical record from Whitehall Glacier (WHG) in northern Victoria Land (Figure 1a) and satellite‐derived SIA data since 1979. Relevant ice‐core proxies are compared with the satellite record to evaluate the sensitivity and robustness of our reconstructions, to quantify decadal‐scale variability, and to discuss the possible drivers of change.

Figure 1

Ice‐core location and key features of the Ross Sea. (a) Landsat Image Mosaic of Antarctica (LIMA) showing the Whitehall Glacier (WHG) ice‐core site with open water directly to the east. (b) Frequency distribution map of 5 day back trajectories from WHG, weighted by daily ERA‐Interim snowfall data [for details see Sinclair et al., 2013]. This shows the dominant flow path of major storms to the ice‐core site; sweeping across the margin of West Antarctica, around the Ross Sea coast and then flowing north over the ice shelf edge and Ross Sea Polynya. (c) Sea‐ice concentrations around Antarctica in December 2004 (from Advanced Microwave Scanning Radiometer–EOS satellite data from the University of Bremen), showing the seasonal extent of the Ross Sea Polynya. Boxes mark the spatial extent of part b of this figure. The East Antarctic Ice Sheet (EAIS) and West Antarctic Ice Sheet (WAIS) are labeled, along with: “1”—the Ross Sea Polynya, and “2”—the Ross Ice Shelf.

2 Ice‐Core Overview

2.1 Ice‐Core Site

The WHG ice‐core site (72°54′ S, 169°5′ E, 400 m asl) lies 12 km from the coast in an area of high snow accumulation [Sinclair et al., 2012]. The regional climate is dominated by the interaction between continental air flowing downslope from East Antarctica and the regular passage of low pressure centers in the circumpolar trough [Sinclair et al., 2013] (Figure 1b). It is also proximal to the Ross Sea Polynya (RSP, Figure 1c), a persistent open‐water region formed by the constant offshore movement of sea ice. The RSP is the largest consistently forming polynya on Earth, a major engine for sea‐ice generation and the most biologically productive region in the Southern Ocean [Arrigo et al., 2008]. The size of the polynya and the rate of sea‐ice production are directly linked to the strength of southerly winds along the western margin of a low pressure center, the Amundsen‐Bellingshausen Seas Low (ABSL), and katabatic outflow from the continental interior [Drucker et al., 2011]. Stronger southerly winds in the western Ross Sea increase the northward advection of sea ice, so that older ice is moved offshore and replaced by newly formed frazil ice [Arrigo et al., 2008; Drucker et al., 2011].

2.2 Ice‐Core Extraction, Processing, and Dating

The WHG ice core was drilled under the framework of the International Trans‐Antarctic Scientific Expedition in 2006 to a depth of 105 m. The ice core was processed at the New Zealand Ice Core Laboratory (GNS Science) (see supporting information for details). Due to the exceptionally high annual accumulation at this site (61 cm water equivalent yr⁻¹), the sample resolution is 62 samples yr⁻¹ from 1979 onward and 44 samples yr⁻¹ prior to 1979. This resolution gives an unequivocal annual layer count and enabled the development of a precise (sub‐seasonal) age model [Sinclair et al., 2012]. Stable isotopes, δ¹⁸O and δD (n = 5738), and major ions (n = 1168) were measured along the length of the ice core. All geochemical analyses were conducted at GNS Science, and analytical procedures are described in the supporting information.

3 Sea‐Ice Area and Geochemical Proxies

3.1 Sea‐Ice Area

This study uses satellite‐derived SIA data for the Ross Sea from the Nimbus‐7 Scanning Multichannel Microwave Radiometer (SMMR) and the Special Sensor Microwave/Imager (SSM/I), [Comiso, 1999]. The time series (1979–2012) of monthly sea‐ice area (Bootstrap algorithm) was obtained from the National Snow and Ice Data Center (http://nsidc.org/data/smmr_ssmi_ancillary/area_extent.html#smmr_ssmi). The Ross Sea geographical boundaries are defined as the area between 160°E and 130°W; the southern boundary is the continental margin or Ross Ice Shelf, and the northern boundary is the sea‐ice edge. Total ice‐covered area is defined as the area of each pixel with at least 15% ice concentration multiplied by the ice fraction in the pixel (0.15 to 1.00).

