Landfast ice (fast ice) defined as sea ice that remains immobilized along coastlines, ice shelves, or iceberg margins, is present throughout the coastal regions of Antarctica. It occurs for periods ranging from weeks to several decades. Fast ice has been observed to extend 200 km from the Antarctic coastline. Fast ice reaches its annual maximum extent during October with an average extent of ∼601000 km2 (Fraser et al., 2021). Based on morphological characteristics and formation mechanisms, fast ice is categorized into two primary types: level fast ice and deformed fast ice. Level fast ice, which dominates in areal extent in the Antarctic (Fedotov et al., 1998), forms in situ through thermodynamically driven processes and is typically thinner. In contrast, deformed fast ice results from the dynamic stacking of drifting ice under wind or wave forcing, being thicker and exhibiting greater stability than level fast ice (Inall et al., 2022). The relative proportion of these two types is governed by coastal geomorphology (Porter-Smith et al., 2021) and advective barriers such as glacier tongues and grounded icebergs (Fraser et al., 2012). On the basis of its persistence, fast ice is further divided into seasonal fast ice, which melts during austral spring or summer, and multi-year fast ice, predominantly occurring in embayments or associated with grounded icebergs (Fraser et al., 2021). The mechanical strength of multi-year fast ice significantly exceeds that of seasonal fast ice (Kharitonov and Borodkin, 2022).

Fast ice plays critical roles in the cryosphere and Earth system. First, it provides mechanical stabilization for vulnerable ice shelves and glacier tongues (Massom et al., 2018) and can respond rapidly to environmental forcing (Leonard et al., 2021). Fast ice also plays a strong role in the defence of weakened Antarctic ice shelves (Bennetts and Teder, 2025; Teder et al., 2025). Second, fast ice modulates the size of coastal polynya and dense water production (Fraser et al., 2019), which influences the formation rate of Antarctic Bottom Water (Ohshima et al., 2013). Third, fast ice serves as a seasonal reservoir of nutrients and its breakup and offshore advection potentially fertilize the ocean (Grotti et al., 2005). Fourth, fast ice also provides a substrate for early-stage primary production (Meiners et al., 2018), which fuels coastal marine food webs, and supports key ecosystem processes (Bluhm et al., 2017).

The formation of fast ice in austral autumn is influenced by oceanic factors including SST variability (Li et al., 2023), vertical oceanic heat flux (Maksym, 2019; Hu et al., 2023), and salinity (Brett et al., 2020). The primary mechanisms for fast ice shrink in late spring include break-out events (Squire, 2020), internal melt (Zhao et al., 2022), wind speed and direction, and tensile failure caused by wind and storm-driven waves (Crocker and Wadhams, 1989; Voermans et al., 2021).

The atmospheric forcing factors also influence the variability of Antarctic fast ice across time scales ranging from hours to synoptic, seasonal, and interannual. Major atmospheric drivers include air temperature (Heil, 2006), wind speed and direction, storms and cyclones (Crocker and Wadhams, 1989), and snowfall (Kawamura et al., 1997; Ushio, 2006). However, most research on atmospheric impacts has so far been conducted regionally. In Prydz Bay, Heil (2006) analysed five decades (1950s–2003) of fast-ice records at Australia's Davis Station and found that the duration of seasonal fast ice depends on winter and spring temperatures, while interannual variability is primarily controlled by wind and storms. Yu et al. (2017) and Liu et al. (2024) identified ice surface temperature as the primary determinant of seasonal fast ice near China's Zhongshan Station. Zhao et al. (2019) demonstrated through measurements during 2012–2016 that local fast-ice thickness decreases with increasing snow accumulation. Zhai et al. (2019) reported that low air temperatures and wind speeds promote fast-ice expansion in the vicinity of Inexpressible Island. In the Ross Sea, Leonard et al. (2021) observed that the extent and duration fast ice near the McMurdo station are influenced by seasonal storm intensity; winter storm intensity additionally affects the following summer's fast ice conditions. On the Antarctic Peninsula and the Weddell Sea, Massom et al. (2018) and Wille et al. (2022) demonstrated the importance of persistent warm northerly winds in triggering anomalous break-up of fast ice near ice shelves, where wind-driven pack ice movement facilitates wave-induced fracturing. Herman et al. (2021) emphasized the role of ocean wave action in fast ice fragmentation near ice shelves. Arndt et al. (2020) identified snow accumulation as a major factor controlling fast ice formation in far eastern part of the Weddell Sea, with strong easterly winds redistributing snow on fast ice and contributing to its summer break-up.

