Oceans and Ice Branch, Laboratory for Hydrospheric Sciences, NASA/Goddard Space Flight Center Greenbelt, Maryland

Scripps Institution of Oceanography, University of California San Diego, La Jolla, California

One of us learned a Russian proverb from Andrey Proshutinsky at the University of Alaska: "When wind blows, ice goes." While there is much wisdom in this saying when applied to the synoptic scale, it appears not to apply to year-to-year sea ice changes. In an earlier study, White and Peterson [1996] reported coherent patterns of interannual variability in the Southern Ocean among sea surface temperature (SST), meridional surface wind (MSW), and sea ice extent (SIE), with warm SST and poleward MSW anomalies associated with SIE retracted back toward Antarctica. The conclusion at that time was that poleward meridional winds drive the sea ice edge back toward Antarctica due to action by the associated wind stress. On further analysis of sea ice concentration (SIC) variability internal to the ice pack, we now find this conclusion no longer tenable. Rather, we find evidence supporting a different hypothesis; that is, the thermal inertia associated with SST anomalies provides a dominant factor influencing the amount of sea ice formed each winter around Antarctica. This does not preclude the wind-driven formation of coastal polynyas [Comiso and Gordon, 1996].

We use two types of data here to demonstrate the relationship between interannual SST and SIC variability associated with the Antarctic circumpolar wave (ACW) [White and Peterson, 1996; Jacobs and Mitchell, 1996; Peterson and White, 1998]. The 1982-1996 SSTs are from the data set prepared by Reynolds and Marsico [1993]. The 1979-1996 SICs were prepared at NASA/Goddard Space Flight Center and obtained from the National Snow and Ice Data Center [1998]. The SICs were processed into SIEs (in degrees latitude, using a 15% ice edge criterion [Gloersen et al., 1992] and average SICs (ASIC, in percent) extending from the sea ice edge to Antarctica in 1° longitude bins.

All time sequences were averaged by month, with residuals (anomalies) about the mean annual cycle formed by subtracting from the monthly averages their respective long-term mean (averages by month over their shared record length). Subsequently, monthly residuals were filtered with a band-pass filter [Kaylor, 1977] with half power points at 3 and 7 years to reduce, among other things, the seasonal cycle remaining in the residuals. Finally, all time sequences were arranged into 6-month averages of June -July -August -September -October -November (JJASON) data, yielding one value for each calendar year.

Time-longitude diagrams(Figure 1)obtained from the time series of SST, SIE, and ASIC residuals from 1979 to 1996 reveal the ACW in all three variables. The ACW is more clearly apparent in SST than in SIE and ASIC, the latter confined to the eastern Indian, Pacific, and western Atlantic Ocean sectors of the Southern Ocean, weak or absent in the eastern Atlantic and western Indian Ocean sectors. This may be related to the observation of Peterson and White [1998] that the ACW in SST residuals follows the Antarctic Circumpolar Current equatorward toward Africa after passing through Drake Passage. But the zonal lag cross correlation computed around the entire globe finds the three variables significantly correlated at 0° phase lag, with warm SST, retracted SIE, and reduced ASIC residuals fluctuating in phase with one another. This indicates that warm SST and poleward MSW residuals are associated not only with retracted SIE [White and Peterson, 1996] but also a reduced ASIC averaged along a line extending from the sea ice edge to Antarctica. Thus the sea ice pack is less concentrated on average in proximity to warm SST and poleward MSW residuals.

In order to obtain a more detailed view of circumpolar wave propagation in the sea ice pack we decomposed the yearly time sequences of JJASON SIC residuals using complex empirical orthogonal function (CEOF) analysis [Preisendorfer, 1988]. We then obtained a revised data sequence (Plate 1)by reconstruction from the first two CEOF modes, representing 55% of the original variance in the residuals. This allows us to focus on the global-scale ACW in these records, independent of shorter-scale interannual variability. The color scale emphasizes positive (red) and negative (blue) SIC residuals. A number of residual SIC crest/trough patterns can be seen propagating most of the way around Antarctica in the winter ice pack. A large residual SIC crest (red) was initiated in 1981 near 90°E; it can be seen to propagate eastward into the vicinity of the Ross Sea by 1984 and continue on into the Amundsen and Bellingshausen Seas by 1986; it moved eastward past the Antarctic Peninsula through Drake Passage in 1987 and on into the Weddell Sea in 1988. This SIC crest propagated most of the way around the globe in 7-8 years at an average speed of ~0.1 m s^-1. A second residual SIC crest formed near 40°E in 1985; it can be seen propagating eastward into the Ross Sea by 1988, on into the Bellingshausen Sea in 1990, and through Drake Passage in 1991 into the Weddell Sea in 1992, taking ~7 years to traverse this path. During these transits and others involving residual SIC troughs (blue), the eastward propagation of the ACW in the winter ice pack finds maximum variability occurring along the sea ice edge and extending 400-600 km into the winter pack. So where the winter sea ice edge is relatively close to Antarctica, these SIC crests penetrate from the sea ice edge to the coast (for example, near 150°E, east of the Ross Sea, and near 60°W over the Antarctic Peninsula); where the winter sea ice edge is relatively far from Antarctica, these SIC crests only weakly penetrate from the sea ice edge to the coast (for example, in the Ross Sea and Weddell Sea). This representation of the ACW in the SIC residuals in the interior of the winter ice pack is characterized by the same global zonal wavenumber 2 pattern found in SST residuals (Figure 1).

