1 Introduction

The majority of Antarctic sea ice is seasonal, forming from early autumn through to early spring, and then melting from late spring through summer. This leads to dramatic seasonal change of sea-ice extent from a minimum of nearly 3.0 × 106 km2 in late summer to a maximum of nearly 20.0 × 106 km2 (Fig. 1) in early spring (Gloersen et al. 1992). In addition to the seasonal change, sea-ice area in the Antarctic also changes over a range of timescales, from daily (e.g., Kimura and Wakatsuchi 2011), to intraseasonal (e.g., Baba et al. 2006), interannual (e.g., Comiso et al. 2017), and multi-decadal timescales (e.g., Turner et al. 2016; Goosse and Zunz 2014; Zhang et al. 2019).

Fig. 1
figure 1

Sea-ice concentration on September 1, 2003 and names of sectors used in this study

Full size image

Processes governing the change in the Antarctic sea ice have been studied extensively using numerical models (Polvani and Smith 2013; Purich et al. 2016; Blanchard-Wrigglesworth et al. 2021; Sun and Eisenman 2021). The model is also a powerful tool to discuss the drivers of local sea-ice variability in relation to changes in meteorological fields (e.g., Fogt et al. 2012; Morioka and Behera 2021). The performance of the numerical model is often evaluated by the reproducibility of sea-ice distributions. However, modeled sea-ice distributions often agree with observations even when the processes of sea-ice change are not reproduced. For example, Uotila et al. (2014) showed that numerical models reproduce sea-ice distributions close to observations even when the reproduced sea-ice drifting velocity differs significantly from observations. To understand the temporal and spatial change of sea-ice cover, it is necessary to reproduce the sea-ice processes in a numerical model (Fichefet and Maqueda 1997). However, improvement of numerical models cannot be accomplished by model-based study alone. The process governing the change of sea-ice cover must first be clarified by analysis-based evaluation using the observational data. It is especially important to reveal the sea-ice process at the ice edge and to clarify the differences in the expansion/reduction process of sea-ice cover by region.

Satellite observation data is very useful for the analyses of sea-ice processes. In particular, the microwave radiometer is the best observation tool, because it is less sensitive than other sensors to interference from weather conditions and solar radiation, and it can continuously observe the sea-ice surface. The Japanese AMSR-E sensor onboard NASA’s Aqua satellite, which was launched in May 2002, provides finer spatial resolution than earlier passive microwave sensors, and enables higher resolution analyses of sea-ice change (e.g., Massom et al. 2006b).

As mentioned above, it is necessary not only to monitor the superficial features of sea ice, but also to understand the processes governing the change of sea ice in various time scales. In general, the area of sea ice changes through both thermodynamic and dynamic processes. Thermodynamic processes involve the formation and melting of sea ice, whereas dynamic processes include ice transport and deformation (e.g., rafting and ridging of ice floe). There are not as many studies that have clarified these processes in the entire Southern Ocean based on observational data (e.g., Holland and Kwok 2012) as those that use numerical models. Kimura (2007) and Kimura and Wakatsuchi (2011) tried to reveal the regional differences of the processes over the Antarctic using satellite-derived sea-ice concentration and velocity. However, those studies were still in their infancy in that they did not adequately detect sea-ice processes in the marginal ice zone. Holland and Kimura (2016) summarized the climatology of the ice concentration budget considering the contribution of both dynamic and thermodynamic sea-ice processes. They calculated the ice budget in places, but does not isolate the ice edge processes, so it also does not clarify the expansion/reduction process of sea-ice cover.

This study divided the processes of the change in ice area into the local change and change by ice transport, based on the same method of Kimura and Wakatsuchi (2004, 2011). This distinction helped us define regional differences in the processes of the seasonal and interannual changes of ice area. This study also distinguishes between the marginal ice zone and inside the ice area as the place, where the ice processes occur. The objective of this study is to clarify the regional differences in the processes governing the change of sea-ice area. We focus only on two-dimensional ice area and not the three-dimensional volume of sea ice, because to-date satellite observations cannot measure the thickness of sea ice in sufficient accuracy and there is insufficient in situ observations. The study aims to identify regional differences in the expansion/reduction process of two-dimensional ice cover, especially the processes occurring in the marginal ice zone.