3.2 Major Ions

In coastal areas, the dominant fraction of inorganic ions, Na⁺, Mg²⁺, Ca²⁺, Cl⁻, and SO₄²⁻, are derived from marine aerosols and can indicate changes in atmospheric transport strength and sea‐ice cover [Wolff et al., 1998; Bertler et al., 2011]. We find a high degree of covariability between inorganic ions (r > 0.74, p < 0.001), with concentrations corresponding to seawater ratios indicating a marine moisture source.

Methanesulphonic acid (MSA) concentrations in snow and ice are derived from dimethylsulphide emissions from marine algae and phytoplankton. In the southern Ross Sea, MSA concentrations reflect ocean productivity and open‐water area in the RSP [Rhodes et al., 2009], while in the Indian Ocean sector of the East Antarctic coastline, MSA appears to originate from sea‐ice algae which results in a positive relationship between sea‐ice extent and MSA [Curran et al., 2003].

The seasonal cycle of MSA and primary productivity (PP; see supporting information) are shown in Figure 2a. While the lag between the October/November peak in MSA concentrations and the December/January peak in PP suggests that some of the spring MSA at WHG may be derived from sea‐ice algae, MSA concentrations are broadly in phase with PP on an annual time scale. The close association between seasonal cycles of MSA and PP indicates that the primary source of MSA at WHG is marine phytoplankton from open polynya waters south of the ice‐core site.

Figure 2

Seasonal cycles in ice‐core geochemistry, primary production (PP), and sea‐ice area (SIA) in the Ross Sea. (a) The seasonal cycle in MSA (1979–2004) and primary production (1998–2004). Information on other inorganic ions (marine aerosols) is provided in Auxiliary Table 1. It is of note that there is no notable difference between the seasonal cycles in MSA, 1998–2004, compared to the longer, 1979–2004 time series. (b) The seasonal cycle in SIA, 1979–2004, and d‐excess for the time periods 1979–1993 and 1993–2004. (c) Mean annual SIA and d‐excess centered on September with the gray line showing the bivariate regression between these variables. One SIA outlier (open diamond) was removed from the regression analysis, and uncertainties were found with a Monte Carlo simulation (see supporting information). Lower (upper) 95% confidence intervals around the mean slope and intercept are −0.0418 (−0.0201) and 2.5442 (2.7285), respectively.

Table 1. Mean d‐excess and SIA and the Relative Difference in SIA, MSA, and Na⁺ From Pre‐1954 Values

Years	Mean d‐excess (‰)	SIA (mill. km^2)	Diff. SIA (%)	Diff. MSA (%)	Diff. Na^+ (%)
1883–1954a^a Time periods correspond to those in Figure 3.	8.2	2.382b^b SIA estimated from the linear regression shown in Figure 2c.
1954–1980a^a Time periods correspond to those in Figure 3.	12.0	2.263b^b SIA estimated from the linear regression shown in Figure 2c.	−5.0	41.4	−7.1
1980–1993a^a Time periods correspond to those in Figure 3.	9.5	2.341c^c SIA from satellite observations 1979–2012. Note that 2012 is the limit of available satellite data.	−1.7	−8.4	8.3
1993–2004a^a Time periods correspond to those in Figure 3.	4.6	2.502c^c SIA from satellite observations 1979–2012. Note that 2012 is the limit of available satellite data.	5.1	−15.9	−19.7
1993–2012		2.503c^c SIA from satellite observations 1979–2012. Note that 2012 is the limit of available satellite data.	5.1

• ^a Time periods correspond to those in Figure 3.
• ^b SIA estimated from the linear regression shown in Figure 2c.
• ^c SIA from satellite observations 1979–2012. Note that 2012 is the limit of available satellite data.

