Multidecadal Fluctuations of Antarctic Sea Ice Revealed
-Ocean deep convection triggered by wind fluctuations melts sea ice-

1. Key findings

• Antarctic sea ice extent has slightly increased from 1979 to 2015, but has declined at a record rate since 2016, possibly related to long-term fluctuations in sea ice. However, the causes of long-term fluctuations are still poorly understood.

• Long-term paleoclimate data^※1 and atmosphere-ocean coupled model^※2 simulations indicate that Antarctic sea ice extent fluctuates with a period of 80 to 100 years.

• Decadal sea ice decline is caused by the upwelling of subsurface (deeper than 200 m) warm and high salinity water due to the strengthening of ocean deep convection triggered by stronger westerly winds. The stronger westerly winds are found to be associated with an atmospheric phenomenon called the positive Southern Annular Mode^※3.

2. Abstract

Dr. Yushi Morioka of the Japan Agency for Marine-Earth Science and Technology (JAMSTEC; President, Hiroyuki Yamato) and his colleagues used long-term paleoclimate data and a series of atmosphere-ocean coupled model simulations to show that Antarctic sea ice fluctuates with a period of 80 to 100 years, and demonstrate that deep ocean convection triggered by westerly wind fluctuations is responsible for the sea ice fluctuations.

Understanding sea ice fluctuations is important because Antarctic sea ice affects not only the surrounding climate, but also global sea level. In contrast to the Arctic sea ice extent which shows a continuous decrease due to global warming, the Antarctic sea ice extent slightly increased from 1979 to 2015. However, in 2016, the sea ice extent declined sharply and has continued to decline at record rates since then. Although atmospheric winds and ocean heat transport are reported to play important roles in the increase and decrease of sea ice extent, the existence and underlying causes of long-term fluctuations in the sea ice extent are still unknown.

In this study, we analyzed 300 years of paleoclimatic data and 3,000 years of coupled atmosphere-ocean coupled model simulations and found that Antarctic sea ice extent fluctuates with a period of 80 to 100 years. During the years when Antarctic sea ice extent decreases, we found westerly winds strengthen in the Antarctic Sea that in turn strengthens deep convection in the ocean and causes the upwelling of subsurface (below 200 m) warm and salty water, resulting in sea ice melting. The strengthening of the westerly winds is associated with an atmospheric phenomenon called the positive Southern Annular Mode.

These results not only improve our understanding of the multidecadal changes in Antarctic sea ice, but also help to predict the future conditions of Antarctic sea ice. As the Antarctic sea ice extent has been decreasing since 2016, we need to pay attention to whether it will continue to decrease over the next few decades and then increase in the coming decades. In this regard, we also plan to incorporate the effects of future greenhouse gas emissions into our atmosphere-ocean coupled model to predict the future of Antarctic sea ice.

The results were published in Communications Earth ＆ Environment on November 8 (19:00 JST). This work was partly supported by the Princeton University and NOAA/GFDL Visiting Research Scientists Program and base support of GFDL from NOAA office of Oceanic and Atmospheric Research (OAR), JAMSTEC Overseas Research Visit Program, and JSPS Grant-in-Aid for Scientific Research JP22K03727.

3. Background

Antarctic sea ice plays an important role in the exchange of heat, water, and gases between the atmosphere and ocean. Changes in the sea ice are also linked to changes in the Antarctic ice sheet, which affects global sea level fluctuations as well as the surrounding climate. In contrast to the Arctic sea ice extent, which shows a continuous decline due to global warming, the Antarctic sea ice extent is seen to increase slightly from 1979, when satellite observations began, to 2015 (Figure 1). However, it sharply declined in 2016 and since then the Antarctic sea ice extent has remained at record lows.

It has been reported that the sea ice increase in the past few decades is mainly related to an atmospheric teleconnection associated with decadal variations in the tropical Pacific (Meehl et al. 2016). It is also suggested that the abrupt sea ice decline in 2016 is mainly related to one of the strongest El Niño-Southern Oscillation (ENSO) in 2015 and the negative Southern Annular Mode in 2016 (Stuecker et al. 2017). Furthermore, it is demonstrated that the persistent sea ice decline since 2016 is related to subsurface (100-500 m) temperature increase in the Southern Ocean (Zhang et al. 2022).