In addition to synoptic-scale processes, numerous studies have examined the roles of large-scale atmospheric and oceanic forcing factors on the formation and decay of regional fast ice. Extensive research has noted linkages between the total Antarctic sea ice extent and large-scale climate modes (Pezza et al., 2012; Clem and Fogt, 2013; Coggins and McDonald, 2015; Irving and Simmonds, 2016; Meehl et al., 2016; Baba and Renwick, 2017; Clem et al., 2017; Simmonds and Li, 2021; Chung et al., 2022; Yu et al., 2022), while fewer studies have assessed their relationships in process level. Using ice records from the South Orkney Islands, Murphy et al. (2014) identified the non-stationary relationship between Weddell Sea fast-ice extent and climate modes, such as the El Nino–Southern Oscillation (ENSO) and the Southern Annular Mode (SAM). Aoki (2017) observed a significant positive correlation between the latitude of fast-ice edges in Lützow–Holm Bay, East Antarctica and eastern equatorial Pacific SST during 1997–2016. However, their studies only focused on a specific region, not the entire Antarctic fast-ice zone. In this study we employ Empirical Orthogonal Function (EOF) analysis to obtain the main modes of Antarctic fast ice variability and analyse SST and atmospheric variables related to the modes to build the relationship between seasonal Antarctic fast ice and climate modes (tropical Pacific SST) for the 2000–2018 period.

The monthly Antarctic fast ice dataset was derived from the Antarctic sea ice distribution data publicly released by Fraser et al. (2021) in the Australian Antarctic Data Centre website. This dataset classifies sea ice in National Aeronautics and Space Administration Moderate Resolution Imaging Spectroradiometer (NASA MODIS) imagery, covering an 18 year period from March 2000 to February 2018, with a spatial resolution of 1 km and a temporal resolution of 15 d. This dataset was utilized to extract the monthly boundaries of fast ice. The latitude of the fast-ice edge at each longitude (1° resolution) was identified by determining the boundary between fast ice and drifting ice or open ocean. The latitudinal difference between the fast ice outer boundary and the corresponding Antarctic continental coastline was calculated to generate a time series of fast ice edge across all longitudes around Antarctica.

Oceanic and atmospheric datasets were derived from the European Centre for Medium-Range Weather Forecasts (ECMWF) fifth-generation reanalysis (ERA5) (Hersbach et al., 2020), which has a spatial resolution of 0.25°×0.25°. Due to its high resolution and advanced data-assimilation system, ERA5 performs well with respect to variables that are of key importance in our study: upper-tropospheric geopotential height fields in the tropics (Hoffmann and Spang, 2022) and MSLP and wind fields in the Southern Ocean (McErlich et al., 2023).

Monthly averaged sea surface temperature (SST), MSLP, 2 m air temperature, 10 m wind fields, and 200 hPa geopotential height fields from ERA5 were employed. In addition, we used the monthly top-of-the-atmosphere (TOA) outgoing longwave radiation (OLR) data provided by the National Oceanic and Atmospheric Administration Climate Prediction Center (NOAA CPC) (Xie et al., 2023). These oceanic and atmospheric datasets were used to investigate the relationship between fast ice variability and atmospheric circulation anomalies.