3.3 Deuterium Excess

Deuterium excess, derived from stable isotope data (d‐excess = δD − 8 ⋅ δ¹⁸O), is a tracer of the origin of precipitation. It is primarily controlled by conditions at the evaporative moisture source, particularly relative humidity and sea surface temperature (SST). These parameters control diffusive processes at the air‐ocean boundary and produce a positive relationship between SST and d‐excess [Merlivat and Jouzel, 1979; Ciais and Jouzel, 1994].

Figure 2b shows the seasonal cycle of SIA and d‐excess. SIA in the Ross Sea reaches its seasonal maximum in late winter/spring (August–October) due to the rapid growth of sea ice in the austral winter and the wind‐driven advection of ice northward. Soon after this time, weakened sea ice around the RSP is susceptible to melt and early breakout, increasing the open‐water area in the southern Ross Sea [Arrigo et al., 2008]. d‐excess shows an inverse seasonal cycle to SIA, and it is therefore possible that the d‐excess signal in the ice‐core record relates partly to the availability of high‐latitude moisture (with relatively low SST). Thus, in years with strong southerly flow and increased ice advection northward, SIA increases alongside the decay and breakout of ice in the southern Ross Sea. Major moisture‐bearing storms that pass over the southern Ross Sea would incorporate additional moisture with a low d‐excess signal that is mixed with moisture from lower‐latitude source regions. Moreover, it has been shown that storm pathways in warmer months are centered over the Ross Sea and in all seasons track along the western margin of the embayment over the RSP region [Sinclair et al., 2013], which is consistent with incorporation of moisture derived from higher latitudes.

We test this hypothesis by investigating the relationship between ice‐core glaciochemistry and SIA between 1979 and 2004. We find a significant negative relationship between mean annual (March–February) SIA and d‐excess (r = −0.53, p = 0.006, n = 26; Figure 2c) with uncertainties in the regression parameters determined by Monte Carlo simulations (see supporting information).

4 Twentieth Century Sea‐Ice Variability

4.1 Sea‐Ice Area, 1883–1980

Between 1883 and around 1950 ice‐core MSA, Na⁺ and d‐excess exhibit relatively low variability (Figures 3a–3c), but from the 1940s d‐excess began to increase, with a sustained period of high d‐excess beginning in 1954. The beginning of this period of anomalously high d‐excess as defined by applying a quadratic fit to data from 1950 onward, and identifying the year when the first derivative of this curve increased above zero and d‐excess increased above the long‐term mean (see quadratic fit in Figure 3c). The 1954–1980 mean d‐excess is 3.8‰ higher than pre‐1954 values (Table 1).

Figure 3

Time series of ice‐core geochemistry, 1883–2004. (a) MSA concentrations (5‐point and 25‐point running mean (heavy line)) and mean values for each time period of interest noted above the curve. The standard deviation of the data set is indicated on the plot (1σ = 27 ppb). (b) Na⁺ concentrations (5‐point and 25‐point running mean (heavy line)) and mean values for each time period of interest noted above the curve. The standard deviation of the data set is indicated on the plot (1σ = 2700 ppb). (c) Monthly d‐excess plotted around the time series mean with concentrations for each shaded time period noted above the curve. The thicker line shows the annual (September‐centered) mean. The standard deviation of the data set is indicated on the plot (1σ = 3.7‰), and the yellow line shows the quadratic fit to data between 1950 and 1962. The blue curve below shows the estimated change in SIA (ΔSIA) relative to pre‐1954 values. (d) Wavelet power spectra of d‐excess normalized by the variance of the data (the square root has been taken before plotting; values above 3 = 1.73 are significant from white noise at the 95% confidence level; and edge effects are negligible below the white dashed lines).

Although d‐excess explains ~30% of the variability in SIA, indicative that other factors are also important, the relationship is highly significant, and our regression model suggests at least a 5% decrease in SIA for this 26 year time period. This decrease is also reflected in the MSA record, which increases by 41% between 1954 and 1980 (Figure 3a; Table 1). This increase in MSA is consistent with the d‐excess record and indicates that open‐water area, and hence marine biological productivity, increased over this time period in the western Ross Sea.