To illustrate more clearly this relationship between SST and adjacent SIC residuals, we mapped SST residuals over the Southern Ocean together with SIC residuals over the adjacent winter ice pack, both averaged over JJASON for the years 1983-1986 (Plate 2). We pick up the ACW in 1983 with negative SST and positive SIC residuals near 0° and 180° and positive SST and negative SIC residuals near 90°E and 90°W. Eastward propagation of the covarying SST and SIC residuals can be seen following from one winter to the next over the 4 years represented; thus colder (warmer) SSTs give rise to higher (lower) SICs. What is remarkable about this Plate 2 is the continuation of the contour patterns in the SST residuals into the SIC residuals across the boundary between the open ocean and the winter ice pack. This is particularly true in the longitude domain extending from west of the Ross Sea into the Amundsen-Bellingshausen Seas, where both SIC and SST residuals are more clearly apparent and perhaps mutually reinforcing. At times when higher SICs are adjacent to warmer SSTs (for example, at 0° longitude in 1985), the SIC residuals appear to be decaying. The entire covarying SST and SIC residual patterns can be seen rotating eastward around Antarctica over the 4 years. A complete rotation is not achieved in 4 years, requiring 8-10 years as indicated in Figure 1 and Plate 1. Thus we find warmer (colder) SSTs over the Southern Ocean associated with lower (higher) SICs over the winter ice pack around Antarctica during JJASON when sea ice extent was near its annual maximum.

Others have inferred that sea ice variability is statistically correlated to the El Niño-Southern Oscillation [van Loon and Shea, 1987; Carleton, 1988; Gloersen, 1995; Gloersen and Mernicky, 1998; Yuan and Martinson, 2000]. We find evidence that the interannual variations in SIC are physically linked to variations in the adjacent SST, in accord with an earlier hypothesis linking sea ice thickness to heat content in the Southern Ocean [Martinson, 1993]. Since the sea ice field disappears at the beginning and reemerges at the end of every austral summer, we hypothesize that upper ocean temperatures carry the memory of the ACW into the sea ice field from one winter to the next. We argue that this memory cannot be maintained in the covarying MSW residuals, the latter observed earlier [White and Peterson, 1996; Jacobs and Mitchell, 1996; Peterson and White, 1998] to be poleward (equatorward) over warm (cool) SST residuals [White, et al., 1998]. In support of this argument, the wind stress associated with poleward MSW residuals would necessarily compact the sea ice in a developing winter ice pack. Because of the relatively high heat capacity of the ocean, we can think of at least two mechanisms where SST fluctuations affect SIC residuals. First, SST residuals may influence the formation of SIC residuals in the interior of the developing winter ice pack simply through its underlying thermal inertia [Martinson, 1993]. Second, warm SST residuals may heat the overlying air carried by poleward winds over the developing winter ice pack. The significance of this covarying relationship between SST and SIC residuals each winter is that it offers the potential for predicting winter SIC residuals from one year to the next. This is similar to the use of SST residuals in the ACW in predicting year-to-year changes in air temperature and precipitation over New Zealand [White and Cherry, 1999] and Australia [White, 2000].

Acknowledgements. We thank A.E. Walker for organizing and analyzing the data sets. The NASA Office of Polar Programs and the NOAA Office of Global Programs supported this work.

P. Gloersen, Oceans and Ice Branch, Laboratory for Hydrospheric Sciences, NASA/Goddard Space Flight Center, Greenbelt, MD 20771.