2 Data

Daily sea-ice velocity and ice concentration were derived from satellite remote sensing data. Data from January 2003 to August 2011 were acquired by AMSR-E. The AMSR-E observations stopped in October 2011 but data from AMSR2 were available from July 2012. Therefore, this study used AMSR-E data from January 2003 to August 2011 and AMSR2 data from August 2012 to December 2019.

Sea-ice drifting velocity data are the same as those used in previous studies (e.g., Holland and Kimura 2016; Silvano et al. 2020). It was calculated from the gridded brightness temperature of 36 GHz horizontal and vertical polarization channels, which were binned into square pixels of 10 × 10 km by the Japan Aerospace Exploration Agency (JAXA). The procedure for detecting ice motion, which has been improved in previous studies (Kimura and Wakatsuchi 2000; Kimura et al. 2013), is based on the maximum cross-correlation method described by Ninnis et al. (1986) and Emery et al. (1991). This method determines the spatial offset that maximizes the cross-correlation coefficient between two brightness temperature arrays in consecutive images separated by 24 h. For this procedure, 6 × 6 grid cell (60 × 60 km) arrays are utilized. As the first step, we detect the ice motion vector on one-grid (10 km) interval. Next, detections of a false match are automatically filtered out based on the value of the maximum cross-correlation coefficient and consistency with surrounding results. After applying the filtering, we constructed a daily ice velocity data over the sea-ice area on a 60 × 60 km grid. The remaining gap in the data was filled through spatial interpolation using the available data from the surrounding grids.

Daily ice concentration data were calculated using a bootstrap algorithm (Comiso 2009) based on the 18 GHz and 36 GHz brightness temperature of AMSE-E and AMSR2. These data, provided on a grid with 10 km spatial resolution, are produced by the Japan Aerospace Exploration Agency and distributed via the Arctic Data archive System. The ice concentration data were averaged over a 60 × 60 km (6 × 6 pixel) area to match the grid size of the ice velocity data.

Both of the ice velocity and ice concentration are binned into a 60 × 60 km polar-stereographic map, which specifies a projection plane or grid tangent to the Earth's surface at 70° S. This translates to a 6% distortion of the grid at the poles but means there is little or no distortion of the grid in near the ice cover. Therefore, in this study, the area of cells surrounded by the grid is treated as 3600 (60 × 60) km2 regardless of latitude.

3 Methods

Fundamental to this study was the use of daily ice concentration and ice velocity data for estimation of the local change of sea-ice area. For this estimation, we focused on any given unit pixel (60 × 60 km size) enclosed by adjacent grids. This study adopted the calculation method used in Kimura and Wakatsuchi (2004, 2011), which assumes that change in sea-ice area is a response to local change (ice formation/melting and dynamic deformation) and ice transport (inflow/outflow). Thus, we could relate an observed increase of sea-ice area (A(t + 1) − A(t)) for a given pixel to the sum of the increased ice area attributable to local change (P(t)) within that pixel during 1 day and to the ice transport (Fin(t)) from surrounding pixels as follows:

$$A\left( {t + 1} \right) - A\left( t \right) = P\left( t \right) + {\text{Fin}}\left( t \right),$$
(1)