Wavelet analysis of the d‐excess record (see supporting information) shows several dominant temporal cycles that support the interpretation of relatively stable SIA in the Ross Sea until the 1950s and variable conditions since then (Figure 3d). Oscillations with a period of ~20 years are evident in the wavelet power in the beginning decades of the record until about 1930 and are also apparent in the smoothed time series (Figure 3c). The increase in d‐excess in the mid‐1950s produces increased power at periods greater than 30 years.

Independent support for the mid‐century change in sea‐ice conditions is found in the analysis of whaling records. Although the use of these data in sea‐ice edge reconstructions has been controversial [Ackley et al., 2003], de la Mare [1997, 2009] found a high correlation between the southernmost whaling positions and direct observations of the ice edge around the continent. The whaling data suggest that the summer sea‐ice edge moved 2.8° southward between the mid‐1950s and early 1970s, which corresponds to a 20% decrease in SIA. Although the scarcity of whaling data precludes a regional estimate of SIA change, these continent‐wide averages correspond well with the shifts observed in the WHG record in the mid‐1950s. A mid‐century change in the sea‐ice regime is also found in sea‐ice reconstructions for East Antarctica (80°E–140°E) using the Law Dome ice‐core MSA record, which shows a 20% reduction in sea‐ice extent between the 1950s and mid‐1990s [Curran et al., 2003].

4.2 Sea‐Ice Change Over the Past Two Decades

Figure 4 shows the d‐excess and SIA time series from 1979 onward. While the mean annual SIA shows an upward trend since 1979, there was a widely reported increase in SIA in this region in the early 1990s [Comiso et al., 2011; Parkinson and Cavalieri, 2012], which is evident in the monthly SIA anomaly data in Figure 4b. Concurrent with the switch to positive SIA anomalies, the WHG d‐excess record shows a sharp decrease in 1993 (Figure 4a) with generally negative d‐excess anomalies for the remainder of the ice‐core record. In the 1970s, a series of oscillations with periods of ~15 years create a peak in wavelet power (Figure 3c).

Figure 4

Ice‐core d‐excess and sea‐ice area anomalies. (a) Monthly d‐excess plotted around mean values, 1979–2005, showing a marked drop in 1993. The heavy line shows the annual (September‐centered) mean. (b) Monthly sea‐ice area (SIA) anomalies for the Ross Sea from the National Snow and Ice Data Center (see supporting information), 1979–2012, with the heavy line showing the annual (September‐centered) mean SIA.

Using mean annual satellite‐derived SIA data and proxy SIA data for the early part of the ice‐core record, we estimate a 5% increase in SIA for 1993–2012 compared to pre‐1954 values (Table 1). MSA and Na⁺ concentrations also decrease in concert with increasing SIA, with a 16% decrease in MSA and a 20% decrease in Na⁺ after 1993, indicating that increased SIA and decreased open water adjacent to the ice‐core site caused reduced marine aerosol input to the record (Figures 3a–3b; Table 1).

A marked increase in SIA in the 1990s [Comiso, 1999] suggests that the low d‐excess is at least partly driven by increased SIA combined with large, more active polynyas supplying additional high‐latitude moisture to the ice‐core budget. To estimate the contribution of high‐latitude waters to snowfall at WHG, we apply a mixing model incorporating the estimated d‐excess of moisture evaporating from polynya waters into the pre‐1954 mean d‐excess signal (see supporting information). Our results suggest that an increased contribution from high‐latitude waters of between 30% and 50% is required to explain the shift observed in the early 1990s.

