The areal extent of summertime ice in the Arctic Ocean has exhibited a downward trend since the last century (e.g., Stroeve et al. 2007; Cavalieri & Parkinson 2008, Cavalieri & Parkinson 2012). Moreover, the sea-ice volume has also reduced radically (e.g., Kwok & Cunningham 2015). Monitoring the thickness of sea ice as well as its area is important to improve our understanding of the present state of the Arctic climate. Nowadays, sea-ice thickness can be monitored by several means using data from satellite microwave sensors (Krishfield et al. 2014) or radar/laser altimeters (e.g., Ricker et al. 2014). However, there are difficulties in the derivation of ice thickness from remote sensing data, especially in summer owing to unstable conditions of the ice surface, for example, melt ponds.

Multi-year ice, which is defined as ice that has survived the summer and includes second-year (and older) ice, is an important factor controlling sea-ice distribution and thickness (Tschudi et al. 2016). Monitoring areas of multi-year ice is important because its distribution and movement can alter the distribution of thick ice (Hutchings & Rigor 2012; Haas et al. 2017; Barber et al. 2018; Mahoney et al. 2019). Many previous studies that have focused on multi-year ice distribution have estimated the age of the sea ice in the Arctic (e.g., Maslanik et al. 2007; Maslanik et al. 2011; Korosov et al. 2018) and reported a reduction of thick multi-year ice (e.g., Comiso 2012). Conversely, some other studies have shown that the reduction of the area of multi-year ice has not been significant during the past 10-year period (Kwok 2018; Tilling et al. 2018). Galley et al. (2016) reported that the multi-year ice fraction in the Beaufort Sea has decreased in the past several decades. To understand the processes controlling the spatiotemporal changes of Arctic sea ice, special attention must be paid to the interannual variability of the multi-year ice cover.

Sea-ice reduction in the Arctic Ocean offers new opportunities for commercial shipping routes across the Arctic. However, rapid unexpected movement of thick ice can obstruct shipping routes for short periods. This study focusses on one such event, in which thick ice blocked the mouth of the AG in September 2018 (Fig. 1). This blockage temporarily prevented the navigation of ships in and out of the western entrance to the North-west Passage, resulting in cancellation of the annual supply barge to three Arctic communities (Weber 2018). Understanding the circumstances of the 2018 blockage will help improve sea-ice forecasting in the Arctic and reduce the impact of sea route blockages. The objective of this study was to reveal both the source of the thick ice in the AG blockage of 2018 and the reasons for the unusual drift behaviour of the multi-year sea ice.

Ice concentration and multi-year ice distribution derived from AMSR-E and AMSR-2 data were prepared for use in this study. Daily ice concentration data were calculated using a bootstrap algorithm (Comiso 2009). These data are provided on a polar-stereographic grid with 10-km spatial resolution by the Japan Aerospace Exploration Agency and distributed via the Arctic Data Archive System.

There are several ways to monitor the distribution of multi-year ice, for example, taking snapshot observations of its surface properties (e.g., Aaboe et al. 2017) and performing long-term tracking (e.g., Korosov et al. 2018). This study estimated the daily distribution of multi-year ice using a Lagrangian tracking method based on daily ice velocity. Ice velocity can be derived from the gridded brightness temperatures acquired via the 36 GHz (December–April) and 18 GHz (May–November) channels of the AMSR-E and AMSR-2. The method of derivation is based on pattern matching, that is, the maximum cross-correlation technique (Kimura et al. 2013). A data set of daily ice velocity without missing data over the sea-ice area on a 60 × 60 km² grid was prepared for 2003–2018, except for September 2011 to July 2012, when there was an observational gap between AMSR-E and AMSR-2. Using this ice velocity data set, the distribution of multi-year ice was estimated by tracking the movement of virtual Lagrangian ice particles, as by Kimura et al. (2013). First, particles were arranged at 10-km intervals over the ice area on the day with the minimum ice cover for each year. The initial distribution of the particles changed annually depending on the interannual difference in the minimum ice area. Then, the displacements of the particles released on 15 September were calculated from the ice velocity in time steps of one day. The ice velocity at the position of each particle was interpolated from grids within 120 km, weighted by a Gaussian function of distance between the particle and the data grid positions. This procedure was used to estimate the temporal change of multi-year ice distribution from 15 September to the end of October the following year. Additionally, each particle includes information on ice concentration. The initial concentration of each particle sets the observed ice concentration on the initial day of minimum ice cover. When the observed ice concentration at the position of a particle was smaller than that of the particle, the concentration was reset to the observed concentration. If the ice concentration was <15%, the particle was deleted. Based on the multi-year ice concentration of each particle, the concentration of multi-year ice on a 60 × 60 km² grid was calculated as the sum of the multi-year ice concentration for particles within a circle of 240-km radius divided by the area of the circle. The multi-year ice data covered the period from 15 September 2013 to 31 October 2018, except for the data-gap period in 2011–12.