where A(t) is the daily ice-covered area in a pixel calculated as the ice concentration multiplied by the area of the pixel (60 × 60 km2), A(t + 1) is the ice area within the same pixel for the next day. Fin(t) is the area of ice advected into the pixel of interest from adjoining pixels. Fin(t) is calculated by multiplying the sea-ice drift velocity of the component perpendicular to the grid boundary by the sea-ice concentration of the upstream grid, length of the time interval and the length of the grid boundary. This calculation is performed for adjacent grids in four directions and Fin(t) is calculated as the sum of them. Figure 2 shows a schematic diagram showing the ice concentration (C0, C1, C2, C3 and C4) of the grid of interest and the adjacent grids and the component of drifting velocity perpendicular to the boundary (V1, V2, V3 and V4). In the case of Fig. 2a, Fin(t) is calculated as (− C0V1 + C2V2 + C3V3 − C0V4) × 86,400(s) × 60,000(m). In another case of Fig. 2b, it is calculated as (C1V1 − C0V2 − C0V3 + C4V4) × 86,400(s) × 60,000(m). Here, the calculation of Fin(t) requires sea-ice drifting velocities in all four directions. However, sea-ice drifting velocities can only be derived, where sea ice is present, so there is no data for the grid offshore from the ice edge. This makes it impossible to calculate the drifting velocity between the ice edge and its offshore grid, i.e., the drifting velocity for the ice edge to move further offshore. This study avoided this problem by extrapolating the ice drift velocity data two pixels off-ice from the ice edge. This was done by giving the average value of the drift velocity data within 2 pixels of the grid to the missing data. This cause the sea ice on the open-water grid closest to the ice edge to move at the calculated drifting velocity of the inside grid closest to the ice edge. This treatment allow the calculation of the area budget according to Eq. (1) at the ice edge and enables us to fully capture the balance of spatio-temporal changes in the two-dimensional quantity of sea-ice area. Here, P(t) is the ice area calculated as the residual of A(t + 1) − A(t) and Fin(t), indicating the local change in the ice area within the pixel. Positive values of P(t) indicate an increase in the ice area due to thermodynamic sea-ice formation, while negative values of it include the effects of both thermodynamic melting and areal reduction due to the dynamic deformation. It is difficult to separate the contribution of dynamic reduction from that of thermodynamic melting. Especially, during the expansion period, the contribution of dynamic deformation cannot be ignored. For example, Williams et al. (2015) observed the ice draft using an Autonomous Underwater Vehicle observations to determine the near-coastal regions of the Weddell, Bellingshausen, and Wilkes Land sectors of Antarctica, and found that 76% of the ice volume is deformed ice.

Fig. 2
figure 2

Schematic diagram of how to calculate the area of sea-ice inflow/outflow at each pixel. C0, C1, C2, C3 and C4 show the ice concentration. V1, V2, V3 and V4 show ice velocity

Full size image

This study distinguishes the area, where the sea-ice change process occurs, i.e., change in the marginal ice zone and inside the ice cover. Revealing what is occurring around the ice edge is important for understanding the temporal change of sea-ice cover. If ice formation occurs in the marginal ice zone, it means that the thermodynamic conditions are favorable for expansion of the ice cover. Conversely, ice advance due to off-ice ice drift, without ice formation in the marginal ice zone, means that the dynamic condition bringing the off-ice motion plays an important role in advance of ice edge.

Domain of the marginal ice zone is defined by following thresholds based on the 10-km interval grid: (1) ice concentration is > 5%, (2) a pixel of open water (ice concentration < 5%) exists within a distance of 3 pixels, and (3) no land pixel exists within a distance of 3 pixels. The marginal ice zone on the 60 × 60 km grid is defined from the detected ice-edge pixels. An example of the extracted marginal ice zone is shown in Fig. 3. The domain is distributed in the area approximately two pixels from the edge of the sea-ice cover.

Fig. 3
figure 3

Extracted ice edge on the same day of Fig. 1, based on the criteria described in Sect. 3