We detrended the fast ice boundary time series together with the sea surface temperature and atmospheric variables. We then removed the monthly climatological mean for each respective month to derive anomaly fields for all variables. Finally, the data were grouped into seasons as follows. The four austral seasons here refer to spring: September, October and November (SON); summer: December, January and February (DJF); autumn: March, April and May (MAM); winter: June, July and August (JJA). Seasonal anomalies were obtained by subtracting the climatological seasonal cycle from each time series. To visualize the spatiotemporal variations in fast ice offshore distance, EOF analysis was performed on the processed data. This method decomposes the variability of fast ice anomalies into spatial patterns (EOFs) and their associated temporal coefficients, called as Principal Components (PCs). The EOFs and PCs are mutually orthogonal, with the variance contribution of each mode indicating the percentage of total variance explained by the mode. The contribution of the mode to the total trend for each season is calculated by the product of the trend in the PCs of the mode and the EOFs of the mode. The data following EOF decomposition were all standardized.

Regression analysis was employed to establish relationships between meteorological variables and fast ice variability. Regression coefficients were calculated by regressing the time series of each meteorological variable at every spatial grid point onto the standardized PCs of fast ice variability. A two-tailed Student's t-test is used to assess the confidence level of all regression results.

Figure 1 presents the seasonal time series of the total anomaly in fast ice extent over the 18 year period from 2000 to 2017, based on fast ice dataset and processing methods described above. The figure clearly shows that although the fast ice extent in all four seasons displays a broadly similar pattern (initial decline, subsequent increase, another decline, followed by a second rise and final decrease) the magnitude and timing of these variations differ substantially among the seasons. In the autumn (Fig. 1a), the fast ice extent overall shows a relatively clear decreasing trend. The largest positive anomaly occurred in 2000, reaching +0.25° latitude in fast-ice extent. It subsequently decreased, turning negative in 2002 and reaching minimum in 2003 with an anomaly of -0.1° latitude. From 2004 to 2007, the anomaly varied but exhibited an overall increasing tendency. During 2008–2011, the anomaly remained consistently negative, but its magnitude stayed within 0.1° latitude. From 2012 to 2016, the anomaly was consistently positive, with the magnitude also within 0.1° latitude. In 2017, the anomaly rapidly decreased down to -0.2° latitude. In winter (Fig. 1b), the trend of the fast ice anomaly resembles that of autumn, but both the maximum and minimum anomalies have a smaller magnitude than in autumn. Unlike in spring, winter showed a more pronounced negative anomalies during 2008–2010, with the minimum occurring in 2010 at -0.15° latitude. The trends of fast ice extent in spring (Fig. 1c) and summer (Fig. 1d) are nearly identical. However, spring 2000 showed a large positive anomaly, while the summer anomaly was almost zero. Unlike autumn and winter, spring and summer exhibited larger positive anomalies during 2012–2015, with the maximum values occurring within this period. To better understand the relationship between the temporal and spatial variability of fast ice, we perform an EOF decomposition on the fast ice data and base our subsequent analysis on the leading modes.

EOF analysis of seasonal Antarctic fast-ice extent anomalies from 2000 to 2017 revealed distinct seasonal differences. The variance contributions of every EOF mode for each season are shown as Table 1. The first mode accounted for a substantial proportion of the total variance, with the largest percentage of 41 % in winter and the smallest of 31 % in summer. The percentage of variance explained by the second mode ranges from 13 % in autumn to 22 % in summer.

In this study, we employed EOF analysis method to examine the main modes of the variation of seasonal Antarctic fast ice extent for the 2000–2018 period. These main modes show distinct temporal and spatial variability depending on seasons. Temporally, the PC1s exhibit a gradually increasing trend with fluctuations during the winter and spring. In contrast, during the summer and autumn, the PCs show an interannual and decadal variability. Spatially, the anomalies in fast ice extent are more pronounced in West Antarctica, particularly in the Antarctic Peninsula, the Bellingshausen and Weddell Seas. This finding is consistent with the changes in the width of fast ice zone (Li et al., 2020). This spatial variability is attributed to the complex coastline of West Antarctica and the presence of numerous peripheral islands, where changes in island-associated fast ice significantly contribute to regional anomalies. Li et al. (2020) also suggests that icebergs play a critical role in the formation and maintenance of fast ice along this coastline. In contrast, the Indian Ocean sector of East Antarctica, characterized by a smoother coastline, exhibits relatively minor variations.