Previous studies have focused only on either increase or decrease of sea ice, and it is still unclear how these are a part of slow and multidecadal changes. One of the reasons for this is that the observational data on Antarctic sea ice is short, covering only the last 40 years, and is insufficient to investigate long-term sea ice variations. Therefore, in this study, we used 300-year paleoclimate reconstructed data and 3000-year atmosphere-ocean coupled model simulations to investigate whether the Antarctic sea ice extent fluctuates on a multidecadal timescale and what the underlying causes are.

Figure 1. a. Annual mean Antarctic sea ice concentration (the percentage of sea ice covering the ocean surface) observed during 1979-2022. b. Monthly Antarctic sea ice extent anomaly during 1979-2022 (unit: 10⁶ km²). Data were taken from the National Snow and Ice Data Center (NSIDC).

4. Results

The Antarctic sea ice extent anomalies from 1700 to 2000 reconstructed from paleoclimate data (Figure 2a) show that the sea ice extent increases and decreases with a cycle of about 100 years. We performed wavelet analysis^※4 to identify the predominant period in the Antarctic sea ice extent anomaly (Figure 2b) and found that the periods of about 40 to 50 years and 80 to 100 years are statistically significant, respectively. However, the relatively short duration of the paleoclimate data (300 years) does not fully explain how representative these periods are. Therefore, we investigated the multidecadal variations in the sea ice extent in detail by conducting a long-term simulation experiment using an atmosphere-ocean coupled model forced with constant pre-industrial greenhouse gases.

Figure 2. a. Time series of Antarctic sea ice extent anomaly (in 10⁶ km²) during 1700-2000, reconstructed by Dalaiden et al. (2023). b. Predominant period (in years) of Antarctic sea ice extent anomaly during 1700-2000. Colors are wavelet power spectral density normalized by the sea ice extent variance (σ²) (※4, unit is σ²). Contours are statistically significant values using Chi-squared test (90% confidence level).

Figure 3a shows the results of a wavelet analysis on the 3000-year Antarctic sea ice extent simulated in the atmosphere-ocean coupled model (SPEAR_LO). The model simulates a statistically significant period of about 80 to 100 years, similar to that found in the paleoclimate data. Using two atmosphere-ocean coupled models (SPEAR_LO and SPEAR_MED) with different atmospheric resolutions, we examined the probability density of sea ice extent anomalies (Figure 3b). We find that the extremely low sea ice extent occurs more frequently than the higher-than-normal sea ice extent. The distribution is skewed toward the years with low sea ice extent (negative skewness). This suggests that compared to the higher sea ice, lower sea ice tends to be associated with greater atmospheric and oceanic variability (e.g., greater wind and ocean current variability) and a greater response of sea ice to the atmospheric and oceanic variability (e.g., thinner sea ice is more susceptible).

Figure 3. a. Period (in years) of Antarctic sea ice extent anomaly simulated in the atmosphere-ocean coupled model (SPEAR_LO). Colors are wavelet power spectral density normalized by the sea ice extent variance (in σ²). Contours are statistically significant using Chi-squared test (95% confidence level). b. Probability density distribution of Antarctic sea ice extent anomaly (in %) simulated in the atmosphere-ocean coupled models (SPEAR_LO and SPEAR_MED). The horizontal axis is the sea ice extent anomaly normalized by the standard deviation (σ) (unit: σ).

In order to investigate the physical processes underlying the multidecadal variability in Antarctic sea ice, we focused on low sea ice years and performed a composite analysis using the simulation results of the atmosphere-ocean coupled model. We find that westerly winds get stronger in the Antarctic Sea about 30 years before the Antarctic sea ice extent decreases (Figure 4a). When westerly winds strengthen in the Southern Hemisphere, the effect of the Earth's rotation (Coriolis force) causes a northward current (Ekman current) in the ocean surface layer (shallower than 50 m). In the Antarctic Sea, the presence of the Antarctic continent to the south provides a condition to induce the upwelling in order to compensate for the northward flow. Since the subsurface layer (depths below 200 m) in the Antarctic Sea contains warmer and saltier water than the surface layer, the upwelling of warm seawater contributes to the melting of sea ice.

To investigate this ocean effect in detail, we performed a composite analysis for ocean variables (Figure 4b). About 30 years before the sea ice extent decreases, the effective potential energy (orange line), which indicates the stability of the upper ocean (shallower than 1000 m), shows a positive anomaly, indicating that the ocean interior is unstable. This is accompanied by an increase in deep convection in the Antarctic Sea (thin blue line) and a deepening of the mixed layer (uniform density layer shallower than 200 m, purple line). The deepening of the mixed layer brings in warmer and saltier water from below the mixed layer, resulting in higher sea surface temperature (red line) and sea surface salinity (blue line) in the Antarctic Sea.