Full size image

Using the two types of ice processes and two types of domain, this study considers four types of sea-ice process: local change inside the ice cover (PI), ice transport inside the ice cover (FI), local change in the marginal ice zone (PE), and ice transport in the marginal ice zone (FE). Physical phenomena to which each process corresponds are summarized in Table 1. The evaluation of the contribution of the four processes is carried out for each of 12 sectors around Antarctica divided by 30° intervals in longitude. Each value was calculated by summing the values of pixels in each longitudinal zone for 1 month based on the daily estimation and then averaging them over the years considered in the analysis. In order to focus on seasonal changes, monthly averages are used in this study. Of course, changes occur on shorter time scales. Each of four processes can take both positive and negative values depending on the day within 1 month. However, we do not focus on changes on time scales shorter than 1 month. Figure 4 presents a schematic diagram of four processes assuming the situation during the sea-ice expansion period. Fin(t) showing the inflow/outflow of sea ice between neighboring pixels, cancels each other out except for boundaries between sectors and inside-ice edge. As a result, FI corresponds to the sum of the zonal (east–west direction) ice transport across the sector boundary and meridional (north–south) ice transport across the boundary of the marginal ice zone and inside. PI, which is calculated as the sum of P(t) inside the ice cover, indicates the increase/decrease of sea-ice area within the sea-ice cover containing the ice formation/melting and decrease due to ice deformation. The sea-ice cover is expanded/shrunk through FE and PE. FE represents the area of expansion/decrease due to the off-ice transport perpendicular to the ice edge, and PE represents the area of expansion due to new sea-ice formation in the marginal ice zone. Because the domain of the marginal ice zone is very narrow, zonal transport in that domain and sea ice decrease due to ice deformation in the domain are negligible.

Table 1 Correspondence table between ice processes considered in this study and physical phenomena
Full size table
Fig. 4
figure 4

Schematic diagram of four types of sea-ice process of local change inside the ice cover (PI), ice transport inside the ice cover (FI), local change in the marginal ice zone (PE), ice transport in the marginal ice zone (FE), and their relations to the ice-edge advance. Dashed lines indicate sector boundaries

Full size image

4 Results

4.1 Processes of seasonal change of sea-ice area

Monthly values of the changed area due to the four processes for each of 12 sectors is shown in Fig. 5. The amount of change in sea-ice area for each month (shown by solid lines) was calculated by subtracting the sea-ice area on the final day of the previous month from the sea-ice area on the final day of the following month. The monthly change in the ice area (solid line) coincides with the sum of the changes due to each process.

Fig. 5
figure 5

Change in ice area during each month (solid line) averaged over 16 years, and the change attributable to the four processes of local change in the marginal ice zone (PE: pale red shading), local change inside the ice cover (PI: dark red shading), ice transport in the marginal ice zone (FE: pale blue shading), and ice transport inside the ice cover (FI: dark blue shading) for 12 sectors around Antarctica divided by 30° intervals in longitude

Full size image

In the western Weddell Sea (Fig. 5a), the speed of sea-ice advance peaks in April and the area continues to expand until August. The ice area expansion until May is largely due to PE (ice formation in the marginal ice zone), whereas the contribution from FI (inflow of sea ice) becomes larger after June. The Inflow is considered to reflect westward drift along the coast, as shown in Fig. 6. The positive FI is sustained from March to December. However, it does not contribute to the increase of the sea-ice area and is counteracted by the negative PI. This negative PI is partly due to sea-ice melting, but may also include a decrease due to dynamic deformation of sea ice during the sea ice.

Fig. 6
figure 6

Monthly mean sea-ice velocity field. Vectors are drawn at thinned 180 km intervals. Background color indicates the drifting speed

Full size image

Processes controlling the seasonal change of sea-ice area in two sectors of the eastern Weddell Sea (Fig. 5b, c) are similar to those of the western Weddell Sea. The increase peaks are coming later than that in the western Weddell Sea (Fig. 5a), i.e., in May for the sector between 30° W and 0° and in June for the sector between 0° and 30° E. The contribution of the PE is substantial throughout the expansion period. As seen in the ice drift field in Fig. 6, both of these sectors are characterized by the significant sea-ice movement westward near the coast and eastward close to the sea-ice edge, but the westward and eastward currents offset each other, resulting in a little FI and small contribution to the overall change of ice area. Moreover, these sectors experience very rapid loss of sea-ice area in December owing to the negative PI. This substantial reduction is due to the ice melting inside the ice cover, which is consistent with Nihashi et al. (2005), who showed that ice melting within the sea-ice cover is an important process in the seasonal decay of sea ice within this area. In the 30° W–0° sector (Fig. 5b), ice reduction due to FE becomes significant since May. Considering that the monthly mean sea-ice motion near the ice edge has no on-ice component, this reduction is mainly due to the intermittent wind-driven on-ice transport.