5 Causal Mechanisms for Sea‐Ice Area Change in the Ross Sea

Atmospheric and circumpolar water temperatures around the margin of West Antarctica have been increasing since the 1950s [Jacobs et al., 2002; Mayewski et al., 2009], and the wind regime on the western margin of the ABSL has been intensifying [Stammerjohn et al., 2008; Holland and Kwok, 2012], providing several potential drivers of the widespread SIA decrease at this time. Although the seasonal bias of oceanic data makes the investigation of the linkages between ocean temperature and sea‐ice cover difficult, previous work [Meredith and King, 2005] has found that strong‐surface intensified warming west of the Antarctic Peninsula has caused a positive feedback whereby surface warming is caused by the reduction in sea‐ice production, which further enhances atmospheric warming and reductions in sea‐ice cover. It is therefore possible that warming of the atmosphere and ocean in the later part of the 20th century may also have played a role in reducing SIA in the Ross Sea.

The increase in SIA since the early‐1990s in the Ross Sea is strongly coupled to atmospheric circulation [Renwick et al., 2012]. There is evidence that the enhanced production of sea ice in this region is linked to a deepened ASL, particularly in autumn [Turner et al., 2009], and atmospheric forcing associated with the Southern Annular Mode (SAM) and the El Niño Southern Oscillation (ENSO), [Goosse et al., 2009]. Increased covariability between the positive phases of SAM and ENSO [Fogt et al., 2011] results in a deepened ABSL, which in turn acts to strengthen southerly flow along the western Ross Sea coast and increase the wind‐driven advection of sea ice northward [Yuan, 2004; Stammerjohn et al., 2008; Holland and Kwok, 2012].

The recent increase in sea ice in the Ross Sea appears to have reversed the trend toward decreasing SIA that began in the 1950s, and is in sharp contrast to the ongoing and rapid declines observed around much of the west Antarctic coastline [Parkinson and Cavalieri, 2012]. This has resulted in a “sea‐ice dipole” between the Ross Sea and the Amundsen‐Bellingshausen Sea coast where dramatic declines are linked to the increased strength of relatively warm northerly winds [Yuan, 2004; Stammerjohn et al., 2008] and increasing atmospheric and ocean temperatures [Meredith and King, 2005].

6 Summary and Conclusions

Our geochemical record (d‐excess, MSA, and marine aerosols) demonstrates a rapid change from stable SIA prior to the 1950s to a highly variable system since then. Other parts of the Antarctic coastline experienced a 20–25% decrease in SIA beginning in the 1950s [de la Mare, 1997; Curran et al., 2003] alongside an ongoing increase in Southern Ocean temperatures and atmospheric warming [Jacobs et al., 2002; Meredith and King, 2005; Mayewski et al., 2009]. In the Ross Sea, we estimate a contemporaneous decrease in SIA of at least 5% between the mid‐1950s and 1980. This decrease in SIA was accompanied by a 41% increase in MSA concentrations, which indicates a rapid and sustained period of increased open‐water area and primary productivity. However, while SIA has continued to decline rapidly in areas such as the Amundsen‐Bellingshausen Sea in recent decades [Comiso et al., 2011], SIA has increased in the Ross Sea since the early 1990s. Our data support the view that the recent increase in SIA in the Ross Sea is caused by changes in atmospheric circulation, most notably stronger southerly winds in the western Ross Sea and enhanced wind‐driven advection of sea ice away from the RSP. The strengthened southerly flow and northward advance of the sea‐ice margin in the Ross Sea appears to have reversed the trend toward decreasing SIA that commenced in the 1950s and has continued in adjacent sectors of Antarctica.

Acknowledgments

This project was funded by the New Zealand Ministry of Business, Innovation and Employment (contracts: VICX0704, CO5X0202, RDF‐VUW1103) and a N.Z. Foundation for Research, Science, and Technology Postdoctoral Fellowship awarded to K.E.S. (contract CO5X0902). Antarctica New Zealand and the U.S. Antarctic Program provided logistical support for ice‐core extraction (project K049), and we thank the Alfred Wegener Institute for the loan of their shallow ice‐core drilling system. We are also grateful to the field team, including Julia Bull, Alex Pyne, and Sepp Kipfstuhl, and Rebecca Pyne, Jessica Dallas, and Gemma Wihare for laboratory assistance, and the staff at the Stable Isotope Laboratory at GNS Science for analytical work.

The editor thanks two anonymous reviewers for their assistance in evaluating this paper.