Figure 4. a. Sea ice extent and wind stress anomalies composited for low sea ice years in the atmosphere-ocean coupled model (SPEAR_LO). The black line is sea ice extent (unit is 10⁶ km² and positive value is sea ice increase), the red line is the horizontal component of wind stress (unit is 10^-3 Pa and positive value is westerly wind), the blue line is the meridional component of wind stress (unit is 10^-4 Pa and positive value is southerly wind), and the thin blue line is the rotational component of wind stress (unit is 10^-9 Pa m^-1 and positive value is counter-clockwise wind). The horizontal axis is lead/lag year, and positive values represent years when sea ice precedes wind stress.
b. As in a, but for the anomaly of ocean variables. The black line represents sea ice extent (unit is 10⁶ km² and positive value is sea ice increase), the red line represents sea surface temperature (unit is 10^-1 ºC with positive value indicates warming), the blue line represents sea surface salinity (unit is 10^-1 PSU and positive value is salinity increase), the purple line represents mixed layer depth (unit is 10 m and positive value indicates deepening), the thin blue line represents deep convection strength in the Antarctic Sea (unit is Sv and positive value indicates stronger convection), the orange line is the effective potential energy (APE; unit is 10¹ J m^-3 and positive value indicate weaker stability) from the surface to 1000 m depth.

To further clarify the vertical processes in the ocean, we computed composite anomalies of potential water temperature (water temperature taking into account the effect of water pressure) and salinity in the Antarctic Sea (Figure 5). We find that water temperature and salinity were higher than normal about 20 years before the sea ice decreases. In particular, salinity (Figure 5b) is higher in the subsurface layer (shallower than 300 m) for more than 30 years, which originates from the subsurface ocean. When westerly winds strengthen in the Antarctic Sea and deep ocean convection occurs, the subsurface warm and saline water starts to upwell. The associated increase in salinity and density in the mixed layer (black line) makes the layer unstable, which causes more deep convection. This positive feedback between salinity and deep convection is identified to bring in the subsurface warm and high salinity water into the mixed layer, resulting in the sea ice melt on multidecadal timescales.

Figure 5. a. Vertical cross-section of potential water temperature (water temperature accounting for the effect of water pressure, in ºC) anomalies composited for low sea ice years in the Antarctic Sea, simulated by the atmosphere-ocean coupled model (SPEAR_LO). The horizontal axis is lead/lag year with positive values indicating years when sea ice precedes potential water temperature. The black line is the depth of the mixed layer (uniform density layer shallower than 200 m). b. Corresponding vertical cross-section of salinity anomalies (unit: PSU).

5. Future Perspectives

This study has demonstrated that the Antarctic sea ice extent fluctuates with a period of 80 to 100 years. The Antarctic sea ice extent has increased slightly from 1979 to 2015, but has decreased significantly since 2016. In light of this observation and the results of this study, it is expected that Antarctic sea ice will continue to decrease in the coming decades. It is necessary to consider the effects of global warming as well as natural variability to assess whether sea ice will continue to decrease or will increase after a few decades,. In this study, the effect of greenhouse gases on the atmosphere-ocean coupled model was assumed to be constant at pre-industrial levels in accordance with paleoclimate data. In the future, we plan to incorporate the expected increase of greenhouse gases into the atmosphere-ocean coupled model to clarify how Antarctic sea ice extent will change in the future.

This study also has revealed that the upwelling of subsurface warm and salty water is responsible for the multidecadal variability of Antarctic sea ice, based on the atmosphere-ocean coupled model simulations. The coupled model simulations by other research institutes have also yielded similar results (see our paper), but the results need to be verified with long-term observational data. Furthermore, in order to clarify the details of how the multidecadal sea ice changes affect the surrounding climate and global sea level, it is essential to verify the results not only through the atmosphere-ocean coupled model simulations, but through oceanographic observations. However, oceanographic data for the Antarctic Sea is scarce in both time and space. Therefore, it is necessary to increase ocean observations in the Antarctic Sea using ships, automatic lifting-drifting buoys such as Argo floats, and autonomous unmanned vehicles (AUVs).