In western parts of the East Antarctic between 30° E and 90° E (Fig. 5d, e), the sea-ice area increases slowly from March to September. In this region, positive PE is dominant in the early stage of expansion but it almost stops in July. From August onward, PI becomes dominant. In the August and September of 30° E–60° E sector, although much of the area is discharged into adjacent sectors as FI, positive PI and FE continue to increase sea-ice area. The discharged ice does not stay in the eastern Weddell Sea but passes through it, causing sea ice increase in the western Weddell Sea (60° W–30° W).

In eastern parts of the East Antarctic between 90° E and 150° E (Fig. 5f, g), the ice area is smaller in comparison with that of other areas, as shown in Fig. 1, even in September when the sea-ice area is at its maximum. Because the domain of the inside the ice cover is not large enough to the analyses, the separation of the ice processes in the summer months is less clear. Both PE and FE are almost nonexistent during the fall and winter months in 120° E–150° E sector. The lack of favorable thermodynamic and dynamic conditions for ice increase in the marginal ice zone results in the small expansion of the ice area.

In the Ross Sea (150° E to 150° W; Fig. 5h, i), the speed of sea ice advance peaks in March and April, i.e., earlier than in other areas. The expansion is caused mainly by PE. In the region between 180° and 150° W, there is persistent positive PI during March–October, which is attributable mainly to the formation of sea ice in Ross Sea polynyas. Since June, almost the same area of PI has been discharged through FI by eastward currents.

The process of ice expansion in the region of the Amundsen–Bellingshausen seas is unique in comparison with the other sectors. Between 150° W and 90° W (Fig. 5j, k), PE, which is dominant in other regions, is minimal and almost disappears by May, whereas FE (off-ice advection) plays a major role in the ice expansion. The contribution of FE becomes prominent from May and continues through to December under the prevailing off-ice ice motion, even when the ice area begins to decline after September.

4.2 Factors controlling sea-ice advance

To clarify those factors that influence interannual variability of sea-ice area in each sector, the seasonal change was divided into two periods: ice expansion and ice reduction. The ice expansion period was defined as the months from March, when sea-ice area begins to increase, to September, when sea-ice area reaches its maximum. The ice reduction period was defined as the months from October, when sea-ice area begins to decrease, to February, when sea-ice area reaches its minimum. Data from March 2003 to September 2003 were considered to represent the 2003 expansion period, and data from October 2003 to February 2004 were considered to represent the 2003 reduction period, and similar representation was considered for other years in the study period.

The changes in sea-ice area during the ice expansion period attributable to the four processes in 12 sectors are shown in Fig. 7a. It is evident that the strength and duration of each process varies spatially around Antarctica. The regional difference can be categorized by the contribution of the processes in the marginal ice zone (PE and FE). As can be seen in Fig. 7a, ice expansion is largest in the Weddell Sea, followed by that in the Ross Sea. It is due to large PE, while the PI is negative in the sector between 60° W and 30° W. This is thought to be due to the dynamic deformation of sea ice rather than thermodynamic melting.

Fig. 7
figure 7

a Change in area of sea ice during the expansion period (solid line) and the area attributable to the four processes of local change in the marginal ice zone (PE: pale red shading), local change inside the ice cover (PI: dark red shading), ice transport in the marginal ice zone (FE: pale blue shading), and ice transport inside the ice cover (FI: dark blue shading) in 12 sectors around Antarctica divided by 30° intervals. b Same as a but for the reduction period

Full size image

The total area of increase during the expansion period in each sector, and the area attributable to PE and FE, are summarized in Fig. 8a. The sum of PE and FE is almost equal to the total change of ice area, indicating that sea-ice processes inside the ice cover contribute little to the change in total sea-ice area. In many areas, including the Weddell Sea and Ross Sea, the total area of sea-ice increase is nearly equal to PE. However, in the Amundsen–Bellingshausen seas (90° W to 30° W), the net PE is almost zero. The increase in sea-ice area can be explained almost entirely by FE (off-ice transport of sea ice) in the Amundsen–Bellingshausen seas.