In addition to the information on sea ice derived from satellite remote sensing, ship-based observations of ice thickness were acquired between 6 September and 2 October 2018 during the JOIS 2018 cruises with the icebreaker CCGS Louis S. St-Laurent. Ice thickness was estimated hourly through visual inspection, which is an approach based on ASSIST observation protocol of the ASPeCT protocol (Worby & Alison 1999).

To estimate the atmospheric conditions, sea-level pressure was obtained on a 0.75° × 0.75° grid from the ERA-Interim reanalysis data set (Dee et al. 2011), provided by the European Centre for Medium-Range Weather Forecasts. Additionally, the daily geostrophic wind velocity was calculated from the sea level pressure using the following equation:

where f is the Coriolis parameter, ρ is density and P is the pressure.

The wind factor (ratio of ice drifting speed to wind speed) was calculated using a least squares technique based on the linear relationship (Thorndike & Colony 1982) between daily sea-ice velocity and geostrophic wind velocity, using the calculation method of Kimura & Wakatsuchi (2000).

The temporal and spatial changes of sea-ice cover and multi-year ice distribution from autumn 2017 to summer 2018 are shown in Fig. 2. The sea-ice area reached its maximum extent in March and then retreated rapidly during May–September. The multi-year ice distribution in the central Arctic shrank, leaving narrow branches, extended to the New Siberian Island and Severnaya Zemlya. Conversely, the area of multi-year ice expanded towards the Atlantic side along the north and west coasts of Greenland via the Transpolar Drift Stream and the East Greenland Current and towards the North American side (hereinafter the “Beaufort branch”), moving across the Beaufort Sea to the East Siberian Sea under the influence of the prevailing clockwise Beaufort Gyre. Although the Beaufort branch disappeared in summer in association with ice retreat, the blockage at the mouth of the AG occurred in September following the migration of the trunk of the Beaufort branch towards the AG. Multi-year ice parcels reached the Canadian coast by the end of September.

The multi-year ice distribution estimated by the particle tracking method agrees well with the observed area of thick ice (Fig. 3). This indicates that most of the ice blocking the mouth of the AG was multi-year ice with thickness >1 m.

To clarify the unusual ice behaviour in 2018, the interannual difference of the sea-ice cover and multi-year ice distribution on 30 September in each of the studied years is presented in Fig. 4. Additionally, Fig. 5 shows the interannual variation of total ice concentration and of multi-year ice concentration on 30 September in each of the studied within circles A and B depicted in Fig. 4. Area A (a few hundred kilometres from the mouth of the AG) was covered by sea ice in September almost once every four years because of the overhang of the Beaufort branch of multi-year ice. In those years, half the sea-ice cover in area A consisted of multi-year ice. However, it is evident that sea ice rarely covered area B, nearer to the AG, that is, the blockage event that occurred in September 2018 was the only case of ice cover in this area since 2003.