(per.gloersen@gsfc.nasa.gov)

W.B. White, Scripps Institution of Oceanography, University of California San Diego, La Jolla, CA 92093-0230. (wbwhite@ucsd.edu)

(Received January 11, 2000; revised October 10, 2000; accepted October 10, 2000.)

__________

Copyright 2001 by the American Geophysical Union

Paper number 2000JC000230

0148-0227/01/2000JC230$09.00

Figure. 1. Time/longitude plots of (a) sea ice extent (SIE) residuals (0.5° longitude intervals), (b) average sea ice concentration (ASIC) residuals in 1° sectors (5% intervals), and (c) sea surface temperature (SST) residuals (0.25°C intervals). All shaded contours are negative values. Residuals were obtained by subtracting the shared 18-year average of June-November; then a band-pass filter further smoothed each parameter (see text). (d) Time-lag correlations shown for the SIE versus ASIC, SIE versus SST, and ASIC versus SST.

Figure. 1. Time/longitude plots of (a) sea ice extent (SIE) residuals (0.5° longitude intervals), (b) average sea ice concentration (ASIC) residuals in 1° sectors (5% intervals), and (c) sea surface temperature (SST) residuals (0.25°C intervals). All shaded contours are negative values. Residuals were obtained by subtracting the shared 18-year average of June-November; then a band-pass filter further smoothed each parameter (see text). (d) Time-lag correlations shown for the SIE versus ASIC, SIE versus SST, and ASIC versus SST.

Plate 1. A 16-year sequence of reconstructed sea ice concentration (SIC) residuals [National Snow and Ice Data Center, 1998] averaged for June-November for each year from 1980 to 1995, depicting the eastward progression of waves in the SIC residuals. These residuals were reconstructed from the first two complex empirical orthogonal functions obtained from band-pass filtering (see text) differences between each 6-month average and their 16-year average.

Plate 1. A 16-year sequence of reconstructed sea ice concentration (SIC) residuals [National Snow and Ice Data Center, 1998] averaged for June-November for each year from 1980 to 1995, depicting the eastward progression of waves in the SIC residuals. These residuals were reconstructed from the first two complex empirical orthogonal functions obtained from band-pass filtering (see text) differences between each 6-month average and their 16-year average.

Plate 2. Four of the 16 years of SIC residuals shown in Plate 1 are replotted here in a polar stereographic projection, along with similarly averaged and filtered sea surface temperatures (SST) residuals [Reynolds and Marsico, 1993]. Here the negative SST residuals are shown as "warm" colors to enhance the remarkable connection between colder (warmer) SST patterns and more (less) dense SIC patterns as they progress eastward from ~90°E to 45°W. The white band separating the two data fields arises from the differing southernmost limit of the SST data and the northernmost limit of the SIC data.

Plate 2. Four of the 16 years of SIC residuals shown in Plate 1 are replotted here in a polar stereographic projection, along with similarly averaged and filtered sea surface temperatures (SST) residuals [Reynolds and Marsico, 1993]. Here the negative SST residuals are shown as "warm" colors to enhance the remarkable connection between colder (warmer) SST patterns and more (less) dense SIC patterns as they progress eastward from ~90°E to 45°W. The white band separating the two data fields arises from the differing southernmost limit of the SST data and the northernmost limit of the SIC data.

GLOERSEN AND WHITE: ANTARCTIC CIRCUMPOLAR WAVE IN SEA ICE

GLOERSEN AND WHITE: ANTARCTIC CIRCUMPOLAR WAVE IN SEA ICE

GLOERSEN AND WHITE: ANTARCTIC CIRCUMPOLAR WAVE IN SEA ICE

GLOERSEN AND WHITE: ANTARCTIC CIRCUMPOLAR WAVE IN SEA ICE

GLOERSEN AND WHITE: ANTARCTIC CIRCUMPOLAR WAVE IN SEA ICE

GLOERSEN AND WHITE: ANTARCTIC CIRCUMPOLAR WAVE IN SEA ICE

GLOERSEN AND WHITE: ANTARCTIC CIRCUMPOLAR WAVE IN SEA ICE

GLOERSEN AND WHITE: ANTARCTIC CIRCUMPOLAR WAVE IN SEA ICE

GLOERSEN AND WHITE: ANTARCTIC CIRCUMPOLAR WAVE IN SEA ICE

GLOERSEN AND WHITE: ANTARCTIC CIRCUMPOLAR WAVE IN SEA ICE