Fig. 8
figure 8

a Total area of sea-ice increase during the expansion period in each sector (black bold line), and increased area attributable to local change (PE: pale red line) and ice transport (FE: pale blue line) processes in the marginal ice zone. b Same as a but for the decreased area during the reduction period

Full size image

4.3 Factors controlling sea-ice change in the ice reduction period

The changes in sea-ice area attributable to the four processes during the ice reduction period in each of the 12 sectors are shown in Fig. 7b. Ice retreat due to PE (ice melting in the marginal ice zone) makes the largest contribution to the overall ice retreat throughout the entire Southern Ocean. In the Weddell Sea, the area of retreat due to PI (ice decrease inside the ice cover) is larger than that in other areas. It is a characteristic of the mechanism of sea-ice retreat in the Weddell Sea. Considering the FE, off-ice advection occurs in western parts of the East Antarctic (30° E to 90° E) and in the Amundsen–Bellingshausen seas (150° W to 90° W), even during the reduction period. It is due to the off-ice motion of sea ice even in the reduction period.

The total area of decrease during the reduction period in each sector and the area attributable to PE and FE are summarized in Fig. 8b. The total decrease is mostly due to the PE in most regions, including the Amundsen–Bellingshausen seas. In the Amundsen–Bellingshausen seas, FE is positive (sign reversed in Fig. 8b), even during the reduction period. In the Weddell Sea, PE is much smaller than the total area of decrease because of the decrease attributable to PI (ice melting inside the ice cover) in this region.

4.4 Factors controlling interannual differences in sea-ice area

An example of the interannual change in the total area of increase attributable to the four processes in the sector between 150° E and 180° is shown in Fig. 9. The area of increase due to PE is larger than the total area of increase. In addition, PI is also positive in all years. Conversely, FE and FI are both negative, i.e., the sea ice moves on-ice direction and exhibits zonal outflow. The change in area through PE, which is the most important contributor, shows similar variability to that of the total area of ice expansion; the correlation coefficient between them is 0.78. In years such as 2010, when FE is extremely small, the large PE offsets it. There is also a negative correlation between PI and FI, indicating that they are balanced, so that when one is smaller, the other is larger and the annual variability in total area is smaller.

Fig. 9
figure 9

Example of interannual time series of the total increase of sea-ice area and that attributable to each of the four processes: thermodynamic change in the marginal ice zone (PE: pale red line), thermodynamic change inside the ice cover (PI: dark red line), dynamic change in the marginal ice zone (FE: pale blue line), and dynamic change inside the ice cover (FI: dark blue line) for the sector between 150° E and 180°

Full size image

To identify the dominant processes responsible for the interannual variability, we considered the correlation between the rate of change in the sea-ice area and contributions due to the four processes. The correlation coefficients between the total increase and the PE and FE around the Antarctic are shown in Fig. 10. The correlation with PE is high in most areas, including the Weddell Sea and Ross Sea, with values in the range of 0.5–0.8. It is also high in the Amundsen–Bellingshausen seas, where seasonal ice advance is controlled largely by FE. The results indicate that the processes that cause seasonal and interannual changes are different in the Amundsen–Bellingshausen seas. Conversely, in the sea areas between 30° E and 60° E and between 90° E and 120° E, PE does not exhibit substantial effect on the interannual change of sea-ice area. In these sectors, interannual variability is controlled primarily by FE rather than by PE.

Fig. 10
figure 10

Correlation coefficients between the interannual change of the total area of ice increase, and increased area attributable to the local change in the marginal ice zone (PE: pale red line) and ice transport in the marginal ice zone (FE: pale blue line) processes

Full size image

The processes controlling the seasonal and interannual variability of sea-ice area in each sector are listed in Table 2. Seasonal advance and retreat of sea ice occur mainly through PE (ice formation and melting in the marginal ice zone) in almost all sectors except the Amundsen–Bellingshausen seas, where ice transport drives seasonal ice advance. Interannual change of the total ice area also occurs through PE in almost all sectors except some parts of the East Antarctic. The cause of interannual change in the amount of decrease of the sea-ice area is not shown in the table, because it is determined by the maximum ice area, not by the conditions during the melting season.