The analysis further elucidates the physical mechanisms by which remote SST anomalies influence Antarctic fast ice. During the autumn, tropical SST anomalies trigger convective activity, leading to upper-level outgoing longwave radiation OLR anomalies. These anomalies excite planetary-scale atmospheric waves, manifested as alternating positive and negative geopotential height anomalies at 200 hPa. As these waves propagate poleward, they perturb near-coastal MSLP and wind fields due to barotropic structure, which directly modulate fast-ice formation or disintegration. Southerly winds, onshore zonal winds, and wind convergence in both direction and speed contribute to the expansion of fast ice extent, whereas northerly winds, offshore zonal winds, and wind divergence lead to reductions in fast ice coverage. Rapid, large-scale reductions in Antarctic fast-ice cover are typically triggered when offshore winds (often in combination with ocean swell and thermodynamic weakening) fracture or detach the ice, leading to its disintegration and opening of coastal polynyas. Additionally, wind anomalies indirectly affect fast ice by altering air temperatures, thereby influencing thermodynamic processes. During the winter and spring, the dominant driver shifts to the subtropical SST dipole mode in the Southern Hemisphere. This dipole pattern similarly influences Antarctic fast ice through the aforementioned chain of processes, though its spatial configuration shifts southward during spring. In the summer, the weakening of the SST dipole reduces the wavelength of atmospheric planetary waves, altering wind speed and direction patterns around Antarctica and resulting in distinct seasonal differences in fast ice variability compared to preceding seasons.

Aoki (2017) established a correlation between the latitude of fast ice edges and sea surface temperatures (SST) in the eastern equatorial Pacific, further describing a teleconnection mechanism whereby anomalously warm SSTs in the tropical eastern Pacific during the summer trigger Rossby waves that propagate toward East Antarctica's Queen Maud Land. Separately, Murphy et al. (2014) demonstrated linkages between Weddell Sea fast ice extent and climate modes (ENSO/SAM), noting that fast ice break-up is modulated by SAM-associated westerly/northwesterly winds. Our results align with their studies, extending previous studies into the entire Antarctic fast-ice zone.

Our study also reveals seasonal differences in the dominant factors controlling fast ice. During the autumn and winter, thermodynamic processes, particularly air temperature, predominantly govern ice extent across most regions. Higher temperatures are associated with less fast ice, while lower temperatures promote its expansion. However, in the spring and summer, when formation of new fast ice is limited, the correlation between air temperature and ice extent weakens, with over half of the regions showing no inverse relationship. Instead, wind direction and speed emerge as the primary dynamical drivers, explaining most of the variability in fast-ice extent. This suggests that dynamic forcing (e.g., wind-driven ice advection, compaction, and disintegration) dominates fast-ice anomalies during these seasons. Wei and Qin (2016) and Voermans et al. (2021) similarly concluded that during the spring, strong onshore winds and gale-driven ocean waves erode coastal fast ice margins.

Although we focused on the novel aspect of remote forcing of tropical and subtropical SSTs on Antarctic fast ice, we recognize that the direct oceanic forcing is local and includes both dynamic and thermodynamic factors. The former are related to currents, tidal effects, and swell (Inall et al., 2022), while basal melt is the primary thermodynamic factor (Kusahara et al., 2021).

The 18 year study period (2000–2018) was insufficient to capture the interdecadal-scale oscillations that may influence Antarctic fast ice dynamics. Long-term observations are required to disentangle these complex interactions between tropical SST anomalies and Antarctic fast ice and validate the proposed mechanisms. The impacts of global warming on Antarctic fast-ice extent should also be investigated in future studies. Further, to achieve more robust process-level attribution of the statistical results obtained, coupled climate model experiments with prescribed tropical SST anomalies are needed. Process-level understanding of the interannual and decadal variability of Antarctic fast-ice extent across seasons enhances our knowledge of the processes governing the variability and future evolution of Antarctic sea ice and ice shelves.