The blockage of the mouth of the AG was caused by southward movement of multi-year ice north of the Canadian Arctic Archipelago. The unusual pattern of the sea-ice distribution in summer 2018 was characterized by this southward movement of the Beaufort branch. To help visualize the causes of this behaviour, Fig. 6 presents the difference of ice concentration between the first and last days of each month, and the anomalies of sea-ice motion and sea-level pressure in June, July, August and September 2018, on the basis of the 2003–2018 climatology. During July–August in 2018, most of the marginal ice areas were characterized by negative anomalies in ice concentration, representing the relatively low Arctic-wide sea-ice coverage in these months relative to that in the 2003–2018 period. However, near the mouth of the AG, ice concentration was characterized by a slight positive anomaly. This increase was maintained by the AG-ward (eastward or southward) motion of the ice associated with the AG-ward wind anomaly. A strong low-pressure anomaly over the Arctic Ocean caused the strong AG-ward motion of ice in July. Although the centre of the low-pressure anomaly moved to the Canadian Arctic Archipelago, the AG-ward wind anomaly continued to push multi-year ice towards the AG in August. The largest anomaly of AG-ward ice motion occurred in September, associated with the combination of a low-pressure anomaly over the Canadian Arctic Archipelago and a high-pressure anomaly over Alaska. In addition, the direction of the ice vector anomaly shifted from eastward (along the continental coast) in July–August to south-eastward (towards the continent) in September. Thick multi-year ice reached the coast of the continent in September following this shift in wind direction. The ice concentration near the mouth of the AG increased over hundreds of kilometres during September. This increase can be explained by the anomalous ice motion in this area. The anomaly reached up to 10–15 cm/s, which is equal to the anomaly in the ice advection distance over 250–400 km during one month (30 days). Figure 7 presents the interannual change of AG-ward wind at the centre of area B in August and September. In both months, the AG-ward wind was dominant. In particular, the AG-ward wind in September 2018 was exceptionally strong compared with the other 13 years. This unusual onshore wind led to the AG-ward ice motion and the AG blockage in 2018. Notable increase in ice in September was also seen in the Greenland Sea, which can be explained by anomalous wind-driven ice drift. It is evident that the AG-ward wind is not related to the AO index (Fig. 7). The wind direction around the mouth of the AG is controlled by the local atmospheric circulation pattern rather than the Arctic-scale pattern reflected by the AO.

A notable anomaly of ice motion can be caused by a change in the mobility of sea ice in addition to a strong wind anomaly. The wind factor in 2018 was not extraordinary in comparison with the other recent years in the study period (2003–2017; Fig. 8). This means that the unusual motion of multi-year ice in 2018 was caused by the unusual wind field rather than by unusual mobility of the ice. However, with regard to the long-term trends, a previous study reported an increase of the wind factor (Kwok et al. 2013). The wind factor of multi-year ice has increased gradually over the past few decades (Fig. 8). Linear trends of the factor for the first-year ice and multi-year ice are 0.75 × 10⁻³ and 2.45 × 10⁻³%/year, respectively. The trend of the wind factor is not significant in terms of the area of first-year ice (99% confidence interval of the trend is −0.13 × 10⁻³ to 1.63 × 10⁻³%/year). However, the increasing trend of the factor is significant in terms of the area of multi-year ice (99% confidence interval is 1.67 × 10⁻³ to 3.23 × 10⁻³%/year).

This study demonstrates that unusual weather conditions over the Arctic Ocean are able to change the distribution of multi-year ice drastically. Blockage of the mouth of the AG by thick multi-year ice in summer 2018 occurred in two steps. First, the area of multi-year ice expanded to the North American side (Beaufort branch), where it remained from autumn 2017 to summer 2018. The appearance of the Beaufort branch overhanging towards the AG is not rare; it can be seen about once every four years. Second, rapid migration of the Beaufort branch towards the AG was propelled by the strong AG-ward wind, especially in September 2018.

Kwok et al. (2013) reported that the wind factor has an increasing trend during both the winter and summer. Our study showed the trend is clear only in the area of multi-year ice, not in the area of first-year ice. This is probably attributable to the thinning of the multi-year sea ice. Owing to the trend, rapid movement of multi-year ice and the blockage of the sea route by thick ice have become increasingly likely year after year.

The AG is an important region of the shipping route along the North-west Passage through the Canadian Arctic Archipelago. Satellite observations can monitor the distribution of multi-year ice such as the growth of the Beaufort branch, whose timescale is from a few months to a year. Such monitoring is useful for determining the potential hazards to maritime navigation. However, rapid movement of multi-year ice and the associated blockage of sea routes can occur on short timescales (less than one month). Therefore, monthly weather forecasts will be essential for predicting such motion and helping safeguard maritime ship navigation through the Arctic Ocean shipping routes.