Table 2 Dominant processes of sea-ice change in each sector on seasonal and interannual time scales
Full size table

5 Summary and discussion

In this study, we evaluated contribution of dynamic and thermodynamic processes of sea ice on the change of two-dimensional sea-ice area using daily sea-ice concentration and ice drift velocity. The change in sea-ice area was attributed to four processes: local changes in the marginal ice zone, ice transport in the marginal ice zone, local changes inside the ice area and ice transport inside the ice area, to examine regional differences in the process governing seasonal and interannual changes of ice area. Although there have been studies on the process of sea-ice area budget (e.g., Holland and Kimura 2016), there is no study that focuses on the ice edge and clarifies the process of its positional change. This study to focus mainly on sea-ice processes in the marginal ice zone and to reveal regional differences in the changing process of sea-ice cover. In almost all of areas in the Antarctic, seasonal ice advance occurs through the ice formation in the marginal ice zone rather than the off-ice transport of sea ice. This result is basically consistent with the estimation by Holland and Kimura (2016). However, this study revealed that sea-ice expansion is caused mainly by the off-ice ice transport in the Amundsen–Bellingshausen seas (60 W to 150 W), which is the same mechanism responsible for seasonal change in the sea-ice area in regions of the Northern Hemisphere, such as the Bering Sea and the Sea of Okhotsk (Kimura and Wakatsuchi 2001). Several previous studies showed that sea-ice variability around the Amundsen–Bellingshausen seas is strongly influenced by wind (Assmann et al. 2005; Massom et al. 2006a; Fogt et al. 2012; Holland and Kwok 2012; Turner et al. 2015). This relation between the wind-driven ice drift and sea-ice distribution is confirmed by the results of this study. The fact that sea-ice distribution is determined by dynamics rather than by thermodynamics is a characteristic of the Amundsen–Bellingshausen seas.

The reliability of estimated contributions of each process in the marginal ice zone depends on the accuracy of the ice drift velocity data. For example, underestimation of off-ice drifting velocities leads to underestimation of the contribution of the ice transport. The accuracy of the drifting velocity data used in this study has been verified by Sumata et al. (2014) and Tian et al. (2022), and shown to have almost no bias in the data compared to velocities derived from buoy motions. The errors in the estimated contribution of each process due to errors in the ice velocity data are considered to be negligible.

In terms of interannual variability, the dominant controlling factor is different to that of seasonal variability. In many areas, sea-ice formation in the marginal ice zone dominates, which is the same as in relation to seasonal change, but an area in which sea-ice transport in the marginal ice zone dominates is evident in the East Antarctic.

Sea-ice area in the Southern Ocean showed a gradual trend of increase until 2014, while that of the Amundsen–Bellingshausen seas and the area around 90E in the East Antarctic showed a decreasing trend (Parkinson and Cavalieri 2012; King 2014). The trend of sea-ice concentration between 2003 and 2019 is shown in Fig. 11. In the majority of the area, there is a decrease trend, which is the opposite of that reported by Parkinson and Cavalieri (2012), reflecting the sudden change of the trend in 2015 (Parkinson and DiGirolamo 2021). Nevertheless, there is agreement that the trend in the Amundsen–Bellingshausen seas and parts of the East Antarctic is the opposite of the trend of the entire Southern Ocean. It suggests that the process of sea-ice change in these regions is different from that in the rest of the Southern Ocean. The Amundsen–Bellingshausen seas, where seasonal changes are driven by sea-ice advection, and the East Antarctic, where interannual changes are more likely determined by sea-ice advection, are both areas, where the opposite trend in sea-ice concentration can be observed. These areas are sensitive to changes in wind speed rather than to thermodynamic factors, such as water temperature, air temperature, and solar radiation; therefore, differences in the processes governing the change in sea-ice area might affect the way in which long-term changes are manifested.

Fig. 11
figure 11

Linear trend of the annual mean sea-ice concentration during 2003–2019

Full size image

Different mechanisms of sea-ice variability have different effects on the way in which sea-ice change affects atmospheric and oceanic fields. Knowledge of such processes is useful not only for understanding the current state of sea-ice variability, but also for predicting future changes of ice area associated with climate change.